Proceedings of the 4th eCAADe International Regional Workshop

4th eCAADe International Regional Workshop

Between Computational Models and Performative Capacities

Editors

Bojan Tepavčević

Vesna Stojaković

Department of Architecture, Faculty of Technical Sciences, University of Novi Sad, Novi Sad


Proceedings of the 4th eCAADe International Regional Workshop

www.arhns.uns.ac.rs/4-ecaade-workshop/conference/

May 19-20, 2016

Novi Sad, Serbia

ISBN 978-86-7892-807-9

Copyright © 2016

Publisher: eCAADe (Education and Research in Computer Aided Architectural Design in Europe) and Faculty of Technical Sciences, University of Novi Sad

72.012:004(082)

COBISS.SR-ID 305457671

eCAADe http://ecaade.org/

Department of Architecutre, FTN, UNS http://www.arhns.uns.ac.rs/

Digital Design Center, FTN, Novi Sad http://www.arhns.uns.ac.rs/cdd/

All rights reserved. Nothing from this publication may be reproduced, stored in computerised system or published in any form or

in any manner, including electronic, mechanical, reprographic or photographic, without prior written permission from the

publisher.

http://ecaade.org/

http://www.arhns.uns.ac.rs/

http://www.arhns.uns.ac.rs/cdd/

Organisation and sponsors of the eCAADe 2016 Conference

Education and research in Computer Aided Architectural Design in Europe

Digital Design Center, FTN, Novi Sad

Faculty of Technical Sciences, Novi Sad

Austrotherm d.o.o.

Evolute GmbH

Pinoles

Organizing committee:

Bojan Tepavčević

Vesna Stojaković

Ivana Bajšanski

Dejan Mitov

Marko Jovanović

Scientific Review Committee:

Joachim B. Kieferle, Hochschule RheinMain, Germany

Jose Duarte, Faculdade de Arquitetura, Universidade de Lisboa, Portugal

Tadeja Zupančič, University of Ljubljana Faculty of Architecture, Slovenia

Aleksander Asanowicz, Bialystok University of Technology, Poland

Marios Tsiliakos, Institute for Experimental Architecture-University of Innsbruck/Foster+Partners, UK

Daniel Lordick, TU Dresden, Germany

Tom Maver, Glasgow School of Art, UK

Milena Stavrić, TU Graz, Austria

Albert Wiltsche, TU Graz, Austria

Birgul Colakoglu, Istanbul Technical University, Turkey

Bob Martens, TU Wien, Austria

Henri Achten, Czech Technical University in Prague, Czech Republic

Gabrijel Wurzer, TU Wien, Austria

Michela Rossi, Politecnico di Milano, Italy

Radovan Štulić, University of Novi Sad, Serbia

Predrag Šiđanin, University of Novi Sad, Serbia

Ratko Obradović, University of Novi Sad, Serbia

Mirko Raković, University of Novi Sad, Serbia

Sonja Krasić, University of Niš, Serbia


Between Computational Models and Performative Capacities Proceedings of 4th International Regional Workshop on Education and research in Computer Aided Architectural Design in Europe May 19-20, 2016 Faculty of Technical Sciences Universtiy of Novi Sad Serbia http://www.arhns.uns.ac.rs/4-ecaade-workshop/ Conference topics: Computational design

Material research аnd digital fabrication

Adaptive architecture and building dynamics

Performance based design

Architectural geometry

Biomimetics and architecture

Procedural modeling and smart cities

Design education

Design theory and digital architecture

Edited by: Bojan Tepavčević Vesna Stojaković

http://www.arhns.uns.ac.rs/4-ecaade-workshop/

Theme:

Contemporary issues in architecture reside between designing complex shapes, increasing

performative capacities of building and efficient fabrication approaches. The new question for the

second digital turn in architecture lies on tangible relations between fabrication, computational design,

and performative capacities of building.

The aim of the 4th eCAADe international regional event is to promote the connection and the

exchange of ideas between the leading experts in the field of digital techniques applied to architecture

and the research groups interested in the relationship between performance based design, fabrication

process and advanced material design. The conference will be a place for testing and discussing

novel ideas and approaches regarding this topic.

Bojan Tepavčević

Vesna Stojaković

Keynote speaker

Matthias Rippmann

Matthias Rippmann studied architecture at the University of Stuttgart and the University of Melbourne

and received his Dipl.-Ing. degree in 2007. He conducted his doctoral research at the Block Research

Group, ETH Zurich, focusing on form finding and fabrication approaches for discrete funicular

structures. Matthias Rippmann developed multiple digital tools, allowing for new, informed design

approaches by incorporating structural and fabrication requirements in order to efficiently create and

realize expressive, nonstandard structures. In this context, his work includes the development of the

form-finding software RhinoVAULT, which was awarded first prize at the ALGODeQ software

competition in 2014. His research and design projects have been published internationally, leading to

invitations to lecture at renowned conferences, offices, symposia and universities around the globe.

His contribution to the ACADIA conference in 2013 was awarded the ‘Best Student Paper’,

recognizing his doctoral research on funicular design explorations.

Prior to his doctoral studies, Matthias Rippmann worked for Behnisch Architects, L.A.V.A and the

Institute for Lightweight Structures and Conceptual Design, before joining the office of Werner Sobek

in Stuttgart, where he was lead-developer of several digital optimization tools used in the planning

phase of the new main station S21 in Stuttgart. In 2010 he co-founded the architecture and

consultancy firm ROK – Rippmann Oesterle Knauss, which he actively co-led for five years. Matthias

Rippmann joined the NCCR Digital Fabrication in September 2015.

Shell Structures Revisited

Learning from the Gothic Master Builders

The Block Research Group develops new computational form-finding optimization tools that enable

architects and designers to explore structural form intuitively. The Gothic Master Builders balanced

stone to create daring cathedrals in structural equilibrium. The innovative graphical analysis

computational methods developed by the Block Research Group extend this historical knowledge to

the design of contemporary efficient shell structures. The applications range from sustainable

construction solutions for developing countries to unique unreinforced vaults in tile or cut stone. The

lecture will conclude with a discussion of how the lessons of the Master Builders can be applied to

design more effective and expressive structures well beyond masonry.

4th International Regional eCAADe Workshop, Novi Sad, May 2016

ISBN - 978-86-7892-807-9

8

CONTENT:

ASSESSMENT OF OPTIMAL SHAPE OF EXTERIOR SHADING FOR OFFICES IN HOT SUMMER CONTINENTAL CLIMATE ........................................................................... 10

BENEFITS OF THE ENVIRONMENTAL SIMULATIONS FOR THE URBAN PLANNING PROCESS ......................................................................................................... 24

GENETIC ALGORITHM APPLICATION AS A SUSTAINABLE URBAN PLANNING OPTIMIZATION TOOL .......................................................................................................... 31

DEVELOPMENT OF PHYSICAL CONTROL AND VIRTUAL NAVIGATION MECHANISM TO ENHANCE ACTIVE COLLABORATION BETWEEN ARCHITECTS AND CLIENTS IN A REACTIVE DIGITAL SPACE ....................................................................... 40

A DATA DRIVEN APPROACH TO LOCATING PROBLEM AREAS IN THE SMART GRID UNDER A DYNAMIC POPULATION ........................................................................... 50

REVISED GENERATIVE STRATEGY IN DESIGN OF ARCHITECTURAL STRUCTURES ..................................................................................................................... 55

IMAGE-BASED MODELING OF COMPLEX GEOMETRIC FORMS IN RESTRICTED SURVEYING CONDITIONS – A CASE STUDY OF THE COACH OF METROPOLITAN OF KARLOVCI IN THE MUSEUM OF VOJVODINA ................................................................... 62

THE APPLICATION OF DIGITAL TOOLS AND SMART MATERIALS IN THE CREATION OF ADAPTIVE SYSTEMS IN ARCHITECTURE ................................................ 75

DESIGNING AND FABRICATION OF ADAPTIVE FAÇADE BASED ON ORIGAMI PATTERN ............................................................................................................................. 81

COST-EFFICIENT APPROACHES IN FABRICATION OF STREET FURNITURE BASED ON SECTIONING DESIGN STRATEGIES .............................................................. 87

CONTEMPORARY METHODS FOR EXISTING BUILDINGS PRESENTATIONS USING MOBILE DEVICES, CASE STUDY OF TEMPLE ON VLASINA HIGHLAND ............ 93

INTERACTIVE GLARE VISUALIZATION MODEL FOR AN ARCHITECTURAL SPACE .................................................................................................................................. 97


ISBN - 978-86-7892-807-9

9

GEOMETRIC FRAMEWORK FOR THE EQUILIBRIUM ANALYSIS OF POINTED

ARCHES ACCORDING TO MILANKOVITCH’S THEORY OF THRUST LINE .................... 108

ARCHITECTURAL REPRESENTATIONS 1 - THE COURSE AT THE FACULTY OF TECHNICAL SCIENCES .................................................................................................... 117

3D MODELING COURSE AT THE COMPUTER GRAPHICS - ENGINEERING ANIMATION STUDIES ........................................................................................................ 124

DIGITAL FABRICATION STRATEGIES IN DESIGN EDUCATION ......................... 139


ISBN - 978-86-7892-807-9

10

ASSESSMENT OF OPTIMAL SHAPE OF EXTERIOR SHADING FOR OFFICES IN HOT SUMMER CONTINENTAL CLIMATE

Sanja Stevanović1*, Viktorija Eremeeva Naumoska2, Miško Ralev2 1University of Niš, Faculty of Sciences and Mathematics (SERBIA) and University of

Primorska, Institute Andrej Marušič (SLOVENIA), [email protected] 2University American College Skopje, School of Architecture and Design (MACEDONIA),

eremeeva @uacs.edu.mk, ralev @uacs.edu.mk

Abstract Window shading influences several important building loads: cooling and heating load through the control of solar heat gains, and lighting load through its influence on available daylight. Its optimal shape is therefore an important factor in analysis of building loads. A number of methods to determine optimal shading shape have been proposed in the literature so far. A large group of them are based on solar path projections and cut-off days and hours, starting with the first such method of Olgyay and Olgyay from 1957. These methods are based solely on solar paths and the underlying choices of when the window surface should be fully shaded or fully exposed, and they usually do not take actual building loads into account. Most of them, in addition, also assume that the optimal shading shape is rectilinear. Kaftan's cellular method and its offspring Shaderade are able to consider arbitrary forms as candidates for optimal shading shape by dividing the shading area into small cells, but due to calculations being performed for each cell independently of the others, they are not able to take lighting load into account.

These deficiencies suggest that in the process of determining optimal shading shape, the shading should be taken as a whole and that simulation of building loads should be performed for each considered shape. As the solar trajectory is curvilinear, it is natural to expect that the optimal shading shape will also be curvilinear. Since non-uniform rational basis spline (NURBS) is widely accepted standard for representation of curves, it is assumed here that the external shading consists of the overhang, the left and the right fin which are placed tightly around the window, but whose outer ends, which determine its shape, are representable as a NURBS curve. Simulations of building loads is performed by EnergyPlus by dividing the shading into trapezoidal segments whose outer ends approximate a NURBS curve. Optimization of shading form then translates to optimization of positions of curve's control points, which is done by genetic algorithm through jEPlus+EA. A cellular office located in Chicago is taken as a case study to demonstrate this approach, and clustering of the Pareto front with respect to heating, cooling and lighting loads helps to identify sets of optimal shading shapes.

Keywords: External shading, Genetic algorithms, Cooling load, Heating load, Lighting load.

1 INTRODUCTION Majority of energy in buildings is used to provide comfortable thermal conditions to their occupants. An effective way to passively reduce cooling load in a building is to provide exterior shading of its windows. The exterior shading, in the form of brise soleils, was popularized by Le Corbusier in the 1950s (see Fig. 1). His aim was to size brise soleils so that the windows are fully shaded at noon in summer, but receive full insolation at noon in winter in order to provide passive heating. This aim was put into theoretical framework by several researchers whose methods propose the optimal shading size using solar path projections and cut-off days and hours, during which complete shading of windows is required. Well known methods of this kind were proposed already by Olgyay and Olgyay in 1957 [1] and by Mazria in 1979 [2], although new solar path methods continue to appear to this date [3-6].

An important deficiency of these early methods is the underlying assumption that the optimal shading form is rectilinear. Solar path is curvilinear, so it is only natural to expect that the optimal shading form may also be curvilinear. Only two methods have been proposed so far that are able to consider arbitrary forms as candidates for optimal shading. The cellular method proposed by Kaftan [8] proceeds by dividing the shading support surface into two dimensional array of cells and by calculating for each cell the amount of beam solar gains it prevents, summed over time periods when shading in


ISBN - 978-86-7892-807-9

11

required. Building upon Kaftan's method, Sargent et al. [9] proposed Shaderade method that for each cell and each time period calculates the desired fraction of solar beam energy transmitted through the cell at that time period and then uses annual calculation to determine the optimal choice for each cell's transmittance value. A particular advantage of these approaches is that sorting the cells of support surface by calculated values allows to identify the areas that are most important for inclusion in the final shading design, and also to reduce the size of existing shading by excluding the areas that are least important. The final shading design still rests on the architect, as the optimal shapes suggested by these two methods tend to be rather serrated.

Fig. 1. Le Corbusier's principle of using brise-soleils at L'unité d'Habitation de Marseille (illustration taken from

[7]).

Another deficiency of solar path methods lies in the fact that they ignore nowadays readily available building energy simulation tools. With high insulation and tight sealing becoming standard in modern construction, heating loads are significantly diminished so that the original Le Corbusier's aim of full insolation in winter is no longer primary factor in the optimal shading design. Furthermore, the magnitude of heating load often becomes similar to that of the lighting load in office buildings, due to both the minimal illuminance level requirements and high internal gains from people and office equipment. Thus, any up-to-date method for optimal shading design necessarily has to take into consideration all of the heating, cooling and lighting loads. In this aspect, Kaftan and Marsh [10] had combined cellular method with Ecotect in order to predict necessity of shading based on both the beam solar gains and thermal comfort indicators. Calculations of desired cell transmittances in Shaderade [9] already include EnergyPlus predictions of heating and cooling loads, but only for the base case of a building without any shading. However, neither of these methods is able to take lighting load into account, due to the inherent construction of optimal shading as a collection of single shading cells, each of which has imponderable influence on the lighting load. In addition to building energy simulation tools, contemporary building studies often rely on the use of genetic algorithms to reach solutions that are close to optimal. Among such shading studies, Castorina [11] encodes the whole shading façade with a particle-spring system and then employs genetic algorithms and EnergyPlus to optimize a single objective function that represents weighted combination of illuminance ratio, lighting load, and the ratio of winter and summer solar gains. Ercan and Elias-Ozkan [12] also consider the whole façade shading by encoding the depth and angle of each shading device and use genetic algorithms and Radiance to optimize daylighting levels, but do not consider interior thermal conditions. On the other hand, Manzan [13] relies on the daylighting tool DAYSIM to estimate lighting load and ESP-r to calculate heating and cooling loads, and then employs genetic algorithm to determine an optimal angle and reveal of a full width rectangular overhang.


ISBN - 978-86-7892-807-9

12

The present paper considers the shape of external shading for south oriented windows in the setting of office buildings. In order to accommodate curvilinearity of optimal shading shape it is assumed that the shading consists of an overhang, eastern and western fin, tightly places around the window, whose outer edges are modelled with a continuous NURBS curve. A genetic algorithm is then employed to optimize NURBS control points' parameters with respect to heating, cooling and lighting loads as objective functions, that are calculated in EnergyPlus. The obtained Pareto solutions are clustered in order to identify groups of shading shapes yielding similar heating, cooling and lighting loads. More details about methodology of this approach are given in the next section. Computational results for the case study of a cellular office located in hot summer continental climate of Chicago are presented in Section 3, and discussed in Section 4.

2 METHODOLOGY

2.1 Building model The case study is undertaken for a single south-facing cellular office in a large office building located in the U.S. city of Chicago (41.98°N, 87.92°W, elevation 201m). In order to facilitate comparison to other studies, the building model is taken from DOE's Commercial Prototype Building Models [14], which cover 80% of the commercial building floor area in the United States for new construction. Materials, constructions and schedules follow the ASHRAE Standard 90.1-2013 version of [14].

Cellular office has dimensions 3m x 4m x 3.5m (w x d x h) and a 2m x 1.4m window on the south side. Its other walls, floor and ceiling are treated as adiabatic surfaces, as it is assumed that studied office is adjacent to similar cellular offices and conditioned corridor. Heating and cooling loads are calculated as district heating and district cooling models in EnergyPlus, with all HVAC equipment removed from the DOE building model. Namely, Pareto solutions for heating, cooling and lighting loads do not depend that much on the exact values of energy use as they do depend on their relative ordering, so that EnergyPlus simulations can be accelerated by calculating district heating and cooling loads only. This is further justified by the fact that EnergyPlus reports a large number of warnings related to HVAC equipment included in the original building model provided by DOE [14].

Other simulation related data are shown in Table 1. To save computation time, EnergyPlus by default recalculates the solar path and its shadowing and daylighting models every 20 days during the year. As the optimal shading shape is obviously directly related to accurate calculation of solar paths, EnergyPlus was instructed to recalculate solar path every single day of the year by setting shadow calculation method in ShadowCalculation object to TimeStepFrequency (important note: this method is wrongly named DetailedTimeStepIntegration in EnergyPlus documentation).

Table 1. Simulation properties of the building model.

Weather file USA_IL_Chicago-OHare.Intl.AP.725300_TMY3.epw Solar distribution FullInteriorAndExterior Shadow calculation method TimestepFrequency Shadow calculation frequency 1 day Constructions ASHRAE 90.1-2013 Natural ventilation Off Cooling setpoint 25°C, Monday-Friday, 6am-10pm Heating setpoint 20°C, Monday-Friday, 6am-10pm, setback point is 15.6°C People activity level 120 W/person Electric equipment gains 100 W/person Lighting load 8.83W/m2 Daylighting illuminance setpoint 375 lux

2.2 Shading geometry Exterior shading of the office window consists of an overhang, eastern and western fin that are placed 0.1m apart from the window edge, so that the fins have height 1.5m, while the overhang is 2.2m wide.


ISBN - 978-86-7892-807-9

13

The outer edge, which defines the shading shape, is modelled as a single NURBS curve as if the fins were rotated for 90° each in order to form 5.2m wide horizontal surface together with the overhang.

NURBS, short for non-uniform rational basis spline, is dominant mathematical technique used for generating and representing curves and surfaces, commonly used in computer-aided design, engineering and manufacturing [15]. NURBS curve C(t), 0 ≤ t ≤ 1, is defined by a series of control points P0, ..., Pk, which act as if they were connected to the curve by springs, weights (spring strengths) w0, ..., wk, and suitably calculated basis functions N0(t), ..., Nk(t):

𝐶(𝑡) =∑ 𝑃𝑖𝑤𝑖𝑁𝑖(𝑡)𝑘𝑖=0

∑ 𝑤𝑖𝑁𝑖(𝑡)𝑘𝑖=0

.

Seven control points P0, ..., P6 are used to define the outer edge in this study. NURBS curve is of order four, indicating that each point on the curve is influenced by four control points closest to it. Basis functions are cubic polynomials obtained from the clamped uniform knot vector (0, 0, 0, 0, 0.25, 0.5, 0.75, 1, 1, 1, 1), which ensures that the starting and ending points of C(t) coincide with the control points P0 and P6, respectively. The x-coordinates of P2 and P4 are placed at the joints of the overhang with the eastern and the western fin, respectively. The x-coordinates of P1, P3 and P5 are placed in the middle of the eastern fin, the overhang, and the western fin, respectively. The y-coordinates of control points represent their distances from the south wall, but it should be noted that these values do not represent actual shading depths, as the interior control points P1, ..., P5 do not necessarily belong to the curve C(t) (see Fig. 2). In order to keep the number of potential curves at a reasonable size, the y-coordinates of control points are confined to the set of alternatives {0, 0.25, 0.5, 0.75, 1, 1.25, 1.5, 1.75, 2}, so that the search space contains 97 =4,782,969 possible curves. Weights of all control points are set to 1.

Despite the usefulness of NURBS in design, EnergyPlus can still model building geometry using rectilinear curves and surfaces only. For this reason, exterior shading is approximated in the building model via a series of 20 trapezoids, whose outer vertices are given by the curve points C(0), C(0.5), ..., C(0.95), C(1), respectively, with the right and the left fin folded back to vertical direction afterwards (see Fig. 2). Calculation of these curve points was hard coded in the building .imf model using EP-Macro language.

Fig. 2. Left: Cubic NURBS curve with seven control points, whose y-coordinates from right to left are: 0.25, 0, 2, 2,

1.5, 1.25, 0.25. Right: Illustration of exterior shading with the same NURBS curve used to define its outer edge, approximated in EnergyPlus with 20 trapezoidal segments.

2.3 Optimization process Optimal design of building components and systems often has to satisfy multiple competing objectives, such as simultaneous minimization of energy consumption, financial costs or environmental impact [16]. As a consequence, candidates for the optimal design are usually very different from each other. The Pareto front consists of those designs for which no other design is better at all objectives simultaneously. Thus, every non-Pareto design is outperformed by some design in the Pareto front. However, for any two Pareto designs holds that if some objective is better met in


ISBN - 978-86-7892-807-9

14

one design, then another objective must be better met in the other design. Hence Pareto front may be used to reveal the trade-off between competing performance objectives, and may serve to guide the designer in selecting proper candidate designs for the problem at hand.

Coupling of building energy simulation tools with optimization methods has become mainstream in the study of energy and buildings in last ten years. Stevanović [16] reviews more than 100 such papers on passive solar design of buildings, while Machairas et al. [17] review different methods used in optimization of building design. Genetic algorithms are mostly used among them. They gained popularity in this community after Caldas and Norford [18] used them to facilitate performance-based façade design. As their name suggests, genetic algorithms are inspired by biological evolution. Population consists of candidate solutions for the optimization problem, with objective functions representing fitness of the candidates. Population then evolves over a number of generations by repeated application of selection, reproduction, mutation and recombination, with the goal of improving objective functions within each generation. The main advantage of genetic algorithms, when applied to a multiobjective optimization problem, is that generations over time approximate the entire Pareto front.

Non-dominated sorting genetic algorithm-II (NSGA-II) is the most well known variant of genetic algorithms, and also used in the current study through its implementation in jEPlus+EA [19]. Parametric study has been prepared in jEPlus by describing the design search space, which consists of alternatives for the y-coordinates of control points P0, ..., P6. The control point positions are used in the building model's .imf file to construct trapezoidal approximation of the NURBS curved shading with EP-Macro commands. NSGA-II was run in jEPlus+EA for 100 generations on a population containing 50 designs with three objective functions being district heating, district cooling and lighting loads. Due to the presence of three objectives, the Pareto front represents a three-dimensional surface, approximated by a large number of candidate designs. In order to cope with large number of Pareto designs, they were further clustered into groups similar with respect to their defining parameters using the agglomerative nesting with Ward's minimum variance method [20]. Agglomerative nesting begins by putting each Pareto design in a separate cluster. It then proceeds by a series of mergings, where at each step two minimally dissimilar clusters are merged into a larger cluster.

3 RESULTS A single EnergyPlus simulation of the cellular office model with shading took between 5-6 seconds on a personal notebook Fujitsu S935, and jEPlus+EA completed running NSGA-II on a population of 50 designs for 100 generations in slightly less than two hours by running four EnergyPlus simulations in parallel.

Fig. 3 shows diagrams of heating, cooling and lighting loads for all considered designs, with red dots indicating 642 Pareto designs, and blue dots indicating non-Pareto designs in the last generation. Note that the Pareto front is three-dimensional, so that Pareto designs do not necessarily appear on the boundary of projected two-dimensional diagrams in Fig. 3.

Fig. 4 shows histograms of frequencies of the y-coordinates of NURBS control points in Pareto designs. Histograms show that Pareto designs tend to have y-coordinates of control points P0, P1, P5 and P6 up to 0.5, and have larger variety of choices for y-coordinates of control points P2, P3, and P4. In order to observe relationship between adjacent control points, Fig. 5 shows color scaled histograms of frequencies of co-appearances of y-coordinates for adjacent control points in Pareto designs.

As evident from diagrams in Fig. 3, Pareto designs span the whole range of heating, cooling and lighting loads. While Pareto designs are defined as such designs for which no other design has simultaneously smaller heating, cooling and lighting loads, this is far from stating that all Pareto designs may be considered as being close to optimal. In addition, there are many Pareto designs that are very similar to each other, and which consequently have very similar loads. This similarity is best revealed through clustering of Pareto designs. Clustering is created by agglomerative nesting with Ward's minimum variance method, using the online interface [20] to statistical R modules. Dendrogram of division into clusters is shown in Fig. 6, together with Pareto designs selected to represent eight largest clusters. The agglomerative coefficient [21] of obtained clustering is 0.9778, indicating a clear and natural structuring. Details of selected representative Pareto designs are given in Table 2, including their shape, y-coordinates of control points, cooling, heating and lighting loads, and equivalent source energy. The equivalent source energy is obtained by multiplying cooling, heating and lighting loads with the respective site-to-source conversion factors of 1.106 for district cooling, 3.640 for district heating and 3.317 for electricity, as reported in [22]. It is evident from Table 2


ISBN - 978-86-7892-807-9

15

that the second cluster contains Pareto designs with minimum equivalent source energy. Details of the 12 designs with minimum equivalent source energy, out of 96 Pareto designs that are contained in this cluster, are given in Table 3.

(a)

(b)

(c)

Fig. 3. Cooling, heating and lighting loads for designs constructed by NSGA-II: (a) cooling-lighting projection; (b) heating-lighting projection; (c) heating-cooling projection. Red dots depict Pareto designs, while blue dots depict

the last generated population.


ISBN - 978-86-7892-807-9

16

Fig. 4. Histograms of frequencies of parameter values in Pareto designs. Histograms are ordered in the way the

corresponding control points appear in the office model when viewed from front.

Fig. 5. Histograms of value pairs appearing as two consecutive parameters in Pareto designs. Darker shades

indicate higher frequencies of co-appearance.


ISBN - 978-86-7892-807-9

17

4 DISCUSSION Diagram in Fig. 3(c) clearly indicates that the heating and the cooling loads are competing objectives, as expected. Diagrams in Figs. 3(a) and 3(b), on the other hand, show that the lighting load of Pareto designs is distributed in isolated levels, with a very little change in the lighting load among designs within the same level. Closer inspection reveals that within each of the levels Pareto designs have similar values of the y-coordinates of control points P3 and P4. These control points largely influence the overhang area, and hence the amount of daylight in the office. Relative permanency of the lighting load within each level further confirms that it is mostly governed by the overhang area, and not much by the actual shape of the overhang. The distribution of Pareto designs in separate levels is then merely a consequence of discrete choices made for the y-coordinates of control points which change in steps of 0.25m. Such choices then cluster Pareto designs in groups with relatively similar overhang area, and hence relatively similar lighting load.

Histograms in Fig. 4 show that Pareto designs tend to have small values of y-coordinates of control points which appear on the eastern and western fins only. At the eastern fin, most Pareto designs have the y-coordinate of the bottom control point P0 in the range 0m-0.5m and that of the control point P1 in the range 0m-0.75m. At the western fin, most Pareto designs have the y-coordinates of the control points P5 and P6 in the range (0m ,0.75m). Clear preference for smaller values 0m-0.25m is evident for both the eastern and the western fin. Histograms in Fig. 4, however, do not reveal much information about the overhang control points. This is especially true for the control point P3, which has its possible value well dispersed within the range (0.5m, 1.5m).

Certain insight into the behavior of the overhang control points may be obtained from the color-scaled histograms in Fig. 5. These histograms show that Pareto designs tend to be split into two basic groups. After a clear preference for small values of the y-coordinates of control points P0, P1 and P2, the y-coordinate of P3 tends to be chosen either in the range (0.5m, 0.75m) or in the range (1.25m, 1.5m). The choice of P3 in the lower range (0.5m, 0.75m) is mostly followed by the choice of the y-coordinate of P4 in the range (0.25m, 0.75m). The choice of P3 in the upper range (1.25m, 1.5m) is mostly followed by the choice of the y-coordinate of P4 in the range (0.75m, 1m). Afterwards, a clear preference for small values of the y-coordinates of P5 and P6 becomes evident again. Pareto designs with P3 and P4 in the lower ranges thus have smaller overhang area, and consequently smaller heating and lighting loads. On the other hand, Pareto designs with P3 and P4 in the upper ranges have larger overhang areas, and consequently smaller cooling loads. It should be noted here that the control point P4 usually has larger y-coordinate P2, so that the western side of the overhang in Pareto designs usually is more pronounced than its eastern side.

Although justified by the color-scaled histograms in Fig. 5, division of the whopping number of 642 Pareto designs in just two groups necessarily has to be rather rough. More refined classification is obtained by agglomerative nesting with respect to similarities among y-coordinates of corresponding control points in different Pareto designs. Agglomerative nesting creates hierarchy from bottom to top, as shown in Fig. 6 on the right. It does not prescribe division of Pareto designs into any predefined number of clusters, as each level in the obtained hierarchy is naturally divided into two sublevels. Actual selection of clusters thus rests with the researcher, and for the purpose of this study eight clusters have been selected as indicated in Fig. 6. As each of these clusters still contains from 50-100 Pareto designs, a representative design has been selected in each cluster (shown in Fig. 6 on the left). Most representative designs have slightly more pronounced western part of the shading, apparently to reduce cooling load from the afternoon sun.

As already mentioned earlier, while optimal designs are necessarily Pareto designs, the opposite does not hold and there are many Pareto designs which are far from being optimal. This is easily visible by comparing with the base case of cellular office without shading, whose heating, cooling and lighting loads are given in the last row of Table 2. As expected, the base case without shading has lower heating load, higher cooling load and lower lighting load than any design with shading. However, when converted into equivalent source energy, it turns out that six of the eight clusters of Pareto designs actually need more equivalent source energy than the base case without shading. This clearly illustrates the problem of excessive shading, present in many current shading methods [9]. While larger shading may reduce cooling load up to 43%, as evident from Table 2, this is not sufficient to cover increases in heating and lighting loads, as the conversion factor for district cooling is much lower than the conversion factors for heating and lighting loads.


ISBN - 978-86-7892-807-9

18

Fig. 6. Agglomerative nesting of Pareto solutions with agglomerative coefficient 0.9778. Selected cluster representatives are drawn on the left side.


ISBN - 978-86-7892-807-9

19

Table 2. Representatives of the main clusters of Pareto designs. The base case without shading is shown in the last row.

Shading shape P0, y (m)

P1, y (m)

P2, y (m)

P3, y (m)

P4, y (m)

P5, y (m)

P6, y (m)

Heating load (GJ)

Cooling load (GJ)

Lighting load (GJ)

Source energy (GJ)

0.25 0.25 0.5 0.5 0.5 0 0 1.783 1.199 0.3190 8.876

0 0 0.25 0.5 0.25 0 0 1.721 1.357 0.3167 8.816

0.25 0.25 0.5 0.75 0.5 0.5 0 1.818 1.113 0.3888 9.140

0.25 0.25 0.5 0.5 1 0.25 0.25 1.829 1.105 0.3684 9.101

0.25 0.5 0.75 1.5 1 0 0.5 1.970 0.9584 0.4316 9.661

0.5 0.25 0.25 1.5 1 0.25 0.25 1.930 0.9869 0.4022 9.449

0 0.25 0.25 1.25 0.25 0.25 0.25 1.829 1.095 0.3747 9.110

0 0.25 0.5 1.25 0.75 0.5 0.25 1.892 1.019 0.4263 9.428

0 0 0 0 0 0 0 1.661 1.692 0.3131 8.956


ISBN - 978-86-7892-807-9

20

Table 3. Twelve best Pareto designs from the cluster with minimum equivalent source energy.

Shading shape P0, y (m)

P1, y (m)

P2, y (m)

P3, y (m)

P4, y (m)

P5, y (m)

P6, y (m)

Heating load (GJ)

Cooling load (GJ)

Lighting load (GJ)

Source energy (GJ)

0 0 0.25 0.5 0.25 0 0 1.721 1.357 0.3167 8.816

0 0 0.25 0.5 0.5 0 0 1.740 1.295 0.3172 8.817

0 0 0 0.5 0.5 0 0 1.722 1.357 0.3165 8.820

0 0 0 0.5 0.25 0 0 1.704 1.421 0.3153 8.821

0 0 0.25 0.5 0.25 0.25 0 1.731 1.329 0.3170 8.823

0 0 0.25 0.5 0.25 0 0.25 1.726 1.346 0.3169 8.823

0 0 0.5 0.25 0.5 0 0 1.730 1.339 0.3161 8.826

0 0 0.25 0.25 0.5 0 0 1.710 1.406 0.3153 8.826

0 0 0.25 0.25 0.75 0 0 1.729 1.343 0.3164 8.827

0 0 0.25 0.5 0.5 0 0.25 1.746 1.283 0.3175 8.827

0 0 0 0.5 0.25 0 0.25 1.709 1.410 0.3154 8.827

0.25 0 0.25 0.5 0.5 0 0 1.745 1.285 0.3176 8.828


ISBN - 978-86-7892-807-9

21

On the other hand, the first two representative designs in Table 2 point out that even the shading of substantially smaller area is able to reduce cooling load by 20-29%, mainly by blocking high midday sun during the summer. These smaller shadings do not increase heating and lighting loads too much, and manage to offer real improvement over the base case without shading, although by relatively humble 0.89-1.56% of the equivalent source energy.

The second representative design in Table 2 has been actually chosen as the Pareto design with the minimum equivalent source energy needs. It represents a cluster with 95 designs, while the first representative design in Table 2 represents a cluster with 76 designs. A total of 130 of these designs have smaller equivalent source energy needs than the base case without shading. Majority of these 130 designs have the y-coordinates of the fin control points P0, P1, P5 and P6 in the range (0m, 0.25m), and the overhang control points P2 in the range (0m, 0.5m), P3 in the range (0.25m, 0.75m), and P4 in the range (0.25m, 0.5m). Table 3 lists shapes and parameters of the 12 Pareto designs with smallest equivalent source energy. Slight prevalence towards more pronounced western part of the shading is visible in this table as well, although some of the designs (and in particular the optimal design) happen to be symmetric.

5 CONCLUSIONS The current study presents the methodology of searching for the optimal shape of fixed exterior shading for cellular office window in the form of a NURBS curve. Coupling of the genetic algorithm NSGA-II with EnergyPlus, offered through jEPlus+EA [19], enables to drive the search for optimal shape entirely by formal parameters, as envisaged earlier by Sargent et al. [9]. Unlike earlier cellular method [8] and Shaderade [9], the present approach is able to take into account lighting load as well.

Representation of the outer edge of shade by a NURBS curve enables the search space to be formally defined by positions of the curve control points. In addition to guaranteeing smooth design of the outer edge of shading, this approach enables good control of the size of the search space (4,782,969 feasible curves in the present study). Instead of returning a single design, the optimization process with respect to heating, cooling and lighting loads produces a collection of candidate Pareto designs. The actual optimal design is then selected among Pareto designs with respect to equivalent source energy.

The case study is performed here for the hot summer continental climate of Chicago, with building construction and schedules set to DOE's Commercial Prototype Building Model [14]. Analysis of produced Pareto designs suggests that the lighting load is less sensitive to the actual shape of the shading, but that it is mainly governed by the surface area of the overhang. Conversion to equivalent source energy further shows that there is a clear danger of excessive shading in considered climate, as Pareto designs with larger shading surface area need more equivalent source energy than the base case without shading. It turns out, however, that even shadings with small surface area may decrease cooling load by 20-29% and improve the base case without shading by 0.89-1.56% of source energy equivalent to total heating, cooling and lighting load. A slight tendency is observed among Pareto designs for the western part of the shading to become more pronounced, serving to block afternoon solar gains when the interior temperature has already risen. Nevertheless, the optimal shading shape in the case study turns out to be symmetric with the overhang control points protruding 0.5m at the middle and 0.25m at the ends, and the western and the eastern fin control points protruding 0.25m at the middles and 0m at their ends, creating a particular wavy design that offers best compromise of heating, cooling and lighting loads.

Further research has to be done to compare optimal shading shape obtained by presented methodology to those created by earlier methods, for this and other common climate types.

ACKNOWLEDGMENTS The author was supported by the research grant TR36035 "Spatial, environmental, energy and social aspects of developing settlements and climate change—mutual impacts" of the Serbian Ministry of Education, Science and Technological Development.


ISBN - 978-86-7892-807-9

22

NOMENCLATURE C(t) NURBS curve defining the shading's outer edge, 0 ≤ t ≤ 1

P0, ..., P6 Control points (y-coordinates) of the curve C(t), m

REFERENCES [1] Olgyay A., Olgyay V., Solar control and shading devices. New Jersey, USA: Princeton University Press; 1957.

[2] Mazria E., The Passive Solar Energy Book. Emmaus, Pensylvania: Rodale Press; 1979.

[3] Arumi-Noe, F., Algorithm for the geometric construction of an optimum shading device. Automation in Construction (1996);5:211-217. [4] Dubois M.-C., A simple chart to design shading devices considering the window solar angle dependent properties. In: Proceedings of EUROSUN 2000: The 3rd ISES Europe Solar Congres; 2000 Jun 19-22; Copenhagen, Denmark.

[5] Marsh A., Computer-optimised shading design. In: Proceedings of Building Simulation 2003: 8th International IBPSA Conference; 2003 Aug 11-14; Eindhoven, Netherlands. IBPSA: 831-837.

[6] Cheung H.D., Chung T.M., Analyzing sunlight duration and optimum shading using a sky map. Building and Environment (2007); 42:3138-3148.

[7] Bösiger W., Le Corbusier, Oeuvres Complètes 1938-1946. Zürich: Grisberger; 1946.

