PhD Defence Topics

Broes De Cat

Department Computer science

PhD defence 16th of May 2014

Supervisor Prof. dr. Marc Denecker

FundingAgentschap voor Innovatiedoor Wetenschap en Technologie

E-mail [email protected]

Introduction / ObjectiveTasks tackled using software are becoming ever larger, involving complex and fast-changing requirements. This evolutioncomplicates software development, e.g., because in most software paradigms the domain knowledge is tightly mixed withhow a task is solved. Recently, the Knowledge Base paradigm was proposed, based on the idea that knowledge is in factnot tied to a specific task. Instead, it proposes to specify knowledge in a truly declarative language; solving different tasksthen comes down to applying the appropriate inference to the specified knowledge.One objective was to put the paradigm into practice by building a knowledge base system and evaluate its applicability onpractical applications. Second, the aim was to develop robust inference engines, to separate performance concerns fromthe modeling of knowledge.

Research MethodologyWe developed IDP, a knowledge base system for the languageFO(.)IDP, a rich extension of first-order logic. Inference engineswere developed for optimal model expansion (closely related tocombinatorial optimization) and deduction. The search algorithmMiniSat(IDP) was designed as the search backend of the system.Experimental evaluation took the form of comparing performanceagainst existing ASP and MiniZinc benchmarks, in combinationwith case studies in the domains of data mining, security andscheduling (see figure).

Results & ConclusionsThe main contributions of the work are:• The knowledge base system IDP, one of the first declarative sytemsnatively supporting multiple inferences.• A demonstration of the applicability of the KB paradigm, withreduced development time and easier maintenance is main results,without sacrificing performance.• The solver MiniSatID, one of the best ASP and CP systems.• Novel techniques to improve automated combinatorial search, thatreduce the need for an initial grounding phase and detect some typesof implicit knowledge.

Major publicationB. De Cat, M. Denecker, P. Stuckey and M. Bruynooghe (2014). Interleaving Grounding and Search. In CoRR,abs/1402.6889, Submitted to the Journal of Artificial Intelligence Research.

Separating Knowledge from Computation:An FO(.) Knowledge Base System and its Model Expansion Inference

Pieter Coulier

Department Department of Civil Engineering

PhD defence 15 May 2014

Supervisor Prof. dr. ir. Geert Degrande

Co-supervisor Prof. dr. ir. Geert Lombaert

Funding Fonds Wetenschappelijk Onderzoek


Introduction Dynamic soil-structure interaction (SSI) often plays a crucial role in structural mechanicsand should be accounted for in numerical models, e.g. for the following problems: damage assessment of buildings subjected to earthquakes design of offshore structures exposed to wind and wave loadings annoyance evaluation of vibrations due to road and railway traffic The numerical solution of large 3D dynamic SSI problems is very challenging from acomputational point of view, however, and often beyond current computer capabilities.Novel techniques are needed to cope with the challenges faced in academia and industry.

Research Methodology Efficient numerical techniques for large scale dynamic SSI problems have been developed:

fast boundary element method based on H-matrices

innovative algorithms for the coupling of finite elements and boundary elements

Challenging applications related to railway induced vibrations have been tackled:

the influence of source-receiver interaction

wave propagation in an urban environment

stiff wave barrier for the mitigation of railway induced vibrations

Results & Conclusions The developed techniques enable the solution of much larger problems than classical methodologies.

Major publications P. Coulier, S. Francois, G. Degrande, and G. Lombaert. Subgrade stiffening next to the track as a wave impeding barrier for railway induced vibrations.Soil Dynamics and Earthquake Engineering, 48:119–131, 2013.

P. Coulier, S. Francois, G. Lombaert, and G. Degrande. Application of hierarchical matrices to boundary element methods for elastodynamics based onGreen’s functions for a horizontally layered halfspace. Engineering Analysis with Boundary Elements, 37(12):1745–1758, 2013.

P. Coulier, S. Francois, G. Lombaert, and G. Degrande. Coupled finite element – hierarchical boundary element methods for dynamic soil–structureinteraction in the frequency domain. International Journal for Numerical Methods in Engineering, 97(7):505–530, 2014.

P. Coulier, A. Dijckmans, S. Francois, G. Degrande, and G. Lombaert. A spatial windowing technique to account for finite dimensions in 2.5D dynamicsoil–structure interaction problems. Soil Dynamics and Earthquake Engineering, 59:51–67, 2014.

P. Coulier, G. Lombaert, and G. Degrande. The influence of source–receiver interaction on the numerical prediction of railway induced vibrations. Journalof Sound and Vibration, 333(12):2520–2538, 2014.

The numerical solution of large scale dynamic soil-structure interactionproblems

without barrier with barrier insertion loss

H-matrix

Tom Redant

Department Elektrotechniek – ESAT


Supervisor Prof. dr. ir. Wim Dehaene


Introduction / ObjectiveThis work is on the design of digitizers for short-range localization systems using chip technology (CMOS). Theconsidered localization systems use estimations of signal arrival times to obtain the range between objects. The mostfamous and commercial example of such a localization system is GPS satellite navigation. The work closes the gapbetween high-level specifications for localization systems and design of the digitizer as part of the receiver.Covering a high bandwidth is a good method to create a high performance localization system. The way it is allocated hasimplications on design of the digitizer. The idea is to allocate a bandwidth as high as possible to get the best performanceover the highest range at the lowest amount of power and the shortest signal durations.