[8] Kaftan E., The cellular method to design energy efficient shading form to accommodate the dynamic characteristics of climate. In: Pereira F.O.R., Rüther R., Souza R.V.G., Afonso S., da Cunha Neto J.A.B., editors. Proceedings of the PLEA 2001 Conference: The 18th international conference on passive and low energy architecture; 2001 Nov 7-9; Florianopolis, Brazil; PLEA: 829-833.

[9] Sargent J.A., Niemasz J., Reinhart C.F., Shaderade: combining Rhinoceros and Energyplus for the design of static exterior shading devices. In: Proceedings of Building Simulation 2011: 12th Conference of International Building Performance Simulation Association; 2011 Nov 14-16; Sydney, Australia. IBPSA: 310-317.

[10] Kaftan E., Marsh A., Integrating the cellular method for shading design with a thermal simulation. In: Santamouris M., editor. Proceedings of the international conference “Passive and low energy cooling for the built environment”; 2005 May 19-21; Santorini, Greece; Heliotopos Conferences: 965-970.

[11] Castorina G., Performative topologies: an evolutionary shape optimization framework for daylighting performance coupling a particle-spring system with an energy simulation tool. In: Johnson J.K., Cabrinha M., Steinfeld K., editors. ACADIA 2012. Synthetic digital ecologies: Proceedings of the 32nd annual conference of the Association for Computer Aided Design in Architecture; 2012 Oct 18-21; San Francisco, California. Riverside Architectural Press: 479-490.

[12] Ercan B., Elias-Ozkan S.T., Performance-based parametric design explorations: A method for generating appropriate building components. Design Studies (2015); 38:33-53.

[13] Manzan M., Genetic optimization of external fixed shading devices. Energy and Buildings (2014); 72:431-440.

[14] US Department of Energy. Building Energy, Codes Program, Commercial Prototype Building Models. Available at <https://www.energycodes.gov/commercial-prototype-building-models> [Accessed 17.2.2016].

[15] Wikipedia. Non-uniform rational B-spline. Available at <https://en.wikipedia.org/wiki/Non-uniform_rational_B-spline> [Accessed 18.2.2016].

[16] Stevanović S., Optimization of passive solar design strategies: A review. Renewable and Sustainable Energy Reviews (2013); 25: 177-196.

[17] Machairas V., Tsangrassoulis A., Axarli K., Algorithms for optimization of building design: A review. Renewable and Sustainable Energy Reviews (2014); 31:101-112.


ISBN - 978-86-7892-807-9

23

[18] Caldas L.G., Norford L.K., Genetic algorithms for optimization of building envelopes and the design and control of HVAC systems. Journal of Solar Energy Engineering (2003); 125:343-351.

[19] Zhang Y., Korolija I., jEPlus - An EnergyPlus simulation manager for parametrics. Available at <http://www.jeplus.org/> [Accessed 18.2.2016].

[20] Wessa P., Office for Research Development and Education. Agglomerative Nesting (v1.0.3) in Free Statistics Software (v1.1.23-r7). Available at <http://www.wessa.net/rwasp_agglomerative hierarchicalclustering.wasp/> [Accessed 14.2.2016].

[21] Kaufman L., Rousseeuw P.J., Finding Groups in Data: An Introduction to Cluster Analysis. Hoboken, New Jersey: John Wiley & Sons; 1990.

[22] Deru, M., Torcellini, P., Source Energy and Emission Factors for Energy Use in Buildings, Technical Report NREL/TP-550-38617, Golden, Colorado: National Renewable Energy Laboratory, 2007.


ISBN - 978-86-7892-807-9

24

BENEFITS OF THE ENVIRONMENTAL SIMULATIONS FOR THE URBAN PLANNING PROCESS

Dragan Milošević1, Ivana Bajšanski2, Stevan Savić3, Igor Žiberna4 1Climatology and Hydrology Research Centre (SERBIA), [email protected]

2Department of Architecture (SERBIA), [email protected] 3Climatology and Hydrology Research Centre (SERBIA), [email protected]

4Department of Geography (SLOVENIA), [email protected]

Abstract Urban areas are exposed to the extreme temperatures and contribute to the creation of outdoor thermal discomfort phenomenon. Due to this, performance based design tools can be used for the simulations of outdoor thermal comfort during the urban planning process. Parametric approach can provide mitigation strategies for the thermal discomfort in urban areas. This paper introduces environmental analysis software tools that enable the interaction between 3D model, parametric design, and outdoor thermal comfort simulations. A parametric model that generates an optimal buildings heights arrangement is created in order to reduce the thermal discomfort in investigated urban areas. Buildings bases and footways were referenced and parameterised in the software. Furthermore, weather data of any city can be used in order to provide adaptability of the parametric model. The results showed direct impact of changed urban design on outdoor thermal comfort during extreme temperature events. Effective method of decreasing thermal discomfort in urban fabrics is by shadowing with buildings. The used parametric approach could support climate-conscious urban planning in cities and results can help in the creation of urban design guidelines in order to create climatically comfortable outdoor urban places.

Keywords: parametric procedures, outdoor thermal comfort, simulations, urban planning process.

1 INTRODUCTION Performance based design of the cities is becoming very important area of research and application. Rapid urbanization and extreme temperatures amplify the need for analysing and evaluating outdoor thermal comfort in open urban spaces. Geometrical characteristics of the built environment, position and orientation of buildings influence thermal comfort and have an effect on the urban planning and design of urban areas [1-4]. Buildings shade is the most effective strategy of decreasing thermal discomfort in open urban spaces [5-6]. Few limitations of the recent scientific methods for urban design modification for improvement of outdoor thermal comfort have been noticed. The software for outdoor thermal comfort simulations cannot use 3D models of the built environment with complex geometrical characteristics. Furthermore, its application is possible only for the stationary thermal comfort conditions of a single person [7] in small urban areas [8]. Integration of Rhinoceros and Grasshopper software applications enables fast processing and powerful modelling capabilities of performance simulation results [9]. LadyBug software enables the user to explore and examine direct relationship between elements of 3D model and environmental data through numerical and graphical data [10].

A complete assessment of the thermal environment can be described through a thermal index such as the Universal Thermal Climate Index (UTCI). The index was developed under ISB (International Society of Biometeorology) Commission 6 by COST (European Cooperation in Science and Technology) Action 730 under the umbrella of the WMO (World Meteorological Organization) Commission on Climatology [11]. UTCI application is particularly significant in the field of urban and regional planning [11]. In the study of Blazejczyk [12], UTCI was compared with some of the more prevalent thermal indices. It was found that present indices express bioclimatic conditions reasonably only under specific meteorological situations, while the UTCI represents specific climates, weather, and locations much better. UTCI worked well in the analysis of urban outdoor thermal environments in Canada and Korea [13] and for subtropical climate of southern Brazil [14]. Integration of UTCI calculation in Grasshopper enables the cooperation among architects, urban planners, urban bio- meteorologists and climatologists on urban planning and street design issues. Together they can create environmentally conscious architectural design that can improve outdoor thermal comfort

mailto:[email protected]


ISBN - 978-86-7892-807-9

25

conditions in cities. In order to make in Grasshopper software a non-stationary model, automated procedure for UTCI calculation is developed and presented in this paper.

The influence of urban morphological characteristics and geometry of buildings on outdoor thermal sensation is evaluated in different cities with different climate conditions [15,16]. Level of thermal sensation is significant parameter which have influence on quality of life in open urban spaces [17]. The height of buildings has a great impact on outdoor comfort condition because the buildings shade can mitigate temperatures of urban areas [18-19].

The aim of this paper is to emphasize the importance of performance based design tools application for the simulations and calculation of outdoor thermal comfort conditions during the urban planning process. For that purpose, a parametric model for evaluation and improvement of outdoor thermal comfort conditions have been created and tested on complex urban environment. The method for improvement of outdoor thermal comfort conditions in open urban spaces is based on the automatic algorithm that creates variations of buildings heights in order to obtain urban design with appropriate heights of buildings. Furthermore, considering factors such as meteorological or climate data, streets orientation, heights and arrangements of surrounding buildings, the algorithm can be used for any urban environment at any location.

2 PARAMETRIC PROCEDURES

2.1 Outdoor thermal comfort calculations in the LadyBug software In this section a method for UTCI calculation is presented. The calculation was performed in the LadyBug software application for environmental analysis. LadyBug is a plug- in of the Grasshopper (graphical algorithm editor) software for generating parametric procedures. Grasshopper uses geometrical parameters from Rhinoceros software for modelling of elements of built environment. With the usage of the three above-mentioned software applications, thermal comfort analysis was made possible.

Meteorological data was imported into the UTCI calculation process as an Energy Plus Weather file (.epw file). Meteorological data file can be downloaded from the internet as Energy Plus Weather Data (.epw file) [20] or can be created in Ecotect software using Convert Weather Data tool. The algorithm for UTCI calculation allows the user to select and input the period of the analysis (day, month and year). The next phase refers to the Solar Adjusted Mean Radiant Temperature (SAMRT) calculation using the Outdoor Solar Temperature Adjustor component in Ladybug within Grasshopper. The hourly SAMRT values were calculated for a predetermined number of body locations in urban area by considering following input data:

• the location data (latitude, longitude and altitude),

• air temperature (T),

• relative humidity (RH),

• wind speed (v) and

• solar radiation for each hour of the year using GenCumulative sky matrix component (generating an annual Perez sky matrix from the weather data) (http://www.radiance- online.org/learning/documentation/manual-pages/pdfs/gendaymtx.pdf).

The calculation also includes:

• the body posture (standing or sitting),

• the rotation angle of the body (from 0 to 360 °),

• context shading (3D model of built environment),

• ground reflectivity (between 0 and 1),

• clothing absorptivity (between 0 and 1) and

• the body location.

The next phase refers to the UTCI calculation (expressed in °C) at a predetermined manikin positions based on T, RH and v, as imported from an epw file, and the calculated adjust solar temperature from the outdoor solar temperature adjustor component. UTCI values are divided into categories regarding


ISBN - 978-86-7892-807-9

26

thermal heat stress [21] (Table 1). The last phase includes the exportation of all numerical results of the UTCI from Ladybug to Excel. The results are exported to Excel by ykTools (Grasshopper add- on) to sort the results in columns and detect all hourly values for a particular analysis period and weather data. The procedure of different software applications in order to obtain the final UTCI numerical values is shown at Fig. 1.

UTCI [°C]

range Stress category

+38 to +46 Very strong heat stress

+26 to +32 Moderate heat stress

Table. 1. UTCI stress categories in summer period [21].

Figure 1. Procedure in different software applications for obtaining final UTCI numerical value.

2.2 Parametric model for evaluating outdoor thermal comfort conditions An automatic procedure was developed in order to assess the non-stationary thermal conditions of the manikin. The procedure calculates UTCI at a predetermined number of positions in an urban area by automatically changing the positions of the manikin. This feature is an advantage over stationary thermal comfort calculations in other models and micrometeorological field measurements. By importing all geometries from Rhinoceros into Grasshopper, the geometry of the buildings and footways are fixed on site, whereas the manikin disposition is automatically changed. The footway represents a one-dimensional line comprising a number of segments that are determined by a given distance. To calculate the UTCI at all predetermined positions of the body at the footway, the algorithm process contains a portion of the code that generates a list of numbers (from 0 to n) representing body positions (Figure 2). Afterwards, the Grasshopper add-on Rabbit automatically changes the manikin position by consulting a list of numbers. After calculating the UTCI value at one position, the position of the manikin is changes, and a calculation of the UTCI value at a new position is performed. The UTCI is calculated until all predetermined body positions are examined [22].

Above +46 Extreme heat stress

+32 to +38 Strong heat stress

+9 to +26 No thermal stress


ISBN - 978-86-7892-807-9

27

Figure 2. View of the street and manikin positions along a specified path after the simulation of outdoor thermal comfort conditions.

2.3 Parametric model for the improvement of outdoor thermal comfort conditions The 3D model of the urban area included the elements of the built environment. Also, model included two dimensional closed poly lines representing the buildings bases which heights were optimized through the procedure. In the procedure the list of all possible permutations of the buildings heights (Figure 3) was automatically generated. Thus, simulations of outdoor thermal comfort were performed for each of the building heights arrangement. The simulations were performed until all buildings heights arrangements were examined i.e. until the list of all permutations is finished (Bajsanski et al., 2015). In the last phase all numerical results of outdoor thermal comfort simulations were obtained and optimal urban design was detected. The final building heights arrangement was selected based on the calculated average minimum UTCI value of the simulations in the analyzed period.

Figure 3. The sample of generated 3D model of buildings in Rhinoceros (left) based on the list of all permutations in Grasshopper (right). 3D model is appropriate for permutation {0} from the final list.

2.4 Sunlight hours analyses The sunlight analyses were performed in LadyBug using Sun Path and Sunlight Hours Analysis components in order to explain UTCI differences between current and proposed urban design. The Sun Path component contains data regarding the location of investigated area. Location data contains longitude, latitude and altitude values that were used for the sun path creation. User can define


ISBN - 978-86-7892-807-9

28

analysis period of the year (1-12 months), month (1-31 days) and day (1-24 hours). Output data is sun vectors that can be used for Sunlight Hours Analysis. This component contains input data as 3D geometry that cast shadow and surface geometry that is shaded by 3D geometry. Furthermore, the user can specify resolution of the result i.e. can set the grid size. Output data is sunlight hours results that can be exported into software for further statistical analysis.

3 APPLICATION OF THE PARAMETRIC PROCEDURES AND RESULTS The procedure for outdoor thermal comfort mitigation was tested on the urban design of the square and the results are presented in this section. The selected urban design is a combination of attached and detached buildings with 12 m, 15 m and 21 m heights. The investigated area is 2796 m². The computer performed 210 different simulations i.e. that was a number of all possible permutations of three heights of 12 m, two heights of 15 m, and two heights of 21 m. The footways were arranged within the square in order to test the procedure for improvement of outdoor thermal comfort (Figure 4). The length of footways is 331 m and the distance between two points is 10 m. The shape of the path and number of potential manikin locations can be arbitrarily chosen. Due to this, in this paper, the application of the method for improving the outdoor thermal comfort is shown in one example with determined path and body location.

To assess the thermal comfort changes in urban designs of the investigated urban area, days from extreme hot periods during 2014 were selected. According to the Republic Hydrometeorological Service Bulletin [23], the most intensive heat wave occurred from the 5th to 10th of July, and the warmest day during this 6-day period was July 7th (a hot and calm summer day). The simulations of outdoor thermal comfort were performed from 09 UTC to 19 UTC when the air temperatures were higher than 30 °C.

Figure 4. Investigated urban design of the square with the determined path for the body locations (from 0 to 33):

current urban design (left); proposed urban design by parametric procedure (right).

The average UTCI result for all thirty-four body locations in current urban design was 32.81 °C. The average UTCI result for the proposed urban design was 32,57°C. Furthermore, the differences between current and proposed urban design at some body location were up to 1.5 °C. Overheating of the proposed urban design and the current urban design were compared in order to illustrate the benefits of this parametric procedure (Figure 5). The presented method allows the user to explore the direct relationship between environmental data and the urban design through numerical and graphical data outputs. Average sunlight hours for current urban design was 8.09 hours while for proposed urban design value was 7.36 hours. Hence, the result was better for 0.73 hours, i.e. 9.03%.


ISBN - 978-86-7892-807-9

29

Figure 5. Visualization of sunlight hours on investigated urban design of the square: current urban design (left); proposed urban design by parametric procedure (right).

4 CONCLUSION Parametric approach for the improvement of outdoor thermal comfort conditions is presented in this paper. Interaction between different software applications gives the user useful information about the influence of changed urban design on thermal comfort in cities. Performance based design of cities is significant for the evaluation of the level of outdoor thermal sensation and of the overheating of urban surfaces. In this research only one parameter (heights of buildings) and its influence on human thermal sensation was tested. The developed procedures can be used for any urban environment. User has to have appropriate weather data with appropriate location information. The results demonstrate how digital technologies can be efficiently used for heat mitigation in any urban fabric. They provide practical instruction for built environment in order to provide comfortable conditions in open urban spaces. Results reveal that surrounding buildings with certain heights have a significant influence on thermal sensation and overshadowing. The proposed procedure integrates Rhinoceros, Grasshopper and LadyBug, which enable fast processing. In future research we aim to use proposed procedure to investigate issues related to the mitigation of overheating in different types of the built environment.

REFERENCES [1] Ahmed, K. S. (2003). Comfort in urban spaces: Defining the boundaries of outdoor thermal comfort for the tropical urban environments. Energy and Buildings, 35(1), 103–110. doi - 10.1016/S0378-7788(02)00085-3

[2] Ng, E., & Cheng, V. (2012). Urban human thermal comfort in hot and humid Hong Kong. Energy and Buildings, 55, 51–65.

[3] Martinelli, L., Lin, T.-P., & Matzarakis, A. (2015). Assessment of the influence of daily shadings pattern on human thermal comfort and attendance in Rome during summer period. Building and Environment, 92, 30–38.

[4] Krüger, E. L., Minella, F. O., & Rasia, F. (2011). Impact of urban geometry on outdoor thermal comfort and air quality from field measurements in Curitiba, Brazil. Building and Environment, 46(3), 621–634.

[5] Ali-Toudert, F., & Mayer, H. (2005). Thermal comfort in urban streets with trees under hot summer conditions. 22nd International Conference, PLEA 2005: Passive and Low Energy Architecture - Environmental Sustainability: The Challenge of Awareness in Developing Societies, Proceedings, 2 (July), 699–704. doi -10.1007/s00704-005-0194-4

[6] Bajsanski, I., Stojakovic, V., & Jovanovic, M. (2016). Effect of tree location on mitigating parking lot insolation. Computers, Environment and Urban Systems 56, 59–67. Doi 10.1016/j.compenvurbsys.2015.11.006


ISBN - 978-86-7892-807-9

30

[7] Matzarakis, A., Rutz, F., & Mayer, H. (2010). Modelling radiation fluxes in simple and complex environments: Basics of the RayMan model. International Journal of Biometeorology, 54(2), 131–139. http://doi.org/10.1007/s00484-009-0261-0

[8] Eliasson, The use of climate knowledge in urban planning, Landscape and Urban Planning 48 (2000) 31-44, doi -10.1016/S0169-2046(00)00034-7.

[9] Shi, X., Yang,W. (2013). Performance-driven architectural design and optimization technique from a perspective of architects. Automation in Construction, 32. pp. 125–135. doi- 10.1016/j.autcon.2013.01.015.

[10] Roudsari, M. Pak, A. Smith, Ladybug: a parametric environmental plugin for Grasshopper to help designers create an environmentally-conscious design. 13th Conference of International Building Performance Simulation Association. Proceedings of BS2013. Chambery, France, August 26-28, 3128-3135.

[11] Jendritzky, G., de Dear, R., & Havenith, G. (2012). UTCI – Why another thermal index? International Journal of Biometeorology, 56(3), 421–428. doi – 10.1007/s00484-011-0513-7

[12] Blazejczyk, K., Epstein, Y., Jendritzky, G., Staiger, H., & Tinz, B. (2012). Comparison of UTCI to selected thermal indices. International Journal of Biometeorology, 56, 515-535. DOI 10.1007/s00484-011-0453-2

[13] Park, S., Tuller, S. E., & Jo, M. (2014). Application of Universal Thermal Climate Index (UTCI) for microclimatic analysis in urban thermal environments. Landscape and Urban Planning, 125, 146-155. doi – 10.1016/j.landurbplan.2014.02.014

[14] Bröde, P., Krüger, E. L., Rossi, F. A., & Fiala, D. (2012). Predicting urban outdoor thermal comfort by the universal thermal climate index UTCI - a case study in Southern Brazil. International Journal of Biometeorology, 56(3), 471-480. doi – 10.1007/s00484-011-0452-3

[15] Yahia MW, Johansson E. (2014). Landscape interventions in improving thermal comfort in the hot dry city of Damascus, Syria - The example of residential spaces with detached buildings. Landsc Urban Plan; 125: 1-16.

[16] Givoni, B., Noguchi, M., Saaroni, H., Pochter, O., Yaakov, Y., Feller, N., & Becker, S. (2003). Outdoor comfort research issues. Energy and Buildings, 35, 77–86. doi- 10.1016/S0378-7788(02)00082-8

[17] Watanabe, S., Nagano, K., Ishii, J., & Horikoshi, T. (2014). Evaluation of outdoor thermal comfort in sunlight, building shade, and pergola shade during summer in a humid subtropical region. Building and Environment, 82, 556–565. doi- 10.1016/j.buildenv.2014.10.002

[18] Simpson, J. R. (2002). Improved estimates of tree-shade effects on residential energy use. Energy and Buildings, 34(10), 1067–1076. doi- 10.1016/S0378-7788(02)00028-2

[19] Akbari, H. (2002). Shade trees reduce building energy use and CO2 emissions from power plants. Environmental Pollution (Barking, Essex : 1987), 116 Suppl , S119–26. doi-10.1016/S0269-7491(01)00264-0

[20] U. S. Department of Energy. Office of Energy Efficiency & Renewable Energy, Washington [Internet], 2015 [cited 2016 February]. Available from: http://apps1. eere.energy.gov/buildings/energyplus/weatherdata_about.cfm.

[21] International Union of Physiological Sciences Thermal Commission, Glossary of terms for thermal physiology, J. Therm. Biol. 28 (2003) 75-106.

[22] Bajšanski, I. V, Milošević, D. D., Savić, S.M. (2015). Evaluation and improvement of outdoor thermal comfort in urban areas on extreme temperature days: Applications of automatic algorithms, Building and environment 94, 632– 643. doi-10.1016/j.buildenv.2015.10.019

[23] RHMZ. Monthly bulletin for Serbia – July 2014. Republic Hydrometeorological Service of Serbia, Department of 462 National Center for Climate Change, Division for Climate forecasts, Information and Trainning, Belgrade


ISBN - 978-86-7892-807-9

31

GENETIC ALGORITHM APPLICATION AS A SUSTAINABLE URBAN PLANNING OPTIMIZATION TOOL

Marko Jovanović1* 1University of Novi Sad, Faculty of Technical Sciences, Serbia, [email protected]

Abstract Urban planning and design is under a wide influence of computation and use of digital tools. Application of algorithms has proven as an efficient approach to sustainable development, and multi-criteria optimization problems. Utilization of genetic algorithms aids in finding an optimum solution, by manipulating certain range of parameters. This research combines the genetic algorithm with the generative approach of urban planning and performative approach of urban analysis. Urban planning is controlled by a set of parameters, while its analysis is conducted through performance simulations. Proper height to width ratio dimensioning and parcel disposition can aid in efficient daylighting systems, various content prone open spaces and in general, a diversity of an urban morphology. In this research, genetic algorithm is implemented as an optimization design tool, to aid in finding the best disposition of built area on a set of specified parcels, based on the adequate exposure to sunlight. Implementation of genetic algorithm provides a floor space index (FSI) for each parcel, as well as optimal solution for daylight system, which can be utilized in master urban plans to provide additional information and justification of existing data.

Keywords: performative, parcel disposition, insolation, generative

1 INTRODUCTION Digital tools have a wide influence on various aspects of urban design and planning. It has become noticeable that computer aided design and management software have allowed for a more profound, and multi criteria approach in this field [1]. Use of GIS (Geographic information systems) technology has provided the opportunity for implementing criteria as various parameters, used to solve the problems on architectural and urban levels both [2, 3, 4]. GIS technology represents a stepping stone to using a parameter based approach to urban planning and management of information. Criteria and information can be introduced as units or driving factors in algorithmic optimization methods as a more advanced approach, therefore enhancing building performance [5, 6].

Implementing an algorithmic approach on urban planning optimization has proven as an indispensable tool altogether and will be researched in more detail. Use of self-organizing systems, such as cellular automata algorithms, has benefited the simulation [7] dynamics of urban areas development and foresight multifold [8]. Genetic algorithms have been used for optimizing various urban design and planning related issues, such as energy calculations [9], energy simulations [10, 11] and land usage planning [12]. Applying genetic algorithms for urban design and planning optimization has been conducted for mid latitude climates on a general building disposition scale [13], the solar energy as an alternative energy source and implementing it as an significant urban design factor has long been the centre of attention in urban design and planning [14].

Aim of this paper is to generate an automated approach to sustainable urban planning, applicable for any geographical location and any climate. The evaluation will be conducted, by applying a genetic algorithm to perform urban planning optimization in regards to sunlight hours. The performance will be rated through the insolation calculation fitness level, with the purpose of maximizing the daily average sun hours on street level.

Many factors influence the insolation of the street, surrounding urban blocks, such as the disposition of the trees for generating an adequate daylighting system [15] or optimum outdoor environment in residential housing [16], height to width street ratio, corner building emphasis [17], floor space index [18], parcel disposition etc. Use of genetic algorithms requires an appropriate problem framing [19], or an adequate comprise between minimizing time consumption and the optimized solution from the chosen search space. For those purposes, in this paper, the influence of the tree disposition on street level has been omitted, in order to generate a more uniform analysis throughout the year, using the above mentioned factors. Factor such as width to height ratio of the street is the dominant influence on


ISBN - 978-86-7892-807-9

32

the urban planning, since it impacts the insolation calculation altogether. Other factors, such as corner height ratio, floor space index contribute to the diversity of the optimum solution, suitable for different needs. These factors, used as parameters, allow for a more automated approach to urban planning, and furthermore are evaluated by their performance to generate a sustainably optimum solution. Utilizing such an approach in early stages of urban planning can benefit the design of each of the associated buildings, in terms of sustainability.

2 GENETIC ALGORITHM BASED URBAN PLANNING OPTIMIZATION METHODOLOGY Use of computation has provided a more automated approach to a wide range of multi-criteria optimization problems. In that respect, a genetic algorithm, as a sustainable urban design tool, is chosen in order to get an optimal solution through manipulating a range of parameters. Genetic algorithms operate on the premises of meta-heurism. Particular sets of data are constantly evaluated in correspondence to a fitness value, in form of a loop. The best ones are chosen and combined amongst each other, in order to produce even better ones. Evaluation is conducted in reference to the fitness value, which determines whether optimum solutions are being found. Mainly, a specific genetic algorithm script needs to be written in reference to a specific problem being researched.

In most cases, an adequate genetic algorithm script is written to suit predetermined needs in building optimization [6] or urban development and planning [20, 21, 13]. Genetic algorithms used in the design process in that way do not provide relationship with the parametric driven design and freedom in manipulating the range of subsequently added parameters.

In this research it is a design algorithm that combines parametric design tool within Rhino- Grasshopper software altogether with sustainable design analysis software Ecotect in order to provide a range of performance design and parametric driven tools. The connection between parametric driven software Rhino-Grasshopper and performance analysis software Ecotect is established via a Grasshopper add-on, Geco, which allocates data between the two programs appropriately. That way, the geometry can be run through performance simulations, and the results can be acquired for the evaluation process back and forth. This approach does not account for an optimal solution, it mainly utilizes the input data it has and procures results suitably. Connection between various software packages and its tools in order to provide performance based and parametric driven sustainable urban planning optimization tool is shown in Figure 1.

Fig. 1. Figure caption

3 ALGORITHM DESIGN The precedence of this research is that Grasshopper has a predetermined tool that serves as a genetic algorithm, Galapagos. As a part of the Grasshopper interface, it is easily implemented in the urban planning algorithm and linked with the performance simulation and evaluation software. Galapagos allows for an automated loop process to be achieved with Grasshopper, until an optimum result is procured. It uses quantifiable parameters, set in Grasshopper, as potential genes, which are intertwined and checked for best layout schemes. The output of such an automated process is the optimum disposition of buildings in urban fragments and blocks, that produces the maximum insolation at the street level.

At the same time, Grasshopper offers an add-on, Geco, which enables the link with the Ecotect software, providing the wide range of performance analysis parameters.

The design of sustainable urban planning optimization genetic algorithm is preceded by two parts: generative and performative algorithm. In the following text, both design phases will be described in detail.


ISBN - 978-86-7892-807-9

33

3.1 Generative Algorithm

3.1.1 Rhinoceros First step in the algorithm requires the provision of geometry in the Rhinoceros software package, consisting of closed polylines, representing parcels and blocks. The geometry can either be generated in the actual program itself, or it can be imported from various other CAD supporting programs, such as AutoCAD, ArchiCAD, CityEngine, 3ds Max etc. The polylines enclose the parcel or block area, and are an obligatory factor in the algorithm's functioning. Once the polylines are checked for consistency, it should be cleaned off of any duplicate lines it may have, to ensure the proper execution of the algorithm. Also, the parcels must be enclosed inside the block polyline, respectively, in order to procure the best results. After the polyline generation and preparation is complete, the geometry is referenced inside the Grasshopper plug-in, in order to be a valid parameter in the urban planning process.

3.1.2 Grasshopper The Grasshopper plug-in operates as a visual programming language, allowing a generative automated approach to architecture and urban planning. It utilizes visual nods instead of written computer language code, thus simplifying the code generation and connection of parameters. The intuitive program writing explores various problems in question, their interrelations, the geometry and without the need to learn how to script. The polylines referenced from Rhinoceros, are presented as input geometry onto which, all the accompanying data will be supervened.

At the start of the algorithm, the user is presented with several manageable parameters, chosen as important and unavoidable in the field of urban design and planning, for this research (Table 1). All are quantified as numeric ranges, allowing for the desired outcome result to be diverse, depending on the predetermined user preferences.

Table 1. The manageable parameters for the algorithm workflow

Parameters Value Type Range

Height / Width Street Ratio Float 0 - 5

Position All the Buildings on the Edge of the Block Boolean 0, 1

Percentage of Inward Positioned Buildings Float 0 - 1

Corner Buildings Higher Boolean 0, 1

Additional Corner Height Percentage Float 0 - 2

Floor Height in meters Float 0 - 5

Floor Count Integer 0 - 3

According to the sustainable design principles, the most important factor chosen is the height to width street ratio [17]. This factor gives an insight into different types of streets and their depending cross sections. The standard solution utilizes the aspect ratio of 1:3, therefore, significantly lowering the height of the surrounding structures and impacting the count of the dwelling units in the area. Low rise habitation allows for better street insolation, which is the predominant aim of the research. Another important parameter is the position of the buildings in the block area. They can either be positioned at the edge of the block or be indrawn. The latter allows for a more imminent connection to the street altogether, in terms of generating open space content and diversity in the form of the street. The amount of just 10 percent indrawn positioned buildings can really impact the block's appearance and provide optimization of open spaces [22]. Landmarks and nodes [23] are an important part of urban planning and are thusly introduced, as important urban content generators. Thus, a manner of emphasizing certain physical spaces or outdoor environment as such is seen in introducing a


ISBN - 978-86-7892-807-9

34

parameter of additional corner height in order to diversify the street longitudinal front. Also, varying the non-corner building's height by adding or reducing floors, and their height respectively, decreases the possibility of monotonous street look altogether. After the process of building's generation, the geometry is prepared for performance analysis in Ecotect. Extension tool Geco is utilized for this approach.

3.1.3 Genetic Algorithm Parameters in the genetic algorithm are mainly oriented towards randomizing the data, such as parcel position or building height, in order to diversify possible solutions, procured by setting initial, manually controlled parameters, stated above. For the purposes of this paper, the fitness value is set to maximum, meaning that the genetic algorithm will continue to run until it drains all viable data sets, and present the optimum solution as a result of the process. Due to an automated approach, once all the desired parameters are set, the algorithm is initiated and left to run, until it procures an optimum solution for a predetermined set of values or is terminated. Also, once the genetic algorithm finishes the optimization process, the end result, in form of a building disposition and height, is accompanied by numerical values pertaining to FSI for each parcel respectively.

3.2. Performative algorithm The geometry is converted into meshes in order to be properly exported to Ecotect. At the same time, a set of parameters is adjusted inside the Grasshopper, which generates a connection with the running instances of Ecotect and controls the precision of the insolation calculation analysis itself (Table 2).

Table 2. A set of parameters for the precise control of the insolation calculation analysis

Parameters Value Type Range

Desired Weather File for Insolation Calculation *.wea

Start Ecotect Analysis Boolean 0, 1

Analysis Grid U Subdivision Integer 0 - 10

Analysis Grid V Subdivision Integer 0 - 10

Sky Subdivision Factor Integer 20 - 50

Perform Insolation Calculation Boolean 0, 1

The insolation calculation analysis is focused towards estimating the average solar hours on the street level in a day throughout the entire year, ranging from 5 am to 8 pm. The results for any climatic and geographical location require a weather file to be set inside the Geco add-on. This allocates the metadata to Ecotect for specific and focused calculation for a desired locality. It also utilizes average values stored in such a file, providing more analytic information. The weather file has the extension *.wea and the information for it can be downloaded from the internet and generated for major cities in any country [24]. The performance calculation itself also requires a planar or spatial grid, for the analysis to be conducted on. Since this research is oriented towards analyzing the insolation on street level, a planar grid encompassing a wider offset of each block respectively is introduced. Depending on the block's size and dimensions, appropriate grid subdivision is selected in both axis, depicted here as U and V, providing higher or lower resolution of data. The precision of such an analysis is controlled by the sky subdivision factor. The lesser the number of the factor, the more precise the analysis is, but at the expense of time. Upon experimentation, a difference of 5 percent was detected in average results obtained from using 20 sky subdivision and 40 sky subdivision in this research. For the purposes of this paper a value of 40 sky subdivision is chosen in order to minimize the running time of the algorithm, with knowledge about the deviation from more precise data.


ISBN - 978-86-7892-807-9

35

Upon adjusting the parameters, the calculation is conducted on the faces of the subdivided mesh analysis grid, pertaining to the exported geometry in Ecotect. The results, in form of numerical values for each face, are then imported back to Grasshopper and averaged for each block's grid respectively. Those values serve as viable markers for evaluating a certain set of parameters, and are utilized as fitness value thresholds in genetic algorithm optimization.

Fig. 2 shows a conceptual model of the entire process stated above, referring to the generative, performative and genetic part of the algorithm. This algorithm is applied on a problematic urban fragment in Novi Sad, Serbia and will be described in detail in the following chapter.

Fig. 2. Algorithm Process Flow


ISBN - 978-86-7892-807-9

36

4 APPLICATION AND RESULTS In order to verify the consistency and the performance of the algorithm, a real world example is introduced. A small urban fragment of Grbavica, located near the city centre of Novi Sad, Serbia is chosen for the application phase of the process. Grbavica, due to its position, proximity to the important urban content and gentrification process has been developing and generating high rise dwellings, at the expense of social and economical degradation [25]. One of the side effects refers to the inadequate height to width street ratio, lack of open public space and insolation. Consequences are seen in overshadowing, green space deprivation and lack of contextualization in the area as well.

The algorithm is used twofold - first - to assess the current state of the situation and second - to optimize the urban planning of this fragment as if it were comprised of vacant parcels. Due to the same uniform approach to both analysis processes, the results are comparable. The parameters for the second approach are set to procure a solution that has 1:2 height to width street ratio, 20 percent of inwards positioned buildings, 50 percent higher corner buildings, and a possible floor height diversification of 2.65 m. The analysis grid is divided by a factor of 5, providing an adequate resolution of the analysis grid and sufficient data for the average fitness value.

The insolation calculation of the current state of Grbavica, containing four urban blocks, surrounding streets and neighbouring blocks, shows that the daily number of sunlight hours for this fragment is averaging at around 2.5 hours. Given the height to width street ratio, the buildings are overshadowing a great deal of the street. Accompanied by the fragment's south orientation deviation of 35 degrees, most of these streets are not getting proper or any sunlight hours, especially during the winter, as can be seen in the certain parts, shaded in blue colour (Figure 3). Also, the average FSI is around 2.7.

Applying the algorithm on the vacant parcel disposition, the urban planning optimization provides a solution which generates an average of 4.8 average sunlight hours daily, which is an increase of 2.3 daily average of sunlight hours. The genetic algorithm ran for 40 hours straight and procured 21 generations of data settings, amongst which the one providing 4.8 sunlight hours was the optimum one. The entire process flow and generation is documented and can be reviewed if proven necessary. The end result shows the building disposition and FSI numeric values for each parcel respectively in a visual representation form (Figure 4). The FSI averaged at 1.7, which is equivalent to a mere reduction of one or two floors floor in general. The parameter settings utilized to generate this solution are not stretched to its upper limitation and provide the opportunity for an even higher increase in daily average of sunlight hours, given the predefined user requirements. This research focused on the adequate solar insolation at the street level, but bared in mind the number of dwelling units, depending on the building's height. As an example, lowering the building's height, by increasing the height to width street ratio to 2.5, the insolation at street level for the same building disposition increases to 5.2 daily average of sunlight hours. This proves that the parameters have a great effect on the outcome of the process and should be set according to in situ conditions and pre-requirements.


ISBN - 978-86-7892-807-9

37

Fig. 3. Grbavica current state analysis grid

Fig. 4. Grbavica optimal parcel disposition and building height with analysis grid


ISBN - 978-86-7892-807-9

38

5 CONCLUSION In this research we designed an algorithm capable of an automated urban planning and design approach in reference to sustainability principles. Urban planning parameters are changeable depending on the desired outcome of the algorithm's utilization. Hence, the optimal solution is procured in reference to the preset search space and parameters, which can serve as a viable underlay for the future urban planning process altogether. The algorithm is applicable to any geographical location and climate area, provided with the appropriate weather file. The utilization of a conjoined performative and generative approach is a great way to insure entities optimal performance rates, and parametric dependency additionally. The algorithm validation is conducted via a case study application approach. The comparison of the results of the two distinctive calculation processes shows an evident increase of average daily sunlight hours after the algorithm's application. This study has illustrated that the urban planning process for the specific urban fragment of Grbavica, used for this analysis, is not adequate and lacks proper insolation. It has also revealed the need to utilize computation and automation in urban planning, through parametric, generative and performative approach. The automated approach applied here, through genetic algorithm optimization methods, contributes to the overall design benefit and time reduction factor as well. The effect of the urban development and planning in the urban fragment of Grbavica in Novi Sad, Serbia, evidently shows clear signs of oversight, regarding sustainability. Such a problem is examined in great depth and a solution is procured, pending the more involved use of computation and automation in urban planning. This implementation was only carried out as a case study, but presents the opportunity for application on any number of worldwide locations, following the settings stated above.