Research MethodologyFirst, a theoretical basis is introduced with respect to localization based on signal arrival times. Two strategies areconsidered regarding the use of bandwidth. These are a single-carrier (SC) approach and a multi-carrier (MC) approach.A prototype digitizer chip for both types will be studied, designed, fabricated and tested.

Results & Conclusions-A figure-of-merit for the specific localization systems isintroduced.-Event-based digitizers have a significant added value withrespect to classic digitizers. Though the effect of inputdepending propagation delays need to be tracked.-The continuous-time digital domain is a useful signalingdomain. For long signal durations, continuous-time digitizerarchitectures are beneficial if sampling clock stability and on-chip oscillator stability improves accordingly.-The design flow of a digitizer using the continuous-timedomain is dominated by long-lasting simulation times. Asolution for reliable spectral analysis based on short simulationdurations has been introduced.-Discrete carrier localization is a valuable alternative to singlecarrier localization, achieving high-update rate, sub-cm rangeprecisions at ranges up to 4 meters. Sub-sampling can beused to reduce power consumption in the receiver’s digitizer.

Top: The principle of estimation of arrival times andthe focus on the receiver’s digitizer. Bottom-Left: Chipphotograph of the digitizer for MC. Bottom-Right: Thesignal’s bandwidth is narrowed down, enablingefficient digitization.

Major publicationT. Redant, T. Ayhan, N. De Clercq, M. Verhelst, P. Reynaert, W. Dehaene (2014). A 40nm CMOS Receiver for 60GHz Discrete-Carrier Indoor Localization Achieving mm-Precision at 4m Range. IEEE International Solid-State Circuits Conference Digest of Technical Papers (ISSCC), 342-343, feb. 2014.

Digitizer Architectures for Short-Range Wireless Localization in Nanoscale CMOS

Both digitizer chips exploit a different method to take advantage ofhigh bandwidths. Both chips use the continuous-time digitaldomain since it fits the typical signals, used in these localizationsystems.

Max Boegli

Department Department of Mechanical Engineering


Supervisor Prof. dr. ir. Jan Swevers

Co-supervisor Prof. dr. ir. Joris De Schutter

Funding EMBOCON, OPTEC


Introduction / ObjectiveReal-time moving horizon estimation (MHE) algorithms are developed to optimally estimate the states and parameters ofadvanced friction models in order to track online friction dynamics in mechatronics systems, and compensate for theestimated friction. Real-time MHE is implemented and experimentally validated on a high-precision positioning system.

Research MethodologyState and parameter estimation via MHE aims to solve an optimization problem over a fixed-size observation window ofpast data, moving in time towards more recent measurements. Despite computation complexity, MHE are gradient-basedestimation techniques that can efficiently be implemented in real-time systems via code generation tools based onautomatic differentiation, such as the ACADO Toolkit developed within OPTEC.

To benefit from gradient-based optimization techniques and automatic code generation; an analytic, i.e. smoothed, version of the generalized Maxwell-slip (GMS) friction model, called S-GMS, is developed. Similar to the GMS model, the S-GMS is a multi-state model that also describes all essential friction characteristics in presliding and sliding motion.

Results & Conclusions Improved accuracy of the newly developed S-GMS model compared to the single-state LuGre model is demonstrated in the presliding hysteresis regime (represented by the friction vs position curves).

Real-time MHE friction observer is implemented for the S-GMS model at a sampling frequency of 1-10 kHz on a high-precision direct-drive linear motor.

Major publicationM. Boegli, T. De Laet, J. De Schutter, J. Swevers (2013). A Smoothed GMS Friction Model suited for Gradient-BasedFriction State and Parameter Estimation. IEEE/ASME Transactions on Mechatronics (in press).

Real-Time Moving Horizon Estimation for Advanced Motion ControlApplication to Friction State and Parameter Estimation

Marijn van de Weijer

DepartmentDepartment of Architecture, in cooperation with Hasselt University, Faculty of Architecture and arts

PhD defence 30 April 2014

Supervisor Prof. dr. ir. Hilde Heynen

Co-supervisor Prof. dr. ir. Koenraad Van Cleempoel

Co-supervisor Prof. dr. ir. Michael Ryckewaert

Funding FWO - Research Foundation Flanders

Introduction / ObjectiveThis research explores whether the intrinsic spatial surplus of the post- war, detached dwellings and low-density,dispersed residential neighbourhoods in Flanders, holds potential for transformation in line with contemporary housingstandards and demands. The ageing population and the decreasing average household size result in a demand for otherresidential typologies: smaller, more compact, and in central locations. Furthermore, high emission levels, trafficcongestion and loss of open space as a result of the low density housing model urge for strategies of transformation.

Research MethodologyThis work adopted an architectural approach in order toinvestigate the feasibility of transformative strategies,revolving around reconfiguration, replacement, andremoval of houses. As main source for data gatheringthroughout the entire project, 10 municipalities acrossFlanders, with a significant share of detached dwellings,were selected as case studies. Design, qualitativeresearch and architectural analysis contribute to atransdisciplinary research, and develop understanding ofdesign strategies intervening on different scale levels –from interior to neighbourhood.

Results & ConclusionsThe dissertation concludes that in low-density residentialenvironments, a search for a more sustainable housing isobstructed by a strong resistance which impedes radical,but necessary transformations. Such transformationswould be most feasible if developed in the form ofbalanced strategies on a municipal level.

Major publicationMarijn van de Weijer, Koenraad Van Cleempoel, and Hilde Heynen, "Positioning Research and Design in Academia andPractice. a Contribution to a Continuing Debate," Design Issues 30, no. 02 (2014): 17-29.

Reconfiguration, Replacement or Removal?

The professional designer should play an important rolein outlining alternative futures, and in deconstructingresilient ideas of what is ‘the ideal dwelling’, like, in thewords of David Harvey, a true insurgent architect, insearch of site-specific solutions for diversification anddensification of existing residential neighbourhoods.

Jordi Everts

Department Department of Electrical Engineering (ESAT)


Supervisor Prof. dr. ir. Johan Driesen

FundingAgency for Innovation by Scienceand Technology (IWT Vlaanderen)


Introduction / ObjectiveThis work investigates the feasibility of a single-stage, soft-switching dual active bridge (DAB) converter for the realizationof single-phase, isolated, and bidirectional AC-DC energy conversions. Thereby it is the principle goal to address thefundamental shortcomings in the existing analyses and circuit implementations of DAB converters. These lead to hard-switching operation and thus result in a reduced efficiency and in the worst case destruction of the switching devices.Application areas are chargers for electric vehicles, inverters for renewable energy sources (e.g. photovoltaic modules),as well as interfaces for residential DC distribution systems and energy storage systems.

Research MethodologyFor the first time it is shown that the single-stage DAB AC-DCconverter can be operated under soft-switching conditions within itswhole operating range. The key to success can be found in combining:

► an extensive circuit analysis and the introduction of new modulationmodes (exploitation of all possible degrees of freedom),

► a correct description of the soft-switching conditions, including themodeling of the semiconductor’s parasitic output capacitances,

► the application of a variable switching frequency,

► the introduction of commutation inductances as en essential circuitmodification.

In order to validate the analyses, a high-efficiency, high-power-densityconverter prototype system is developed using state-of-the-art designmethods/procedures and multi-objective optimization algorithms.

Results & ConclusionsConversion efficiencies higher than 95 % within themajor part of the output power range, with a very flatefficiency curve and thus a high partial-load efficiency,are reported. The peak efficiency is around 96 % andthe efficiency at nominal power approximately 95.6 %.Moreover, a high power density of 2 kW/liter isobtained. From a comparison with several (similar)dual-stage prototype systems found in literature, it isclear that the achieved performance is state-of-the-art.

Major publicationJ. Everts, F. Krismer, J. Van den Keybus, J. Driesen, and J.W. Kolar, ”Optimal ZVS Modulation of Single-Phase Single-Stage Bidirectional DAB AC-DC Converters,” IEEE Trans. on Power Electronics, vol.29, no.8, pp.3954-3970, Aug. 2014.

Modeling and Optimization of Bidirectional Dual Active Bridge AC-DC Converter Topologies

This work 230 VAC, 400 VDC

3.7 kW 2 kW/liter ! Bidirectional

Reference system 230 VAC, 400 VDC

3.33 kW 0.66 kW/liter Unidirectional

Measured conversion efficiency

Converter prototype system

Claus Claeys

Department Mechanical Engineering


Supervisor Prof. dr. ir. Wim Desmet

Co-supervisor Prof. dr. ir. Paul Sas

Funding IWT-Vlaanderen / FWO


Introduction / ObjectiveIn a context of noise control engineering, the mass of conventional panels and structures drives their acoustic insulationperformance. Silence comes typically with heavy constructions, but conflicts with the emerging tendency towardslightweight design. Therefore, there is a very challenging quest for innovative and effective noise control measures of lowmass and compact volume.Vibro−acoustic metamaterials come to the fore as possible candidates for lightweight material systems with superiornoise and vibration insulation, be it at least in some targeted and tuneable frequency ranges, referred to as stopbands. Inthis dissertation a novel method of metamaterial implementation is proposed, designed and analysed.

Research MethodologyThe potential of resonant metamaterials with vibro-acoustic stop band behaviour isinvestigated through a combined numerical-experimental approach. Unit cell modellingallows characterisation of wave propagation in infinite periodic structures built by arepetition of this unit cell. The knowledge and insight gained from studying these models isused to design and analyse equivalent finite size structures and investigate the linkbetween predicted stop band behaviour and the resulting vibro-aocustic behaviour.Through design of and measurements on a set of demonstrators, the potential ofmetamaterials to improve acoustic behaviour is proven and numerically predictedsensitivities for design parameters are confirmed.