The algorithm is limited to polylines as input parameters. Due to this fact, a certain number of large parcel areas are not properly propagated, dimensioned and may appear vast in size. The parcel's ground space index is set to a range from 0.35 to 0.65 at random. This may cause very sharp angled floor plans in certain areas towards the centre of the block, but does not influence the insolation calculation phase of the algorithm. The sky subdivision factor can dramatically increase the calculation time and should be set accordingly. When the optimum solution is procured by the algorithm using a high number of sky subdivisions, that same solution can be run manually through a more precise analysis phase for improved results.

Future work will focus on resolving the limitations of the algorithm and the optimization of the buildings in an urban fragment, pertaining to facades, material application, greenery etc. Visualization of the results, derived by utilizing genetic algorithms is also considered a viable research area and is applicable in the genetic algorithm optimization approach.

ACKNOWLEDGEMENT This research was supported by the Serbian Ministry of Education and Science (project no. TR36042).

REFERENCES [1] Chakrabarty, B.K., Computer-aided design in urban development and management—A software for integrated planning and design by optimization, Building and Environment, Volume 42, p 473494, 2007

[2] Stevens, D., Dragicevic, S., Rothley, K., iCity: A GIS-CA modelling tool for urban planning and decision making, Environmental Modelling & Software, Volume 22, p 761 - 773, 2007

[3] Choi, Y. et al, PV Analyst: Coupling ArcGIS with TRNSYS to assess distributed photovoltaic potential in urban areas, Solar Energy, Volume 85, p 2924 - 2939, 2011

[4] Guney, C. et al, Tailoring a geomodel for analyzing an urban skyline, Landscape and Urban Planning, Volume 105, p 160 - 173, 2012

[5] Turrin, M. et al, Performative skins for passive climatic comfort - A parametric design process, Automation in Construction, Volume 22, p 36 - 50, 2012

[6] Rakha, T., Nassar, K., Genetic algorithms for ceiling form optimization in response to daylight levels, Renewable Energy, Volume 36, p 2348 - 2356, 2011


ISBN - 978-86-7892-807-9

39

[7] Garcia, M. et al, A comparative analysis of cellular automata models for simulation of small urban areas in Galicia, NW Spain, Computers, Environment and Urban Systems, Volume 36, p 291-301, 2012

[8] Al-Ahmadi, K. et al, Calibration of a fuzzy cellular automata model of urban dynamics in Saudi Arabia, Ecological complexity, Volume 6, p 80-101, 2009

[9] Turrin, M., von Buelow, P., Stouffs, R., Design explorations of performance driven geometry in architectural design using parametric modeling and genetic algorithms, Advanced Engineering Informatics, Volume 25, p 656 - 675, 2011

[10] Shi, X., Design optimization of insulation usage and space conditioning load using energy simulation and genetic algorithm, Energy 36, p 1659 - 1997, 2011

[11] Kabolizade, M., Ebadi, H., Mohammadzadeh, A., Design and implementation of an algorithm for automatic 3D reconstruction of building models using genetic algorithm, International Journal of Applied Earth Observation and Geoinformation, Volume 19, p 104 - 114, 2012

[12] Porta, J. et al, High performance genetic algorithm for land use planning, Computers, Environment and Urban Systems, Volume 37, p 45 - 58, 2013

[13] Panao, M., Goncalves, H., Ferrao, P., Optimization of the urban building efficiency potential for mid-latitude climates using a genetic algorithm approach, Renewable Energy, Volume 33, p 887 - 896, 2008

[14] Kanters, J., Horvat, M., Solar energy as a design parameter in urban planning, Energy Procedia, Volume 30, p 1143 - 1152, 2012

[15] Hongbing, W. et al, Optimal tree design for daylighting in residential buildings, Building and Environment, Volume 45, p 2594 - 2606, 2010

[16] Ooka, R., Chen, H., Kato, S., Study on optimum arrangement of trees for design of pleasant outdoor environment using multi-objective genetic algorithm and coupled simulation of convection, radiation and conduction, Journal of Wind Engineering and Industrial Aerodynamics, Volume 96, p 1733 - 1748, 2008

[17] Davies, L., Urban Design Compendium, English Partnerships, The Housing Corporation, London, 2007

[18] Patrick, Brooks. SPACEMATE® November 12, 2011 | Visually and Mathematically Examining Space, Density, and Urban Form, The Architecture Report, accessed October 23, 2012 [Online] <http://thearchitecturereport.com/spacemate-visually-examining-space-density-and-urban-form>

[19] Bleil de Souza, C., Contrasting paradigms of design thinking: The building thermal simulation tool user vs. the building designer, Automation in Construction, Volume 22, p 112 - 121, 2012

[20] Huang,G., Gao, W., Simulation Study on CA Model Based on Parameter Optimization of Genetic Algorithm and Urban Development, Procedia Engineering, Volume15, p 2175 -2179, 2011

[21] Feng, C., Lin, J., Using a genetic algorithm to generate alternative sketch maps for urban planning, Computers Environment and Urban Systems, Volume 23, p 91 - 108, 1999

[22] Neema, M.N., Ohgai, A., Multi-objective location modeling of urban parks and open spaces: Continuous optimization, Computers, Environment and Urban Systems, Volume 34, p 359 - 376, 2010

[23] Lynch, K., The Image of the City, Massachusetts, M.I.T. Press, 1972

[24] EnergyPlus Energy Simulation Software n.d., accessed 18 of July 2013 < http://apps1.eere.energy.gov/buildings/energyplus/weatherdata_about.cfm>

[25] Nedučin, D., Carić, O., Kubet, V., Influences of gentrification on identity shift of an urban fragment - a case study, Spatium, Volume 21, p 66 - 75, 2009


ISBN - 978-86-7892-807-9

40

DEVELOPMENT OF PHYSICAL CONTROL AND VIRTUAL NAVIGATION MECHANISM TO ENHANCE ACTIVE

COLLABORATION BETWEEN ARCHITECTS AND CLIENTS IN A REACTIVE DIGITAL SPACE

Odysseas Kontovourkis1*, Ioanna Dimitriou2, Konstantinos P. Chatzimanolis3, Michaelina Stylianou4, Philippos Michael5, Styliana Gregoriou6

1*23456Department of Architecture, University of Cyprus (CYPRUS) 1*[email protected] [email protected] [email protected]

[email protected] [email protected] [email protected]

Abstract In this work, a computational design methodology to enhance a productive relationship between architects and clients during the design decision-making process is presented. The aim is to allow clients to be involved in architectural design procedure and collaborate with the architects during the conceptual stage of the design development within a reactive digital space. In order to test this hypothesis, a case study that involves the development of different living space typologies and navigation within these, as well as the selection of openings that allow variation of façades geometry is presented. Gesture-oriented human behavior and lighting values are used as the input data that control the digital outcomes. Results are further analyzed according to the qualitative criteria of clients’ spatial experience and aesthetics, on the one hand, and according to the quantitative criteria of lighting performance and area of living space, on the other hand, establishing a communication and control between architects, clients and digital products.

Keywords: Physical control, virtual navigation, gesture-oriented behavior, architects, clients, reactive digital space.

1 INTRODUCTION Nowadays, radical advances in technology, which has been used in different fields of research including architecture, aims to improve the living conditions of users, allowing in parallel a reactive relationship to be established between users of space and architectural products. This adaptation led to a flexible and transformable design, which might response directly and extensively to the needs and desires of users. Although, current innovative mechanisms can provide users with more realistic experience of virtual space, a few of them have incorporated their advantages to help the relationship between architects and clients. The participation of users in the process of developing and controlling the digital architectural space using pre-set hand gestures and physical 3D objects, opens new directions in the area of digital design allowing a more intuitive and natural way of virtual space manipulation [1]. Within this context, the client might turn to an architect who is able to create his/her own space, using simple body movements, an approach that might allow cost-effective tools that involve an active participation of users during the design decision-making process [2]. Within this framework, this paper investigates reactive architectural spaces where architects and client can be productively involved in order to enhance collaboration during the conceptual design stage using integrated computational design mechanisms. This projects lies into the research framework of the Digital Architecture and Prototype Systems Development Lab (dAPSd Lab) in the Department of Architecture at the University of Cyprus, and it has been realized within the agenda of the postgraduate course ARH-522 - Advanced Computer-Aided Design Topics during the Fall semester 2015. Similar works of the Lab in this area can be found in [3, 4, and 5].

The current research investigates two discrete and correlated categories of examples that refer to physical to digital gesture-oriented control and digital navigation. Regarding the physical to digital space manipulation and control of spaces, examples of mechanisms can be found, mostly in regard to their ability to provide an environment for quick understanding and easy manipulation. For instance, in the game ‘Geek Run’ [6], an important element is the recognition of several blocks using the Kinect device and then these are translated into virtual space elements. To achieve similar results, several sensory devices are incorporated, constituting the main part of this study. In another example, a








ISBN - 978-86-7892-807-9

41

prototype receives information from the physical environment, using specific sensors, in order to change the openings of a building envelope. Simultaneously, through specific sensors, the detection of different lighting level of the designed space can be achieved [7]. A key feature of the installation is the translation of the physical movements of users in digital space, which has been used in a variety of game designs such at the Woodland Wiggle [8]. In this case, children play with their bodies to create different ‘stories’ on a screen by using a Kinect device. Different disciplines have used technologies similar to ’Techviz‘ [9] in order to ’live in their creations‘ using devices like Oculus and joystick. The ability to navigate in any space designed for the user is an important aspect towards such investigation since it provides an understanding of the space under investigation through perception. It allows identification of potential problems and suggests solution for further improvement. These possibilities could not be granted otherwise to users, prior to project implementation.

In this work, an integrated computational methodology is introduced and experimentally proceeded using simple and freely available computational mechanisms, which allow transferring of input data, processing and computational control and finally projection as visual outcomes. This might establish a user-friendly platform of operation among the architectural design ‘actors’, architects and clients, in order to develop and navigate in design space respectively. Initially, a series of experiments have been carried out in physical space to test each process and then integrated to produce the overall digital design methodology. Description of the methodology involved as well as the experimental results is presented in the following chapters.

2 METHODOLOGY

2.1 Data processing It is well know that new digital tools used by architects to communicate their work and ideas are rapidly developed, offering capabilities that move beyond typical CAAD (Computer-Aided Architectural Design) software. These tools allow innovative ways for design investigation, enhancing creativity and flexibility during the design decision-making process. In addition, such tools might allow an active involvement of non-architects, for instance clients, who are rarely involved during the first stages of design development.

As it has been mentioned, this paper introduce a methodology where such involvement can be achieved, examining in parallel the ability of tools to be implemented in an experimental case study. In order to achieve an interactive relationship between architect and client, namely user-architect (UA) and user-client (UC), a common point of reference has been found that allow their mutual involvement through gesture-based control of the proposed digital space. Devices capable to track and analyze physical data are introduced and then information is translated in digital level, allowing development and control of morphologies (Fig. 1).

Fig. 1. Example of UA (1st picture) and UC (3rd picture) in action and their monitor output presented on each

screen respectively (2nd and 4th pictures)

In order to achieve the translation from physical to digital space, the input data acquisition devices Kinect camera and Leap Motion controller as well as a light sensor incorporated in Arduino board are applied. Also, the experimental and algorithmic control is achieved in Rhino 3D modeling software using the visual programming environment of Grasshopper (parametric-associative design plug-in for Rhino).

Analytically, a dual relationship is established, which refers to the reactivity of investigated digital outcomes with the UA, on the one hand, and with the UC, on the other hand. In terms of the first relationship established, the camera based device Kinect is used, which is able to recognize the user of space and indentify twenty basic points of the human body, as well as the position displacement of


ISBN - 978-86-7892-807-9

42

these points in real time. The data is transferred through the Quokka and the Firefly plug-ins to Grasshopper, where different gestures are recognized connected with respective rules that activate control of the geometry. Another hardware that is implemented in this project is the Leap Motion controller. It is capable to trace the hands of the user, at a short predefined range and identify points and joints, as well as their movement in physical space. In this case, the Primate and the Firefly plug-ins are used to connect the devices with Grasshopper. Finally, a light sensor embedded in an Arduino board is used as another input data acquisition device, which identifies lighting levels in the interior space of the projected outcome. In this case, the Firefly plug-in is used to import in Grasshopper numerical values derived from the sensor. Finally, by using the Horster plug-in for Grasshopper, the reactive digital result is extracted and projected back to the physical word as perspective views. This is done via projector that transmits visual images onto a screen, allowing optical stimulus to be created (Fig. 2).

Fig. 2. Establishing communication between various physical input data acquisition devices, parametric-

associative plug-ins and projected outputs

2.2 Physical data acquisition The phase of experimentation, as well as the phase where realization of the project occurs, requires the interaction of the user-architect (UA) with the user-client (UC), who are simultaneously acting in real time. The first user (UA) is responsible for the production of shape in the virtual space by manipulating physical geometries (two cubes, with dimensions 4 x 4 x 4 cm), which can be freely moved on a transparent surface. The geometries produced are projected on a monitor while at the same time they are translated into an actual virtual space that is displayed on the screen so that the UC can experience the interior space. The Leap Motion controller device is responsible to capture the physical behavior of the UA, which is placed on a physical model beneath the transparent surface. The second user (UC) navigates in a limited digital space, utilizing particular body movements that take place in physical space. Furthermore, by using various body gestures, the UC can affect and alter the final form of the structure that is projected on the monitor. The Kinect camera device is responsible to recognize this behavior due to its ability to track movement in space by calculating 15 points of a person’s skeleton [10]. For this project, seven points are used; right palm, left palm, waist, the two heels, right shoulder and the head, which are processed through parametric and algorithmic principles, manage to transfer the physical movement of the user into digital navigation mechanism.

2.3 Algorithmic analysis In the level of computational mechanism applied as well as in the level of algorithmic development, both UA and UC users behavior, after this is transferred in the digital environment via data acquisition devices, is algorithmically controlled through a number of rules that aim to influence a reactive relationship with the digital geometrical results. Also, the simulated variations that are produced in each case are directly perceivable by the users influencing their next action behavior in a cyclically iterated feedback loop process. In terms of UA behavior, the position and movement of user’s palm center as well as his/her thumb are recorded creating a common space in between that is defined by the two physical cubes. Using the initial cubes and the middle of space, the unified morphology is achieved. In addition, by moving the two cubes, the UA is able to produce various typologies like linear and L shape morphologies. Simultaneously, the UC observes the changes of the space created by UA and with specific movements, he/she is able to create openings on the surfaces of proposed morphology that is projected on the screen, navigating at the same time in the interior space with the assistance of Kinect camera. Openings are represented as modules, divided in each surface, which


ISBN - 978-86-7892-807-9

43

are randomly positioned by the UC based on his/her observation in regard to the amount of light entering the interior space. The UC is able to create openings by keeping the left hand steady, at shoulder height, while the movement of the right hand controls the quantity of the openings. The user (UC) can ‘save’ the created openings by raising the right hand above the point of the head and lowering the left hand in a position below the point of the waist. Changes on the UC’s position are achieved through heels tracking. Analytically, by moving the right foot forward, the UC is able to move forward while by moving the left foot backward the user is able to move backward. The lighting inside the structure is affected by changing on the openings but these alterations are hardly perceived by the user. In order to overcome this, a light sensor that detects the levels of lighting is used to mitigate the alteration from digital to physical space. Values are then projected to the screen in the form of a diagram as a percentage of lighting, while the user navigates in the space. In addition, all rules of user behavior are presented on the screen, assisting their movement and hence his/her navigation within the proposed morphology. Additional information regarding the position of user, amount of lighting, total openings inside the space, amount of openings that are ‘saved’ and the total surface area of the space, are also presented so that the user can have real time information in regard to the digital results obtained (Fig. 3).

Fig. 3. (a) UC movement instructions, (b) UA movement instructions, along with their parametric activation info

3 EXPERIMENTAL RESULTS

3.1 Experimentation settings A series of studies and experimentations are carried out before the integration of various computational mechanisms is realized, in an effort to test and find the optimal way of introducing technologies that are used to transfer and control data from the physical to the virtual space. Within this framework, the experimentations regarding the use of Leap Motion controller results the specification of movement's range and the simplification of the fingers’ movement. This is achieved by using two boxes of specific dimensions in order to resolve the limited ability of device to capture the finger movement range, which in turn, influences the hands movement of the UC. Likewise, the set-up of the Kinect camera, due to its limited movement recognition range in physical space, is also simplified. Finally, the light sensor values acquired through the use of Arduino board is also experimentally tested and its position is decided (Fig. 4).

Fig. 4. The process of experimentation


ISBN - 978-86-7892-807-9

44

Following figure (Fig. 5) demonstrates the set-up of the scene, which shows the position of each data acquisition device as well as the screen where the digital output is projected. Analytically, a custom large screen measured 3m x 1.65m and a projector placed within a distance of 4m behind, are used in order to display the results. The Kinect camera is placed at the bottom center of the screen, in 0.5m height, which is used to recognize the user's head, waist and centers of each palm. Additionally, a base measured 40cm x 40cm x 15cm, with the Leap motion controller positioned in the middle, at the bottom of a transparent plexiglas surface, is placed on a table positioned next to the screen (outside the range of the Kinect camera). Finally, two cardboard cubes measured 4cm x 4cm x 4cm, which are used by the user UA to create the 3D digital space are presented (Fig. 5).

Fig. 5. 3D Illustration of scene set-up

3.2 Case studies A number of case studies are presented and compared, while these are varying according to the parameters applied. Specifically, parameters include the space area of the produced morphology, the number of openings, the user's location and lighting levels. In the first case, the openings, the user’s position and the area of space are constants while the intensity of light is a variable received by the light sensor. In the second case, the openings, the position and the lighting are constants, while the area of space is a variable. In the third case, the openings, the area and the lighting are constants, while varying the position of the user. In the fourth case, the location, the size and the lighting are constants, while the numbers of openings are variables. As the parameters and the user's location change, different spatial qualities in the digital environment are produced. Possible morphologies include linear and L shape spaces, that are changed depending on the amount of lighting. Also, the results produced in digital space offer an experience of the space developed while this is investigated in real time and according to users’ needs and desires, accelerating in parallel their participation in the conceptual design stage (Fig. 6).


ISBN - 978-86-7892-807-9

45

Fig. 6. Sample of case studies representation showing the plan view of the morphology along with multiple

internal views and lighting variations

As it shows in the sample of morphologies produced in the Fig. 7, the UA creates two types of digital spaces, the linear and L shape. Then, correlations as well as comparisons between different spaces are investigated either based on their minimal or their maximal length. In the first case, the minimum length of the linear shape A is compared with the maximal length of the linear shape B. The two spaces are controlled by the UA, leading to different area values in the range of 94 m². Then, the UC determines the number of openings as well as their position leading to specific light values that affect the space quality of the interior. In this case, the maximum lighting value presented in linear shape A is 87% with 40 openings, while the maximum light value in the linear shape B is 90% with 155 openings (Fig. 7).

In the second case study, the experimental comparison of a linear A and an L shape space A is carried out, where two cases with similar area values are selected, with a difference of 16 m². The number of openings is used to determine the quality of the developed spaces regarding to light conditions in the interior. Thus, the maximum lighting value presented in L shape space A is 94% with 64 openings, while the maximum light value in the linear space A is 87% with 40 openings (Fig. 7).

The comparison between the two L shape spaces with different area values and different side length appears to have reasonable difference too. Because of the different size, it is observed that the longest one, even if it is more exposed, meaning more openings, the lighting output values are lower. This is an important observation although obvious one. For instance, in the L shape A with a total area of 144m2 the maximum lighting output value is 94% with 64 openings, in comparison with the L shape B with 240m2 area in which the maximum lighting output value is 91% with 126 openings (Fig. 7).


ISBN - 978-86-7892-807-9

46

Fig. 7. Linear and L shape output space analyses

As it has been mentioned, the users’ experience within the digital spaces is affected by four variables: openings, position, area and light value. Based on these variables, three matrices are created, where each matrix alters different parameters. Matrix 1 (Fig. 8) illustrates six results with different light values by changing the pattern of the openings. More specifically, the number of the openings is 30, while the UC's position is 0 and the area is 80 m². The light values range from 15% to 55%. In Matrix 2 (Fig. 9), the area of space and the light values are altered, while the number of the openings is 30 and the UC's position is 0. As expected, the light value is decreased when the area is increased. When the area value is at the minimum of 32 m², the light value is at the maximum value of 93% and when the area is at the maximum of 112 m² the light value is at the minimum of 15%. In Matrix 3 (Fig. 10) different views are illustrated, obtained by the variation of the UC's position and the light value. In this case, the number of openings is constantly 30 and the area is 112 m². The UC moves from position 0 to position 5, while the light value ranges from 15% to 46%.


ISBN - 978-86-7892-807-9

47

Fig. 8. Matrix 1: Position= 0, Area= 80 m2, Openings= 30

Fig. 9. Matrix 2: Position= 0, Openings= 30

Fig. 10. Matrix 3: Area= 112m2, Openings= 30

Results show that computational technique applied for the production and navigation of an interior space can also be used for interpretation as well as for experimentation in relation to the actual conditions of the space created in digital environment. Although, the movement of the user inside the space is limited due to the external walls, it allows the creation of apertures therein and hence gives a first impression of the relationship that can be established between users, space and openings in actual scale According to the number and/or pattern of the openings and/or the shape of the whole morphology, the space that is projected, is apprehend differently each time. The openings develop different internal lighting conditions, which are presented by a color bar and a percentage on the screen allowing constant interaction with the users. In order to achieve an ultimate control of the integrated technology and of the obtained results, large number of experimentations and case studies are necessary to be undertaken so that a continuous feedback loop control can be achieved, In addition, a series of further experimentation is essential part of the process in order to ensure the reactive integration of the collected data from the physical to the digital space.


ISBN - 978-86-7892-807-9

48

During the unified operation of the systems and the complete setup of the installation, random users were summoned to experience the interactive installation. The first impressions where positive due to the capability of mechanism to transfer physical human behavior and directly correlating this with the digital environment as well as due to the immediate reaction occurs during the design decision-making process. The users’ feedback reaffirms that the proposed system can be applied efficiently to satisfy the architect – client relationship and establish a more direct way of communication in comparison with conventional approaches. This reactive and integrated relationship might be applied during the conceptual design stage of the design process, where the architect can invite the client to familiarize the space, making certain adjustments according to personal desires and choices. The experimentations presented in previous paragraphs lead to a series of observations, both at the level of the production and experience of interior space as well as at the level of the capabilities offered by the specific computational mechanisms.

4 CONCLUSION This work aims to provide to the architects as well as to the clients an alternative way of communication, on the one hand, through physical to digital space creation and, on the other hand, through navigation and selection of openings, both integrated in a single reactive platform of operation. Furthermore, it aims to allow a more effective manipulation of morphology and space, especially by clients, as there will no longer be necessary for the clients to have special skills in order to be involved in the design and to understand spaces that are created for their needs. Finally, this mechanism might allow the real time intervention of users according to their desires and preferences and present results of human-driven alteration in real time.

Further work will be concentrated towards two parallel directions. Firstly, towards the improvement of the existing computation mechanism, keeping in parallel the simplification and easy usage of the platform, and towards the addition of extra parameters, criteria and objectives in order to allow additional possibilities of conceptual design development to be established under a holistic design approach.

ACKNOWLEDGEMENTS This work has been developed based on the results obtained within the framework of the postgraduate course ARH-522 – Advanced Computer-Aided Design Topics in the Department of Architecture at the University of Cyprus during the Fall semester 2015 taught by Odysseas Kontovourkis, Assistant Professor of Architecture. The team members were Ioanna Dimitriou and Styliana Gregoriou, Ph.D. students, and Konstantinos P. Chatzimanolis, Michaelina Stylianou and Philippos Michael, Diploma students. The methodology, the results and the conclusion discussed in this paper are the authors’ view.

REFERENCES [1] Teng, T., Johnson, B. R. (2015). Transformable Physical Design Media. Conference Proceedings, Real Time - Extending the Reach of Computation, 33th eCAADe Conference Proceedings, Vol. 1, TU Wien, Wien, pp. 45-54.

[2] Heydarian, A., Carneiro, J. P., Gerber, D., Becerik-Gerber, B., Hayes, T. and Wood, W. (2014). Immersive Virtual Environments: Experiments On Impacting Design And Human Building Interaction. Rethinking Comprehensive Design: Speculative Counterculture, CAADRIA 2014 Conference Proceedings, Kyoto Institute of Technology, Kyoto, pp. 729-738.

[3] Kontovourkis, O. (2013). Physical Data Computing in Adaptive Design Process. Proceeding of International Conference of Adaptation and Movement in Architecture (ICAMA 2013), Ryerson University, Toronto, pp. 463-475.

[4] Kontovourkis, O. (2014). Computational mechanisms in architectural adaptation and interaction process through human behavior physical inputs. Architecturae et Artibus, Politechnika Bialostocka 19(1), pp. 29-33.

[5] Kontovourkis, O., Alexandrou, K., Vassiliades, C., Frangogiannopoulos, S., Dimitriou, N., Petrou, C. (2015). Adaptive building skin development and wireless gesture-oriented controll. [PowerPoint


ISBN - 978-86-7892-807-9

49

slides] Presented at the 3rd Regionsal International Workshop Computational Morphologies at Polytechnic University of Milan, Milan, 14 May 2015.

[6] Beltran, M. (2011). Geek Run. Accessed: October, 2015 http://mariamari-a.com/?ds-gallery-category=interaction#geek-run-2

[7] Interactive Facade - School of Architecture Potsdam, YouTube Video, Posted by Soheil Bonakdarzadeh, December 6, 2010. Accessed: October 2015 https://www.youtube.com/watch?v=7riee1FWCYA

[8] O'Shea, C. (2013). Woodland Wiggle. Accessed: October2015 http://www.chrisoshea.org/woodland-wiggle

[9] Techviz, Techviz XL. Accessed: October, 2015 http://www.techviz.net/techviz-xl

[10] Lange, B., Rizzo, S., Chang, C. Y., Suma, E. A. and Bolas, M. (2011). Markerless full body tracking: Depth-sensing technology within virtual environments,. Interservice/Industry Training, Simulation and Education Conference (I/ITSEC 2011), Orlando, Florida.

http://mariamari-a.com/?ds-gallery-category=interaction#geek-run-2

https://www.youtube.com/watch?v=7riee1FWCYA

http://www.chrisoshea.org/woodland-wiggle

http://www.techviz.net/techviz-xl


ISBN - 978-86-7892-807-9

50

A DATA DRIVEN APPROACH TO LOCATING PROBLEM AREAS IN THE SMART GRID UNDER A DYNAMIC POPULATION

Vuk Malbasa, Po Chen 1University of Novi Sad, Faculty of Technical Sciences, (Serbia), [email protected]

2Power System Control and Protection Lab, Texas A&M (USA), [email protected]

Abstract As an integral part of any infrastructure, the traditional power grid is designed relying on static

assumptions about the situational, grid and user context. However, grid designs that rely on static assumptions of human behavior imply exploitation limitations because the operational context may change in time. Because power grids are operated near their collapse point, for optimal efficiency, the smart grid demands that the dynamism of moving populations is taken into account during the design phase. In addition to the growing complexity, heterogeneity, and cognitive management logic of smart grids, all driven by market demand, they have become more intricate with rising exploitation of distributed generation devices, such as solar or wind power generators. Here, a grid using cognitive management logic may dynamically allocate otherwise dormant resources and adjust operational policies to compensate for performance variations caused by changes in the situational, grid and user context, in a manner which is transparent to the population. Taking into account the dynamic nature people’s behavior and therefore of power consumption may require the analysis of historical information which is descriptive of the grid’s operation. Our work focuses on modeling the movement of populations for the purposes of intelligent cities while taking into account developing technologies such as distributed generation. We show how to identify problem locations based on population movement data and changes in user behavior, as an application which is transparent to the population.

Keywords: Smart city, smart grid, power systems, fault location, dynamic population.

1 INTRODUCTION The global population continues to grow steadily and a significant amount of growth is happening in cities, thus straining traditional infrastructure. With advancing technological opportunities, a significant amount of resources have been invested in leveraging new applications in order to improve infrastructure within the urban fabric. A critical component of this Endeavour is the concept of smart cities. The potential for adapting the infrastructure to the dynamism of moving populations is a strong opportunity to both increase reliability and security while optimizing efficiency. One facet of adaptation is the utilization of emerging technologies, and an area in which this potential is highly exposed is the transformation of the traditional manually operated power grid towards a more the smart grid.

Urban populations have been academically analyzed as early as 1938 as being a significant segment with features which are unique [1]. The ecological impact of daily movement patterns of urban populations has been studied in [2]. The dynamic nature of urban populations has been recognized as an important aspect of design to urban planners [3]. Here, the authors describe how the mobility of populations should change the traditional static assumptions about “zones” into dynamic knowledge about “flows”.

The Network Dynamics and Simulation Science Laboratory are dedicated to create tools which can be used to illuminate societal problems by developing synthetic information products [4]. Their Portland, OR, data set contains data about the daily movement patterns of the entire city with a population of 1.6 million. The data is realistic in as much as it matches the actual census data exactly, but is completely synthetic. This data set is popular with researchers because the extensiveness, detail and accuracy of information presented are able to match the granularity of automatically harvested data from personal communication devices, and other sensors, while retaining anonymity of the studied population [5]. In order to illustrate this data set, in Figure 1. we show the total number of hours spent at each location in the movement and residential data sets [5].

This kind of analysis may prove useful for differentiating between daily movement patterns such as the commute to work or going to school, and anomalies which may occur such as shifts in the availability




ISBN - 978-86-7892-807-9

51

of DG resources or faults on the transmission or distribution grids. By focusing on the differences between people which are affected and identified, either manually by reporting trouble or automatically, and people who are not affected it should be possible to reveal the trouble location using the an approach similar to overlapping of movement patterns of only those who are affected.

Figure 1. Movement and residential log-density for the EpiSims Portland, OR, data set, overlaid with a map of

counties in the region

2 BACKGROUND Several key technologies allow for the processing of such large data, such as the Portland data set. Mobility is a key aspect of urban populations, whether it is the daily commute, or other kinds of periodical travel.

First, to harvest an enormous amount of data at the population level it is necessary for the data gathering technology to be transparent to the population. With the advent of personal communication devices, such as smart phones and tablets, which are portable and thus often carried around by individuals and also both connected via wireless or cell technology to the internet and location enabled by GPS or cell tower triangulation. The most apparent application of this technology can be seen in Google maps. In this case the application is implicitly used for navigation, while also providing speed and position data to the server. The gathered data can then be used to elucidate actual speeds at which cars using the navigation service are traversing the roads, and thus can be used to show which roads are congested and where the slowdowns occur [6].

The second technology is more closely related to machine learning in that it deals with data representation. Our choice in representing the data is to overlay a regular square lattice across the entire studied area. Then, each person is represented with the time they spend on each lattice location. Even though the city area contains 1.6 million individuals, and we used a 256×256 lattice for over 65 thousand locations, each particular person only visits a handful of locations. In the Portland data each person visits an average of 3 locations and each location is visited by an average of 25 people. For each person, therefore, most locations contain mostly zeros, for places which the person does not spend time in. Places where the person has spent time have values greater than 0.This structure inherent in the data leads to a sparse representation. Therefore while the attributes of our problem are sparse the solution is not. This type of data representation has been studied in the literature and adequate methodologies have been developed in order to exploit this facet of the problem to reduce computational complexity [7].

Thirdly, in the case of generalized linear regression problems which are large and sparse instead of relying on closed form or exact methods, such as the Gaussian elimination with an improbable execution complexity of O(n3), it is possible to obtain useful results using approximate iterative methods [7].

Finally, even though our model is sparse in attributes it is not sparse in the parameters. A large amount of parameters in a model may cause that model to over fit the data. This condition occurs when the model performs well on the gathered data but is of low predictive power for unseen data. This is often termed as the generalization power of the model, or how applicable the results are in the operational context, which is often plagued by noisy and uncertain data from measurements. To


ISBN - 978-86-7892-807-9

52

alleviate this issue we used a regularization method on our model. Regularization in statistical models is a technique to force the model to be conservative in its predictions. In spatial statistics, and within a spatial context, it is useful to introduce a Gaussian Markov random field prior. It shrinks parameters towards the values of their neighborhood on the lattice. The effect is that of smoothing the resulting lattice, while also maintaining conservative estimates in predictions on unseen data.

3 RESULTS We aggregated 1.6 million examples each with 65 thousand attributes into a sparse data set show in Figure 1. In sparse form this data takes up less than 90 megabytes compared to approximately 100 gigabytes it takes in its full form. Under the static assumption, in which the model assumes that all people spend all of their time at their place of residence it is not possible to correctly articulate the dynamic energy needs of a population.

In our first experiment we observe how much the population shifts during the day, according to our residential and mobility data from Portland. Using the residential data as a baseline we note the difference between the two population densities during the day.

Figure 2. Difference in population density obtained by observing the difference between residential and movement

data in the EpiSims Portland, OR, data set

In Figure 2. we note that in the central parts of the city, the maximum change in population density occurs such that approximately 6000 persons who otherwise do not reside in the city center lattice squares move there during the day. Similarly some of the lattice squares which represent places where people reside but do not visit may on average decrease in density by approximately 1000 persons.

For our second experiment, let us assume that we may observe an alteration in the population behavior which is caused by a disturbance in the power supply. There are many such cases which can be identified in various ways, which are beyond the scope of this paper. The alteration is not implicitly reported via distress calls, but it has noticeable effects on the population such that each person is either behaving in their normal fashion, or has altered behavior. In this case the problem can be cast as a regression problem because it should be possible to identify the affected locations from user behavior and population movement data.

In this experiment we artificially disturbed the behavior of people at a location which many people visit, but no one resides in. These areas can occur in areas such as the city centers, airports or industrial zones. Then by matching affected people with their movement data and comparing against the unaffected part of the population we identify the problem locations. In the following figure we compare the result of using the full movement data to analyze those affected with that of assuming no dynamism in the population. We clearly observe in Figure 3. that by using movement data to analyze potential issues with the power grid it is possible to significantly narrow down the location of the


ISBN - 978-86-7892-807-9

53

problem, as only a few most significant locations are highlighted on the right side. In comparison to using movement data, using only static assumptions about the population does not reveal the problem area, but highlights the residences of all of the affected people who’s behavior has been affected. It is significantly cheaper for companies trying to locate problem areas to deal with only a handful of locations revealed by the movement data in the right panel than the rather uninformative locations highlighted in the left panel, which uses a static assumption about populations.

Figure 3. Differences in analyzing data under static (left) and dynamic (right) population assumptions

in order to find the affected region

4 CONCLUSIONS AND FUTURE WORK By taking into account the correct representation of data we have reduced the population density problem size from 100 gigabytes to less than 90 megabytes. This significant reduction is possible by using a sparse representation, which also brings computational advantages. In our second experiment, by leveraging automatically harvested mobility data about populations it is possible to identify problem areas with more precision than if only static assumptions about the population are made. In this case the mobility data reveals the exact location of the problem area. The same problem area is obscured when making static assumptions about the population. This can be significant to the smart grid by taking advantage of mobility data compared to static assumptions traditionally applied. By taking into account the dynamism of moving populations we have shown that it is possible to significantly narrow down the search for problems in the smart grid.

Future work includes using improved computational tools such as the popular multi-core graphics processing units, improving detail in the lattice used for our experiments and including the evaluation of different approaches to detecting unusual behavior in the population.

REFERENCES [1] Wirth, L. (1938). Urbanism as a Way of Life. American journal of sociology, 1-24.

[2] Goodchild, M. F., & Janelle, D. G. (1984). The city around the clock: Space—time patterns of urban ecological structure. Environment and Planning A,16(6), 807-820.

[3] Bertolini, L., & Dijst, M. (2003). Mobility environments and network cities.Journal of urban design, 8(1), 27-43

[4] Bisset, K., Atkins, K., Barrett, C. L., Beckman, R., Eubank, S., Marathe, A., ... & KUMAR, V. (2006). Synthetic Data Products for Societal Infrastructures and Proto-Populations: Data Set 1.0 (No. 06-006). NDSSL Technical Report.

Problem area is obscured using static assumptions

Problem area is revealed by taking into account the

dynamism of moving populations


ISBN - 978-86-7892-807-9

54

[5] Malbasa, V., & Vucetic, S. (2011, August). Spatially regularized logistic regression for disease mapping on large moving populations. In Proceedings of the 17th ACM SIGKDD international conference on Knowledge discovery and data mining (pp. 1352-1360). ACM.

[6] Xu, Y., & Jin, Y. (2002). U.S. Patent No. 6,401,027. Washington, DC: U.S. Patent and Trademark Office.

[7] Saad, Y. (2003). Iterative methods for sparse linear systems. Siam.


ISBN - 978-86-7892-807-9

55

REVISED GENERATIVE STRATEGY IN DESIGN OF ARCHITECTURAL STRUCTURES

Miodrag Nestorović1*, Jelena Milošević2, Aleksandar Ćopić3 1University of Belgrade - Faculty of Architecture (SERBIA), [email protected]

2University of Belgrade - Faculty of Architecture (SERBIA), [email protected] 3University of Belgrade - Faculty of Architecture (SERBIA), [email protected]

Abstract This paper researches the potentials and implications of the application of generative computational tools in the process of architectural design. Assimilation of digital technologies caused significant transformations in the field of architectural design, both in aesthetic and technical terms. The new mediums for design exploration, communication and production, opened up possibilities for propositions of new design strategies. Within this paper, we explore revised generative strategy. In this hybrid approach generative computation is used in the conception phase of the design for production of the first sketches, which are then refined in the process of digital sculpting. In this semi-automated design process, architectural structure emerges as a result of designer-computer interaction, i.e. computational automation is combined with a designer's revision. While automated computational procedures facilitate searching for the solutions by production of alternatives and gradual tuning, the designer maintains its controlling role, eliminating restrictions of completely automotive processes, and intensifying designer-computer relation. The effectiveness of the proposed strategy was tested through pragmatic, action oriented design research. More than an illustration, application in practical design situation represents useful experience.