Results & ConclusionsSandwich structure based metamaterials with embedded resonant structures exhibit vibro−acoustic stopbandbehaviour, resulting in an improved acoustic insertion loss and paving the way for a new class of light and compactnoise control solutions.

Major publicationClaeys C., Vergote K., Sas P. and Desmet, W., On the potential of tuned resonators to obtain low-frequency vibrationalstop bands in periodic panels. Journal of Sound and Vibration, 332, 6 (2013), 1418-1436.

Design and Analysis of Resonant Metamaterials for Acoustic Insulation

Frequency band:700-1000Hz

Additional noise reduction:15 dB

No added weightDemonstration: http://youtu.be/hMCfRHshjXc

Metamaterials

Valentijn De Smedt

Department Elektrotechniek - ESAT


Supervisor Prof. dr. ir. Wim Dehaene

Prof. dr. ir. Georges Gielen


Introduction / ObjectiveOne of the results of the quickly emerging technological possibilities in electronics industry is the appearance of WirelessSensor Networks in which hundreds of autonomously operating nodes are wirelessly connected in a network. Since acost decrease of these sensor nodes is necessary to become economically feasible, a growing need for fully-integratedwireless solutions is observed. This integration often goes at the cost of accuracy under influence of temperature andchanging supply voltages. This work investigates the possible techniques to overcome the temperature- and supplyvoltage-sensitivity of low-power fully-integrated time references for ultra-wideband transmitters.

Research MethodologyIn a first step, the basic needs to build a frequency reference are studied. Furthermore, a closer look is taken to thecauses of short-term as well as long-term frequency instability in an oscillator. This results in a design strategy, which isapplied to six design cases: 3 free-running oscillators, 2 injection-locked solutions and a sensor interface. During thedesign process, 2 major steps are alternately taken:

- Theoretical study of the circuit topology combined with numerical calculations and/or finite-element simulations.

- Circuit design and simulations to obtain more accurate results on previously estimated parameters.

This eventually results in a circuit implementation which is processed and measured. At measurement time, simulationsare used to fully understand the underlying mechanisms.

Results & ConclusionsDifferent circuit techniques are demonstrated to obtainseveral highly stable analog building blocks. The mainhighlights are:- A novel highly temperature- and supply voltage-

independent pulsed oscillator topology is developed.The frequency only depends on a bondwire LC tank.

- A novel injection-locked receiver topology isdemonstrated. This receiver, which is combined withthe clock generator, allows network coordination.

- A sensor interface is developed employing relativetiming accuracy between oscillator stages to obtainan absolute accurate and stable sensor readout.

By combining these blocks into a wireless sensor nodethe feasibility of fully-integrated timing solutions isdemonstrated. The presented designs surpass the stateof the art when looking at power and frequency stability.

The sensor interface is characterized using its measuredoutput spectrum (left). A high-quality bondwire inductor isused to obtain a highly-stable frequency tank (right).

Major publicationDe Smedt, V.; Gielen, G.G.E.; Dehaene, W., "Transient Behavior and Phase Noise Performance of Pulsed-HarmonicOscillators," Circuits and Systems I: Regular Papers, IEEE Transactions on, 2014

Temperature- and Supply Voltage-independent Time References for Wireless Sensor Networks

An injection-locked clock and wake-up receiver is developed.

Qingge Xie

Department Department of Metallurgy and Materials Engineering (MTM)


Supervisor Prof. dr. ir. Paul Van Houtte

Co-supervisor Prof. dr. ir. Albert Van Bael

Funding M2i in the Netherlands

[email protected]@126.com

Introduction / ObjectiveAnisotropic behavior of single phase metals is mainly a result of the morphological texture and the crystallographictexture. The ALAMEL model and the VPSC model with input of them can predict the metal forming operations. It is thenimperative that the two models (in comparison with the Taylor model) must make a sufficiently accurate calculation of theplastic anisotropy and the texture evolution. A problem of the ALAMEL model (giving oscillated slip rates of active slipsystems) was revisited and a grain boundary driven (GBD) model was proposed to solve it.Research MethodologyThe experimental data of four industrially produced steel sheets were used for validation of the three models in predictionof :

The flow stress anisotropy (flow stresses of tensile tests, plane strain tensile tests, simple shear tests and compressiontests. Each test was repeated three times and was tested in 2 or 3 directions) of a low carbon steel (DC06).

The initial grain shape effect on the initial q-value, the evolution of the q-value and the evolution of the deformationtextures for three low carbon steels (T52, T57 and T61).

Results & ConclusionsA number of effects of the texture, the initial grain shape,the slip activities, the short/long range interaction, and soon were investigated : The prediction of the flow stress anisotropy by thethree models combined with an isotropic hardening law isnot sufficiently accurate for DC06, which has a strongtexture.

Both ALAMEL and VPSC give the same initial grainshape effect on the q-value evolutions for three lowcarbon steels (T52, T57 and T61).

The newly proposed GBD model is more stable thanthe ALAMEL model in predicting the work hardeningbehavior.