Keywords: architectural design, design-driver, design-exploration, finding form, generative design.

1 INTRODUCTION This paper reviews the application of generative computational tools in design process, and reports on application of modified generative strategy as design research method. The proposed method combines the application of computational algorithm for performing specific task within the design process and designer's revision of the obtained results. The assessment of the suggested strategy was carried out in the context of the specific design. The design task was revitalization of the currently devastated space bordered by Bridges Gazela and Novi Železnički, Streets Jurija Gagarina, Vladimira Popovića and Bulevar vojvode Mišića in Belgrade.

The revised generative approach is a hybrid process in which application of automated generative systems is restricted to the part of the design process, or on performing particular tasks. In this set, the essence is to master relations between design object, specifications, and constraints. The proposed hybrid strategy is based on computer-designer and objective-subjective relations. Respectively, the design research includes the collection, analysis, and testing of the objective and subjective information. The information is used for construction of dynamic simulation context, they are implemented in the design exploration system. Contrary to the conventional practice, this process has moved from the creation of a design to a creation of a customized tool that meet needs of particular design problem solving process [1, 2]. Definition and implementation of design-tool for generation of forms represents part of the creative design process. The diverse design situations that emerge as the products of the simulations are the basis for the proposition of architectural concept design. This process continues the parametric logic, making possible creation of formal complexities by application of the formation method based on computer code.

Proposed approach is based on the inclusion of information that influence diverse aspects of architecture. It is practically impossible to express parametrically all aspects of jagged design problem environment. To overcome this limitation revised generative strategy implies the application of human-centric interactive design systems. In these systems the computation is limited to performing certain tasks, while designers perform hierarchical definition of parameters, interpretation of the program, comparisons of diverse results, selection, as well as manipulation, control and adjustments of form.


ISBN - 978-86-7892-807-9

56

2 MATERIALS AND METHODS Creation of the structure was realized in the process that included spatial analysis of the context, analysis of the program scheme, the definition of the constraints, synthesis, alternatives, and selection of the most acceptable.

Our first assessment of the context was mental map. It facilitated us identification of connections, and interpretation of the situation characterized by disperse physical structure, spatially and ambientally dominated by natural environment. Insufficient articulated space is characterized by dynamic, arbitrary flow of the activities, lucking in the medium for the critical level of connection (Fig. 1).

Fig. 1. Existing state: dynamic, arbitrary flow of activities

In the further researches of the location, we analyzed and activated fact that it should accommodate and adjust future, potential values. We observed actions of the users in the current context with the aim to comprehend diverse aspects of their interactions. The information about the users' expectations form the location were obtained by questionnaire. The questionnaire introduced a way of an overview from the perspective of users, suggesting problems and potentials. We described physical information, connecting them with the location and intensity of their influence. In this way we obtained a map with coordinates of the attractors and repellers, and their intensity. Collected data were processed, and translated into parameters of the program algorithm used for form exploration.

2.1 Computation and Sculpting The design exploration was realized by the application of automated and manual design explorers (Fig. 2). In the first step the form was generated by developed, automated computer code based on the smart particles. Under the influence of the defined factors, emerged traces described by the motion of the particles. They were inputs for the second sculpting phase, that represents the process of digital modeling of the architectonic structure by manual tunings. This is the phase of adjustments, modifications and refinements of the formal solution obtained by simulation.

The design was generated by the application of the swarm behavior simulation [3]. In simulation we used smart particles, representing persons or groups, and factors that influence them: forces that are products of social behavior, attraction or repulsion, desire, goals, etc. Influence factors represent environmental influences, they initiate and express motions of the particles, which are described by the traces in space and time, while their density suggested potential locations for the new structures.

For computation, we used Processing, open platform based on the Java programming language [4], [5]. This open-source application facilitated modifications and adaptations in the process of creation of custom made design tool. It is important to point on the advantage of openly-shared computational tools, the products of open societies of programmers and users, which mutually exchange and upgrade codes [6]. Concept of ad hock digital tools is strange to the conventional architectural practice which still does not enough exploit the potentials of this recourse.


ISBN - 978-86-7892-807-9

57

Fig. 2. Two stage design process: (a) generating phase (b) sculpting phase

In order to create the generative system, it was necessary to translate ideas into a computer language, to integrate and interconnect in the program existing libraries. Respectively, the most important library is plethoraLibrary. This library deals with the wide spectra of algorithms, including swarm algorithm applied in this design research. The basic elements of the generative systems are agents, attractors, repellers. Agents simulate behavior of a person. Under the influences they move, interact and exchange information. Three main influence factors are cohesion, alignment and separation. They are defined by the radius and the power of the influence. Cohesion is the capability of the agent to generate swarms, separation is opposite function, demonstrating divergency of the agents, while alignment represents ability to exchange information between the agents. Variables are defined by the range from which they can randomly take arbitrary values, enabling us to obtain natural results considering that each person is an individual.

The motions of the agents are defined by a vector. Attractors and repellers are classes of the agents with the certain parameters. They exhibit cohesion or separation behavior in dependence of their purposes. Contrary to other agents that are moving in space they are fixed. Attractors (a�⃗ ) act on the agents in the motion when they enter in their action field. Since agents are represented by the vector of motion, when they enter the attractor vector field we obtain new vector of motion as a result of the summation of these vectors. While it is in the radius of the attractor vector field, the vector of the motion of attractor changes and update each frame. Repeller functions are same as attractor, with the difference that the value of its vector is (r⃗ = −1a�⃗ ).

The final code program has six parts: (1) main program, (2) GUI, (3) MyIsoPt, (4) myPleAgent, (5) myAttractor, (6) myRepeler. The basis of the program are information necessary for its operations,


ISBN - 978-86-7892-807-9

58

related to the graphical representation, line colors, position of the elements, etc. These are the variables that could be defined and that facilitate managing the code. In the program we could define the population of agents and the borders of their emergence and motion. The code contins main functions, variables and supporting functions. Main finctions enable reading agents, attracrors, repellers, they also define the borders of the simulation. Variables represent numerical values and supporting functions enable realization of the functions. The obtained results, path lines, could be exported as video, images, and geometry. Produced forms represent the results of a dynamic process which is unique in its response to set constraints.

The results of the simulations performed by the Processing code, i.e. spatial path lines are used as sweep lines in the process of 3D modeling. In order to generate spatial structure we had to define cross section curves. The defined curves are anthropometric and related to the spatial needs of the future design. The spatial structure was obtained by sweeping cross section curves along sweep lines. The obtained 3D structure, in the phase of digital sculpting, was subjected, adjustments, modifications and refinements. Contrary to the previously automated process this phase represents haptic and intuitive design exploration, which provides a solution that will satisfy demands that cannot be numerically defined.

3 RESULTS The goal of design process was spatial intervention, and creation of the new identity of the location by proposing innovative design of sports and recreation complex. The form and the performance of the architectural structure is regulated by the information on context, their organization, placing and sorting [7]. Design solution explores and interprets the concepts of dynamics and fluidity. It emerges through the empirical research and computer simulation of influences of the location and potential users based on smart particles - agents and forces that determine their motions.

Several basic starting points determined the expression of the design solution. Primary, reaction on the current relation in the space. Dynamic activities and experiences in the space are materialized. By diverting and directing of non-articulated, arbitrary flows, they are translated into architecture, forming enclosure of the building. Their dynamic essence influences the entire design process, since they are reviewed through parameters by which building's shape and structure is generated.

Dynamism and fluidity are processes that are transferred in the design concept. But, dynamism and fluidity though interdependent represent different concepts. The dynamism is created by the forces which influence acceleration and deceleration of the particles. Their flows are further transferred to the architectonic structure. Fluidity could be marked by the easy motion transfer and constant changes of the substrate enabled by its form, composition and physical structure. While dynamics is defined by parameters, precisely determined and controlled by application of graphical programs in the design process, fluidity represents symbolic, spatial, functional and aesthetical characteristic of resulting design solution, striving to be environmental supplement, touristic attraction, unique place of gathering, communication and connection between the Old and New Belgrade (Fig. 3).

Fig. 3. Rendered preview of the design solution


ISBN - 978-86-7892-807-9

59

The building is in symbiotic relation to its surroundings. The imperative was to achieve balanced relation between urban landmark and its surrounding. That was achieved by shaping the building envisaged as a complement of its environment. The communication of the centre and its surroundings is realized by penetration of the public space in the volume of the building. The design represents a material trace of the tension between dichotomies centre/environment, inside/outside, solid/transparent. (Fig. 4)

Fig. 4. Ambient previews

Volumetrically, the complex is composed of two free-form structures. The first structure (Fig. 5), which was subject of detail elaboration, connects Old and New Belgrade. This structure incorporates Stari Železnički Bridge, facilitating transportation and communication between two riversides. This object is planned to accommodate diverse land sports. The second structure is situated on the New Belgrade side. It represents the centre for water sports. Form and the functional adaptability influenced the selection of the structural systems, that responds to the flexibility of the technological concept.

Fig. 5. Model of the architectonic structure

Obtained results confirm the usefulness of the revised generative approach in the design process. The combination of objective and subjective parameters facilitated produstion of innovative design solution. On the other hand, this approach demands considerable time for the analysis, data processing and creation of the code. Although application of computer code in this reserach was restricted, in terms that it was not the only tool and products of the program were not end products of the design process, the attitude pleaded by this research is that there are great potentials for the application of computation as design conception tool.


ISBN - 978-86-7892-807-9

60

4 CONCLUSION Simulation and parametric design tools augmented expression of design autonomy [8], [9]. A number of authors have taken the position that allowing objective external information to determine the values used to generate a design solution alters the very definition of design. On the other hand, standpoint that these technologies should not function as stand-alone devices, limits their role to a sort of sub-system with a specific function, which operation contributes to the functionality of a system for design production.

In order to position role of computational tools in design process, and summarize our intentions, we will exploit notion embedding [10]. Embedding, in this context, represents an insertion of a structure referred to as an embedded system, designed for performing a specific task, within the design process. Herein applied design-driver for computational form-finding of free-forms, represents the embedded system, that operates within the design process.

Though challenging, imposing bottom-up objectivity to design process could be too presume. The manifold nature of the design [11] makes awkward its straightforward algorithmic formulation, or definition of the universal computational design problem-solvers. Even in well defined computationally implemented form-finding processes myriad design choices and decisions remain for the designer during the process. Instead, production of computational tools that handle a specific task within a design process seems appropriate and, currently, more realistic approach.

In the proposed hybrid platform, creation of embedded system represents an integral part of a multi-stage design process. Designers are assigned to identify a problem, set a logic system, and create tool specialized for resolving that specific problem. There are two advantages of creation custom-made design tools. First, in terms of operating since those tools are optimized to handle a specific task, and second, their application enables automated production of variant solutions consequently leading to optimization of the design process itself.

In this semi-automated design process, architectural structures emerges as a result of designer-computer interaction, i.e. computational automation is combined with a designer's revision. While automated computational procedures facilitate search for the solutions by production of alternatives and gradual tunings, the designer maintains its controlling role, eliminating restrictions of completely automotive processes, particularly expressed in the design aspects and situations that could not be exactly formulated.

Approximating design solely in terms of embedding is insufficient because it would consider the architectural formation as an entity that is passively shaped by technical limits and/or environmental constraints. The deliberately limited performability of the embedding should not cause design inclination exclusively to a specific design aspect. Even in design approach sub-ordered to a specific feature that favorites certain attribute, architecture reveals its degree of performance as a function of integration of multiple design aspects.

Maybe, entire design process cannot be reduced to an algorithm, but an algorithm can be embedded to manage certain sequences of design production, to perform particular tasks, without necessarily altering the nature of design. In this hybrid approach computation represents sort of rational core in the compound design process, that influences and directs designer's reflections and decisions. The idea is to preserve the advantages of design research process while using the potentials offered by computational tools.

The application of computational algorithms exceeds traditional CAD paradigms [12], considering that the potential of the computers are more efficiently used in the processes of generation of design solutions on an interactive basis, creating automated systems which anticipate states and respond to their conditions. The optimal architectural solution is the result of the adequate combination of parameters. In this set the task of the designer, positioned between formal precision, soft-control and management, is to define sets of parametric relations, not with the aim of model production, but with the aim of designing through the model. Results are design solutions that are close to figuration and that closely build different ecosystems within social or natural context.

ACKNOWLEDGMENTS The authors are supported by the Ministry of Education, Science and Technological Development of Republic Serbia, Project No. TR36008.


ISBN - 978-86-7892-807-9

61

REFERENCES [1] Galanter, P. (2003). What is Generative Art? Complexity Theory as a Context for Art Theory. Retrieved mart 20, 2014, from http://philipgalanter.com/downloads/ga2003_what_is_genart.pdf

[2] Kilian, A. (2006). Design Exploration through Bidirectional Modeling of Constraints. PhD dissertation. Cambridge, MA: Massachusetts Institute of Technology.

[3] Beni, G., Wang, J. (1989). Swarm Intelligence in Cellular Robotic Systems, Proceed. NATO Advanced Workshop on Robots and Biological Systems, Tuscany, Italy.

[4] Reas, C., Fry, B. (2007). Processing. Cambridge, Massachusetts, London, England: MIT press.

[5] Processing. (n.d.). Retrieved November 7, 2013, from https://www.processing.org/

[6] Milošević, J., Šobić, Z., Nestorović, M. (2014). Towards Generative Convergence in Design of Architectural Structures. Proceedings of 1st International Academic Conference Places and Technology, University of Belgrade Faculty of Architecture, pp. 744-751.

[7] Grobman, Y. J., & Neuman, E. (2012). Performalisam: Form and Performance in Digital Architecture. New York: Routledge.

[8] Mitchell, W. J. (1990). The Logic of Architecture - Design, Computation, and Cognition. MIT Press.

[9] Kolarevic, B. (2005). The Architecture in Digital Age. New York and London: Spon Press.

[10] Milošević, J., Nestorović, M., Šobić, Z. (2013). Computational Morphogenesis: Performance-Oriented Architectural Design Concept. On Architecture – International Conference and Exhibition, Conference Proceedings, Belgrade: STRAND, pp. 555-565.

[11] Lawson, B. (1980). How Designers Think. London: The architectural Press Ltd.

[12] Terzidis, K. (2006). Algorithmic Architecture. Oxford: Elsevier, Architectural Press.


ISBN - 978-86-7892-807-9

62

IMAGE-BASED MODELING OF COMPLEX GEOMETRIC FORMS IN RESTRICTED SURVEYING CONDITIONS – A CASE STUDY OF THE COACH OF METROPOLITAN OF KARLOVCI IN THE MUSEUM OF

VOJVODINA

Isidora Đurić 1*, Jelena Letić 2 1Faculty of Technical Sciences, University of Novi Sad (SERBIA), [email protected]

2Faculty of Technical Sciences, University of Novi Sad (SERBIA), [email protected]

Abstract In this paper, a multi-technique (IBM) image-based modeling approach for creating a realistic 3D model of the highly complex geometric structure of the Coach of Metropolitan of Karlovci was presented. The subject of this research is the Coach of Metropolitan of Karlovci in the Museum of Vojvodina, as an object of a highly complex form, which dates from the 18th century, and represents the only preserved coach on the territory of the former Yugoslavia from that period. In attempt to determine the efficient and inexpensive method for the 3D reconstruction of the complex geometric form, we designed a surveying plan and the workflow, which are based on combining different approaches of object surveying and modeling. The research was based on a bottom up approach, which focuses on the analysis of individual parts of the Coach of Metropolitan of Karlovci and their connection into a whole entity. Combining different surveying and 3D modeling methods, the 3D models of the individual parts of the coach were produced and further they were connected into a whole object. Using the proposed IBM approach for complex geometric forms, the complete 3D model of the Coach of Metropolitan of Karlovci with realistic details was created. In addition, the reconstructed 3D model was analyzed, whereby the current and the original state of the parts of the coach were compared. Furthermore, the demonstrated approach can be adapted according to different needs of the required application.

Keywords: image-based modeling; complex geometric form; 3D reconstruction; surveying

1 INTRODUCTION Image-based modeling (IBM) and three-dimensional (3D) reconstruction of various forms and objects have a wide variety of applications, including visualization, animation, analysis, valorization and conservation. Complex geometric forms with a high level of details are often objects of great cultural and historical value, and as such, they are very important for preservation in terms of 3D reconstruction and virtual representation. Digital documentation of the existing historical and archeological structures [1,2,3], but also of the ruined monuments [3,4] has a great significance for cultural heritage. On the other hand, such complex structures are usually difficult and the most challenging tasks for 3D reconstruction. Many different 3D surveying techniques of complex geometric forms [2-9] have been used over the past years. Related research works concern various methods of image-based modeling and proposals for innovative techniques and technologies, while other focus their attention on their practical application to real cases. However, there is still no fully comprehensive approach that can be applied for a number of different purposes at the same time. In particular, for surveying and 3D reconstruction of complex scenes, the combination and integration of different techniques and technologies orften provides best results [3,5,8]. We noticed that the standard IBM approach fails when applied to surveying of highly complex geometric forms in restricted surveying conditions. In this paper, we designed a surveying plan and the workflow in order to acquire accurate data of the highly complex geometric form of the Coach of Metropolitan of Karlovci in considering its specifics and the restricted surveying conditions.

In order to develop and test the method, we focus on the Coach of Metropolitan of Karlovci from the 18th century, the subject of the permanent exhibition of the Museum of Vojvodina in Novi Sad, as an object of a highly complex structure. The research is based on combining different approaches of the object surveying and modeling, in an attempt to determine the efficient and inexpensive method for 3D reconstruction of a complex geometric form. Object surveying implies measuring in three-dimensional space, where point clouds, as products of surveying, are then employed for object 3D modeling. The


ISBN - 978-86-7892-807-9

63

problem discussed in this paper is based on photogrammetry, as the primary technique for the processing of image data, in order to define the most appropriate method for creating a complete 3D model of a highly complex geometric structure.

1.1 Case study: The Coach of Metropolitan of Karlovci, in the Museum of Vojvodina The subject of this research is the Coach of Metropolitan of Karlovci, as an object of a highly complex form. The Coach of Metropolitan of Karlovci is an enclosed, four-horse-drawn carriage, dating from the 18th century, and it represents the only preserved coach on the territory of the former Yugoslavia from that period. The object is located in the Museum of Vojvodina in Novi Sad, as a part of the permanent exhibition.

The main feature of this coach is its complex form and construction, as well as the high level of detail. As depicted in Fig. 1, the coach consists of several key parts - four wheels, a front and a back part of the coach, and a cabin as a central part of the coach. A particular feature of construction of the coach is its timber structure with highly detailed woodcarvings, as well as the complexity of the interconnections between individual parts of the entire construction. Also, considering that the object dates from the 18th century, it is apparent that the current structural state of the coach is characterized by certain damage and deformations of individual parts. Even a large number of cracks and holes in the structure are clearly visible in photographs.

Fig. 1. The Coach of Metropolitan of Karlovci from the 18th century, Museum of Vojvodina in Novi Sad, Serbia

1.2 The surveying problem with the standard IBM approach Standard IBM approach uses multiple overlapped images and photogrammetric bundle adjustment to find the global shape of the environment or object. The standard IBM approach entails creating a plan of surveying that is designed by taking into consideration the maximum possible distance of camera from the object, geometry of the coach and camera parameters, such as the size of the sensor, the focal length and the direct light, as well as the required conditions for the IBM in the Agisoft PhotoScan software. In this case, camera poses relative to the object were calculated and photographs of the whole coach were taken according to the previously obtained camera parameters. The coach was recorded with a fixed camera height, the maximum possible distance from the object (h = 2m) and moving around it (b = 52.2cm). Camera parameters, computed by a standard IBM approach, were used for the recording: f-stop - f / 3.5, ISO - 400, exposure time - 1/60 sec, and location constraints were taken into account as well. A single set of photographs was obtained for the entire coach.

The complex structure of the building, made up of compound parts that are mutually obstructed, as well as the location constraints, has caused the results obtained by the standard IBM approach to be unsatisfying. Negative results of the 3D reconstruction of the coach demonstrate that the standard method of surveying the whole object does not provide satisfying results. Fig. 2 below shows a 3D


ISBN - 978-86-7892-807-9

64

model of the coach, reconstructed by using the standard IBM approach in the Agisoft PhotoScan software.

Fig. 2. Image-based 3D model of the Coach of Metropolitan of Karlovci automatically generated in Agisoft

PhotoScan software

The result obtained using the standard IBM approach is not satisfying due to:

- The reconstructed geometry of the coach appears to be cut and some important parts, such as back wheels, are missing.

- The reconstructed model has a very low level of details.

By default, PhotoScan uses an automatic reconstruction volume selection algorithm, which can produce undesirable selections in some cases [10]. In this particular case, too large reconstruction volume selection has led to incorrect results, where certain parts are cropped and are not included in the final 3D model of the coach. It is obvious that large dimensions and a complex form of the object contribute substantially to the negative results of the image-based 3D modeling. Furthermore, the low level of details is the result of the unreliable data received after the automatic 3D model processing in Agisoft software. This phenomenon is caused by the location constraints and restrictive surveying conditions, in terms of the lack of space to maneuver around the object, as well as due to the uneven illumination on one side of the coach compared to the other. According to this, it is concluded that the complex geometric form in restricted surveying conditions need to be treated in respect of its key parts and their mutual connections.

1.3 Method Since the standard IBM approach does not provide desirable results for the complex geometric form reconstruction, the object is logically divided into connected segments which are then separately surveyed. The process of surveying and 3D reconstruction of key parts of the object is designed. The proposed IBM approach for complex geometric forms is split into the following main procedure phases, which are based on combining several photogrammetric techniques:

• Surveying the individual parts of the object.

• Surveying the 3D coordinates of the whole coach, in order to determine key connection points between parts of the object. In this step, the PhotoModeler software is used.

• 3D reconstruction of the individual parts of the object. The 3D model of each single part of the coach is automatically generated in Agisoft PhotoScan software.

• 3D reconstruction of the whole object. From the step 2, the referent positions of individual parts have been obtained, after what individual coach segments are manually connected to a single entity, by using 3D modeling software – 3ds Max.


ISBN - 978-86-7892-807-9

65

2 SURVEYING OF THE COACH OF METROPOLITAN OF KARLOVCI Further research is based on a bottom up approach, which focuses on the analysis of individual parts and their connection into a whole entity. In the first phase, the Coach of Metropolitan of Karlovci is surveyed. The surveying of the coach is split into these two steps:

• Surveying the individual parts of the coach. • Surveying 3D coordinates of the whole coach, in order to find key connection points for joining

the parts into a single object.

2.1 The surveying of the individual parts of the coach

The whole structure of the coach is logically divided into these key connected segments (Fig. 3):

• the front and the back wheels, • the cabin – the central part of the coach, • the front and the back part of the coach, which form interconnections between the wheels

Fig. 3. The coach with marked characteristic segments

The main feature of each individual part of the coach is a complex form, as well as a specific state of the structure, characterized by certain damages and deformations. Wheels are elements with the highest degree of deformation and a large number of cracks and holes, clearly visible in photographs. The central part – cabin, is characterized by specific materialization, which largely consists of painted surfaces and a transparent glass area. The front and the back part of the coach are elements with the most complex construction (timber structure with highly detailed woodcarvings), as they also form interconnections between the wheels.

Each of the divided parts of the coach was surveyed separately according to the previously obtained camera orientation and the position relative to each of them. In the project, the tripod and the camera NIKON D7000 (pixel size - 4.78μm, sensor size - 23.6x15.6mm, focal length - 18mm) were used, with the manual settings of the following parameters: F-stop, ISO speed, Exposure time. For each single part of the coach, the camera parameters, such as distance from the object (h) and moving around it (b), were calculated, while, due to the size and the complexity of construction, all parts were captured with two different camera heights. Table 1. displays the camera parameters for the individual parts surveyed.

Table 1. Camera parameters for the individual parts surveyed

Camera parameters: The front and the back wheels

The front and the back part of the

coach

The cabin

F-stop f/8 f/8 f/8 ISO speed ISO – 250 ISO – 250 ISO – 250

Exposure time ½ sec 1/1.6 (0,62) sec ½ sec h 1 m 1.5 m 2 m b 25.6 cm 38.4 cm 52.2 cm


ISBN - 978-86-7892-807-9

66

The values of F-stop – f/8, ISO – 250, and Exposure time – ½, are optimum settings for the most suitable picture brightness, while the settings for the Exposure time were occasionally changed in accordance with the shifting luminosity. The values for the distance from the object (h), moving around it (b), and camera height were calculated according to sizes and dimensions of individual parts of the coach. Therefore, lower parts of the object such as the front wheels, were captured with a closer distance and less degree of moving around it, while the highest part of the structure – the cabin, was captured with the maximum possible distance of the camera from the object. We noticed that the greater freedom in using and changing different parameters (h, b, and Exposure time), had an influence on better final results of the object surveying and modeling in the comparisons of the negative results obtained by the standard surveying method, earlier applied on the entire coach.

Furthermore, a considerably higher number of photographs for each individual part of the coach have been produced, as opposed to the surveying method previously applied in the section 1.2. That has significantly contributed to a more precise data acquired during the further work on processing the photographs. For each series of photographs, masks were manually applied to the photographs in the Photoshop software, in order to specify the areas on the photographs which can otherwise be confusing to the Agisoft PhotoScan software or lead to incorrect reconstruction results. The series of photographs, taken for every individual part of the object, are in the further work used for their automatic modeling.

2.2 The surveying of 3D coordinates of the whole coach In this step, the photographs of the whole coach obtained by the standard surveying method, applied in the preceding section 1.2., are imported into the photogrammetric modeling software, PhotoModeler. In order to determine key connections between parts of the objects, reference points that indicate accurate relative positions of the parts of the coach have been recognized on the images. Fourteen key points in total are manually marked, including four points on the axis of each wheel, six points on the cabin, and two points both on the front and the back part of the coach. In order to avoid mistakes in estimating the accurate relative positions of the parts of the coach, the points which are uniformly arranged across all parts of the structure and easily identifiable and separable on the photographs were selected. The same tie points are marked around the construction of the coach in several photographs, while each point is recognized in the minimum of two photographs. The measured image correspondences (tie points) between the parts of the coach are used to determine key connection points for joining the individual parts into a single object. In this way, the produced frame of points is exported to the .obj format required for the 3D modeling software, 3ds Max. The frame of points and the automatically generated 3D models of the individual parts are further used for the 3D reconstruction of the entire coach.

Due to the policy of object conservation, in the process of tie point identification, coded targets have not been placed to the object, yet the points were selected manually in PhotoModeler software, which affected the value of the error in the process of estimating the accurate positions of the points (max Residual > 1, 5). After completing the data acquisition and image processing phases of surveying, the next step of the proposed workflow is 3D model generation which is described below.

3 MODELING THE COACH OF METROPOLITAN OF KARLOVCI 3D model generation is a process of converting image correspondences (2D coordinates of points) into 3D coordinates through a mathematical model, and in that way producing 3D point cloud. Furthermore, a polygonal model (mesh) is generated to produce the best 3D representation of the surveyed object. In addition, texture-mapping is performed for a photo-realistic visualization of the 3D results. The procedure of modeling individual parts of the coach and their linking into a whole object is divided into two main steps, based on combining different techniques for 3D reconstruction, which are described below.

3.1 3D reconstruction of individual parts of the coach The series of photographs for each part of the coach, gained by previously applied surveying method, are separately imported into photogrammetric modeling software, Agisoft PhotoScan. In order to avoid incorrect reconstruction results during the process of 3D model generation, the masks previously applied to the photographs in Photoshop software are also imported. Afterwards, 3D models of


ISBN - 978-86-7892-807-9

67

individual parts of the coach are automatically generated. Also, the textures from the images are extracted automatically. In that way, the textured mesh for each individual part of the object is created. The 3D models of individual parts of the coach are shown in Figs. 4-6. In order to clean the models of redundant faces, each produced mesh is manually filtered by removing polygons in Agisoft PhotoScan software. The final models are thus exported to the .obj format required for 3D modeling software, 3ds Max, where they are manually connected into a single entity. The following sections present the results of the reconstruction of the individual parts of the coach including the estimation of error.

Fig. 4. 3D reconstructions of the front and the back wheel

Fig. 5. 3D reconstructions of the front part and the back part of the coach

Fig. 6. 3D reconstruction of the cabin

3.1.1 The wheels The results of modeling the wheels show a significant level of accuracy and details. A damage on the structure and visibly expressed cracks and holes, which greatly correspond to the real state of the object, can be clearly notice on the reconstructed 3D models of wheels. Since the wheels are the most


ISBN - 978-86-7892-807-9

68

extended and most clearly defined part of the coach, with the lowest level of interweaving with other elements, it can be concluded that the complexity of structure and surrounding conditions significantly influence the final results of modeling. Specifically, surrounding conditions, in terms of space limitations to maneuver around wheels, as well as different illumination for each individual wheel, have influenced the fact that the accuracy varies between these elements i.e. better results were obtained for the small wheel on the right side of the structure in comparison to the wheel on the left side (Fig. 7). In addition, lower estimated errors are calculated for the wheels on the right side of the object, while the lowest estimated error is obtained for the front wheel on the right side of the coach, which indicates that the surveying conditions, such as adequate lighting, contribute to better outcomes.

Fig. 7. 3D models of a small wheels

The problem with a limited movement around elements of the coach during the surveying process is most clearly expressed on the 3D models of large wheels. From these results it can be noticed a distinct accuracy of the reconstruction of the front parts of the wheels, as well as of a part of the back side, where it is possible to access during surveying. On the other hand, the 3D reconstructions of the back parts of the wheels that are obscured by the rest of the structure are quite incomplete (Fig. 8).

Fig. 8. 3D models of the large wheels

3.1.2 The front and the back part of the coach 3D representations of the front and the back part of the coach greatly correspond to the real state of the object. The front and the back parts of the coach are characterized by highly detailed woodcarvings and the most proficient level of detail in the structure, in terms of links between structural joints, which can be seen from the obtained 3D models. In this case, the accuracy of the reconstruction, as well as the low estimated error, can be linked to the most favorable surveying conditions, including a greater distance from the object and available space to maneuver around it.


ISBN - 978-86-7892-807-9

69

3.1.3 The cabin Cabin, the central part of the coach, surrounded with other individual elements (e.g. fabric in front of the cabin), represents the element with the most unfavorable surveying conditions. Therefore, the estimated error for this element is the highest compared to other parts of the coach. Unfavorable surveying conditions such as limited distance and movement around the object are conditioned by the area of the room and the central position of the cabin in relation to the overall structure. In addition, the resulting model was also influenced by its materialization. The cabin consists mostly of a painted structure and transparent glass surfaces which, due to the reflection of light, does not represent a favorable element for the surveying and reconstruction process. According to that, glass surfaces, as a limiting factor in the surveying process, are removed from photographs, whereas the painted surfaces greatly correspond to the real state of the object.

3.2 3D reconstruction of the whole object The creation of the 3D model of a complex geometric structure is usually the most difficult task for the automation process. Therefore, the image-based 3D modeling typically requires some user`s interaction. In this step, the automatically generated models of individual parts of the coach are manually connected into the whole complex form. The 3D model of the coach is created using 3D modeling software, 3ds Max. The referent key connection points (tie points) used for joining the individual parts into a single object are obtained by previously surveying the 3D coordinates of the whole coach within software PhotoModeler. The file with the tie points is imported into 3ds Max software, and further, each individual part of the coach is added to the scene by using the Merge option. Fig. 9 shows the imported frame of tie points and individual parts of the object in 3ds Max software. The pivot points of all parts are adjusted to the positions of tie points by using the Align option. Finally, all parts of the coach are joined into a single mesh (Fig. 10).

Fig. 9. Imported tie points and the individual parts of the coach in 3ds Max software


ISBN - 978-86-7892-807-9

70

Fig. 10. The mesh of the coach created by joining individual parts into single entity using tie points

Dimensions of each single part are scaled and adjusted in regard to dimensions of the most accurately generated element, e.g. element with the smallest value of error – the front wheel on the right side of the coach. 3D model of the whole coach is created through this process, and it represents the starting point for potential subsequent analysis and implementation (Fig. 11).

Fig. 11. 3D model of the coach, represented in 3ds Max

Low level of precision occurs in the process of adjusting the dimensions of all parts to the scale of the obtained 3D model (small wheel on the right side) and in the process of placing the parts into the obtained frame of reference points where minimal changes in position and the degree of rotation of each individual part contribute to the fact that relations between the parts on the reconstructed 3D model of the coach do not fully correspond to the real state of the object. It is assumed that this phenomenon can be linked to the methodological procedure of surveying the individual parts, whereby elements in focus give more accurate results compared to the surrounding; thus, there is a divergence of correct positions of elements in the connection process. In addition, the accuracy of the resulting frame of reference points is not guaranteed due to the limited performances of equipment used for the surveying, i.e. the lack of equipment such as laser scanner, projector for mapping object with the grid of points and additional adequate lighting.

4 RESULT AND ESTIMATED ERROR 3D model of the coach is obtained by using the IBM approach for complex geometric forms. 3D models of the individual parts of the coach are gained by the process of automatic modeling within the


ISBN - 978-86-7892-807-9

71

software Agisoft PhotoScan. The result is a model with a significant level of accuracy and congruence with the real state of the object.

Based on results of the 3D reconstruction of the parts of the coach, it has been observed that individually generated models provide far more precise results compared to the model of the whole coach. Using the comparative analysis of the real state of the individual parts of the coach and the obtained 3D models of these parts, errors were estimated after the entire 3D reconstruction of the object. Recognizable points of the structure (holes in the structure or targets set on certain parts) were determined on each of the individual parts of the coach, which were used as a reference points for measuring the real distance between them. These points were also marked on the reconstructed models of the individual parts within software Agisoft PhotoScan, which we used to obtain scale i.e. ratio between the size of the pixel on the sensor and the pixel size in nature. GSD (Ground Sample Distance) value is calculated for each of the parts of the coach in the introductory calculation, while the pixel size is obtained after the reconstruction of each element. The difference between these two values, expressed as a percentage, constitutes an error which varies according to the part surveyed. Estimated error for surveying the entire coach is 48.33%, while the values of estimated error are significantly lower for the individual parts of the coach (Table 2).

Table 2. Estimated errors for the reconstructed parts of the coach

Part of coach Estimated error (%) Cabin 11.49 %

Front part 1.11 % Back part 0.48 %

Small wheel on the right side 0.14 % Small wheel on the left side 0.83 %

Large wheel on the right side 2.09 % Large wheel on the left side 3.02 %

From Table 2. it can be noted that there are differences between percentages for the parts which were surveyed with the same camera parameters (front and back wheels). This phenomenon can be linked to the lack of uniform illumination on one side of the coach compared to the other side. Considering this, it can be concluded that the higher intensity of illumination with an unfavorable light direction, in this case - a combination of natural and artificial light on the left side of the coach, adversely affects the process of surveying the object, while the right side with the focused, non-diffuse lighting provides better results.

5 ANALYSIS OF THE RECONSTRUCTED 3D MODEL 3D reconstruction of the whole coach is created after the process of connecting individual 3D models into a single entity. This reconstructed 3D model can be used for various purposes, depending on the needs of the application. Obtained orthophoto images, in contrast to photographs that have a perspective deformation, allow direct measurement and analysis of the object (Fig. 12).

The value for the pixel size, which is obtained in the exporting process of orthophoto images from software Agisoft PhotoScan, in the case of precisely generated model, should be equal to the value of the GSD, obtained in the introductory surveying calculation. On the basis of the known value of the pixel size and orthophoto images, it is possible to measure deformations of the individual elements.

In this particular example, where the parts of the coach are separately generated, the difference between those values, and therefore the accuracy of the orthophoto images, varies from 0.14mm (for the front right wheel) to 11.49mm (for the cabin), depending on the part of the coach. According to this, it is concluded that the maximum accuracy that could be achieved for the model of the whole coach is 0.14mm deviation from the real object.


ISBN - 978-86-7892-807-9

72

Fig. 12. Orthophoto image of the connected parts of the carriage

Furthermore, for the purpose of the evaluation of the current state of the object in comparison to its original state, a comparative analysis for two front small wheels is made, within software Compare Cloud. If it starts from the assumption that the constructions of these two wheels were the same size (the range and volume), by comparing the generated 3D models (point cloud or mesh), it is possible to obtain values that indicate the deviation of certain parts of the elements. Using four reference points, with the semi-automatic option Match Scale and the automatic option Align inside Cloud Compare, 3D models are scaled and aligned. The mean square error is obtained in this process, i.e. the root-mean-square deviation RMS=0.0129076. Deviations for the points i.e. meshes of these two wheels were measured with the option Distances (Cloud/Mesh). The results are presented in the diagram for the overall mean value of the error (Fig. 13) as well as with the graphic representation of overlapping wheels with highlighted points that do not match (Fig. 14).

Fig. 13. Diagram for the overall mean value of the error


ISBN - 978-86-7892-807-9

73

Fig. 14. Graphic representation of the overlapping wheels

Analyzing these results, it can be concluded that the points on different parts of the wheels (around the flange and on the bars) deviate in values up to 0.08m, i.e. if it starts from the assumption that the wheels used to be identical, eventually it came to different deformations of one wheel compared to the other.

6 CONCLUSIONS The produced model of the Coach of Metropolitan of Karlovci demonstrates the use of IBM in the accurate 3D reconstruction of a complex geometric form. A multi-technique IBM approach for creating a realistic 3D model of a highly complex geometric structure is presented. It combines different methods of object surveying and modeling, with the aim of determining the efficient and inexpensive method for the 3D reconstruction of a complex geometric form in restricted surveying conditions.

Firstly, we applied a standard IBM approach for creating a 3D model of the whole object. Negative results of 3D reconstruction of the coach demonstrated that the standard method of the surveying of the whole object does not provide satisfying results. Therefore, the research was based on a bottom up approach, which focuses on the analysis of individual parts of the object and their connection into a whole entity. Using the demonstrated IBM approach for complex geometric forms, the complete 3D model of the coach with realistic details was obtained.

We noticed that the complexity of the structure, as well as the restricted surveying conditions, including location constraints, the illumination or ambient light problems, and the position of the object in relation to other elements, significantly influence the level of accuracy and the value of error of the final 3D reconstruction. The element with the smallest value of error is a front wheel on the right side of the coach. The wheel takes advantage of its position, which is characterized by the most suitable ambient light in comparison to positions of other elements. In addition, wheels are the most extended parts of the coach construction, with visual features clearly visible in the images, which also contributed to the better 3D reconstruction result of this particular part in relation to others. Furthermore, in the 3D models of the front and the back part of the coach, the high level of correspondence with the real state and small value of error are obtained. The front and the back part of the coach are characterized by the highest level of complexity and detail in their construction, while the good location conditions, such as a sufficient distance of camera from the object and an open space to maneuver around it, are in this case most suitable. Considering that, it is obvious that appropriate surveying conditions largely influence the precise results of 3D reconstruction. The part of the coach with the highest value of error in 3D reconstruction is the cabin. This phenomenon is mainly caused by restrictive surveying conditions and the unfavorable position of this element in relation to others, but also due to the complex materialization of the object, featured by painted surfaces and transparent glass area.

Afterwards, by combining different 3D modeling methods, the individual parts were connected into a whole and the 3D representation of the Coach of Metropolitan of Karlovci was created. The obtained 3D model provides much better final results than the attempt of modeling by using the standard IBM


ISBN - 978-86-7892-807-9

74

approach. According to that, we concluded that in the case of modeling of complex geometric form the use of various softwares gives better outcome than the standard IBM approach.

The goal of future research is to include more software for deformation measurements in order to gain more precise referent points for joining individual parts into a whole object. In this way, an improved 3D model that corresponds perfectly to a real state can be optimized and adapted for various applications.

REFERENCES [1] Yilmaz, H., Yakar, M., Gulec, S. & Dulgerler, O. (2007). Importance of Digital Close-range Photogrammetry in Documentation of Cultural Heritage. Journal of Cultural Heritage 8(4), pp. 428 - 433.

[2] Krasić, S. & Pejić, P. (2014). Comparative Analysis of Terrestrial Semi-automatic and Automatic Photogrammetry in 3D Modeling Process. Nexus Network Journal 16(2), pp. 273–283.

[3] Stojaković, V. & Tapavčević, B. (2011). Image-based modeling approach in creating 3D morphogenetic reconstruction of Liberty Square in Novi Sad. Journal of Cultural Heritage 12, pp. 105 - 110.

[4] Lingua, A., Piumatti, P. & Rinaudo, F. (2003). Digital Photogrammetry: A Standard Approach to Cultural Heritage Survey. The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences XXXIV(5), pp. 210-215.

[5] Remondino, F. & Rizzi, A. (2010). Reality-based 3D Documentation of Natural and Cultural Heritage Sites – Techniques, Problems, and Examples. Appl Geomat 2, pp. 85-100.

[6] Voltolini, F., Remondino, F., Pontin, M. & Gonzo, L. (2006). Experiences and Considerations in Image-based modeling of Complex Architectures. IAPRS XXXVI(5), Dresden, pp. 25-27.

[7] El-Hakim, S.F., Beraldin, J.-A., Gonzo, L., Whiting, E. (2005). A Hierarchical 3D Reconstruction Approach for Documenting Complex Heritage Sites. XX CIPA Int. Symposium. Torino, Italy.

[8] El-Hakim, S.F., Beraldin, J. A., Picard, M. & Godin, G. (2004). Detailed 3D Reconstruction of Large-scale Heritage Sites With Integrated Techniques. IEEE Computer Graphics and Applications 24(3), pp. 21-29.

[9] El-Hakim, S.F., Beraldin, J. & Picard, M. (2002). Detailed 3d Reconstruction of Monuments Using Multiple Techniques. Proceedings of the CIPA WG 6 International workshop, Corfu, Greece, pp.58-64.

[10] Agisoft (2013). Agisoft PhotoScan User Manual, Professional edition, Version 1.0.0. Available at: http://www.agisoft.com/

http://www.agisoft.com/


ISBN - 978-86-7892-807-9

75

THE APPLICATION OF DIGITAL TOOLS AND SMART MATERIALS IN THE CREATION OF ADAPTIVE SYSTEMS IN ARCHITECTURE

Jelena Kicanovic1 1University of Novi Sad (SERBIA), [email protected]

Abstract The present study deals with possibilities of application of adaptive systems and their materialization with smart materials, evoking the evolutionary processes that can be in the natural environment. A special segment of research deals with implementation of adaptive systems on the facades of buildings and design strategies based on properties of electro-active polymers.

Keywords: adaptive architecture, smart materials, electro-active polymer

1 INTRODUCTION Kinetic structures have greater application in architecture. The reason for this lies in the fact that with such structures we can achieve not only the dynamics of the facade, but also a better response to the impacts from the environment. The concept of interactivity, indicates a specific direction in the development of architecture as a new media practice that insists on systems that learn from their environment and users with whom they have a conversation, while programs represent only a virtual link. The application of adaptive systems, includes ability to react on specified terms, whose users - entrants are generating a new physical structure. The aim of this research is to explore possibilities in the application of smart materials in contemporary architecture and its implementation in adaptive facade systems.

In most cases development of adaptive structures in architecture used elements of mechatronics - sensors and actuators [1]. This commits better control of the movement. On the other hand, Reichart and others [2,3] used natural systems to generate the adaptive systems. The advantage of such systems [4] is that they are independent of the engine, following biomimetic intelligence [5] rather than technology principles. Some authors use smart materials for ultra-lightweight structures [6,7]. On the contrary, this paper shows the possibility of integrated application of smart materials with parametric control of moves.

This paper provides an overview and description of the independent adaptive system, which adjusts to changes in the environment, using the characteristics of the used material. The research is based on the use of smart materials, which do not require the application of mechanical principles. In this research design strategy based on parametric modeling and energy performance optimization are used for creating an adaptive facade system.

2 MATERIALS AND METHODOLOGY With unique combination of properties, certain materials provide a wide range of functional characteristics, without losing its performance, covering all the defined requirements, as input parameters, performing the integration with the environment, representing the link between a traditional context and creative pragmatic cases in order to simplify various physical systems. Despite recent innovations in the field of smart materials, the concept of a dynamic flexible architecture has remained on conventional approaches, which consist of the use of external mechanisms, transforming the structure. Using the engine, means dynamics, which is created based on the articulation of rigid parts. With today's innovations in the field of engineering materials [8], created an opportunity, examines the way in which kinetics can be applied to architecture, so that it becomes a truly flexible in their midst. Create adaptive, dynamic system that meets the needs of their potential users, implies a certain approach to solving problems, such as the materialization. So one of the proposed solutions was a electro-active polymer (EAP) - an actuator, which transforms electricity into mechanical power. EAP is defined layered structure, where the acrylic layer located between the two electrodes by changing the


ISBN - 978-86-7892-807-9

76

voltage, changing the form of the generated modules. In particular, the application of lightweight biodegradable material, such as soft dielectric electro-active polymer [9] is achieved by creating energy efficient and similar structures for microclimate in which it is located, but also beyond. This study raises further research, creating the possibility of direct control and activation of the material as a tool. Research directly involves in the environment, morphology of materials, defined characteristics. Materials due to their characteristics provide new opportunities and possibilities and solving the problem, using the environmental parameters like initiators. This work includes a case study, a practical application of adaptive systems, which are the structures generated through predefined input parameters of the evolutionary system. For derived facade, implementation of adaptive algorithms presents the best solution. Derived facade has dynamic segments, which enables collecting a small and large amounts of sunlight. To do this, it is necessary to assign the facade appropriately and devise a mechanism of displacement.

The definition of research is divided into several stages. Each of them solves a specific problem in a most efficient way, to come to the desired form.

Difficulty begins by defining surfaces, which clearly highlights the form of physical structures. Following the division of the tread, ie. surface on a rectangular grid is defined by a number of panels, as well as their dimensions. Identifying a reference point or surface, defines the creation of sectional panels as modules of generated form. Further decomposition of modules, allocates the reference geometric shapes, which interact with the point attractors - which define the criteria for displacement of facade.

In this paper is proposed parametric model for adaptive facade that can be applied on various buildings and whose application to objects depends on variable parameters, enabling the development of interactive architectural systems, whose constituents - panels shown through parameter define dimensions and shape.

Parametric model was made with the help of an algorithm for generating Voronoi cells [10]. The density of the points, the distance between and the size to which it is applied, determines the shape and size of the cells (Fig. 1). The next step is to create an axis around which panels bend. This is the only fixed part of each element, which incurves along the defined axis.

Fig. 1. Variation of facade cells, depending on predefined parameters

Panels, defining the form of the grid network, which is modified in relation to the impact of direct solar radiation, according to the parameter which defines the rotation of angles of integrated modules.

The process of generating the form of facade panels, presents the idealization of the physical system, which weighs in nature - black body, absorbs and emits all the energy it receives. It creates a physical system of facade panels, which absorbed solar energy converts to kinetic, forming movable modules whose mobility, as well as the angle of opening or closing, depends on the amount of absorbed solar energy (Fig. 2).


ISBN - 978-86-7892-807-9

77

Fig. 2. Performance analysis

Surface, defines the finality of a physical space, divided in the following way - planar surface, means creating rectangular grid, where the actuators in the form of defined points, are changing gradually the morphology. There is a clear transition from properly divided surface into the irregular surface, characterized by newly Voronoi cells (Fig. 3).

Fig. 3. Visualization of the concept

The logic of generating specific structures can be defined using an algorithm, as a digital media in architecture, by which the transfer made formulation of the evolutionary process of adaptation in nature, the case of specific physical system, seeking its form in real time. Striving for the interaction of architecture and its users, thus obtained, using new materials, creating a biomimetic models driven by passive systems and improve access to the design of architectural form (Fig. 4).


ISBN - 978-86-7892-807-9

78

Fig. 4. Spacious display adaptive system through visualization materialization of the structure and its integration with the environment

3 PERFORMANCE SIMULATION One of the most important aspects, which have been taken into account in the design of the physical system is the reduction of energy consumption and increase of the comfort space, which interacts with both its potential users and with microclimate that surrounds it. Accordingly, the emphasis is put on different approaches such as implementation of the alternative sources of energy - solar energy, as discussed in this paper and the principles of interaction with the environment, manipulating the structure which becomes an equal participant. In this way, the physical systems are formed with intention to create a more intimate whole. To be designed by independent physical system of this type, access to modern architectural context, involves the use of adaptive systems like software platform, as well as the receptors, which react to changes in the environment, thereby creating a defined vision of the input data, where after processing they performed interaction with the environment, mainly using electricity as a driver-generated forms. Specifically, this work is based on the use of smart materials - independent structures, whose characteristics allow its operation as an independent, self-sustaining system. The newly formed facades, represents a dynamic system, which will have movable panels, according to the amount of incident solar radiation if they are opened or closed. This will be created indoors to provide the necessary illumination. Depending on the input parameters, data analysis is performed, generating the desired form, which meets defined criteria.

Dynamic physical structure finds purpose through the controlled lighting inside the building. It is necessary to analyze the amount of illumination on the outside of the facade, as well as the amount of light, which is obtained due to the openness. Then, it is necessary to define the obtained amount of brightness in the interior. Based on the data obtained, it should allocate the necessary openness of the facade. The resulting amount of light, should serve to further analysis, which will be followed by defining an algorithm to get the rotation angles of the facade panels. The application of alternative sources, such as solar energy is needed to show how the analysis of insolation gets interactions of physical structure - facade panels, with the environment as well as its environmental values, given through the graphic visualization, distribution of temperature values, for the longest and the shortest


ISBN - 978-86-7892-807-9

79

day of the year (Fig. 5, 6). Plugin - DIVA, performs simulation of insolation of defined area - 'the cube', certain characteristics, referring to the materialization of decomposed surfaces, such as walls, ceiling, floor and glass windows. During the longest day of the year - in summer period, the results for a given position of the reference curve, which is defined by the angle of rotation and openness facade panels, show a large percentage of insolation in room (Fig. 5), over 60%, while during the shortest day of the year - in the winter period (Fig. 6), as a result of a small angle of incidence of sunlight, compared to the shifted reference curve - attractor, presents the obtained results, which define the necessary comfort in terms of sufficient amounts of insolation during the day.

Fig. 5. Analysis of insolation of the interior during the longest day of the year, through the graphical display false color scale and their distribution of the generated surface

Fig. 5. Analysis of insolation of the interior during the shortest day of the year, through the graphical display false color scale and their distribution of the generated surface, where the observed uniform distribution of insolation

analyzed area

4 CONCLUSION In this research, application of smart materials and parametric design are examined. This paper studies the application in the different ways - the use of digital tools and smart materials in architecture, which create an adaptive systems, as a means of interaction of architecture and its users is developing a dynamic integration in the process of creating physical models, such as the surface of the facade, whose cell or facade panels can be parametrized and thus respond to various architectural requirements. Under the dynamic approach, it is understood, not only as living motion patterns, but also the implementation of biomimetic models, which is an essential factor in the creation of the final structure. Justification and importance of form, stems from the fact that seeks to establish equilibrium with the natural environment and the time and necessary energy efficiency, thanks to which the projected structure, referred to as justified self-sustaining. The idea is, the creation of a common space while respecting the context in which it occurs, since the entire natural environment, gives the


ISBN - 978-86-7892-807-9

80

story, the status of reality. Generating the form of facade panels, is a symbolic representation, clamps of man and nature that surrounds it.

REFERENCES [1] LIFT architects The Air Flow(er), 2013. http://www.liftarchitects.com/air-ower/ (last ccessed: 23 January 2015 18:14)

[2] S. Reichert, A. Menges, D. Correa Meteorosensitive architecture: Biomimetic building skins based on materially embedded and hygroscopically enabled responsiveness, Material Ecology, 2015; 60: 1427

[3] J. Duro-Royo, K. Zolotovsky, L. Mogas-Soldevila, S. Varshney, N. Oxman, M. C. Boyce, C. Ortiz MetaMesh: A hierarchical computational model for design and fabrication of biomimetic armored surfaces, Material Ecology, 2013; 60: 5069 [4] N. Oxman Material-based design computation, Massachusetts Institute of Technology (2010) [5] M. Hensel, A. Menges, M. Weinstock Emergent technologies and design towards a biological paradigm in architecture Routledge, Oxford (2010) [6] S. Reichert, T. Schwinn, R. La Magna, F. Waimer, J. Knippers, A. Menges Fibrous structures: An integrative approach to design computation, simulation and fabrication for lightweight, glass and carbon fibre composite structures in architecture based on biomimetic design principles, Computer Aided Design, 2014; 52: -3927 [7] Jim Rhone A Deegre of Freedom, 2014. http://dk-digital-knowledge.com/2014/07/diploma-a-degree-of-freedom-jim-rhone/ (last ccessed: 13 October 2015 11:28) [8] J. Knippers, J. Cremers, M. Gabler, J. Lienhard Construction manual for polymers + membranes Birkhuser-Verlag (2011) [9] Kretzer M. Towards a new softness. In: Proc. int. conf. adapt. archit. 2011. [10] A. Okabe, B. Boots, K. Sugihara, S.N. Chiu Spacial tesselations: concepts and applications of Voronoi diagrams Wiley, Chichester (2009)


ISBN - 978-86-7892-807-9

81

DESIGNING AND FABRICATION OF ADAPTIVE FAÇADE BASED ON ORIGAMI PATTERN

Marko Vučić University of Novi Sad, Faculty of Technical Sciences, Department of Architecture and Urban

Planning, Novi Sad, Serbia BSc., Post-graduate student, [email protected]

Abstract Facades are an equally important part of a building as its construction and function. It should be simultaneously thought through, as it has the most influence on its surroundings. Nowadays, new attention has been brought to improvement of building’s value. New materials enhanced by new technologies give the opportunity to improve the functionality and aesthetics of built structures.

The main issue for traditionally designed facades is their immobility and invariability. The utilization of new technologies such as robotics, mechatronics and smart materials can solve these issues and can give contribution in energy efficiency. The most valuable contribution is the interactive versatility of form and in that manner the change of its conventional structure.

This paper indicates how implementation of parametric design, kinetics and scale model can improve research and producing of new foldable forms inspired by origami patterns.

Keywords: Façade, folding, parametric design, mechatronics, origami pattern

1 INTRODUCTION Adaptive shading systems in architecture often refer to deployable structures which have the ability to respond to changing environmental conditions (e.g. solar radiation) by mediating external loads and internal demands (e.g. light incidence). In most cases, they were conceptualized as modular elements to fit perforated facades [1]. Implemented in building envelopes in simple small-scale as valves and flaps, medium-scale like windows, blinds or louvers and large scale applications like retractable roofs or folding bridges, the design of these technical systems, however, is subject with many variations [2]. Small-scale designs are easier to fabricate, but the effect remains on the level of texture. Bigger scale adaptive facades are more complex but have a bigger impact on the shape of the building. Shape is important in way of heat exchange (FT) factor which depends on geometry of exchanger [3] and generally for building environment and urban genesis.

For solving the problem of complex rigid-foldable structures, origami was the role model that is used in many various situations [4]. Origami technic is very useful for designing kinetic and deployable structures, such as space exploration needs, packaging issues and foldable building structures as well. Because of the specific feature of these kind of systems, the semi-automated deployment of structure, this can be good approach for improving control of the façade.

Inseparable part of exploring such complex forms is fabrication and testing on a scale model. Nowadays design production is changing rapidly, as architects increasingly use computers to generate designs and then digital fabrication to build models for design review [5]. CAD/CAM is incorporated into their practices in purpose of rapid and higher quality prototyping, and CNC fabrication [6, 7].

Aim of this paper is to explore possibilities of implementing complex structure that can fold, adapt and give different forms to the building. In section 2 we use principles from [1-3] that are the basics in the process of creating facades, but the approach for design is more parametric in purpose of getting façade that assimilates better on different shapes of the building. As well, utilizing the origami pattern, part of the research is solving the problem of material thickness, but keeping the rigidness. Last section examines the process of fabrication as an inseparable part of the whole design process.



ISBN - 978-86-7892-807-9

82

2 ORIGAMI PATTERN ALGORITHM AND MATERIAL THICKNESS PROBLEM As stated, most of the facades are fixed or with a minimal range of movement. The façade gives form to a building, and it lets the building communicate with the environment and its surroundings. The general approach is to make a façade that is flexible, that can change its form and that can make an impact on the building’s insolation. Folded structures, like origami, can serve as an inspiration for making a form that can be used in such a manner. Origami is a Japanese art of paper folding, which was founded in the 17th century. The main principle is that without the use of glue or cutting tools, is possible to create spatial forms, mostly from a square or rectangular sheet of paper. The methodology consists of utilizing a vast number of valley and mountain folds, as well as the interdependency between the same, to generate a specified form [8]. In this case we tried to apply the origami pattern designed by Ron Resch as a closed spatial structure when folded into an expanded form that stays connected through the entire façade’s envelope (Figure 1). This pattern consists of a square matrix where some squares are split with a diagonal. When it comes to folding, diagonals will represent valleys and the orthogonal lines will be folded as the mountain. The rule for splitting squares is that 2x2 matrix is split with two diagonals, and center of the next 2x2 matrix for splitting, is one row differed and further on. In that way there are some squares that are not split and those squares generate the building’s envelope after folding. This feature is adequate, since the facade has the ability to convert to a flat surface.

Fig. 1. Water bomb origami pattern designed by Ron Resch

For exploring this form we used a software for parametric modeling with specific plug-ins for testing the physics of the model. The algorithm for modeling is basically the same as the procedure for folding the paper. Firstly, the square matrix is established, and in order to acquire the best solution, with software restriction, a matrix that has one system of squares divisible with 4 is used. With that information, the software can count and set certain squares for splitting in the way that is defined in the workflow. The rule is that every fifth square has the same characteristic (1-plain, 2-ascending diagonal, 3-descending diagonal, 4-descending diagonal, 5-ascending diagonal and repeating the same rule) and if it’s combined with the rule before we can get the desired pattern.


ISBN - 978-86-7892-807-9

83

Fig. 2. Phases in algorithm for creating pattern

Next step is to sort the novel lines into three groups by folding and angle. First group contains mountain folded edges with 90° angle and those edges are plain squares. Next group is the rest of the orthogonal lines which are mountain folds as well but with 180° angle. Third group contains all diagonal edges that represent valley folds with a 180° angle. As well, those edges are the boundaries for polygons which create the polygonal mesh. Further on, the physics simulation components are included in the process. The software needs to input all square or triangular mesh polygons compared to each polygon’s edge that is used for the rotation for the two adjacent polygon faces. When that structure of two polygons and one edge is detected, two end points of the particular edge define the axis vector for rotating and the rotation angle is already assigned to every edge when they are sorted. Physics plug-in uses that rotation rule and applies the force in certain time intervals. The result is folding the flat mesh surface to a desired form. One great option that the plug-in offers is making anchor points that remain in their predefined position and in that way the control of the new form is better.

Fig. 3. Phases of model generation in different time intervals

The result is the same in a paper form, but it can be tested with different parameters, such as different angles of folding, positions of anchor points and number and size of pattern tiles. Both, the paper model and the computer aided designed model are still two-dimensional, but due to the need for thickness, complexity of folding becomes more demanding.

Next step is to create a three-dimensional form. When mesh polygons in reality become thicker that can violate hinges between them. With the empiric approach it’s determined how the hinges can be generated and how the polygons can rotate without any difficulties and deformations. That empiric approach includes testing different scale models made of cardboard, defining the hinge obstacles and solving them. One exception is made in this stage; the plain square polygons could not remain rigid, so we will observe them as flexible from now on. Other triangular polygons and hinges between them are solved in that way that on two shorter edge hinges is on the one of two sides and longer edge hinge is on the other side. If triangular polygons are made as frame with cavity, edges of that element are axially loaded and in that manner can be thin, so it can be cost-effective. If two of the triangles are connected via longer edge, and four of them are connected via shorter edges, we can get a modular structure in form of a flower that can be multiplied as many times as necessary.


ISBN - 978-86-7892-807-9

84

Fig. 4. Flower structure that enables opening Fig. 5. Multiplied flower structure

Those flower structures are connected via triangle shorter edges hinges and in that way the entire structure is developable from folded stage to stage of flatness. As it is mentioned, plain square parts will be made of flexible thin material such as textile, rubber textile or any other material with similar properties. An important characteristic of that material is its partial translucency, because if the façade form is in 100% folded stage, the building still needs interior insolation (Figure 6). On the inside of the façade the structure is generated from spikes (peduncle of flower structure) that are connection points to the building. In those points we need to place joints that allow two degrees of movement freedom. The first, joint makes the flower structure capable of opening, and the second, offsets it from the building along the normal of the facade. Because it is hard to control this kind of structure, its movement is limited to particular blob forms.

Fig. 6. Facade structure with flexible translucent textile covering

3 FABRICATION AND TESTING ON SCALE MODEL Part of the research is to test this theory on scale model and confirm it in reality. In order to have an optimal usage of material, and to create stronger hinges, an entire structure is made of one sheet of cardboard. The scale is 1:20 and the cardboard that is used for this model is 3 mm sandwich board made of two thin layers of millboard and foam between them. To stick to the rule for disposition of hinges it was necessary to cut the board on both sides, orthogonal lines on one and diagonal lines on other side. Plane square parts are cut off entirely. When every line is cut and the whole pattern is folded it comes connecting for the solid structure (building). Given the size and scale of the model it is impossible to make the joints on every spike. Therefore, the form is positioned and connected in nine points (2x2 matrix) with joints that allow opening only but not moving in the normal direction. The middle of four square parts is connected with joints that allow opening and moving in the normal direction. Those joints are made of plastic tubes 3 mm in diameter. In place for the full movement of joints the applied mechanism is similar to the steam engine mechanism. Furthermore, mechanism is connected to the electric servo motor and as the motor rotates, the mechanism makes the movement in the normal direction to the initial flat surface. Those servo motors are wired with an Arduino


ISBN - 978-86-7892-807-9

85

microcontroller which makes the controlled movements. In order to explore different ways of controlling the origami form, four different inputs are involved in the whole process, such as analog potentiometer, photoresistor, visual control on display and Kinect - motion sensing controller. For all of these inputs it was necessary to write a code, so input parameters can be translated to output signals that motors can understand. A lot of different input types are also a part of the research in that respect that people can be involved in different ways. For example, photoresistors leave no connection between people and the façade, they are involved in automated process that opens the structure in shaded parts. Potentiometers create a limited effect, depending on the position of an active user that can affect only one part of the façade. The real influence is provided by the Kinect, since it gives the opportunity to interact with the façade in more interesting way.

Fig. 7. Scale model for testing kinematics

Fig. 8. Mechanisms for joints and implementation of Arduino controlled kinematics

Beside the physical model, virtual 3D representation of the facade is created, based on the construction of Radnicki university building in Novi Sad. That building is devastated since it was caught on fire, and until today has no facade. This was just a conceptual approach which has a purpose to observe how it appears in the urban nucleus compared to its surroundings (Figure 9).


ISBN - 978-86-7892-807-9

86

Fig. 9. Photorealistic visualizations of origami facade structure

4 CONCLUSION Utilizing parametric design software, scale models and Arduino controlled mechatronics result is the origami façade that gives interesting spatial relations with the urban genesis. The facade is capable to create a form that can change itself depending on the input parameters, and it is designed with the use of new age materials and robotics that enable folding. Several technics are utilized for this research. Parametric design gives us the possibility to explore form, pattern and size of elements in different situations, scale model gives the results in testing possibility of folding and relations between all elements in structure, microcontrollers and servo motors enable testing ability for adapting in different cases, all in purpose to produce better form that fits desired.

Limitation of this paper approach was the deformability of the origami structures that paper as a material suffers, but it can be a problem for the rigid geometry. Future work can be focused on finding more optimized origami structures or solving problem of joints that is connected to the deformability as well.

REFERENCES [1] Schleicher, S., Lienhard, J., Poppinga, S., Masselter, T., Speck, T., & Knippers, J. (2011). Bio-inspired kinematics of adaptive shading systems for free form facades. In Proceedings of the IABSE-IASS Symposium, Taller Longer Lighter, London, UK (Vol. 9).

[2] Schleicher, S., Lienhard, J., Poppinga, S., Speck, T., & Knippers, J. (2015). A methodology for transferring principles of plant movements to elastic systems in architecture. Computer-Aided Design, 60, 105-117.

[3] Geletka, V., & Sedlakova, A. (2011). Energy consumption conditioned by shapes of buildings. Budownictwo o zoptymalizowanym potencjale energetycznym, (8)), 46-53.

[4] Tachi, T. (2010, November). Geometric considerations for the design of rigid origami structures. In Proceedings of the International Association for Shell and Spatial Structures (IASS) Symposium (Vol. 12, pp. 458-460).

[5] Sass, L. (2007). Synthesis of design production with integrated digital fabrication. Automation in Construction, 16(3), 298-310.

[6] Ryder, G., Ion, B., Green, G., Harrison, D., & Wood, B. (2002). Rapid design and manufacture tools in architecture. Automation in construction, 11(3), 279-290.

[7] L. Iwamoto, Embodied fabrication: computer-aided spacemaking, in: P. Bessley, N. Cheng, R. Williamson (Eds.), Proceedings of the 2004 AIA/ACADIA Fabrication Conference, Cambridge and Toronto, 2004, 270–278.

[8] Jackson, P. (2011). Folding techniques for designers: from sheet to form. Laurence King Publishing.


ISBN - 978-86-7892-807-9

87

COST-EFFICIENT APPROACHES IN FABRICATION OF STREET FURNITURE BASED ON SECTIONING DESIGN STRATEGIES

Dejan Mitov1* *1Faculty of Technical Sciences, University of Novi Sad (Serbia), [email protected]

Abstract Street furniture is a very inspiring theme for many contemporary designers because it gives a lot of space for the design and its realization itself is significantly cheaper than with larger buildings. This is also the reason why some design approaches are primarily tested in urban furniture. Three key approaches to the realization of free forms are using: linear, surface, and cubic materials. The sectioning design strategy is very common in the design of parameter generated objects of urban furniture because it is based on the board materials and simple preparation of drawings for fabrication. However, one of the problems that often occurs by this method is a great loss of material. The aim of this paper is to analyze and examine various design approaches based on sectioning strategies in order to minimize the vast of materials.

Keywords: digital fabrication, street furniture design, sectioning design method.

1 INTRODUCTION Sitting, as one of the most common functions in public space, requires carefully designed furniture that matches the ergonomics of the human body and allows ease of use. The development of CAD and CAM technology enables forming of more irregular lines, in a way that it is easy to allow continuity in their complete form. Generating free surfaces based on a small number of NURBS curves allows the controlled generating of these surfaces, with whom he later allowed further management. After obtaining the initial design of surfaces starts its materialization with existing conventional materials that are already widespread in everyday use. The realization of free forms in urban furniture design may be carried out in three directions regarding usage of materials: using linear elements, board materials or three-dimensional blocks. The first approach requires the processing of plastic materials (bending), whose efficiency depends solely on the quality of the material and how it provides such a deformation that will allow obtaining the pre-defined shape. Bending of plastic materials has been used for plastic Fluxx chair designed by Douwe Jacobs, where curved-line folding was used as a method [1]. Rigid materials is possible to curve just by making a small cut as it was used on Dukta pattern [2]. In the last approach, i.e. design of furniture from the three-dimensional cube, it is necessary to use cutting-edge technology, such as CNC machine with 5-axis or an industrial robot, taking into account the dimensions of the element, depending on the working field of the machine and sizes of blocks from which the elements are cut. On the other hand, plate materials are commonly used in architecture, and their application in digital fabrication can be inspired by the folding principle [3], tessellation method [4], and sectioning method [5]. In the first two approaches, CNC machines with 5-axis or industrial robots are needed for processing of their contour while the method of intersection only needs CNC machine with 3 axes and simple 2D drawings.

In this approach, designers first generate a free surface with different methods, later to be subdivided into plate elements using the method of intersection. In this manner, sectioning can be considered a geometric approximation of the given form, and not a design strategy of shape defining. However, with regard to its prevalence, this method is used from the beginning of the design process, and can be considered a designed strategy. Regarding the direction of the cross-section (horizontal, vertical and at a free angle), functionally and structurally different results are obtained. Finally, the distance between the layers significantly affects its final perception. Designers of "Mafoombey" music box created a single enclosed and acoustic structure by abolishing the distance between the horizontal layers [6], as opposed to the "Digital Origami Emergency Shelter" where horizontal layers with gaps are used [7], which makes it quite different compared to the previous. Sectioning method can be implemented in one direction, as shown in aforementioned projects, or can be double – transversal. In this case, the subject of the special research can be the places of intersection of the panels, which sometimes have to be custom designed, as for DRL-10 Space pavilion [8] or Metropol Parasol pavilion


ISBN - 978-86-7892-807-9

88

in Sevilla [9]. If the elements have larger dimensions it is necessary to design and implement ways of continuing separate parts as in the case of One Main [10]. The principles of these large-scale projects have been taken and further developed, and as the final result puzzle-joint connections are obtained.

2 METHODOLOGY A sectioning method in parametric design includes automated processing of cross-sections of free surfaces with a range of levels at specified distances. Because of the planar sections, which are usually vertical, it is possible to use plate materials which are very widespread in the market and there is a large selection of quality finish, various colors, textures etc. Depending on continuity of the section, or the space between two adjacent plates, as well as the thickness of plates, the design freedom is obtained in transferring the free surfaces that were previously defined. For these reasons "sectioning" as the principle in the design of urban furniture was accepted by a large number of designers. However, the main problems with this principle are the use of material, the total weight of the furniture (which is directly dependent on material consumption), and continuous interconnectivity of the elements (which prevents easy repair of elements). Flat materials are produced at predefined dimensions that have to be fit into when it comes to the production of furniture. Depending on the shape of cross sections it can be easier or harder to fit into boards of plate materials when cutting, which may result in residues which cannot be used, and, therefore, a large material consumption. Another problem is the weight of the furniture which can be very large, making it impossible to efficiently maneuver with it but also requires stronger fasteners that are burdened with the weight of the material. This problem can be solved by cutting the openings in each of the cross sections, which can follow the forms of the outer contour (offset) or it can be differently shaped. However, in this case, the consumption of material is not reduced. More advanced level of saving material is an "arched cross-section" that allows better fitting because the material elements enter into each other. Although this approach is a lot more efficient than the previous one in terms of material savings and reduction of weight, yet it can result in irrational material consumption especially when cross-sections are large-sized and difficult to fit due to the breaking in the "key points". Therefore, it is possible to perform disruptions in these key points in order to avoid sudden turns in the direction of the cross-section, i.e. to reduce the elements to the approximately straight line (Fig. 1).

Fig. 1. Different construction of the same section

3 THE METHOD OF INTERSECTION IN THE CASE OF BARABARKA BENCH Barabarka bench (Fig. 2) is a piece of urban furniture on which the preliminary analysis is directly carried out in order to measure the difference in material consumption and weight reduction. After defining the spatial curves that generate the free surface, surfaces are offset inwards for a distance of 6 cm which also represents the width of each rib from which the bench is assembled. Then the vertical cross-sections are made with an intermediate distance of 4.8 centimeters which is the quadruple material thickness of 12 mm, or a "tact", each of which includes one "A element" (large arc), "B element" (segments of arcs) and two circular spacers. “A” elements define the shape of the bench and represent its basic structure while “B” elements are used to increase the distance between “A” elements in order to reduce material consumption and are located in places which are essential for comfort during use. The material used for the realization of this bench is waterproof plywood because


ISBN - 978-86-7892-807-9

89

its cross-laminated layers carry theame load in both directions; therefore, it is not necessary to take into account the orientation of elements while stacking for cutting. The plywood with a thickness of 12 mm is chosen to achieve more efficient cutting in terms of time and ability to use cutters of smaller cross-sections because they give fine finishing of the material, and which have a length of up to 12mm. Standard dimension of table plywood with thickness of 12 mm is 2440x1220mm, which was taken as the basis for the analysis of material consumption. In the case of full cross section of “A” elements 12 plates of material are required, as is the case with perforated sections. When using continuous arched “A“ elements it is necessary to use 8.5 boards, which is by 3.5 less than in the previous case. Finally, with the principle of fragmented “A” elements, it takes only 4.5 plywood plates, which is quite less than in the two previous cases (Fig. 3).

Fig. 2. Barabarka bench

Fig. 3. Construction of puzzle joint concetion


ISBN - 978-86-7892-807-9

90

The main design challenge when it turns to the segmentation of cross sections into smaller parts is to facilitate the necessary structural stability, which in this case is disrupted due to disruption in the “A” elements. “A” segments of the elements are interconnected with stylized carpentry ties that prevent their mutual withdrawal. This connection is parameterized with 6 key points which define a curve that represents the cutting line (Fig. 4). It is essential for this connection that the cutting line is away enough from the edge of the “A” element in order not to weaken it, and to achieve greater curvature within ties to prevent extraction and simultaneously to tessellate cutting line in that way that no osculating circle on it is smaller than the radius of the diameter of the milling cutter. The last of the mentioned conditions is essential in order to achieve continuous fitting of two segments, i.e. to avoid cavities on this connection and, therefore, jeopardize the stability of the connection. Due to the CNC processing, it is possible to precisely fit the elements to obtain a final “A” element.

Fig. 4. Shapes arrangement on plywood boards

These connections between segments enables smooth transferring of vertical loads, and what remains a risk is the reaction of these connections on horizontal loads. For this reason, it was decided to set the circular spacers and “B” elements on the positions of these connectors in order to additionally harden these bonds by horizontal connections. On the other hand, in order to easily connect the 'tacts' and eventually disassemble them due to the need for reparation of some of them, it was decided to connect the entire system with a horizontal steel reinforcement, which is a kind of clewline. In order to enable the straight transfer of loads in ties, it is essential that it remains a straight line, i.e. to avoid its bending due to the effects of the loads, which may cause negative effects on the assembly and its deformation. Therefore, in the elevation, all cross-sections intersect in these points where the ties are (Fig. 5). In terms of design, straight ties are in contradiction with the amorphous lines of cross sections, and the role of “B” elements and its dynamics is to neutralize the visual line of ties and to increase the contact area of seating.


ISBN - 978-86-7892-807-9

91

Fig. 5. Points of frames cutting and puzzle joints

Along the entire bench, there are three points with the steel reinforcement bars and none of these points is in contact with the ground. At these points, the plywood ribs are put crosswise to the existing “A” elements, to allow their irregular arrangement in order to achieve greater stiffness of the bench, so that the vertical load on the bench pressures its central part to the interior, and the ends outwards. When fitting vertical “A” elements and the horizontal elements plug-formed carpentry ties are used, which ensure equal equidistance between the ribs and prevent their move. Due to the use of CNC processing in nonconvex polygons, such as the ends of the “A” elements, small radiuses in the material remains, depending on the thickness of cutters. The algorithm "Dogbone" for Grasshopper is designed to ensure good fit of vertical and horizontal ribs. It has provided additional cuttings of materials at critical points, depending on the radius of the cutters to be used (Fig. 6).

Fig. 6. Dogbone algorithm solution

4 CONCLUSION Barabarka bench is an example of analyze the consumption of materials, depending on the fulfillment of cross-section, or whether the section is continuous or divided into segments. Huge savings of material due to segmentation profiles are clearly shown, but it is necessary in this case to ensure a quality connection at the points of continuing to accept the effects of the loads in all directions. Thanks to parametric design various analyzes and tests can be carried out concerning the design or production. With further use of CNC machines it is possible to precisely transfer of what was previously drawn. This approach to the optimization of material consumption enables contemporary design of amorphous shapes to become closer to reality and realization. Further guidelines for research with the method of cross sections are reflected in the reduction of the offset thickness, different types of extensions and different ways of connecting segments.