Major publication

Q. Xie, P. Eyckens, H. Vegter, J. Moerman, A. Van Bael, P. Van Houtte. (2013). Polycrystal plasticity models based oncrystallographic and morphologic texture: Evaluation of predictions of plastic anisotropy and deformation texture.Materials Science and Engineering A, 581, 66-72. Q. Xie, A. Van Bael, J. Sidor, J. Moerman, P. Van Houtte. (2014). A new cluster type model for the simulation oftextures of polycrystalline metals. Acta Materialia, 69, 175-186.

Validation of the Taylor, ALAMEL and VPSC models for plastic anisotropy modeling of sheet metals

Figure 1: The q-value prediction by the VPSC modelwith (dashed line) or without (full line) considering theinitial grain shape. The open circles indicate theexperimental values.

Marcos Souza Lenzi

Department Department of Mechanical Engineering - PMA

PhD defence 31 March 2014



Introduction / ObjectiveFrequency sweeps over a wide frequency range and for a large number of frequencies are required in many applications inacoustics. When dealing with large systems, like those arising when tackling mid-frequency applications, this becomescomputationally expensive, if not prohibitive. This thesis proposes the development of computational algorithms that lead toFast Frequency Sweep approaches for the Finite Element Method (FEM) and Indirect Boundary Element Method (IBEM).

Research MethodologyThe two approaches developed in this thesis are based on reduced-order models built via Padé approximations. The Well-Conditioned Asymptotic Waveform Evaluation (WCAWE) is selected as Model Order Reduction (MOR) method of choice.

The fast frequency sweep for FEM (FFSFEM) is able to treat systems of more complicated wavenumber dependency, likely to be encountered when applying frequency-dependent boundary conditions and/or loadings.

The fast frequency sweep for IBEM (FFSIBEM) consists in a more involved process. The matrices are fully populated and inherently frequency dependent. Therefore, the approach accelerates the assembling of the system matrices and solving the associated linear system.

Results & ConclusionsMany numerical examples are used for the validation ofthe proposed approaches. In the FFSFEM, exterioracoustics problems involving PML, such as a sphere withvibrating cap, provided overall speed-up above 30.Interior acoustics problems, such as a car cavity withrealistic acoustic treatment provided speed-up factorsfrom 14 up to 21. For all test cases, frequency responsesobtained by the FFSFEM are in good agreement with theones obtained by the direct approach.

In the FFSIBEM, exterior and interior/exterior problemsare tackled. In particular, a car engine compartment isevaluated, on which a

Major publicationSouza Lenzi, M., Lefteriu, S., Beriot, H., and Desmet, W. A fast frequency sweep approach using Padé approximationsfor solving Helmholtz finite element models. Journal of Sound and Vibration (2012).

Non-modal acceleration methods for broadband acoustic simulation

100 200 300 400 500 600 700 800 900 100020

40

60

80

100

120

Frequency (Hz)

Sou

nd P

ress

ure

Leve

l (dB

)

FFSDirect approach

Numerical examples have shown approximate soundpressure levels (at the chosen field points) with goodaccuracy when comparing to the direct solver.

sensitivity analysis onmany MOR parametersis performed in order toverify the robustness ofthe approximations.

IntroductionThe field of natural language processing has made great advances in the last decades. Most of us have used Googletranslate, and have heard of Watson, the computer that won Jeopardy! Luckily, there are still many open problems innatural language processing, specifically with the machine learning algorithms behind it.

In this thesis we apply methods from operations research to natural language processing (NLP), and more specificallyinteger linear programming. Linear programming is an optimization method developed during WWII, to plan production inan optimal way. The goal is to maximize a linear objective function, subject to constraints.

Research MethodologyWe apply these optimization methods to 4 different problems in NLP:

�Sentence compression: summarizing individual sentences

�Rhetorical structure classification: dividing a document into segments that convey the same type of information (e.g. definitions, procedures, ..)

�Coreference resolution: determining which expressions in a text refer to the same person, location, company, ..

�Text simplification: reducing the difficulty of a text to a quantifiable level

Results & ConclusionsSolving NLP problems with integer linear programmingachieves excellent results

�By combining multiple local decisions into a globallyoptimal solution

�By reducing the NLP problems to well-known andintensively studied problems in operations research

Major publicationDe Belder J., de Kok D., van Noord G., Nauze F., van der Beek L., and Moens M.-F. (2013) Question answering ofinformative Web pages: How summarization technology helps. Essential Speech and Language Technology for Dutch:Resources, Tools and Applications, pp. 339-357, Berlin: Springer