ISBN - 978-86-7892-807-9

92

REFERENCES [1] Vergauwen, A. De Temmerman, N. Brancart, S. (2014): The design and physical modelling of deployable structures based on curved-line folding; 4th International Conference on Mobile, Adaptable and Rapidly Assembled Structures; 11 - 13 June 2014; Ostend, Belgium

[2] Ohshima, T. Igarashi, T. Mitani, J. Tanaka, H. (2013): Wood Weaver - Fabricating curved objects without moulds or glue; Computation and Performance – 31st eCAADe Conference; 18-20 September 2014; Delft University of Technology, Netherland

[3] Buri, H. Weinand, Y. (2008): ORIGAMI – Folded Plate Structures, Architecture; 10th World Conference on Timber Engineering; 2-5 June 2008; Miyazaki, Japan

[4] LI, J.M. Knippers, J. (2015): Pattern and Form - Their Influence on Segmental Plate Shells; IASS2015 Annual International Symposium on Future Visions; 17 - 20 August 2015; Amsterdam, The Netherlands

[5] Iwamoto, L. (2009): Digital Fabrications – Architectural and Material Techniques; Princeton Architectural Press, China

[6] Marcos, C. L. (2011): New materiality: ideation, representation and digital fabrication; eCAADe 2011 Representing Fragile Places; 21-24 Septembre 2014; Ljubljana, Slovenia

[7] Teixeira, S. A. Campos, D. R. (2011): Origami cientifico: a linguagem das dobraduras no design contemporaneo; Revista Faculdade de Arquitetura, Artes e Comunicação; Bauru; Brazil

[8] Bruckermann, O. and Alberdi, J. (2010): Stuctural Design of the DRL-10 Space Pavilion; Journal of Architectural Engineering, Volume 16, Issue 3

[9] Schmid, V. Koppitz, J.P. Thurik, A. (2011): Neue Konzepte im Holzbau mit Furnierschichtholz – Die Holztragkonstruktion des Metropol Parasol in Sevilla, Bautechnik, Volume 88, Issue 10.

[10] Website: http://www.decoi-architects.org/2011/10/onemain/ - date February 12, 2016

http://www.decoi-architects.org/2011/10/onemain/


ISBN - 978-86-7892-807-9

93

CONTEMPORARY METHODS FOR EXISTING BUILDINGS PRESENTATIONS USING MOBILE DEVICES, CASE STUDY OF

TEMPLE ON VLASINA HIGHLAND

Petar Pejić1*, Sonja Krasić2, Milica Veljković3 1Teaching assistant, Department of Visual Communications, Faculty of Civil Engineering and

Architecture, University of Niš, SERBIA, [email protected]

2 PhD, Associate Professor, Department of Visual Communications, Faculty of Civil Engineering and Architecture, University of Niš, SERBIA, [email protected]

3Teaching assistant, Department of Visual Communications, Faculty of Civil Engineering and

Architecture, University of Niš, SERBIA, [email protected]

Abstract The rapid technological evolution, enabled the creation of contemporary mobile devices with a great processing capabilities. This facilitates presentation of complex architectonic designs and three dimensional models of buildings on mobile devices, using methods of virtual and augmented reality. This paper presents a case study of the existing church of St. Prophet Elijah (Crkva Svetog proroka Ilije) located on Vlasina highland presentation using smart mobile devices. The process of creating an augmented reality presentation of the church 3D model and virtual reality presentation of locations around the building is presented. For the purpose of the paper, mobile applications for devices with android operating system were created and tested.

Keywords: Presentation, mobile devices, existing buildings, virtual reality, augmented reality.

1 INTRODUCTION The traditional methods of presentation, perspective images and videos, are characterized by stable performance and simple use. However, they are limited in terms of freedom of viewing content. The user often sees only parts of the facility which is presented. With this method of presentation, it is not possible to show all the facility details interesting to the user. [1].

The rapid technological evolution, enabled the creation of contemporary mobile devices with a great procession capabilities. Together with the integration of various sensors which allow interaction between a mobile device and the real world, the potential for use in architecture is great. This facilitates presentation of complex architectonic designs and three dimensional models of buildings on mobile devices, using the methods of Virtual and Augmented reality. The methods of virtual and augmented reality exceed the limitations of traditional ways of presentation of the existing buildings on mobile devices.

Virtual reality represents a computer environment that can simulate a physical presence in a virtual or imaginary world [2]. Virtual reality is an evolutionary advancement of digital methods for presentation of 3D building models. As with animation and two-dimensional image, aspiring authors is to get a more realistic view of virtual objects, which implies the convergence of end user experience in a virtual world, to those of the real world. Therefore, virtual reality allows the user real time interaction and freedom of movement within it.

Augmented reality (AR) is an emerging computer technology where the perception of the user is enhanced by the seamless blending between a realistic environment and computer-generated virtual objects coexisting in the same space [1]. The resulting mixture supplements reality, rather than replacing it [3].

A more comprehensive definition of AR would be as a system that has the following characteristics: (1) combines real and virtual world, (2) interactive in real time and (3) registered in 3D [4]. Augmented Reality enhances a user's perception of and interaction with the real world. The virtual objects display information that the users cannot directly detect with their own senses [5].





ISBN - 978-86-7892-807-9

94

This paper presents a case study of the existing church of St. Prophet Elijah (Crkva Svetog proroka Ilije) located on Vlasina highland presentation using smart mobile devices. The process of creating an augmented reality presentation of church 3D model and virtual reality presentation of locations around the building is presented. For the purpose of the paper, mobile applications for devices with android operating system were created and tested.

2 MATERIAL AND METHODS The Church of St. Prophet Elijah “Svetog Proroka ilije” (Figure 1) was built in the XIX century in Vlasina Rid, near the Vlasina lake. Medieval monastery was at the site of the present-day church, which was destroyed during the period of the Ottoman rule. The existing church was built at the same location in 1838. Since 2010. The Church of St. Prophet Elijah is under the reconstruction.

Fig. 1. Hram Svetog Proroka ilije

Core Material for the production of three-dimensional model and virtual tours from locations close to the church are photos. At the site, a large number of photographs of the building showing all parts of the exterior of the church, was collected. Based on the collected photos, a three-dimensional model of the Temple was created using the method of semi-automatic terrestrial Photogrammetry of one shot (Figure 2).

Fig. 2. 3D model of Hram Svetog Proroka ilije

For the purpose of the church presentation in the actual environment, the method of Virtual Reality is used. The photographing of the entire environment at three locations near the temple was done. Textures for spherical Virtual Tours were created on the bases of the collected photos for each location (Figure 3).


ISBN - 978-86-7892-807-9

95

Fig. 3. Texture for the spherical virtual tour

3 MOBILE APPLICATION Mobile application is developed using Unity development engine (Figure 4). 3D model of The Church of St. Prophet Elijah is imported and connected with the marker in order to present object using Augmented Reality technology. Previously created textures for virtual 360° tour are applied to the sphere in order to create virtual reality scenes. The final design is compiled and tested on mobile devices with android operating system.

Figure 4. Application info screen Figure 5. Augmented reality in work

The created android application is tested using “Project Tango” tablet. First “Augmented Reality” part of application is tested with appropriated marker (Figure 5). After starting the application, the device camera records real surrounding, while the application is searching for predefined markers. When the application detects the location of the marker, on the display of the device we can see a 3D model of the Church of St. Prophet Elijah connected with the marker. Moving the marker will cause a coordinated move of both the marker and the 3D model of the church on the device display.

The second part of application presents 360 degree spherical panoramic images used in building photographic "Virtual Reality" tours (Figure 6). This tours allows users of mobile devices to interactively "be there and look around" with the full realism that digital photography can capture. Moving and rotating of device causing movement and rotation of a virtual scene on display in the same manner.


ISBN - 978-86-7892-807-9

96

Figure 6. Virtual reality in work

4 CONCLUSION The goal of this paper is to demonstrate a contemporary method for presentation of the existing buildings using mobile devices. Application of the Augmented Reality for the presentation of the existing architectural objects is a cutting-edge tool. This approach enables the spatial presentation of the 3D model within the real environment. It allows the user to view the existing buildings as small scale models inside the real surrounding. Virtual Reality tours allow users preview of 360° visual environment that offers far more contextual information than a series of static images or a video clip.

By using commercial mobile devices and free Android software the preview can be seen. The use of the application itself is completely intuitive. The quality of the 3D model presentation is at a good level, but it is worse than the Virtual Reality tours. It is caused by the limited processing capabilities of the devices and complex calculations needed for the proper functioning. Augmented reality presentation of a 3D model represents a great tool for presentation of existing buildings. It provides a better understanding of architectural structures because users can choose which part of object and from which point of view they want to see. On the other hand Virtual Reality tour provides a better quality of photo realistic representation of a location, but it is limited because users can see space only from predefined points.

This paper presents a case study of the existing Church of St. Prophet Elijah (Crkva Svetog proroka Ilije) located on Vlasina highland using both Virtual and Augmented Reality methods in order to present positive and negative sides of bothof them. Future development of technology will allow the development of novel methodology for presentation of existing buildings, which will be the topic of future research.

REFERENCES [1] Pejić, P., Krasić, S., Jovanović, N. (2014). The application of augmented reality in the presentation of existing architectural facilities, International conference MoNGeometrija 2014, Vlasina, pp. 74-81.

[2] Rizov, T. (2014). ГЕОМЕТРИСКО ПРЕТСТАВУВАЊЕ НА ОБЈЕКТИ ВО ИНТЕРАКТИВНА АУГМЕНТНА РЕАЛНОСТ, Skopje: Univerzitet "Sv. Kiril i Metodij", Mašinski fakultet.

[3] Pejić, P., Krasić, S., Petković, D., Veljković, M. (2015). Application of augmented reality in facade redesign presentation, Journal of Industrial Design and Engineering Grapphics, 10/4, pp. 45-49.

[4] Rizov, T., Tashevski, R. (2014). Technical vizualization in the process of vehicle identification using augmented reality," International conference MoNGeometrija 2014, vol. 1, pp. 51-61.

[5] Pejic, P., Rizov, T., Krasic, S., Tashevski, R. (2015). Presentation of Existing Architectural Objects Using Augmented Reality: Case study - Ada Bridge, Belgrade, Serbia, South East European Journal of Architecture and Design, 2015/1, pp. 1-4.


ISBN - 978-86-7892-807-9

97

INTERACTIVE GLARE VISUALIZATION MODEL FOR AN ARCHITECTURAL SPACE

Florina Dutt*1, Subhajit Das2, Matthew Swarts3 1Georgia Institute of Technology (USA), [email protected]

2 Georgia Institute of Technology (USA), [email protected] 3 Georgia Institute of Technology (USA), [email protected]

Abstract Lighting design and its impact on indoor comfort conditions are an integral part of a good interior design. The impact of lighting in an interior space is manifold, and it involves many sub-components like glare, color, tone, luminance, control, energy efficiency, flexibility, etc. While other aspects of light have been researched and discussed multiple times, this paper discusses the research undertaken to understand glare from an artificial lighting source in an indoor space. The paper discusses a parametric model to convey instantaneous approximate glare level in an interior space to users. We foresee Architects as one of our main end user’s and likewise for them it is of utmost importance to know what impact the proposed lighting arrangement and proposed furniture layout will have on the indoor comfort quality. Essentially, the designer would need to know the ramification of the ‘discomfort glare’ at the early stage of the building design, when he can afford to recommend changes to the design scheme by selecting other potential alternatives for his client. Unfortunately, most of the current lighting analysis tools, offer lighting simulation studies based on rigorous computation and analysis. Evidently this makes it difficult for the designer to quickly analyze and interpret the interior discomfort glare level inflicted due to the design of the interior space. Intending to address this problem, this paper, explains a novel approach to approximate interior glare data. Furthermore, we visualize this data, as a color coded overlay on the 3d model of the interior space, assisting users to understand the implications of their proposed interior design layout. Our focus is to make the analysis and the visualization relatively fluid, fast & computationally scalable. Our system embeds real-time user interaction by varying user inputs and editing the 3d geometry of the space. Additionally, we tested our proposed parametric model on a case study space - A Computer Lab interior layout in our college facility.

Keywords: Parametric glare model, Interior design, Lighting Analysis, Design simulation.

1 INTRODUCTION Our intent is to develop a relatively fast algorithm which will not only compute glare from an artificial light source in an indoor space but also will visualize the same. As a result, the designer/ architect can use it fluidly at the conceptual/ schematic design stage to quickly validate design proposals. Adding our system to the current workflow would facilitate production of quality interior space by enriching the indoor comfort level. Researching along similar lines, we realized that there are many rendering engines available with sophisticated graphical algorithms, representing an aesthetic aspect of lighting, comprising of indirect illumination, global illumination, ambient occlusion, daylighting, etc. However, there’s lack of any tool which can visualize the performance of interior lighting fixtures regarding glare for the comfort of the user almost instantaneously. We developed the strategy, to deliver fast and almost real time results, at the cost of slight precision. Our research venture addresses these issues and presents a solution which can successfully implement such a system for varied design layouts and furniture arrangements.

2 MOTIVATION Not only is glare an important component of discomfort for occupants both in indoor and outdoor space, but it also has a serious impact on people’s psychological and physiologically health. Linewise, one of the driving reasons to pursue this research, was to enable Architects design spaces, which do not transcend discomfort due to glare from indoor lighting.


ISBN - 978-86-7892-807-9

98

Recent medical and biological study enlightened the fact that light entering human eyes not only have a visual effect but also a significant non-visual biological implication on the human body (Bommel, July 2006). Evidently any space with good lighting design has positive impacts on the human health, well-being and alertness. It also enriches sleep/wake cycles, performance patterns, core body temperature and production of hormones. Certainly, this commands that new rules governing the design of good and healthy light installations are essential. Not only does good lighting arrangement attains expected the level of visual performance, but it also regulates the surrounding spatial appearance. (L. Bellia, October 2011).

Recent discoveries in photobiology are creating a link between lighting, health and well-being. It fosters lighting design, which deals with indoor glare problem by analyzing and processing the luminance pattern, along with providing adequate visual quality to space. Visibility is still an essential aspect of any lighting installation, but good quality of lighting installations are being judged based on other important virtues including quantity and quality of required light for wellbeing, health, interpersonal relationships and aesthetic appeals. (L. Bellia, October 2011).

Light defined as that part of the electromagnetic spectrum (∼380–780 nm) that gives rise to a visual sensation. Lighting in buildings, whether through the-the use of daylight or by artificial means, is designed primarily for the visual needs of the occupants and their expected tasks within a given space. Solar radiation, daylight, artificial light has a range of influences on the human. In addition to vision, it controls the circadian rhythm of hormone secretions and body temperature with implications for sleep/wake states, alertness, mood, and behavior. Symptoms of the disruption of these cycles through changes to the natural light/dark cycle can range from temporary jet lag to severe depression. Considering lighting is such an important part of healthy human life, an interior space design endowed with good lighting conditions, adequacy to meet aforementioned human comfort level is at the highest level of importance (Webb, July 2006). Noteworthy to mention that unlike other e-m radiation, light is sensitive to our eyes due to a certain range of wavelengths, and thus we can see it. Light has biological effects like stimulating or relaxing or supporting circadian rhythm, in addition to its more commonly known visual functions like the illumination of workspace conforming to relevant standards. Above all, light has emotional effects, like enriching space quality, creating memorable effects, etc. (The Lighting Handbook, Oct 2013)

3 RELATED WORKS In the domain of computer graphics and computational geometry, some studies have been successful in analyzing and visualizing interior artificial lighting or glare for the user. Most of them were very precise, but not very fast computationally to allow the end user quickly test his ideas in almost real time and get a near precise feedback in an ad-hoc manner. Consequently, this motivated our research to facilitate architects and interior designers enough information to select appropriate design option. Having said that, we found, related research ventures in this area helped us understand the nature, significance, and the complexity of the problem.

Christoph F Reinhart et al. developed a fully integrated design analysis method that simultaneously considers annual daylight availability, visual comfort and energy use. Their work also explains how their tool could be practically implemented using technologies that were available back then. Their work requires that the information needed to carry out a fully integrated lighting/thermal analysis must be available in building information models namely scene geometry, materiality, thermal zones, program, and schedules (Christoph F Reinhart, 2010). It could be a drawback when the designer does not have an adequately detailed model for the project. Often, during the early stage, designers tend to keep the model light regarding data to test different design alternatives and directions. Eduardo Fernandez et al. proposed a method by providing optimal light source positions as well as optimal shapes for skylight installations in interior architectural models. They facilitated the same by exploiting the scene coherence to compute global illumination using a meta-heuristic technique for optimization. Their method provides a fast and accurate method for inverse lighting that enables designers to browse solutions in short time (Eduardo Fernandez, December 2012).


ISBN - 978-86-7892-807-9

99

Figure 1. Shows the existing furniture arrangement in the Computer Lab

Hirning et al. conducted an impressive investigation on discomfort glare with 493 surveys collected from five green buildings in Brisbane, Australia. The study consisted of a specially tailored questionnaire to assess potential factors relating to discomfort glare, conducted on full-time employees. The luminous environment of the occupants was captured by the luminance maps extracted from high dynamic range (HDR). Over 49% of occupants reported some discomfort at the time of the survey. It further revealed occupants were more sensitive to glare than any other component of light (Hirning, et al., 2014).

Interior working environment should not impart any visual fatigue to the users. Glare from artificial lighting is one of the causes of that creates visual fatigue. Research by Kim Wonwoo et al. suggested that background luminance is one of the main factors affecting the degree of discomfort glare. They conducted two experiments including visual sensitivity test and glare sensitivity test to investigate the effect. They concluded that the luminance of the immediate background of a source should be considered before the average background luminance (Wonwoo & Yasuko, 2004).

Figure 2. Shows the existing lighting arrangement in the Computer Lab


ISBN - 978-86-7892-807-9

100

Figure 3. a. Shows the 3d model developed in Rhino. b. Shows the layout of the Computer Lab with furniture arrangements.

4 UNDERSTANDING LIGHT AND GLARE

Out of many, most important quality criteria of a proper lighting system are: a. Glare limitation b. Correct light color c. Avoid reflections d. Uniform brightness distribution e. Sufficient illumination level f. Appropriate color rendering g. Personal control h. Energy efficiency i. Daylight integration j. Light as an interior design element

As we focused on alleviating glare effect, we dived deep into the topic of light and glare to understand their role on each other. Clearly, glare reduces user’s performance in the room and inhibits his vision. The paper discusses our work to find zones in interior space with high glare value on reflective surfaces due to interior lighting design. When we mention glare, we mean ‘Discomforting Glare,' which is the amount of glare, physically and physiologically harmful and uncomfortable for the user. It creates an irritation or at times pain due to unsuitable distributions of brightness (significantly higher than the luminance to which the visual system is adapted) in the person's field of view. Making it even worse, disability glare (the reduction of visual performance) can accompany discomfort glare (CIE, 1987). To calculate the amount of glare at a certain point we studied and implemented UGR, unified glare rating (CIE, 1995):

UGR = 8 log 10 ( [0.25/Lb ] ∑ [ L 2 * w / p 2 ] ) where, Lb - background luminance (cd m-2 ) ∑ - summation of all of the separate glare sources present L - source luminance, measured at the observer’s eye (cd m-2 ) w - solid angle of each source at the observer’s eye (steradian) p - 'Guth' position index


ISBN - 978-86-7892-807-9

101

UGR values range from 10 to 30. Higher values represent significant discomfort glare, and lower values represent relatively comfortable glares. Thus, values less than 10 are considered to be in the comfortable glare range. According to our study, for reading, writing, training, meetings and computer-based work UGR values should not exceed 19. (The Lighting Handbook, Oct 2013). Discomfort glare can be further subdivided into the direct glare and reflected glare. Direct glare is that glare which travels directly from the source of the light to the user’s eyes. Reflected glare occurs when glare is reflected from any reflective surface, upon which light strikes from its source. Direct glare is mainly caused by direct observation of high luminance in the visual environment of the observer. Reflected glare is caused by the reflection of light from a surface (Mark S. Sanders, 1993).

Figure 4. Persona development of a sample user, quantifying the metrics to be analyzed

Figure 5. Shows the point grid arrangement of pixel space, user & lighting fixtures. Also shows the 3D Model of the interior space


ISBN - 978-86-7892-807-9

102

Figure 6. Flow chart describing the work flow process to build the parametric model.


ISBN - 978-86-7892-807-9

103

5 PROCESS AND METHODOLOGY

Implementation details of our system is outlined below:

5.1 Test Space

To test our case we studied the glare effect from artificial lights at the computer lab of Georgia Institute of Technology’s College of Architecture at Atlanta, USA. It’s a room 42 feet by 42 feet (approx.) in length and breadth with a clear height of 12 feet (approx.). There’s no direct daylight access, as there’s no external window looking outside. The room equipped with around 20 desktop workstations arranged in a grid-like fashion as shown in Figure 1. There is a presenter’s standing space at the front of a projector screen facing the desktop computers. Under the false ceiling, the interior lighting fixtures are hung down as shown in Figure 2. It is noteworthy to add that these light fixtures also follow, grid-like arrangement. Before developing the algorithm, we developed a detailed 3d model of the Lab space to understand its geometry and internal arrangement, just like a typical architect would do while designing the interior room layout. The 3d model was developed partly in Rhino (NURBS-based 3D modeling program) and partly in Revit leveraging the building information modeling workflow in practice to help layout furniture as shown in Figure 3a. For the parametric modeling of the lighting analysis, we used Autodesk’s Dynamo Visual Programming tool. The whole setup rendered a very powerful framework to implement and test our glare visualization algorithm. We developed the first proof of concept in Python, embedded in Dynamo as custom nodes, which leverages its pre-built objects and nodes. Work is underway to translate and implement the same algorithm in C# programming language to develop a Lighting analysis library for Dynamo.

Figure 7. Shows the Dynamo script, having custom nodes written in Python

5.2 Computational Model in Revit We envisaged that designers shall use our parametric model in real time while designing spaces and interior layouts. Considering Revit as a relatively powerful 3d BIM tool having unprecedented popularity in the realm of building design, our parametric model fits very well with the current workflow. We conceptualized that our users would make a very basic layout (internal space planning and furniture layout) in Revit as shown in Figure 3b, represented as simple line elements. As aforementioned, for our case, the model is the computer lab space, with a set of desktop machines and lighting fixtures arranged in a regular orthogonal grid. The model is kept very light having a basic line drawing in Revit, containing two important elements.

1. An outer periphery representing room boundary and its dimension

2. An internal furniture layout represented by simple lines.


ISBN - 978-86-7892-807-9

104

For the test case, we represent the desktop machines as simple lines (whose length depict the width of the desktop monitor.

5.3 Computational Model in Dynamo

Next, we open Dynamo (from Revit) and import the layout designed in Revit as Line objects. Our script (Refer Figure 7) uses Python to build custom nodes in Dynamo to separate the input line lists into two separate lists 1. List of Lines representing room boundary. 2. List of Lines representing desktop monitors (in general it can be any furniture for any generic case). First, we handle the list of lines representing the room boundaries. We make a bounding box, which represents the room itself. We make a 2D grid of points inside the bounding box. We take the corner points of the bounding box and use its coordinates as the limiting factor for the point list formed inside. They are used as the maximum and minimum values of the x,y,z coordinates for the point lists. Also, the points near the edges of the room boundary, are offset a bit inside so as to help compute the glare value later on in the script. We use each of these points to form a cuboid whose length and breadth equals half of the distance between the neighboring points and the selected point. The height of the cuboid is kept at the 2-inch dimension. Essentially these cuboids would behave as pixels over which the glare level can be mapped or visualized. We can programmatically change the resolution of the pixels by changing the number of points (in x and y directions) inside the room boundary.

Figure 8. a. Detailed floor layout with room section elaborating the furniture arrangement with the lighting plan. Pink dots represent the lights while black dots represent the user. b. Schematic sections explaining the lighting

computations studied to conduct the glare calculation for the user.

5.4 Points & Lines for Computation To represent the desktop monitor (or any input furniture element which can affect or get affected from or by glare) in our computational space, we store lines representing desktop monitors in a separate node as a python list. Making the system interactive, we allow the users to input the desired height of the desktop monitors, which they can change later on to understand the effect of glare. Based on the input we translate the desktop monitor line object at that level. Other indoor furniture related user inputs are taken including the desk height, varying which the user can test and compare the influence of glare at the different height the monitor is placed. Next, we find its mid-point, used to compute all the computations with the light source. Our idea is to find the level of glare at each pixel space in the room. The glare would be affected by the neighboring desktop computers around each pixel space


ISBN - 978-86-7892-807-9

105

(cuboids placed at each grid points as mentioned before). We enable another level of user interactivity by allowing the user to feed in the minimum threshold distance from each pixel space as shown in Figure 8a. This threshold distance shall determine how many desktop computers shall be accounted for while deriving the glare value. Likewise, in Dynamo Python script, the program iterates through all the desktop machine (represented as a line) and calculates their distance from the light source, It makes a new list of selected desktop machines which are below the threshold distance. These desktop machines shall be used for computing the said glare.

Figure 9. a. Shows the information visualization of room glare at each pixel space mapped over the floor. Also shows, the interior furniture layout represented as simple line elements.

Figure 9. b. Shows the impact of glare at different room heights with the same layout. It also shows the lighting configurations tested. Dark blue points show lights which are disabled while cyan colored points show enabled

lights.


ISBN - 978-86-7892-807-9

106

5.5 Lighting Elements We make another 2d grid of points inside the room boundary in a similar way as the one we made before to build our pixel space on the floor. Though these points are not placed on the ground, they are set at the height of the light fixture level, representing the lights in the computer lab. The height parameter is input by the user adding another level of interactivity for the user. By varying the height of the light, the parametric model can be used to test and compare the impact of glare on the user while placing the light fixtures at varying heights. The light points have two states enabled or disabled. Out of all the points, only those light points with enabled status shall be used for shooting light rays around the room for making glare computations as shown in Figure 9a and 9b. The parametric model includes three different kinds of lighting arrangements. One of them is regular grid-like while the other one is the hexagonal arrangement. The third configuration is the random placement of the lights over the ceiling. The list is extendable by adding custom light arrangement enabling added leverage and flexibility to the designer to test new design ideas and directions. Besides, each light point is accompanied by a rectangle of size 3inch by 3inch representing the light fixture surface. It is used to compute direct glare angles.

5.6 Glare Computation At this point for each pixel space, we have a list of selected desktop machines and a list of the enabled light source as point objects close to the false ceiling. We compute two kinds of glare as explained below:

5.6.1 Direct Glare It is the light cast directly on the eyes from the light source. To compute direct glare, we record the angle the light fixture (rectangle object) makes with the Pixel space point location. It’s as if trying to compute how much would be the glare at that pixel space if a user is standing at that point. We also record the distance (position index in UGR formula) from the light source to the pixel space point.

5.6.2 Reflected Glare It is light reflected off of the desktop screen into user’s eyes. To compute the reflected glare at that particular pixel space, we iterate over all the selected desktop machines. For each machine, we pick only those lights whose rays will hit the desktop screens from the side facing them as shown in Figure 10. A ray is cast from each of the selected light source points to the midpoint of the desktop machine line object. Also, the distance from the selected desktop machine midpoint to the pixel space point is recorded. Both of these values, the angle of the reflected light and the distance of the reflected light to the pixel space point is needed to calculate the UGR value.

5.6.3 Computing UGR value. For each pixel space, the UGR value is summed up for each selected desktop machine for all light objects casting light rays on its surface. Further, the UGR values are summed up for all the selected desktop machines allowing one UGR value for each pixel space. Clearly the angle ‘w’ computed and position index ‘p’ retrieved from these computations are two key components in computing glare value (by UGR formulae as shown above). Once the UGR values are computed for each pixel space, then they are mapped to the pixel space by a selected color encoding. The list of UGR values obtained for all pixel space gives us the maximum and minimum UGR levels for every pixel space. The color values set for mapping are then interpolated between these maximum and minimum values. As we built the pixel space as a list of cubes able to be overlaid with color and visualized in Dynamo.

6 RESULT AND OUTPUT Figure 10 shows some of the output results obtained from our parametric model. Different type of lighting configurations resulted in the different visualization of glare level at each pixel space in the computer lab. The color map used represents glare level from the maximum value (darker shade of red) to the minimum value (lighter shade of red.). The regular arrangement of desktop machines was compared with irregularly placed desktop machines in the room. The results varied when the users changed the threshold distance for neighboring desktop machine or when the user changed the lighting configuration or the height of the desktop machine. Another user input which changed the result of the visualization was the height parameter of the light fixtures. Clearly, the way the setup of


ISBN - 978-86-7892-807-9

107

the parametric model is framed, it is built to handle any generic case, with any interior design scheme, furniture layout, room dimensions, and lighting fixtures. Even though the result obtained is an approximation of industry standard glare computation, the primary intent of this venture was to significantly speed up glare computation and visualization by employing fast and scalable visualization algorithm at the cost of some level of precision. It allows the end user the liberty to validate their design layout quickly to make an informed decision. The computation time is controllable by the end user up to a certain extent by controlling the dimension of each pixel unit. Higher pixel resolution took little more time but unquestionably yielded more precise results. The average time for average resolution pixel space model (81 cuboids) for ‘the computer lab’ test case took around 20-25 seconds. Higher resolution model (225 cuboids) generated glare resolution visualization at around 45-50 seconds per visualization.

7 CONCLUSION This attempt to visualize near precise real time glare data was successful, and it did provide us quick and robust information visualization of glare data from artificial lighting overlaid on the 3d model of the architectural space. Throughout the process, our main intent was to have the system assist the architect/ designer to select, validate and improve their existing design space. Providing democratized access to rapid interior glare visualization, the said parametric model also emphasizes the impact of lighting design on indoor comfort level and spatial quality, which at times gets unnoticed or unsolved. In future, we are working towards including daylight along with artificial interior lighting to gauge the level of glare in the space. This functionality would render the system more useful for designers making it a more realistic visualization of the impact of the interior furniture layout on lighting design and vice versa. Interactivity was one of the key components and drivers of the visualization model. Certainly ability to set furniture layout, room dimensions, custom floor layouts along with the ability to customize lighting arrangement, selectively setting lights on & off, specifying threshold distance to other furniture’s for glare computation, etc. are few of the model’s user input, enriching its interactive quotient. Besides, building off of just simple line geometry on popular modeling platforms like Revit, it is designed to conveniently get absorbed in the current workflow.

REFERENCES [1] Bommel, W. J. v., July 2006. Non-visual biological effect of lighting and the practical meaning for lighting for work. Applied Ergonomics, Volume 37, Issue 4, pp. 461-466. [2] Christoph F Reinhart, J. W., 2010. The Daylighting Dashboard - A Simulation based Design Analysis. Cambridge, MA, Harvard GSD.

[3] CIE, 1987. International Lighting Vocabulary, Paris: CIE Publication.

[4] CIE, 1995. Discomfort Glare in Interior Lighting. Technical Report. CIE , p. 117.

[5] Eduardo Fernandez, G. B., December 2012. Inverse Lighting Design for Interior Buildings Integrating Natural and Artificial Sources. Building and Environment, pp. 1096-1108.

[6] General Electric Co., Oct 2013. The Lighting Handbook. Cleveland: General Electric Co..

[7] Hirning, M. B., Isoardi, G. L. & Cowling, I., 2014. Discomfort glare in open plan green buildings. Energy and Buildings, February, Volume 70, pp. 427-440.

[8] J. Alstan Jakubiec, C. F. R., 2012. The ‘adaptive zone’–A concept for assessing discomfort. Lighting Research and Technology, pp. 149-170.

[9] L. Bellia, F. B. G. S., October 2011. Lighting in indoor environments: Visual and non-visual effects of light sources with different spectral power distributions. Energy and Buildings, pp. 1984-1992.

[10] Mark S. Sanders, E. J. M., 1993. Human Factors In Engineering and Design. s.l.: McGraw-Hill Science/Engineering/Math.

[11] Webb, A. R., July 2006. Considerations for lighting in the built environment: Non-visual effects of light. Energy and Buildings, Volume 38, Issue 7, pp. 721-727.

[12] Wonwoo , K. & Yasuko , K., 2004. Effect of local background luminance on discomfort glare. Building and Environment, December, 39(12), pp. 1435-1442.


ISBN - 978-86-7892-807-9

108

GEOMETRIC FRAMEWORK FOR THE EQUILIBRIUM ANALYSIS OF POINTED ARCHES ACCORDING TO

MILANKOVITCH’S THEORY OF THRUST LINE

Dimitrije Nikolić1* 1Department of Architecture and Urbanism, Faculty of Technical Sciences

University of Novi Sad, Novi Sad (SERBIA) [email protected]

Abstract More than a century ago, Serbian scholar Milutin Milankovitch in his remarkable work set the complete and correct theory for the equilibrium of the arch of general shape according to thrust line analysis, and computed the minimum thickness of semicircular arch subjected to self-weight. Although pointed masonry arches, beside circular and elliptical, are very common in historic structures, particularly in Gothic architecture, their structural behaviour according to thrust line theory is still insufficiently researched. In this paper, employing radial stereotomy, the characteristic elements of analysis as well as computation, which diverse the pointed from semicircular arches, are noticed. To be more precise, eccentricity, being the measure of pointedness, position of the application points of relevant forces and the critical sections containing rupture points, are particularly treated. Hence, the appropriate correlation between the shape of an arch and collapse mode are provided. It has been concluded that, when limit equilibrium state is attained, there are four admissible collapse modes and their order of the occurrence is pointed out. In addition, the geometry of optimal pointed arch, having maximum use of its thickness, is indicated. Furthermore, the correlation between eccentricity, embrace angle and minimum thickness is graphically presented, and the numerical values for the most common pointed arches are provided. The framework set in this paper can be used for the various analysis of mechanical behaviour of different circular based masonry arches containing the pointed crown.

Keywords: Gothic arches, geometric analysis, limit equilibrium analysis, thrust line theory, collapse mode.

1 INTRODUCTION In order to predict and prevent possible collapse of vaulted masonry structures, being a large part of architectural heritage, scholars provide various models, which can be used in the stability and safety analysis. However, before the end of the 17th century, even the greatest arches, vaults and domes were built following purely geometrical rules of construction [1] which had been based solely on experience and intuition [2], i.e. the correct dimensioning was the result of empirical observations over a long period of time, rather than the result of the knowledge of structural mechanics [3]. Although considered earlier, it was Couplet [4] who first stated precise assumptions about material behaviour: masonry has no tensile strength, has infinite compressive strength and sliding cannot occur [5]. Introducing the collapse modes, i.e. the possible mechanisms of collapse caused by the formations of rupture joints or so called hinges, as the basis for the analysis of arch failure, he tried to find the minimum thickness of a uniform semicircular arch subjected to its own weight, which is today known as the Couplet problem. Thereafter, Coulomb [6] derived, from static equilibrium, method of maxima and minima (see [7] as well), showing the existence of two limits for the value of the horizontal thrust which arch produces.

According to [8], it was apparently Young who in 1817 proposed theory of arches based in the line of thrust concept. However, Moseley published series of articles on the theory of the stability of structures, particularly arches [9], where thrust line (i.e. line of resistance) is elaborated in detail. Accordingly, representing the load path, it is the locus of the application points of the resultant thrust forces at the joints (beds) between the voussoirs of the arch. In addition, he introduced the principle of least resistance [10], establishing that the actual thrust line was that of the minimum thrust. Following the aforementioned assumptions, the problem of the stability of arch reduces to mainly geometrical task. Namely, a self-weight of the arch or its portion is substituted by the area of the arch ring, limited by extrados and intrados curves as well as by the particular joints between voussoirs, and is applied in the centre of gravity i.e. in the centroid of the limited area. Therefore, the important task is to


ISBN - 978-86-7892-807-9

109

determine this area and its centroid. Even later, in the 20th century, when the elastic theory completely took primacy, the usual procedure in practice still was to check the stability by graphical statics [8].

However, the most general theory of thrust line, concerning remarkable mathematical elaboration, not recognized until nowadays, and particularly applied to arches, had been developed by Serbian scholar Milutin Milankovitch [11] (see [12] as well), and is of great influence to this paper. Milankovitch introduced in the computation the true location of the centre of gravity of each ideal, generic voussoir, which was until then assumed to be located along the centreline of the arch. Hence, he was the first who gave the correct solution for the minimum thickness of semicircular arch.

Despite the fact that pointed masonry arches, beside circular and elliptical, are very common in historic structures, and their adoption represents the basis of distinction between Romanesque and Gothic architecture [13], their structural behaviour according to thrust line theory is still insufficiently researched. Aita et al. [14] considered pointed arches applying the non-linear elastic analysis and so-called method of stability area, De Rosa and Galizia [15] modelled and analysed masonry pointed arch as discrete systems of rigid blocks, and Romano and Ochsendorf based analysis on a work-balance equations (see [16] and [17]). In recent years, in the frame of limit equilibrium analysis, Alexakis and Makris computed the minimum thickness of elliptical arches [18]. However, the researches considering minimum thickness of pointed arches are barely present (in [16] and [17] one can find the values for only a few different shapes of pointed arches).

Since the 18th century, scholars have also inspected the effect of stereotomy on the shape of thrust line, and concluded that different stereotomies yield to the different distinguishable physically admissible thrust lines (see [19] and [20]). Commonly considered rupturing i.e. cut directions are radial (Fig. 1 (a)), where directions of the joints between voussoirs are concurrent to the centre of the arch (as used in Italian pointed arches), and normal (Fig. 1 (b)), where the joints are normal to the intrados (in the case of circular arch these two stereotomies coincide). A number of authors (e.g. [21], [22], [23], [11]) show that the calculation can be simplified by assuming the vertical sections, and what is often the case in the analysis of cross vaults (see for example [24], [25], [26], [27] and [28]).