Integer Linear Programming for Natural Language Processing

De Belder Jan

Department Department of Computer Science


Supervisor Prof. dr. Marie-Francine Moens


Alemayehu Haddis Getahun

Department Chemical engineering


Supervisor Prof. dr. ir. Bart Van der Bruggen

Co-supervisor Prof. dr. ir. Ilse Smets

Funding IUC-JU-VLIR


Introduction / ObjectiveMost urban centers in Ethiopia do not have wastewater treatment facility. Few onsite systems that exist in institutions andindustries do not work properly because of inappropriate technology selection. Constructed wetlands and wastestabilization ponds are reported to be the best options for developing countries. However, the functionality andapplicability of these systems in low income countries is not well studied. This PhD focused on evaluating low costwastewater treatment systems from the perspective of functionality, sustainability and cost effectiveness.Research MethodologyFirst a laboratory based water quality assessment was done followed bywastewater management survey in the country (emphasis on Universities).This led to technology choice and the constructed wetland was identified usingselection criteria. Before moving to the study on constructed wetlands a benchscale (pot) experiment with 8 indigenous macrophytes was performed. Two ofthe best performing macrophytes were used in the pilot scale horizontalsubsurface flow constructed wetland which was studied for performance,nutrient uptake, and root zone microbial respiration under irregular flowconditions.Results & ConclusionsCyperus papyrus and Scirpus validus were found to bethe macrophytes of choice for use in constructedwetlands. Plant harvesting seems to be of littleimportance in terms of P removal but significant for Nremoval. The performance of the horizontal subsurfaceflow constructed wetland was good, highly tolerant toirregular flow conditions and load fluctuations.

Major publicationHaddis, A., De Geyter, A., Smets, I. and Van der Bruggen, B. (2012). Assessment and selection of low cost onsite wastewater treatment systems in Africa: The case of Ethiopia. In Daniel, J.A. (ed). Advances in Environmental Research, Vol. 23. pp. 115–141. NOVA publishers, USA.

Performance of horizontal subsurface flow constructed wetlands for treating irregular wastewater flows in a tropical climate.

The conventional rule of thumb design criteria (5m2

/PE) can not be applied for a tropical low incomecountry like Ethiopia which needs only 3.6m2 /PE. Itis highly recommended that wastewater treatmentdesigns should consider local conditions.Universities, institutions and small communitiesare encouraged to use constructed wetlands as asecondary wastewater treatment option.

01020304050

CP CA TL CL CI SV% Plant uptake

Macrohpytes

Nutrient uptake by plants

N

P50

70

90

110

8 16 24 32 40 48 56 64 72 80 88 96% BOD5removal

Sampling period (days)

CPCATLCLCISVC

Veerle Cox

Department Architecture


Supervisor Prof. dr. ir.-arch. Bruno De Meulder

Funding IWT


logo funding agency if applicable

Introduction / ObjectiveThe “Spindus” research project, which is the abbreviation of “Spatial Innovation,Planning, Design and User involvement” was a four-year research projectintending to broaden the concept of spatial quality. It’s aim was to developpractical and pedagogical planning and design methodologies to assess, evaluateand implement spatial quality by adopting an interdisciplinary (involving differentresearch disciplines in a shared methodology) and transdisciplinary (involvingdifferent types of users) approach. In the area of urban design, Spindusresearchers searched for and analyzed methodologies that would allow projectsto be participatory and inclusive. The PhD research, framed within the Spindusproject, aspired to achieve insights into the capacity of designerly research withinparticipatory urban design practices.

Research MethodologyThe research methodology of the overall PhD was action research as a “learningby doing” process was initiated through organizing a series of urban designworkshops. The outcomes of a workshop would be verified and based on theseresults changes were proposed to eventually launch a new workshop. As a result,four 8-day workshops took place in Marke, Hoog Kortrijk, Luchtbal andRoeselare-Hooglede (see pictures). The urban design workshops made us ofdesignerly research, and in-depth interviews and questionnaires were conductedwith key actors, designers, residents and users of the case study sites.

Results & ConclusionsDesign may not only serve to bring about a shared vision for the future – and tomerely have a product as an outcome – but can be versatile and inclusive as itmay integrate different aspirations and opinions in space. The manuscriptpresents an altered workshop methodology working towards making thedesignerly research process more participatory and inclusive. The adaptedmethod suggests how urban design and social innovation can be betterintegrated and go beyond their usual scope, namely by shortening feedbackloops, introducing an iterative design process and stimulating social innovationwhile using design as a tool for facilitating dialogue and communication. As thecity and the territory are always in development, the mutual enforcement ofdesign and research is needed to envision – in an inclusive manner by theinvolvement of different types of users – how we want to live in the future.

Major publicationV. Cox*, M. Goethals*, B. De Meulder, J. Schreurs, F. Moulaert (2014). Beyond design and participation: the Thought forFood project in Flanders, Belgium. Journal of Urban Design, (forthcoming). *joint first authorship

Investigating urban design workshops in Flanders: Researching designerly research

Marke

Hoog Kortrijk

Luchtbal

Roeselare-Hooglede

Johan Rimez

Department Electrical Engineering (ELECTA)


Supervisor Prof. dr. ir. Ronnie Belmans

Co-supervisor Prof. dr. ir. Dirk Van Hertem


Introduction / ObjectiveThe contribution of this work is the extension of the optimal power flow formulation of electrical systems with an integratedDC grid using HVDC Voltage Source Converters. Not only the base case calculations, but also curative and preventivesecurity constrained cases and a wide range of optimisation objectives have been integrated in the developed algorithms.

This research is particularly important to Transmission System Operators for grid planning and operation of point-to-pointDC connections (e.g. the Belgian NEMO and ALEGrO projects), multi-terminal schemes and even complexly meshed DCsupergrids with multiple infeeds from offshore wind power plants.