Fig. 1. Different types of stereotomy present in pointed arches: (a) radial, (b) normal and (c) vertical

As pointed out by Ochsendorf [29], the exercised slices should reflect the construction of the masonry. Thus, there are several possible types of joints direction concerning pointed arches, among which the radial and normal are mostly used [26]. The aim of this paper is to provide the framework for the equilibrium analysis of pointed arches, according to thrust line theory, with employed radial stereotomy. Hence, specific geometric parameters which affect the failure of pointed arch are identified. In addition, general guidelines for the detection of minimum thickness of pointed arches are provided.

2 POINTED ARCH GEOMETRY AND ITS SPECIFIC PARAMETERS In order to carry out the appropriate correspondence between the shape of the arch and the mode of collapse, in this paper, pointed arches are considered regarding the eccentricity being the measure of pointedness, and in further discussion according to the number of hinges formed when limit state is attained. There are three types of pointed arches, as classified in common architectural dictionaries or manuscripts (see e.g. [30], [31] or [32] ): (a) slightly pointed arches having small eccentricity (also known as drop, depressed or obtuse arch), including the traditional third-point and fifth-point (’quinto acuto’) arch) (Fig. 2 (a)), (b) equilateral or three-pointed arch (’terzo acuto’) (Fig. 2 (b)), and (c)


ISBN - 978-86-7892-807-9

110

strongly pointed arches having big eccentricity (also known as lancet, acute or narrow pointed arch) including for example ’recto acuto’ (Fig. 2 (c)). According to the second order classification, arches can be incomplete (segmental) or overcomplete (horseshoe), if the angle of embrace is less or greater than 90°, as shown in Fig. 2 (d) and Fig. 2 (e), respectively.

Fig. 2. Different shapes of pointed arches: (a) drop arch (e.g. ‘quinto acuto’), (b) equilateral arch (‘terzo acuto’),

(c) lancet arch (e.g. ‘recto acuto’), (d) incomplete or segmental pointed arch, (e) overcomplete pointed arch

Due to the symmetry of arch, in following discussion only half-arch is considered. In Fig. 3 (a) relevant geometrical parameters are shown: R and t denote the mean radius and the thickness of the arch ring, respectively. The minimum value of thickness to radius ratio, t/R, i.e. the minimum possible thickness of the pointed arch normalized by the radius is of particular interest in this paper.

Fig. 3. (a) Characteristic geometric parameters of pointed arch, (b) free-body diagram of the top portion of the

arch with corresponding thrust forces acting on it, and (c) force polygon

Further, the value e, which measures the deviation from the circular shape, is the horizontal distance between the circular axis’ centre O and the centre of the pointed arch C. The angle α represents the angle of embrace, which is the complement of the springing angle. The angle φ is angular coordinate measured from the vertical axis of the symmetry of the arch, and defines the generic section. The substantial parameter of pointed arch is its eccentricity, ξ, being the measure of pointedness, and following [17] represents the ratio between e and the difference between R and e (i.e. the mean radius or semispan of the corresponding circular arch). The parameters of particular importance are the application points B and S of the horizontal thrust H acting at the crown joint, and the reaction force R acting at the springings, since, beside the eccentricity by default, their position affects the position of


ISBN - 978-86-7892-807-9

111

thrust line through the arch and therewith the value of horizontal thrust, changing the location of critical sections as well, as one can see in Fig. 4. The critical section regarding extrados and intrados refers to the joint where the thrust line approaches closest to extrados and intrados, respectively (see Fig. 3 (a) and Fig. 4).

Fig. 4. Geometric parameters which affect the position of thrust line: (a) eccentricity, (b) application point of the

horizontal thrust force at the crown, (c) application point of the reaction thrust force at the springing

The rigid arch is indeterminate to the third degree such that for any arch there is a family of possible equilibrium solutions, which can be visualized with lines of thrust obtained through graphical statics methods [17]. Accordingly, the force polygon expresses graphically the equilibrium of the system (see [20]). The lines of action of the resultant thrust forces generate the funicular polygon, and the lines of action of the weights of the voussoirs meet at the corners of the funicular polygon to satisfy moment equilibrium [19]. Therefore, knowing the weight W being the area of the half-arch as well as its centre of gravity, and assuming the application points B and S of the forces H and F, respectively, from rotational equilibrium about point S one can determine the value H of horizontal thrust (see Fig. 3 (a) and (c)). Furthermore, the resultant thrust T at generic section together with its point of application A is uniquely determined from the force and moment equilibrium, either graphically with the force polygon (see Fig. 3 (b) and (c)) or analytically by solving equilibrium equations. Accordingly, one can obtain the closed form expression for the thrust line of pointed arches (traced with the dashed line in Fig. 3 (a)). Hence, for each generic section, the distance of the thrust line, i.e. of the application point of the resultant thrust force, from the extrados and intrados, can be computed.

3 GENESIS OF ADMISSIBLE COLLAPSE MODES From Couplet onward, researchers have considered different collapse modes of the various shapes of arches, searching for the appropriate hinges arrangement and the corresponding minimum thickness. Some authors provided different collapse modes regarding the limit thrust line of the pointed arches (e.g. Méry [33]), but there is no complete set of collapse modes with precise correlation to the eccentricity and embrace angle.

In Fig. 5 (a) and (d) slightly and strongly pointed arch, respectively, of thickness greater than the minimum are shown, and the corresponding minimum (having maximum rise and minimum span) and maximum (having minimum rise and maximum span [34]) thrust lines are traced. It is known that for circular arches (where eccentricity equals zero) limit thrust line passes through extrados at the crown and at the springings; but if the eccentricity is slightly increased, resulting in a pointed arch, for the absence of concentrated loads thrust cannot pass through the apex of the crown, and therefore the minimum thrust line touches extrados on the both sides near the crown. In addition, it touches intrados at some (rupture) point, and passes between intrados and extrados at the springings. It is evident that if there is a slight asymmetry, either geometrically or in the loading, then one of the hinges near the crown will not be formed [17]. As the thickness decrease, the application point of the horizontal thrust at the crown departs from extrados approaching intrados, and the reaction at the springings approaches extrados.


ISBN - 978-86-7892-807-9

112

Fig. 5. Decreasement of arch thickness: (a) and (d) arch of thickness greater than the minimum – stable arch,

(b) and (e) arch of minimum thickness – limit equilibrium state, at the point of collapse, (c) and (f) arch of thickness smaller than the minimum – impossible state

When the thickness of the arch is sufficiently reduced and the limit state is assumed, minimum and the maximum thrust coincide [17] producing the unique, limit thrust line, which touches the intrados and extrados in more than four points. Thus the arch reaches a limit equilibrium state, i.e. it reaches the point of collapse [35], as shown in Fig. 5 (b) and (e). Limit thrust line passes through extrados at the springings and between intrados and extrados at the crown, forming six hinges, or through intrados of the crown and between intrados and extrados at the springings, forming five hinges, for slightly or strongly pointed arches, respectively. One can conclude that for the particular value of eccentricity, thrust line touches simultaneously both extrados at springings and intrados at crown, as well as extrados and intrados in another two points each, resulting in the limit collapse mode containing seven hinges. Such arch represents theoretically the thinnest possible (in the range of common shapes of pointed arches, excluding very large eccentricities for the greater embrace angle), having maximum use of its thickness, and therefore is optimal for the chosen angle of embrace. Furthermore, for the great value of eccentricity regarding embrace angle, collapse mode having five hinges occurs. However, if the thickness was yet reduced, the thrust line would tend to come out from the arch ring and the arch would become unstable (in Fig. 5 (c) and (f) imaginary position of thrust line is assumed). For that reason, builders, for strongly pointed arches, expanded the key-stone or added massive blocks, acting as the additional load at the crown (see [13] and [26]). Accordingly, for a fixed angle of embrace, one can notice the correlation between eccentricity and the order of the occurrence of collapse modes. Namely, starting with the segmental arch (having zero eccentricity and collapse mode with five hinges), and gradually increasing the eccentricity, four different collapse modes containing six, seven and five hinges (appearing in two different patterns) successively could be formed. It should be noted that it is a general rule but not all modes are present for all embrace angles.

4 DETERMINATION OF MINIMUM THICKNESS AND LIMIT COLLAPSE MODES After remarkable mathematical elaboration, searching for the expression more appropriate for the iterative procedure, Milankovitch computed numerically the minimum thickness of semicircular arch [11]. Since iterations may be easily done by using computer programs nowadays, the mathematical calculation which concerns the finding of the minimum value of equation by the mean of differentiating can be omitted, and the iterative procedures can be done at the earlier stage of the computation, i.e. with the more complex expressions. In order to obtain the limit thickness, the thickness of the arch and the position of relevant application point have to be modified according to critical sections, in which thrust line simultaneously has to reach extrados and intrados. Hence, the sections neighbouring to the


ISBN - 978-86-7892-807-9

113

critical sections, defining the critical area, have to be thoroughly analysed. In pre-processing step, arch is initially divided into five portions having equal angle of embrace, as shown in Fig. 6 (a). Furthermore, each arch portion is subdivided into appropriate number of segments regarding generic sections, where the greater number of segments should correspond to the arch portion which will contain critical section. For example, in Fig. 6 the subdivision 3–7–3–7–3 is shown, and regarding the critical section, the second preceding and the second succeeding section are adopted as the bordering sections which determine the critical area for the next iteration, along with the thickness modification, and so on, until thrust line reaches extrados and intrados up to a satisfactory precision.

Fig. 6. Appropriate subdivision of the arch with respect to the critical sections and the gradual narrowing of the

critical area selection

As shown in the previous section, when minimum thickness is assumed, thrust line passes through one extremity (i.e. endpoint on extrados or intrados) of springing or crown joint. Hence, one of two application points, regarding horizontal thrust at crown and reaction force at springing, is known, and the other one can be set as the variable whose change affects the position of thrust line along the arch. On the other hand, the thickness of the arch is modified with respect to the other critical joint.

When the limit collapse mode with seven hinges is considered, both relevant application points are known, and therefore eccentricity and thickness should be varied with respect to the critical sections. Regarding the limit collapse mode comprising five hinges, there exists only one critical section, and that regards extrados. By gradually increasing eccentricity smaller than the limit one, along with the adjusting of thickness, critical section regarding intrados approaches the springing, and when it reaches it, limit eccentricity is obtained. Thus, to determine the eccentricity of limit collapse mode, its value needs to be known only approximately.

5 RESULTS On the basis of the material presented in the section 3 of this paper, one can define the characteristic positions of the application points of horizontal thrust acting at the crown and reaction force acting at the springings. Thus, the overview of derived collapse modes concerning the order of occurrence, as well as the corresponding position of theoretical hinges is given in Table 1.

Table 1. Admissible collapse modes of pointed arches with the position of theoretical hinges

Collapse mode 0 1 2 3 4 Hinge position extrados intrados extrados intrados extrados intrados extrados intrados extrados intrados

Crown ● ● ● ● Springings ● ● ● ● ● ● ● ● Additional ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

Theoretical number of hinges 5 6 7 5 5

In accordance with the appropriate collapse mode, regarding the eccentricity and embrace angle, the iterative procedures based on the guidelines presented in section 4 can be developed, and the


ISBN - 978-86-7892-807-9

114

computational analysis can be conducted. Hence, the correlation between eccentricity and minimum thickness is traced in the graph presented in Fig. 7, that contains the family of curves regarding the various embrace angles.

Fig. 7. Graph showing the correlation between eccentricity and minimum thickness for various embrace angles

One can notice the trend of each curve having cusp (in the range of shown eccentricities and embrace angles) which corresponds to optimal pointed arch. In addition, numerical values for the three most common shapes of pointed arches, having right embrace angle, are pointed out. In the case of overcomplete arches, minimum thicknesses result in much greater values, and are here omitted for the lack of space.

6 CONCLUSION One of usual approaches employed in the stability and safety analysis of vaulted masonry structures concerns thrust line theory developed during the 18th and 19th century and which has been revisited several times since then. Considering the arch analysis, thrust line, representing the load path, is the locus of the application points of the resultant thrust forces which develop at the joints between the voussoirs of the arch. Serbian scholar Milutin Milankovitch was the first who provided the complete theory for the equilibrium of the arch of general shape according to thrust line analysis, as well as the correct mathematical elaboration, concerning the true location of the centres of mass of generic voussoir. In addition, he computed the minimum thickness of semicircular arch subjected to self-weight. Recently, in the frame of limit equilibrium analysis, the minimum thickness of elliptical arches has been computed. However, such researches considering pointed arches are barely present. Traditional assumptions eliminate the possibility of failure due to material strength, but permit only failure due to instability, i.e. by rotation of arch portion about rupture point on intrados or extrados. Since collapse mode of circular arch is not valid for pointed arches due to pointed crown, it was necessary to identify appropriate collapse modes.

In this paper the framework for the equilibrium analysis, according to thrust line theory, particularly applied to pointed arches, has been provided. Employing the radial stereotomy, the particular elements of analysis as well as computation, such as the eccentricity, being the measure of pointedness, position of application points of relevant forces or the critical sections forming the rupture points, that diverse the pointed from semicircular arches, have been noticed. It has been shown that they affect the global position of the thrust line along the arch, and hence have been correlated with the admissible collapse modes and the corresponding minimum thickness of the arch. It has been concluded that when limit equilibrium state is attained, there are four admissible collapse modes of pointed arches and their order of occurrence has been pointed out. In addition, the optimal arch having maximum use of its thickness has been indicated. Furthermore, the correlation between eccentricity, embrace angle and the minimum thickness has been graphically presented, and numerical values for the most common pointed arches have been provided.


ISBN - 978-86-7892-807-9

115

The framework set in this paper can be used in the computational analysis of mechanical behaviour of pointed masonry arches, as well as other types of circular based arches containing the pointed crown, particularly Tudor or foiled arches. Moreover, regarding the further possible developments of the present research, different types of exercised stereotomy, such are vertical or normal, and different loading conditions could be considered, with the respect to corresponding collapse modes.

ACKNOWLEDGEMENTS The paper was done within the Project No. TR36042 supported by the Ministry of Education, Science and Technological Development of the Republic of Serbia.

REFERENCES [1] Radelet-de Grave, P. (2003). The use of a particular form of the parallelogram law of forces for the building of vaults (1650–1750). In: Becchi, A., Corradi, M., Foce, F., Pedemonte, O. (ed.) Essays on the history of mechanics, Springer Basel AG, pp. 135–163.

[2] Benvenuto, E. (1991). An introduction to the history of structural mechanics. Springer-Verlag New York

[3] Cowan, H. J. (1981). Some Observations on the Structural Design of Masonry Arches and Domes Before the Age of Structural Mechanics. Architectural Science Review. 24(4), pp. 98–102.

[4] Couplet, P. (1729 ,1730). De la poussée des voûtes. Histoire de l’Académie Royale des Sciences, pp. 79–117 and pp. 117–141.

[5] Heyman, J. (1998). Structural Analysis: A Historical Approach. Cambridge University Press

[6] Coulomb, C. A. (1773). Essai sur une application des regles des maximis et minimis à quelques problémes de statique relativs a larquitecture. In: Mémoires de mathematique et de physique présentés lacadmie royal des sciences per divers savants et lus dans ses assemblées. 1, París, pp. 343–382.

[7] Heyman, J. (1972). Coulomb’s memoir on statics: An essay in the history of civil engineering. Cambridge university press, Cambridge

[8] Huerta, S. (2008). The Analysis of Masonry Architecture: A Historical Approach. Architectural Science Review, 51(4), pp. 297–328.

[9] Moseley, H. (1843). The Mechanical Principles of Engineering and Architecture. London

[10] Moseley, H. (1833). On a new principle in statics, called the Principle of least Pressure. The London and Edinburgh Philosophical magazine and journal of science, Vol. III, pp. 285–288.

[11] Milankovitch, M. (1907). Theorie der Druckkurven. Zeitschrift für Mathematik und Physik, 55, pp. 1–27.

[12] Foce, F. (2007). Milankovitch’s Theorie der Druckkurven: Good mechanics for masonry architecture. Nexus Netw. J. 9, pp. 185–210.

[13] Fitchen, J. (1961). The construction of Gothic cathedrals. The university of Chicago press, Chicago

[14] Aita, D., Barsotti, R., Bennati, S., Foce, F. (2004). The statics of pointed masonry arches between limit and elastic analysis. Arch Bridges IV, Advances in Assessment, Structural Design and Construction, P. Roca and C. Molins, eds., CIMNE, Barcelona, Spain, pp. 354–362.

[15] De Rosa, E., Galizia, F. (2007). Evaluation of safety of pointed masonry arches through the Static Theorem of Limit Analysis. In: ARCH07 – 5th International Conference on Arch Bridges. pp. 659–668.

[16] Romano, A. (2005). Modelling, Analysis and Testing of Masonry Structures. Doctoral Thesis, University of Naples Federico II, Faculty of Engineering, Naples, Italia

[17] Romano, A., Ochsendorf, J. A. (2010). The mechanics of gothic masonry arches. Int. J. Archit. Herit. 4, pp. 59–82.

[18] Alexakis, H., Makris, N. (2013). Minimum thickness of elliptical masonry arches. Acta Mech., 224(12), pp. 2977–2991.


ISBN - 978-86-7892-807-9

116

[19] Alexakis, H., Makris, N. (2014). Limit equilibrium analysis of masonry arches. Arch. Appl. Mech. 84(5), pp. 757-772.

[20] Makris, N., Alexakis, H. (2013). The effect of stereotomy on the shape of the thrust-line and the minimum thickness of semicircular masonry arches. Arch. Appl. Mech. 83(10), pp. 1511–1533.

[21] Fontana, M. (1794). Della dinamica libri tre. Parte II. Pavia

[22] Lamé, M.G., Clapeyron, E. (1823). Mémoire sur la stabilité des voûtes. Annales Des Mines 8, pp. 789–836.

[23] Lorgna, A. M. (1782). Saggi di statica e meccanica applicate alle arti. Vol. 1. Verona

[24] Block, P., Thierry Ciblac, T., Ochsendorf, J. (2006). Real-time limit analysis of vaulted masonry buildings. Computers and Structures 84, pp. 1841–1852.

[25] Körner, C. (1895). Gewölbte Decken. In: Durm, J., Ende, H., Schmitt, E., Wagner, H. (eds.) Handbuch der Architectur, Third part, 2. tome, 3. chapter, pp. 141–553. Verlag von Arno Bergsträsser, Darmstadt

[26] Ungewitter, G. (1890). Lehrbuch der Gotischen Konstruktionen. Vol. 1, T. O.Weigel Nachfolger, Leipzig, Germany

[27] Wittmann, W. (1879). Zur Theorie der Gewölbe. Zeitschrift für Bauwesen 29, pp. 61-74.

[28] Wolfe, W. S. (1921). Graphical analysis: a text book on graphic statics. McGraw-Hill book Company, Inc. New York

[29] Ochsendorf, J. (2002). Collapse of masonry structures. PhD Thesis, Department of Engineering, University of Cambridge, Cambridge

[30] Curl, J. S. (2007). A Dictionary of Architecture and Landscape Architecture. 2nd edition. Oxford University Press

[31] Davies, N., Jokiniemi, E. (2008). Dictionary of Architecture and Building Construction. Elsevier/Architectural Press

[32] Branner, R. (1960). Villard de Honnecourt, Archimedes, and Chartres. Journal of the Society of Architectural Historians, 19(3), pp. 91–96.

[33] Méry M. E. (1840). Sur l’èquilibre des voûtes en berceau. Annales des ponts et chaussées, 1(1), pp. 50–70. and plates CLXXXIII–CLXXXIV

[34] Como, M. (2013). Statics of Historic Masonry Constructions. Springer-Verlag, Berlin Heidelberg

[35] Heyman, J. (1966). The stone skeleton. Int. J. Solids Struct. 2, pp. 249–279.


ISBN - 978-86-7892-807-9

117

ARCHITECTURAL REPRESENTATIONS 1 - THE COURSE AT THE FACULTY OF TECHNICAL SCIENCES

Marko Lazić1*, dr Predrag Šiđanin2, dr Radovan Štulić3, dr Dejana Nedučin4 1Teaching assistant, Faculty of technical science, University of Novi Sad (SERBIA),

[email protected] 2Full professor, Faculty of technical science, University of Novi Sad (SERBIA),

[email protected] 3Full professor, Faculty of technical science, University of Novi Sad (SERBIA),

[email protected] 4Assistant professor, Faculty of technical science, University of Novi Sad (SERBIA),

[email protected]

Abstract The obligatory course Architectural representations 1 belongs to the group of courses that concern applied 3D modeling, visualization and architectural documentation management, being offered to students of architecture and urbanism at the Faculty of Technical Sciences of the University of Novi Sad. In this paper we present the structure of this course, pointing out the way how our students are brought into the state of art of the technology for successful handling of all common engineering aspects same as aspects of aesthetics during the whole design process of the built environment. One of the very important aims of the course is the fully understanding of the principles of BIM technologies, as well the capability development of their proper applications, particularly in the aspects of the precise drawing, creation of the whole project documentation and of the appropriate visualization. As the result of this course student works are presented and analyzed in this paper.

Keywords: Architectural representations, BIM technology, educational methodology.

1 INTRODUCTION For implementation of theoretical knowledge into practice information technologies are necessary in architectural education. Architectural representations 1 is one of obligatory courses in Architecture study program at the Faculty of Technical Sciences. It is a new course attended in 3rd semester of the undergraduate academic studies. Aim of this course is enabling students to realize basic interpretations and representations in architecture and urbanism using basic computer techniques. This aim is achieved through the use of Building Information Model (BIM) technology. BIM technology enables students to make conceptual architectural designs from scratch.

Mastering different areas of knowledge are necessary before this course is attended. Good spatial abilities i.e. high capabilities for fully understanding 3D properties out of 2D representations are among other abilities important for students. Good base for this are courses that students attend prior to course that is described in this paper.

2 METHODS Overview of the one of the obligatory courses of Architecture study program is main focus of this paper. Teaching methods in this course is presented as a process integrated into other important processes from first year of study to graduation.

In first part need for BIM technology usage in education of architects is analyzed. After this analysis, context for course development is presented. In chapter 5 teaching methodology is shown in detail. Some of the results, their interpretation and conclusions are presented in chapter 6.


ISBN - 978-86-7892-807-9

118

3 BIM TECHNOLOGY IN ARCHITECTURAL EDUCATION BIM technology is important for the architectural profession. This technology is intended to be exclusively used for the construction industries and therefore is developed according to their needs. Often the BIM is viewed as a set of tools, but it is a dynamic technology that is still in development. Eastman et al. [1] defines BIM as "modeling technology and an associated set of processes to production, communication and analysis of building model", where BIM is seen from the perspective of addition to profession, rather than as a set of tools that are available for users. In the literature BIM is also described as catalyst of changes in construction industry [2] and factor that will increase efficiency of it [3]. It is used in to improve the design, construction and facility maintenance processes in the built environment [1, 4, 5]. BIM has been demonstrated in several applications within the construction industry including better collaboration between stakeholders, collision detection between design disciplines, cost estimation, site planning and others [6, 7, 8].

The great potential in the field of development of BIM technology is confirmed by many researchers. The BIM have been recognized to be nascent but evolving rapidly within the construction context [9] and estimates that the annual value of BIM products and services will increase from 1.8 billion dollars in 2012 to 6.5 billion in 2020. There is a number of research focused on how the economic system will be changed through the full implementation of BIM in construction industry. Examples include greater collaboration between participants [10]; implementations of intelligent system [11], the relative proportions of the system to increased interoperability by bringing digital form [12] and visible improvements in terms of accuracy, interactivity, productivity, cost management and increase of the project overall quality [13].

Adoption of BIM technology is increasingly encouraged by the government and government institutions. United States are the global leaders in the adoption of this technology [14]. In 2007, the Administration of General Services (GSA), which is responsible for all federal projects, introduced the BIM technology as standard [15]. This has led to the implementation of BIM technology on approximately 9,000 facilities and had a great influence on other countries. The European Union in 2014 recommended all its members to use BIM (European Union Public Procurement Directive). There are laws enacted in the Netherlands, Denmark, Finland and Norway that already involve the use of BIM in certain areas. The UK government also has a BIM strategy as it would be made a obligatory part of public procurement projects from this year.

There are a number of commercial software solutions available for of BIM technology that are constantly improving. In Europe the most used software packages are Autodesk Revit, Graphisopht ArchiCAD, Nemstchek Vectorworks and Allplan Bently [16]. ArchiCAD software is used on Architectural representations 1 course because it has user friendly interface that is easy to apprehend by students and because improved versions of this software are available for students, teachers and laboratories for free since 2006.

4 COURSE IN RELATION TO OTHER COURSES IN STUDY PROGRAM The freshmen students of architecture attend two obligatory descriptive geometry courses in order to get fundamental knowledge in geometry and train visualization abilities of 3D space. The sequence of the topics (starting from basic geometric elements and finishing with various complex forms and their shadows) gives students opportunity to achieve good spatial abilities i.e. high capabilities for fully understanding 3D properties out of 2D representations what is crucial for any further practical computer aided either drawing or design. Items architectural drawing are essential in order to overcome the art display work in BIM software technology. More items from the first year of study includes this aspect and students are trained to work with the development of the conceptual design, as well as documentation and paper formats that can be used for this purpose. Thus, students can better understand and learn lesions in less time on analyzed course.

Blank coordinate system is first drawn or displayed in most of 3D drawing software, and also in descriptive geometry (DG). Geometry of any 3D object is presented by its projections. In ArchiCAD tools are working similar to drawing in DG. In next chapter this is illustrated through roof drawing analysis. Many tools in ArchiCAD are automated, but full understanding of DG principles are necessary for drawing and presenting complex geometrical structures.

Basic usage of computer technology, as well as usage of other software tools for 2D and 3D drawing are not a necessary, but they make a good foundation for BIM technology usage. In the first year, students receive basic information and skills of CAD technology and mesh modeling, as part of the


ISBN - 978-86-7892-807-9

119

study program. Interoperability and compatibility of CAD and BIM technology through used software is good and students can work on different parts of the architectural project in different technology.

Knowledge gained on the course presented in this paper can be applied directly to the 10 compulsory courses from 4th to 8th semester, where it can be used for production and preparing documentation for architectural and urban projects.

The course Architectural representation 1 is fully integrated with other courses on same study program. Knowledge gained in first year of study is necessary for understanding and learning and it can be used for many courses after 2nd year of study.

5 COURSE STRUCTURE The course introduces a new knowledge base for further studies and utilized efficiently the knowledge gained by students on previous courses. It is performed with one-time lectures and 2 hours exercises. Lectures are organized in the way to give students a deeper theoretical knowledge in the development and possibilities of BIM technology, and define the connections and relationship to other technologies used in architectural design (CAD Mesh modeling, NURBS modeling, etc.). Commercially available software packages are analyzed in the field with emphasis on the core similarities and differences between them. The material for the lectures is based on the state of the art in this domain and follows the new trends in the rapid development of BIM technology and the constant improvements made in this area. The literature about BIM, available on Serbian language is theoretically oriented, it doesn't include the practical implementing of the software. For that reason all the information from the lectures including the guidelines for the practical software application is available in Serbian language for the students in PowerPoint presentation and updated annually. Additional literature in English about BIM is also given [1, 17, 18] as optional.

The exercises consist of practical training where students learn to use Graphisopht ArchiCAD software. The structure of the exercise is organized as follows: Introduction to the functioning of the software and its tasks and possibilities (2 weeks), Developing the geometry of objects (7 weeks), Creating the planning documentation (3 weeks), Revision of the course material and skill testing (3 weeks). The main part of the course is dedicated to working with geometry and mastering basic aspects of drawing elements in 3D space. The application of the theoretical knowledge in practical work is always a greatest challenge for students when learning a new software. The way of drawing elements variable in length and height (roof, shell, terrain, morph) requires well mastered principles of working in the projections. The procedure for drawing a roof level in the software is shown in the Figure 1. The first step is the setting of the inclination and the reference amount (which represents the horizontal plane) shown in the Figure 1a. After that, the first trace of the roof deck level is inflicted. When inflicted by principals of geometry, the f trace of the plane and its angle of inclination toward the horizontal plane there is two solutions for defining the exact deck position level - Figure 1b. It is necessary to choose the selected direction of the slope of the plane (at least one point on a level above the horizontal is selected). Only after these procedures, the first projection of the roof level borders are drawn - Figure 1c. Subsequent manipulation of this drawn level requires a good understanding of the relationship of the first trace level to level position, as compared to the first projections limit level to the trace on the horizontal plane - pictures 1d and 1e.

The material students have access to on the practical exercises as well as the material for lectures is drawn up in the form of a script or two PDF. files that are updated annually. The material is supplemented and also in the form of video tutorials.

Students are graded in accordion to the conditions defined by the Bologna Declaration, they can gain a total of 100 points, including 30point for the pre-exam and 70 points for the test results. The regular atendancy at the exercises and lectures carry a total of 10 points, and the seminar carries 20 points. Students who collect the minimum of 15points are candidates for the final exam.


ISBN - 978-86-7892-807-9

120

a) b)

c)

d)

e) Fig. 1. The modeling of the roof level in ArchiCAD software

Students are graded for the level of their knowledge by doing their graphical work project at home independently and by software knowledge testing by resolving a graphical project task on the final exam. For their graphical homework project they get the material in a form of 2D drawings of small housing facilities. A ground floor level drawing and a number of perspective drawings of the facility are included in the material. Students apply the knowledge gained on the practical exercises and compile all the necessary parts of the documentation according to the criteria defined by the course program. In addition to the basic criteria such as proper dimensioning, making the display section and perspective view of the object, it is necessary to satisfy the criteria of working with multi-scale interoperability of various software, work with project design, project documentation in PDF format and apply different aspects of 3Dmodel visualization.


ISBN - 978-86-7892-807-9

121

Unlike the graphic work project students do at home, the testing included in the final exam imply a precise and detailed criteria set of software tools knowledge in each examination in the time interval of 1 hour. Detailed instructions are given for every task in a form 7 checkpoints that need to be met when working in the software.

Fig. 2. The example of exam task

By implementing these two different approaches, the homework and the final exam testing, the students who pass the subject are prepared to draw architectural projects from the preliminary design to the preparing a basic set of documentation needed for the realization of the final project. The program of this course prepares the students for work in the practice, for drawing quick and precise design solutions and doing project corrections if needed.

6 RESULTS The success of the presented teaching methodology can be viewed from two aspects. The first is the quality of student projects, and the other is analysis of statistics on final exam. The most of the students have drew their architectural project manually on other courses before attending this course. After attending it, they are trained to design and present projects that are complex in terms of geometry and architectural program. This can be seen in the following examples on Figures 3 and 4.

Fig. 3. Example of a student final project (floor plan and section)


ISBN - 978-86-7892-807-9

122

Fig. 4. Example of a student final project (3D rendered representation)

The statistics of students passing exam shows good results. On average 80% students that listened this course have passed its final exam in first term. This is complemented with statistics that all this students have had 1.2 week absence per student on average for 15 working weeks. Other 20% have had 3.33 week absence per student on average. In last generation of students that listened this course 110 out of 116 students have passed after 1 year after they finished the classes.

Results in both aspects of the analysis are showing that the course is well integrated in the study program and that the majority of students are ready to apply the knowledge gained by attending this course on other courses in their studies.

7 CONCLUSIONS In this paper importance of BIM technology in contemporary architectural practice, and in the architectural education, is presented. Knowledge necessary for using it with its full potential is also analyzed. Descriptive geometry is among other courses attended prior to analyzed course, the most important one.

Some examples of student work that illustrate the results of this course are presented in this paper. The course is well established and is an integral part of the architectural undergraduate study program. After the course is completed, students are able to draw and present complex architectural projects, that are required on courses that they attend in higher semesters and also have a good foundation for practical professional work.

REFERENCES [1] Eastman, C., Teicholz, P., & Sacks, R. (2011). BIM handbook: A guide to building information modeling for owners, managers, designers, engineers and contractors. John Wiley & Sons.


ISBN - 978-86-7892-807-9

123

[2] Bernstein, P.G., 2005. Building information modeling. Structural Engineer(Atlanta, Ga), 6(7), pp.18-21.

[3] Hampson, K. D., & Brandon, P. (2004). Construction 2020-A Vision For Australia's Property And Construction Industry. CRC Construction Innovation.

[4] Barlish, K., & Sullivan, K. (2012). How to measure the benefits of BIM—A case study approach. Automation in construction, 24, 149-159.Bazjanac, V. (2008). IFC BIM-based methodology for semi-automated building energy performance simulation. Lawrence Berkeley National Laboratory.

[5] Azhar, S. (2011). Building information modeling (BIM): Trends, benefits, risks, and challenges for the AEC industry. Leadership and Management in Engineering.

[6] Gu, N., & London, K. (2010). Understanding and facilitating BIM adoption in the AEC industry. Automation in construction, 19(8), 988-999.

[7] Boktor, J., Hanna, A., & Menassa, C. C. (2013). State of practice of building information modeling in the mechanical construction industry. Journal of Management in Engineering, 30(1), 78-85.

[8] Yan, W., Culp, C., & Graf, R. (2011). Integrating BIM and gaming for real-time interactive architectural visualization. Automation in Construction, 20(4), 446-458.

[9] Pike Research, (2012), web page: http://www.navigantresearch.com/wp-content/uploads/2012/05/BIM-12-Executive-Summary.pdf (accessed 21. 02. 2016.)

[10] Mao, W., Zhu, Y. & Ahmad, I. (2007). Applying metadata models to unstructured content of construction documents: A view-based approach. Journal of Automation in Construction, 16(1), 242-252.

[11] Lin, F., Yang, J. & Skitmore, M. (2003) The integration between design and maintenance of office building automation: A decision support approach. In, Procs 5th Asia-Pacific Structural Engineering and Construction Conference, Johor Baru, Malaysia, 1-7.

[12] Mihindu, S., & Arayici, Y. (2008, July). Digital construction through BIM systems will drive the re-engineering of construction business practices. InVisualisation, 2008 International Conference (pp. 29-34). IEEE.

[13] London, K., Singh, V., Taylor, C., Gu, N., & Brankovic, L. (2008). Building information modelling project decision support framework. In Proceedings of the Twenty-Fourth Annual Conference Association of Researchers in Construction Management (ARCOM). Association of Researchers in Construction Management (ARCOM).

[14] Wong, A. K. D., Wong, F. K., & Nadeem, A. (2009). Comparative roles of major stakeholders for the implementation of BIM in various countries. Hong Kong Polytechnic University.

[15] Khemlani, L. (2012). Around the world with BIM. AECbytes feature.

[16] European Architectural Barometer (2011), web page: http://uspdate.usp-mc.nl/uspdate.php?maand=mrt&jaar=2012&page=Bouwsector:-AutoCAD-is-de-populairste-CAD-software-onder-architecten-in-Europa (accessed 21. 02. 2016.)

[17] Deutsch, R. (2011). BIM and integrated design: strategies for architectural practice. John Wiley & Sons.

[18] Kymmell, W. (2008). Building Information Modeling: Planning and Managing Construction Projects with 4D CAD and Simulations.


ISBN - 978-86-7892-807-9

124

3D MODELING COURSE AT THE COMPUTER GRAPHICS - ENGINEERING ANIMATION STUDIES

Boris Stajić1, Nenad Šunjka2, Jovan Mijatov3, Ana Perišić4, Ratko Obradović5 1-5University of Novi Sad, Faculty of Technical Sciences, SERBIA

Email: [email protected]; [email protected]; [email protected] [email protected]; [email protected]

Abstract This paper shows the content of the course on 3D modeling, which students of Computer Graphics –Engineering Animation attend during the second year of their studies. By describing two assignments and a course project which students completed during the semester, we present the processes of developing the skills of texturing, setting the light environment and rendering. The course emphasizes correct modeling, topology and different way of modeling rigid and organic models. The subject 3D modeling is a resumption of the subject Spatial Shape Design from the beginning of this study programme. Studying the subject Spatial Shape Design, students have the opportunity to cover the basics of modeling in 3D space. After completing these two courses, students acquire enough skills and knowledge for studying the basics of animation of rigid bodies.

Keywords: Computer Graphics, Texturing, Lighting, Rendering.

1 INTRODUCTION The programme for academic studies in Engineering Animation was established at the Faculty of Technical Sciences ([1], [2], [5] and [7]),University of Novi Sad as an interdisciplinary programme between mathematics and several engineering disciplines: electrical engineering and computer science, electronics and telecommunications; computer graphics and computer geometry; mechanical engineering within the frame of theory of mechanism and machines and engineering drawing with descriptive geometry, engineering communications and the basic principles of constructions; architecture and civil engineering, through the theory and interpretation of geometrical space in architecture, urbanism and civil engineering, wherever visual presentation can contribute to solving problems.

Computer Graphics can be used for educational interdisciplinary presentations and visualization is an ideal means to teach any discipline that could use visual presentation. Everyone needs visualization because it is the most natural way in which people view the world. It represents an excellent choice for presentation – visual presentation in studying and teaching, as well as in information transfer. We are all familiar with the expression which says that "a picture is worth 1000 words".

Engineering Animation is used as a presentation technique. It can present an important link between an idea and its realization like building a model/object. Engineers always needed a method to explain a project to the public, especially to the potential investors without whose support the project would have no chance to be realized. The methods of presentation changed through history from drawings, impressive scale models, to Multimedia Digital Objects displays. The goal of every presentation is to give plenty of information about the project. The data should be readable to professionals but also to anybody else not from technical or space designing professions. A high-level presentation is not only easy to comprehend regardless of the level of viewer’s education and professionalism but should also impress the potential sponsors.

It also has an important place in the film industry, especially since 3D movies have been created. It is also used for the development of computer games and WEB design, the industries which owe much of their attractiveness and propulsiveness to the sophisticated use of computer animation [9].