Research MethodologyThe methodology used for this research is the algorithmic creation of a set of specific state variables, equality andinequality constraints which are integrated into an existing optimisation case:

• the AC/DC converters are modelled by expressing their loss function as equality constraints;

• two independent methods are developed to model the DC grid topology: the first is analogous to the classic AC gridapproach, the second creates the equations for the DC branches and grid topology separately;

• for the preventive security constrained OPF, the base case and the distinct systems taking into account the requestedoutages are assembled into one global optimisation case.

Results & ConclusionsThe research resulted in the creation algorithms and methodologieswhich integrate• the necessary constraint equations for the DC grid and AC/DC

converters, in a generic way, so even the most complex topologiescan be modeled,

• the creation of a solvable OPF formulation integrating allcontingency states of the hybrid AC/DC system,

• an extended set of objective and user-customisable objectivefunctions allowing also multi-objective OPF,

• the development of an open-source calculation tool

The algorithms have been tested on a range of small and largenetworks and are successfully benchmarked with other calculationtools.

Major publicationJ. Rimez, R. Belmans, “A combined AC/DC optimal power flow algorithm for meshed AC and DC networks linked by VSCconverters”, approved for publication in International Transactions on Electrical Energy Systems, Wiley, 2014

Optimal operation of hybrid AC/DC meshed grids

© Friends of the Supergrid

Hariharsudan SivaramakrishnanRadhakrishnanDepartment Department of Electrical Engineering


Supervisor Prof. dr. ir. Robert Mertens

Co-supervisor Prof. dr. ir. Jozef Poortmans

Funding KU Leuven


1014 1015 1016100

150

200

250

300

350

400

2-layer stack (reference) 3-layer stack (low) 3-layer stack (high)

Eff

ectiv

e lif

etim

e [

s]

Excess carrier density, p [cm-3]

Introduction / ObjectiveIn silicon photovoltaics, the cost of the silicon material itself forms a major fraction of the final solar module cost. In orderto reduce the amount of expensive high quality silicon to fabricate solar cells, two silicon solar cell concepts have beenenvisaged, namely the wafer-equivalent epitaxial silicon solar cell (WE-epicell) and the layer-transferred epitaxial siliconsolar cell (LT-epicell), which are only 30-50 µm thick, compared to today’s solar cell, which is ~170 µm thick. For both cellconcepts, porous silicon (PSi) is crucial. In WE-epicells, PSi is used as a Bragg reflector and a gettering layer. In LT-epicells, PSi enables the silicon layer transfer process itself. In both cell concepts, PSi also acts as the epitaxial template.The importance and role of PSi as a gettering layer in WE-epicells and as an epitaxial template in both cell concepts arestudied in detail and improvements are also demonstrated in this thesis.

Research MethodologyPorous silicon as a gettering layer: Firstly, thegettering mechanism and gettering efficiency of PSi isstudied theoretically and correlated experimentally.Secondly, an increase in PSi gettering efficiency isdemonstrated by reducing the void size. Maintechniques used in this work include: ab initiosimulations using density functional theory (DFT), totalreflection X-ray fluorescence (TXRF), secondary ionmass spectroscopy (SIMS) and microwave-detectedphotoconductance decay (µ-PCD).

Results & ConclusionsLeft: An example map ofFe concentration in theepitaxial layer in an Fe-contaminated sampleobtained using µ-PCD. Inthe area where there isembedded PSi, no Fe isdetected in the epitaxiallayer, while high Feconcentrations can bedetected outside this area,proving the high getteringefficiency of PSi.

Major publicationsH. Sivaramakrishnan Radhakrishnan et al., “Gettering of transition metals by porous silicon in epitaxial silicon solar cells,” Phys. StatusSolidi A, vol. 209, no. 10, pp. 1866–1871 (2012).H. Sivaramakrishnan Radhakrishnan et al., “Improving the Quality of Epitaxial Foils Produced Using a Porous Silicon-based LayerTransfer Process for High-Efficiency Thin-Film Crystalline Silicon Solar Cells”, IEEE J. of Photovoltaics, vol. 4, no. 1, pp. 70-77 (2014).

Silicon epitaxial layers grown on buried porous silicon templates for solar cells: detailed electrical and chemical understanding

Porous silicon as an epitaxial template: Firstly, methodologiesfor quantifying the quality of an epitaxial layer grown on PSitemplates using minority carrier lifetime measurements aredevised, both theoretically and experimentally. Secondly,methods to improve the porous silicon template to allow thegrowth of higher quality epitaxial layers is demonstrated. Maintechniques used in this work include numerical simulationsusing PC-1D, photoluminescence (PL), microwave-detectedphotoconductance (µ-PCD) and quasi-steady statephotoconductance (QSSPC).

( ) g [ ]

7.5x1011cm-3 5x1012

cm-3 A novel PSi stack which results in a surface zone of void-free

Si (left) is shown to increase the quality of the epitaxial layer,as evidenced by the improved minority carrier lifetime (right).