Not the least important is a place it has in education as a frame for digital learning in general. Engineering Animation is often used for the simulation of production processes, unavailable or insufficiently visible elements (underground and underwater installations, geological mapping, mechanical elements, anatomic parts etc.), risk simulations (earthquakes, floods, fire, etc.) but also for the visualization of different types of data/information.


ISBN - 978-86-7892-807-9

125

All this gives this profession a significant social importance and justifies the investments both in the development of the required technology and in the training of the professionals who need be able to "professionally answer" to the demands of this widespread profession, necessary at present but also in the future. In achieving these goals with our students independent work is accentuated, participation in the professional and developmental projects is encouraged, the abilities to solve specific problems are emphasized and developed, teamwork is nursed and a variety of ideas and approaches are required.

2 COURSE CONTENT 3D modeling course deals with the following topics [6]:

Mapping. Mapping procedure, Interpolation, Interpolation parameters, linear interpolation in 3D. MIP map. Bump mapping. Dislocated map. Illumination map. Illumination. Definitions: light, lighting, shadows. Visualization of lights. Reflection: Phong model, Specular reflection, Diffuse reflection. White and colored light. Tinted light. Light and water, Caustics. Light sources: point and cone light. Cylindrical light and area light. Ambient and linear light. Basic components of light sources: position and orientation, color and intensity, shadow. Lighting scenes: Key light, Fill light, Kick and Rim light. Light position: frontal and laterally. Camera. 3D view: synthetic camera. Visible surfaces. Synthetic camera: position, orientation, Look and Up vectors. Aspect Ratio. Angle of view, camera lens. Front and Back clipping planes. Focal Length. Camera types: Target and Free. Camera view control. Depth of Field. Camera on Path. Types of Camera Shots. Dynamic camera.

3 CONTENT OF THE PRACTICAL PART OF THE COURSE The practical part of the course consists of two assignments and a Course Project.

3.1 Course assignment 1 (CA1) Part I

The task is to model, in detail, the objects from the set (High Poly = HP) and their surroundings. The measures of the surroundings are given in Fig.1.The backyard of each house should not be shorter than 3m and not longer than 10m.The road in front of each house is 3m long. The lot on which the house is situated exceeds the dimensions of the house by 1m on each side. Generate Topology modification can be used only in this part of the assignment.

Fig. 1. Measures of surroundings

It is necessary to modify and add models in the surroundings, in the backyard:

Wavy part of the yard or a hill; Make 1-3 stumps; Make1-3 trees (object from software); Create 1-2 water surfaces (fountain, bowl with water, muddy pond, small lake); Set 1-3 bushes or flowers beds (objects from software); Create a stone path (min 0.3m wide and 2m long); Set three smaller and one bigger stone in geometrical shape; Part II Modeling optimized (Low Poly = LP) version of the house from the first part of the assignment. The size of the object must be the same as those from the High Poly version. The dimensions of the yard


ISBN - 978-86-7892-807-9

126

on which is the house is situated stay the same as in the first part, there are no more objects to set. The modification of field set in motion must be done in both versions. Part III Modeling one room (whose interior fits with the houses which are on the set). Objects which must be in the room are: A wooden table, a chair, and a cushion; A drape (it may be a curtain, a canopy or a garment that is hung on the hanger); A carpet or fur on the floor that covers only part of the floor (alternatively fur or tapestry on the wall); A copper cup, a cup, vase or bowl; A weapon that consists of metal and wood work (a sword, a knife, an ax ... ); The entire hull or part of the armor modeled on images from the exercise; A glass bottle of wine or a glass crucible with liquid or a glass bottle with perfume/drug/poison; A wooden object (damaged) – a bucket or a shield or a decorative object; Figurines or a small sculpture; Three pictures with carved or damaged frame; A fireplace or fireside; A piece of parchment or an open book; Candlesticks with candles (min 1, max 5); The room must have at least two windows with wooden frames and at least one wooden door with wooden frames (the doors should have metal locks inside and out); A wooden cabinet with two drawers (drawers should have metal handles); A wooden wardrobe with double door (the doors should have metal handles).

3.2 Course assignment 2 (CA2) Set materials on all objects in the scene from the course assignment one. The use of Mental Ray materials (Arch&Design) is required for each object you need to texture. The objects in one .max file are divided into three independent parts: HP Exterior, LP Exterior and Interior as CA1 demands. Specific requirements necessary to fulfill in this part are:

HP Exterior:

• use modifier UVW map for manipulation of texture • make a material in which at least 2 channels for maps are used • make at least one Multi/sub object material • use procedural map Noise and one more procedural map of your choice • use viewport canvas as a way of drawing at least one mask for material (black and white

colors) • make at least one Mix and Composite map and Blend material • make trees using maps for leaves and bark of a tree

LP Exterior:

• It is necessary to put all geometry in one object (Attach command) • This part of the assignment is done with one Unwrap map for color and one for bump

(normal map). Dimensions of maps are 2048x2048 Interior:

• On the objects, make at least one material metal, glass, wood, textile, and water • Make unwrap of an object with organic form (for example shell or figure) • Make unwrap of an object with straight geometrical forms (you can model a box) • The use of bump maps and/or normal maps for objects is obligatory when required • The use of cutout maps for objects is obligatory when required (at least on one object)

With each material characteristic features of the material (color, reflexivity, transparency etc.) should be considered.

3.2 Course Project (CP) The project consists of two parts. The first part involves interior visualization, and the second part exterior visualization.

3.2.1 Interior visualization The model of a room made in the CA1 is modified as follows:

• The model should have two windows on one wall and the third window on one of the two adjacent walls

• Define curved arched model – background on the outside that can be seen through all three windows


ISBN - 978-86-7892-807-9

127

• It is necessary to add the ceiling in the room • Light sources can be oil lamp, fire in the fireplace or candles on the table.... • It is necessary to add at least one Target camera in the corner of the room through which you

can see at least one window and, at least, one artificial light source. Adjust, if necessary, the parameters of the camera so that the interior can be seen as well as possible.

• Each model needs to be covered by material. • Light up the model (room) with standard lights in two scenarios: day and night; • Assume that sunlight enters into the room from the direction of the wall of the room where

there are two windows; • In the Day light scenario, it is necessary to light up the scene to fit the model of the real

behavior of light, which includes direct lighting, indirect lighting, shadows • In the Night light scenario, it is necessary to use at least one light for each artificial light

source; each light should be named in a way which clearly suggests what kind of light it is Assume that an oil lamp, a candle or a fireplace emits a different color of light than the other two light sources. Pay attention to all the parameters of lighting depending on the features and scenarios:

1. Position, intensity, and color of light 2. Volume and attenuation of light 3. Lighting components: Diffuse, Specular, and Ambient 4. The presence, type, intensity, color, quality and softness of the shadows

Pay attention to the background/external environment so that it can be seen or visible through the windows from different sides of the room, and that its brightness should be stronger from the direction where the light is coming from. Also, pay attention to the background with a different color tone in day and night scenario.

Advice: It is desirable to use the reference images, photos or concept art as an inspiration for setting up the basic visual characteristics of render.

Use Mental Ray renderer as render engine [10]. It is necessary to deliver at least two render images of the interior in high resolution from two different viewing angles and two wireframe render images for the same or different viewing angles.

3.2.2 Exterior visualization The model of a building with a yard built in CA1 and textured in the CA2 is modified as follows:

• Scene should have a background with an appropriate map or global Environment map • Define and map materials onto the models; Choose textures according to own taste; Each

model needs to be covered by material • Use Mental Ray renderer as a render engine • Scene needs to be lighted up with standard lights in two scenarios: day and night. In the night

scenario, use interior lighting as several light sources. Glow effect can be used on artificial light sources in the night scenario (lantern, lamp, candle, campfire, moonlight, light from a window...)

It is necessary to create at least two Target cameras and set up their parameters so that a scene can be viewed from at least two perspectives. It is desirable to do additional image processing (post-production) in Photoshop. The following pictures show examples of reference material that have been given to students during the course and the work on the projects.


ISBN - 978-86-7892-807-9

128

Fig. 2. Reference examples

4 STUDENTS RENDERING PICTURES The topic of the Course Project was Middle Ages.

We have chosen a single topic as part of our attempt to create an algorithm for evaluating students' works in a way which will enable us to minimize or eliminate the influence of subjective experience of students' works.

The work on the course is divided into three stages, accompanying the three tasks (CA1, CA2 and CP) which are evaluated individually. There is a gradation in relation to the complexity of the subject material but also in the evaluation of student work because more complex tasks are more highly scored.

1. The first topic involves only 3D modeling of objects in the scene based on 2D images that students get as reference and it is scored with 15 points.

In the CA1 students have practiced the basic 3D shape modeling. In the version of the house with a high number of polygons (High Poly Modeling) it was necessary, primarily, to set the basic form of house and roof. Then they made openings for doors and windows by setting up new edges to the appropriate places. Adding Shell modifier would change the house from and empty shell by giving thickness to the walls. After these major parts, they were able to continue adding details of medium size such as doors, frames, chimney, beams etc. The final phase of modeling a house with a high number of polygons consisted of modeling minor details such as fences, ladders, rope, tiles. The whole procedure is shown in Figure 3.

In order to raise the level of realism the objects were modified in such a way to break perfect geometrical shapes. To achieve this new edges were introduced. An example is a wooden beam whose geometry was modified so that its perfect rectangular shape was slightly curved to resemble hand carved wood. Chamfer modifier was placed on top of some objects. It is used to cut or round up hard edges of an object. This is used to quickly and easily gain a fine finishing effect which enhances visual quality. The wooden beam creation example can be seen in Figure 4.

The criteria which have to be met include: optimization of the geometry in relation to the level of detail that is modeled, geometry regularity – requires work with a square mesh of polygons, the naming of objects in the scene and the form and size of objects related to the scene. The works that received the highest scores had no problems such as the appearance of Ngon and open geometry and met the other specified conditions. These are the works that reach production level. The aim of CA1 is to train the students for the level of 3D modeling that is required for practical work.

During this course, the students also have learned and practiced Low Poly geometry creation. This has been achieved by using Box Object, on top of which Edit Poly modifier was placed. Using Edit Poly modifier tools a new house, without doors and windows, was modeled. The only condition that the new house had to meet was to follow the contour of the old, High Poly, house. This is needed so we can correctly extract High Poly details into the simpler, Low Poly house texture.

2. The second stage (CA2) includes the basic principles of texturing models and is assigned 20 points. This was students' first experience with the basic principles of texturing and with unwrapping. They textured the scene they had previously worked on for CA1 and after the evaluation and comments they corrected the flaws in it so that it was prepared for adding texture. The evaluation criteria focused on quality / realism and texture mapping as well as the use of all the options shown in the exercises,


ISBN - 978-86-7892-807-9

129

for example: proper application of blend and mix maps, procedural map, dirt map, cutout map, normal map and correct unwrapping of organic forms / geometries. In preparation for the assignment the students were given numerous examples of mapping out practices and worked on models that they have textured and unwrapped. After mastering the basic principles the students started texturing their project works and also had the opportunity to consult with teachers if they had any questions.

Fig. 3. 3D Scene Modeling Process

The second task contained specific sets of challenges which were all related to correct mapping and texturing of objects. Under the term “mapping” we consider a procedure in which a polygon wire is projected and stretched into 2D space so that the texture can be correctly applied to the object. Three modifiers were generally used to achieve this. These are UVWMap, MapScaler and UnwrapUVW modifiers. UVWMap and MapScaler are used on regular geometric forms such as tables, beams, walls etc. These modifiers stretch textures procedurally in 2D space using some if the standard primitives such as a Box, Cylinder or Sphere. The modifier will try to project a texture to geometry in 3D space using the aforementioned primitives. The example of UVWMap modifier use can be seen in Figure 5. An orange wire box, which can be seen around the beam, is used to project the texture to geometry.


ISBN - 978-86-7892-807-9

130

Fig. 4. Wooden Beam Modeling Preview Fig. 5.UVW Map Modifier Preview

Unwrap UVW is a special and advanced tool within the 3DStudio Max software package. It contains the editor in which it is possible to exactly specify which part of the texture will be mapped to a specific part of geometry. This modifier contains tools and functions that make this process easier. It is also possible to manually move the vertices and polygons to achieve the desired result. Working with UnwrapUVW modifier is more time consuming than using UVWMap and MapScaler modifiers, so we use it in cases where it is absolutely necessary. For example, we use UnwrapUVW on objects with organic forms, like sculptures, trees, vegetation in general and many other objects whose form is not geometric or rectangular. Figure 6 shows an example of the use of UnwrapUVW modifier on 3D teapot object.

3. The third phase is the final preparation of project task (CP) which involves the application of base rendering [8] (in this case Mental Ray render), setting up lights ([3] and [11]), environment and cameras on the scene. This assignment is worth 30 points. On the basis of assessments and comments of CA2 the students did corrections on textures and maps and adapted them to their final projects. This task (CP) represents the sum of everything that they have learned during this course and carries the most points within the evaluation of the subject. The results of their work and the skills the students have mastered this term can be seen in figures 7-23 which show their progress and creative approach to solving the task. The goal is to get the work done on the final production segment level. Within the course we also did showreel / presentation of the most successful work that has been placed on the website [7], to promote and support our best students.

The course project is a continuation of the previous two class assignments. The students here illuminate their modeled and textured scenes. The lighting system consists of the following lights:

1. Key light, directional light with parallel rays projection that mimics sun or moon light. 2. Sky light, set to simulate shades of blue ambient light obtained from the scattering of light in

the Earth’s atmosphere. 3. Sky portals, which are used to reinforce the influence of external lights on the interior. 4. Artificial light sources, such as oil lamps, candles, and a fireplace. 5. Additional, fill lights, to illuminate places of the scene that are too dark.

The students placed the appropriate Environment maps according to ambient and light in their scenes, e.g., a map for a sunny day, clouds, landscape photo that was taken at night for night renders etc.

Global Illumination and Photon Mapping were used in order to obtain realistic lighting.

Renders were then enhanced in programs for 2D picture editing. Contrast and colors were changed, vignetting, dust particles and backgrounds were added.


ISBN - 978-86-7892-807-9

131

Fig. 6. Unwrap UVW Preview

The course and student evaluation are designed so that students cover a broad and demanding field such as texturing, unwrapping and rendering gradually, step by step. This enables them to grasp the demanding and difficult material. The objective is achieved by repeatedly providing students with insight into their progress during the semester and enabling more corrections. The work on the course is supervised by teachers and teaching assistants so students can always consult them if they encounter problems in completing the project. The final results of the students' projects show that this practice, which is implemented in the course 3D modeling, proved successful. At the end of the course the students were trained to prepare the presentation of projects on a production level.

The total number of points on the course is 100, where CA1 is worth 15 points, CA2 is worth 20 points, CP carries 30 points, the activity in class exercises is worth 5 points and the theoretical (written) part of the exam carries 30 points [4].

In the next paragraph we give a student’s opinion on the course 3D modeling:

"The course 3D modeling has proven to be very useful and is a good example of the production of animated films, or any CG content. It is directly related to the subject Design of Spatial Form from the first year and it would be good if the two subjects were in the same year of study. It clearly differentiates between modeling, texturing, lighting and rendering. Here we meet for the first time the problems related to topology and optimizing the number of polygons. Then, by setting realistic lighting and rendering using Mental Ray we achieve a high level of realism to the scene. Using Mental Ray we were able to accurately show the physics of light and recreate effects such as caustics and other. In addition to all the technical aspects of software that are offered, we are left with enough space to present our ideas and imagination. Within each of the tasks, we are given the freedom to create, within the given topic, the content of our CG."

Comments on students’ works:

The following figures (7-23) are representative examples of students’ works. The grading system can be divided into three parts: The complexity and regularity of geometry, the texture and quality of


ISBN - 978-86-7892-807-9

132

materials, and lighting and presentation of the final render. In some images of students' works we show the geometry wireframe; in this part of the assessment it was important that the geometry wireframe is correct and not to cut itself. Also, the wireframe had to be square, we avoided triangles and polygons. The objects had to be properly arranged in the area and were not supposed to cut each other (e.g., a beam should not go through a wall, unless there is an opening in that section of the wall). An example of mesh geometry is shown in Figure 10, in the middle.

Since this year’s theme was the Middle Ages, the scenes are mostly dominated by wooden and stone objects. Therefore, it was necessary to select the proper textures, downloaded from the Internet, and use them to properly place the appropriate materials on the geometry. Students were instructed not to produce objects and materials that give the impression of modern machining, but instead to create the impression of manual processing. Also, we made sure that objects should not look new but used, shabby and worn. Figure 7 shows a wooden log cabin with weathered wooden beams.

In the pictures of the interior it can be seen that students are encouraged to put external light so that it illuminates the room under an interesting angle and creates volumetric effect. The night-light outside has cold, blue moon tones, which is in contrast with the warm, yellow and red light of lamps, candles and fireplaces (Fig. 12). In the set-up of exterior light the students tried to match it to the background of their choice so that the 3D scene and the background fit well. This is clearly illustrated in Fig. 9 where the castle nicely fits with the landscape in which it is located.

Fig. 7. Interior by day, final rendering by Vanja Ivanović

Fig. 8. Exterior by day, final rendering by Vanja Ivanović


ISBN - 978-86-7892-807-9

133

Fig. 9. Exterior by day and Exterior at night, combined rendered pictures with wireframe parts, final rendering by

Đorđe Vidović

Fig. 10. Exterior by day, combined rendered pictures with wireframe parts, final rendering by Živica Ranisavljev

Fig. 11. Interior at night, final rendering by Filip Mirčeski


ISBN - 978-86-7892-807-9

134

Fig. 12. Interior at night, combined rendered pictures with wireframe parts, final rendering by Đorđe Vidović

Fig. 13. Interior at night, combined rendered pictures with wireframe parts, final rendering by Bogdana Bojović

Fig. 14. Exterior at night, combined rendered pictures with wireframe parts,

final rendering by Dragica Ćurčin


ISBN - 978-86-7892-807-9

135


final rendering by Danilo Nikić


final rendering by David Danji

Fig. 17. Interior by day, final rendering by Filip Mirčeski


ISBN - 978-86-7892-807-9

136

Fig. 18. Interior by day, combined rendered pictures with wireframe parts,

final rendering by Živica Ranisavljev

Fig. 19. Interior by day, combined rendered pictures with wireframe parts,

final rendering by Slaviša Kostić

Fig. 20. Interior at night, final rendering by Dušan Njegovanović


ISBN - 978-86-7892-807-9

137

Fig. 21. Interior at night, combined rendered pictures with wireframe parts,

final rendering by Dragica Ćurčin

Fig. 22. Exterior by day, combined rendered pictures with wireframe parts,

final rendering by Slaviša Kostić

Fig. 23. Interior by day, final rendering by Marko Šašić


ISBN - 978-86-7892-807-9

138

5 CONCLUSION This paper shows in detail the practical part of the 3D Modeling course exemplified by two course assignments and a course project.

In the first task, it was necessary to model multiple objects, which are enumerated precisely, in High and Low Poly versions. In the second task, materials were placed on all objects from the first task, and three independent units were observed: High Poly Exterior, Low Poly Exterior and the third was Interior. The Course Project consisted of two parts: the first part involved visualization of the interior and the second part the visualization of the exterior. Under visualization, we mean appointing multiple lights, creating shadows and setting up the camera, image rendering and post-production.

Students attend the course in 3D Modeling during the third semester of their studies. In the fourth semester, they take the course on Character Animation continuing where 3D Modeling class finished. The students then learn about rigid body animation, animal rigging with a relatively simple system of movement, as well as the creation of short films on these two topics.

Under Rigid body we mean Solid body in which, in the terminology of classical mechanics, deformation is neglected. That means the distance between any two given points of a rigid body remains constant in time regardless of external forces exerted on it.

This is the fourth generation of students attending this course. With the first two generations we did not do unwrapping and the textures were done more trivially. Also, initially we did not do a detailed rendering technique using Mental Ray renderer. Now all these elements are packed in a 15-week course with 4 hours per week and the results are shown in the paper.

REFERENCES [1] Obradović, R., Vujanović, M., (2012). New Curriculum at the Faculty of Technical Sciences: Computer Graphics - Engineering Animation, 3rd International Scientific Conference moNGeometrija 2012, University of Novi Sad, Faculty of Technical Sciences, Novi Sad, Serbia, pp.481-486.

[2] Obradović, R., Popkonstantinović, B., Šidjanin, P., Vujanović, M., Milojević, Z. (2010). Computer Graphics and Computer Animation Studies at Serbian Faculties, 2nd International Scientific Conference on Geometry and Graphics moNGeometrija 2010, University of Belgrade, Belgrade, Serbia.

[3] Dutre Philip, Bala Kavita, Bekaert Philippe (2006). Advanced Global Illumination, A K Peters.

[4] Obradović, R., Aleksić, J., Vujanović, M., Banjac, B., Kekeljević, I., Dimitrijević, M., Malešević, B. (2014). Dominant Pre-Exam Obligations as a Basis for Successful Students Work, 4th International Scientific Conference on Geometry and Graphics moNGeometrija 2014, University of Niš, Vlasina, Serbia, pp.337-347.

[5] Faculty of Technical Sciences, Undergraduate Academic Studies Engineering Animation http://www.ftn.uns.ac.rs/2028806618/engineering-animation (accessed on March, 2016)

[6] Course 3D modeling

http://www.ftn.uns.ac.rs/88239028/3d-modeling (accessed on March, 2016) [7] Computer Graphics - Engineering Animation

http://www.racunarska-grafika.com/ (accessed on March, 2016) http://www.racunarska-grafika.com/index.php/studentski-radovi (accessed on March, 2016)

[8] Pharr Matt, Humphreys Greg (2010). Physically Based Rendering - from theory to implementation, Elsevier.

[9] Obradović, R., Vujanović, M., Popkonstantinović, B., Ivetić, D., Šiđanin, P. (2015). Study program Computer Graphics - Engineering Animation and their relation with modern Serbian CG Industry, WBCInno2015 International conference – September 18th, Novi Sad, Serbia, pp. 54-57.

[10] Alan Watt, Fabio Policarpo (2001). 3D Games Real-time Rendering and Software Technology, ACM SIGGRAPH Series.

[11] Jeremy Birn (2006). Digital Lighting & Rendering, New Riders, USA.


ISBN - 978-86-7892-807-9

139

DIGITAL FABRICATION STRATEGIES IN DESIGN EDUCATION

Bojan Tepavčević1*, Milena Stavric 2, Albert Wiltsche 3 , Vesna Stojaković 4, Mirko Raković 5

1,4 Department of Architecture, Faculty of Technical Sciences, SERBIA, [email protected], [email protected]

2,3 Graz University of Technology, Institute of Architecture and Media, AUSTRIA, [email protected], [email protected]

5 Department of Mechatronics, Faculty of Technical Sciences, SERBIA, [email protected]

Abstract The rapid development of parametrical tools for architectural design has created a big challenge for contemporary architectural education. Mathematics and geometry play again an important role in understanding these new tools. Recently, many universities have been introducing digital design and fabrication into their syllabus to provide and teach a broader understanding of parametrical design. In order to make virtual parametric models also buildable, for architectural usage, a huge amount of knowledge and skills are needed, which induces a big task for teachers and a challenge for students.

This paper presents our teaching approach to the design process through parametrical modelling strongly based on geometry, mathematics, programming, hardware computing and material behavior. Our teaching approach will be presented with four design projects that represent different strategies for digital fabrication. All described strategies involve different fabrication tools, methods and material logic into the design process. The core of our teaching approach is an understanding of geometrical and material properties of objects and its translation into a geometrical and mathematical language in terms of computer algorithms.

1 INTRODUCTION The role of digital fabrication in architectural and design education has been increased in the past decade. Digital fabrication courses became common in many architectural schools providing knowledge and fostering innovation in design by research. Digital fabrication was a very rare subject in architectural education at the beginning of the 21st century, but it has become a common standpoint in education practice because it can be used as a catalyst for design instead of just a means for production [1]. Despite that fact, in many architectural schools, digital fabrication labs only affect the increase in the number of physical models produced [2]. The reason lies in the fact that fostering innovation in design by digital fabrication can be achieved only by integrating with other areas into the design curricula such as computer science, geometry, mathematics and mechatronics. Moreover, experiments with different materials in the process of fabrication are also of extreme importance in order to achieve different material effects and to change our common way of thinking and perceiving materials. This interplay between material design and digital fabrication has been recently defined as new architectural phenomenon called digital materiality [3], material ecology [4] or new structuralism [5].

While there are many design fabrication experiments done in many academic institutions, only few papers have been published regarding the methods of teaching digital fabrication through different design strategies. The goal of this paper is to show different digital fabrication strategies collected from two years of teaching at the master program „Digital Techniques, Design and Production in Architecture and Urbanism“ at the University of Novi Sad, where the topic Digital fabrication is taught in four different courses. Four different design-to-fabrication strategies are taught in order to develop different skills and sensitivity for geometry, design, materiality and fabrication tools. Those four aspects will be further explained in detail through four case studies.







ISBN - 978-86-7892-807-9

140

2 RELATED WORK The first digital fabrication laboratories in architectural schools came up in the late 90’s and they were a result of the collaboration with mechanical engineering laboratories [6]. Rapid prototyping and CNC routers were the first machines that were utilized for fabricating models in architecture. In the following years, techniques and strategies have been rapidly expanded.

Many books and papers have been written regarding digital fabrication strategies, their influence on design thinking in digital age, material and design strategies. The role of digital fabrication in architectural education is examined and discussed, but only relating specific topics. The role of digital fabrication laboratories and a historical review are discussed by Celani [6], while Hemsath focuses his research on didactics of digital fabrication pedagogy, teaching methods and curriculums [7]. Some books and texts provide valuable insight on design experiments based on digital fabrication done in some schools [8, 9], while others are focused on specific fabrication strategies. Çokcan and Braumann discuss the use of robots in architectural education and their experiments on their courses at TU Vienna [10], while Greenhalgh explores the role of rapid prototyping in architectural education comparing to traditional scale models [11]. In this paper various design to fabrication strategies in education based on class experiments will be discussed.

3 METHODOLOGY The master program “Digital Techniques Design and Production in Architecture and Urbanism“ at the University of Novi Sad consists of 6 courses. Four of the courses are directly connected to the issues regarding digital fabrication. During the study students meet with different requirements, problems and topics in order to develop a sensitivity for various material, design and fabrication conditions. Moreover, design optimization tools are also used in order to achieve cost-efficient solutions and reduce scrap materials.

Students obtain basic skills and knowledge in the program course “Digital Design in Architecture and Urbanism“ where basic computation design strategies are taught. In the same way, geometric requirements for the basic approaches to fabrication of planar sheet materials are examined. The course subject “Digital Fabrication“ covers different approaches to digital fabrication by use of Laser cutting and CNC milling. Those two courses provide the knowledge for advanced techniques of digital fabrication. In the third course, “Generative Design“, students meet with basic requirements pertaining with designing interactive kinetic structures. The fourth course is focused on gaining basic skills in robotic fabrication based on mastering the programming language for industrial robots and integrating it with the design from the CAD software. In the four mentioned courses students meet with a diverse set of fabrication strategies based on different types of materials, fabrication tools and design requirements, as it is shown in table 1. In the following text, different fabrication strategies are described in more detail.

Table 1. Courses and fabrication strategies based on different types of materials, fabrication tools and design requirements

Course name Type of fabrication Type of Material Additional skills

Digital Fabrication Laser cutting,

CNC milling

Planar sheet (paper),

thick panels (XPS)

Geometry, Programming (Grasshopper3D)

Generative Design Laser cutting Planar sheet (paper), Programming (Processing),

Hardware and control (Arduino)

Sensors integration (Kinect)

Interactive systems Robotic assembly, Hotwire cutting

Volumetric (MDF bricks, EPS blocks)

Programming (Grasshopper3D, C#, Rapid),

Mathematics

Digital Design optional optional Programming (Grasshopper3D),

Geometry, Optimization


ISBN - 978-86-7892-807-9

141

3.1 Laser cutting and fabrication with planar sheet materials Laser cutting has been extensively used in architectural studios since it can be used for model fabrication without advanced knowledge in parametric design or 3D modelling. In the course “Digital fabrication” taught in the winter semester 2015 and 2016 two projects “Folded space structure” and “Digital ornament” are taught, exploring different design strategies.

The goal of the first project was to investigate folding techniques and to transfer geometrical and mathematical folding rules on buildable folding design. Students designed at first virtual parametrical models. Then they fabricated paper models which were finally presented in several exhibitions in Europe. The initial idea for this project originated from folding techniques for façade panels. Such panels can be manufactured from any foldable thin material like sheet metal or waterproofed paper (Figure 1).

Fig. 1. Folding structures made of “Fedrigoni paper” (170 grams, size 70x70 cm)

The first objective of the “Folded space structure” project was the definition of a developable parametrical model, which can be built from paper (thickness 130, 170 or 250grams). It was limited to the size of a laser cutter machine which cut the folding pattern. Hence, an optimization of the used material was necessary. By varying the design parameters and changing the design rules, it was possible to get a huge number of variations of the folded design. A special focus of the project was the efficient use of engraving and cutting the material in order to reach efficiency in the assembling process which is done by hand. Namely, it is well known fact that the construction of such folding structures is connected with very precise handwork. The costs and the working hours increase tremendously if the folding pattern is not appropriate prepared. This is why we try to include the material properties and rules of folding from the start of design process and to translate the geometrical rules of the transformations in the parametrical model.

3.2 Digital ornament and CNC milling of thick XPS plates In the past few years, computational design and digital fabrication had a great influence on rethinking ornaments in architecture. Today, an ornament is not considered only as an element of décor, but also as an element of structure or surface treatment and finishing. Moreover, parametric design tools allow greater control in designing repetitive elements of ornaments going beyond basic geometric transformations and wallpaper groups.

In the second project called “Digital ornament”, in the course Digital Fabrication, students were introduced to basic material/design/fabrication strategies made out of XPS plates. The goal of the project was rethinking the design of ornaments based on traditional patterns and its materialization in a novel way. The importance of the CNC milling in the architectural curricula lies in the fact that students have the opportunity to meet with challenges of design and fabrication strategies of full-scale models. Mill bit sizes and types, milling machine tolerance, size, thickness and properties of materials must be considered during the design phase. Design constraints are size and thickness of the XPS


ISBN - 978-86-7892-807-9

142

plates (5cm thick, 46cmx46cm size), size and type of the milling bit (6mm and 8mm size, chamfer and ball nose) and time for design and fabrication. Grasshopper definitions that automatically generate toolpaths and G-code for a CNC milling machine were designed. In the projects, students explore diverse possibilities in subtractive fabrication approaches based on the creation of ornamental design with different milling depth of cut and toolpaths. Such milled styrofoam can be used on facades or serve as concrete formwork.

Fig. 2. “Digital ornament” made of 5cm thick XPS plates

3.3. Responsive structures and hardware computing Responsive architecture is an emerging field in design and it incorporates adjustable elements that can adapt to the changes in environment conditions, user activities or social contexts [9]. In most cases responsive structures consists of sensors that measure environmental conditions (e.g. change of light or temperature intensity), actuators and microcontrollers for programming interaction and movable elements. One of the most challenging tasks in designing responsive facades is the integration of design strategies based on the geometry of kinetic elements and hardware computing in order to provide response from environmental stimuli or respond to design requests to behave a certain way. During the master course “Generative design”, students are introduced with basics in the programming language “Processing” as well as with using an Arduino microcontroller, used for signal processing from various types of sensors, and the Kinect sensor, etc.

The goal of the course project is the creation of adaptive structures that are able to respond with kinetic change to different environmental stimuli. Students have the opportunity to experiment with different sensors (light, movement, touch, humidity) and design movable patterns whose change of shape is directly affected by different environmental conditions.

Figure 3: Responsive façade model and sliding panels.


ISBN - 978-86-7892-807-9

143

3.4. Robotic fabrication and assembly In the past 10 years, robotic fabrication has been rapidly increased in academic institutions. In 2016, there are more than 30 robotic fabrication architectural labs in Europe. The reason for such a rapid development lies in the fact that industrial robots can be used for various tasks regarding fabrication and form-findings such as assembly with gripping tools, 3d printing, 5 axis milling, hot wire cutting etc. The most challenging task to the implementation of robotic fabrication in architectural design is creating a link between the CAD environment and programming robot for complex fabrication task.

In the master course “Interactive Systems”, students get experience in programming and fabrication. The goal of this course is designing a library for CAD software that can generate “RAPID-code“ for controlling the ABB industrial robot, based on the geometry defined in Grasshopper visual programming language. The focus of this course is not only the fabrication process but also developing the components for Grasshopper and software for ABB robot. In the course, three programming languages are used: RAPID programming language for industrial robot, Grasshopper for the implementation of parametric architectural design and C# in order to develop a library of Grasshopper components that are used to automatically generate RAPID code for the robot. In the course students work on two projects in order to meet with different robot arm tools (finger pneumatic parallel gripper and hotwire cutting tools), different fabrication strategies (brick laying assembly and subtractive fabrication) and material properties (MDF bricks and styrofoam).

Fig. 4. Brick laying assembly with industrial robot

4 CONCLUSION In this paper we discussed and described different approaches regarding the digital fabrication in architectural education described on course examples in the master program “Digital Techniques, Design and Production in Architecture at the Department of Architecture”, University of Novi Sad. Different methodologies based on various material types, engineering skills (programming and geometry) and fabrication tools are presented. It is shown that digital fabrication has a great potential in design research experiments in architecture that can affect our way of thinking about design in the digital age. Various examples presented in this paper show that diversity in design solutions could only be achieved with a broad understanding of various fields which makes this program specific. At the same time, fostering innovation in digital fabrication, can be result only through interplay between geometry, programming as well as a deep understanding of material properties and fabrication tools.


ISBN - 978-86-7892-807-9

144

ACKNOWLEDGMENT The authors would like to thank our students for their enthusiasm and work. Projects by Igor Nađ, Nataša Jovanović (Figure 1), Nikola Šćekić (Figure 2), Milica Milanović & Stefan Ilišković, Jovana Kovačević & Simo Ristanić, Ines Vladislav & Nemanja Jaćimovski, Milan Trninić & Stevan Ašćerić (Figure 3 from left to right) and Mladen Papović, Mirjana Živanov (Figure 4 from left to right) are presented.

REFERENCES [1] Cheng, NY and Hegre, E 2009, ‘Serendipity and Discovery in a Machine Age: Craft and a CNC Router’, Proceedings of the 29th Annual Conference of the Association for Computer Aided Design in Architecture, Chicago, Illinois, pp. 284-286.

[2] Duarte, J. P., G. Celani and R. Pupo. 2011. Inserting computational technologies in architectural curricula. Pp. 390-411 in Computational Design Methods and Technologies: Applications in CAD, CAM and CAE Education. Ning Gu and Xyan Wang, eds. IGI Global.

[3] F. Gramazio, M. Kohler, Digital Materiality in Architecture, First ed. Lars Muller, Zurich, 2008.

[4] N. Oxman, Towards a Material Ecology, Proceedings of the 32nd Annual Conference of the Association for Computer Aided Design in Architecture (ACADIA), San Francisco, 2012, pp. 19-20.

[5] R. Oxman, R. Oxman, 2010. New Structuralism: Design, Engineering and Architectural Technologies, Architectural Design, 80: 14–23

[6] G. Celani, Digital Fabrication Laboratories: Pedagogy and Impacts on Architectural Education, Nexus Netw J – Vol.14, No. 3, 2012

[7] T. Hemsath, Searching for Innovation Through Teaching Digital Fabrication, 28th eCAADe Conference Proceedings, 2010, pp.21-30

[8] M. Stavrić, Dr. P. Šiđanin, B. Tepavčević, Architectural Scale Models in the Digital Age: Design, Representation and Manufacturing, Springer-Verlag, Vienna, 2013.

[9] L. Iwamoto, Digital Fabrication Architectural and Material Techniques, Princeton Architectural Press, 2009.

[10] S. B. Çokcan, J. Braumann Industrial Robots for Design Education:

Robots as Open Interfaces beyond Fabrication in: Global Design and Local Materialization (Zhang and Sun Eds), Springer Berlin Heidelberg, pp 109-117

[11] S. Greenhalgh, 2009 Rapid Prototyping in Design Education: A Comparative Study of Rapid Prototyping and Traditional Model Construction, Master Thesis, Utah State University



This is the conference proceedings of the 4th eCAADe conference, held from 19–20 May 2016 at the

Faculty of Technical Sciences, University of Novi Sad, Novi Sad, Serbia.

Contemporary issues in architecture reside between designing complex shapes, increasing

performative capacities of building and efficient fabrication approaches. The new question for second

digital turn in architecture lies on tangible relations between fabrication, computational design, and

performative capacities of building.

The aim of 4th eCAADe international regional event is to promote the connection and exchange of

ideas between leading experts in the field of digital techniques applied to architecture and the

research groups interested in the relationship between performance based design, fabrication process

and advanced material design. Conference will be a place for testing and discussion novel ideas and

approaches regarding this topic.

eCAADe

eCAADe — the association for Education and research in Computer Aided Architectural Design in

Europe – is a non-profit making association of institutions and individuals with a common interest in

promoting good practice and sharing information in relation to the use of computers in research and

education in architecture and related professions. eCAADe was founded in 1983.

Digital Design Center

Digital Design Center is founded in 2014 at Faculty of Technical Sciences, University of Novi Sad,

Serbia.

Digital Design Center is dedicated to develop and investigate application of recently developed

technologies in architecture, urbanism, and design. With the rapid expansion of digital technologies

and simultaneous adjustment of industrial tools and techniques, architecture become perceivable in

the new way that was not possible before. Many opportunities that consider analysis, study, design

and product architectural structures exists today, and new ones constantly appear. Our goal is to

obtain, define, and solve current architectural problems, concentrating to the fields of architectural

geometry, robots in architecture, digital fabrication and production, applications for 3d modeling,

surveying and visualization, programming, and other approaches.

ISBN 978-86-7892-807-9

Documents

Proceedings of the 4th eCAADe International Regional Workshop