PSi

Epitaxial growth surface

Josep Balasch

Department Department of Electrical Engineering – ESAT / COSIC


Supervisor Prof. dr. Ingrid Verbauwhede

Co-supervisor Prof. dr. Bart Jacobs


Introduction / ObjectiveElectronic devices are becoming ubiquitous, mobile and integrated with their environment. This shift is however notwithout risks. On the one hand, the security of the devices can be compromised by malicious adversaries performingphysical attacks. On the other hand, the privacy of the users can be threatened by the monitoring nature of new servicesand applications. It is in this context that this research is situated. It addresses aspects related to the security ofembedded devices and the privacy of the humans interacting with them.

Research MethodologyIntegration of (physical) security and privacy into an embedded system implies to follow asystematic approach considering all abstraction layers of the design space:

System: security and privacy requirements

Algorithm: cryptographic blocks

Architecture: hardware, software

Circuit: transistors and logic gates

Major publicationJ. Balasch, A. Rial, C. Troncoso, C. Geuens, B. Preneel, and I. Verbauwhede, “PrETP: Privacy-Preserving Electronic TollPricing”, 19th Usenix Security Symposium, pages 63-78. USENIX, 2010.

Implementation Aspects of Security and Privacy in Embedded Design

EVALUATE

IMPLEMENT

DESIGN

Results & ConclusionsEmbedded design for embedded security:

In-depth study and characterization of the effects of setup time violations in embedded microcontrollers with a two-stage pipeline.

Security evaluation of a widely used commercial family of secure memories.

Development of a masking-based countermeasure against side channel attacks implementable at any order and based on inner product constructions.

Embedded design for privacy:

Design, implementation and evaluation of a privacy-preserving solution for Electronic Toll Pricing

Steven Op de beeck

Department Computer Science


Supervisor Prof. dr. ir. Wouter Joosen

Funding iMinds, IWT, and FWO


Accountingserver

Newspaperserver

Contentstorage

Profilestorage

Introduction / ObjectiveDesigning the software architecture is an essential part of the development of distributed systems. Yet, to be useful to developers, the architecture needs to be documented. Architects document software architecture using Architecture Description Languages, amongst several other ways. However, there is no ADL that supports the concepts specific to distributed system, like element instantiation and allocation to networked hosts. We developed an ADL that solves these and other problems that we uncovered in the state of the art of architecture description.

Research MethodologyFirst, this research was driven by an industry-grade case study on e-Media. An e-Media platform supports news in different media formats and sizes, to be delivered through different communication channels, onto multiple end-user devices. The case study was essential in the creation of an environment that was complex enough to reveal the problems that we set out to solve.

Second, we have created and maintained a set of tools for architects and developers to validate our research results and put them into practice.

Results & ConclusionsOur work consists of three contributions:

MViewADL, an ADL for distributed systems that supports the description of components and connectors, the specification of their runtime instances, and the allocation of instances onto distributed hosts.

ReVew, a concept and technique for the modularization of the contributions of architects to the architecture description. The concept introduces the generic idea of stepwise refinement of architecture description. The technique implements the concept in a generic way, to allow its use in various ADLs.

Tool support for the production of MViewADL descriptions, and subsequent code generation to multiple middleware platforms, such as JBoss and Spring.

Major publicationSteven Op de beeck, Marko van Dooren, Bert Lagaisse, Wouter Joosen, Modularity and variability of distributed softwarearchitectures through multi-view refinement of AO-connectors, LNCS Transactions on Aspect-Oriented SoftwareDevelopment, volume 7800, 2013.

Multi-view Architecture Description for Distributed Systems

Joris Van Ostaeyen



Supervisor Prof. dr. ir. Joost Duflou

Funding IWT


Introduction / ObjectiveOver the last decade, insight has grown that between a pure product manufacturer and a pure service provider, variousbusiness model options exist, in which products and services are combined to varying degrees. This concept is describedby the term Product--Service Systems (PSSs), which is the subject of a considerable amount of recent research attention.The overarching goal of the presented work is to analyze the business potential of a PSS from the point of view of amanufacturer of investment goods.Research MethodologyThe main research method applied is case study research. Five in-depth case studies were performed with Belgianmanufacturers who currently have a product-centric business model, but who would like to gain insight into the potentialbenefits of a PSS. The business potential of a PSS is investigated for a manufacturer of elevators, a provider of lightingcontrol systems, a provider of fire detection systems, a developer of diamond polishing systems and a manufacturer ofwind turbine gearboxes.

Results & ConclusionsIn this thesis a theoretical foundation for PSS research is proposed, including a new PSS definition, representationscheme and typology. Moreover, Functional Hierarchy Modeling is presented, a theoretical framework that allows torepresent the functions of an investment good on different levels of abstraction. Three complementary approaches forPSS ideation, that support manufacturers in identifying a broad set of PSS options, are proposed and illustrated.The second part of this dissertation presents a generic methodology to evaluate the business potential of a PSS. Thismethodology focuses on the innovation potential of a PSS in cost and value and allows to analyze the impact of risks anduncertainties. It is validated through its application on five in-depth case studies.

Major publicationVan Ostaeyen, J., Van Horenbeek, A., Pintelon, L., Duflou, J. (2013). A refined typology of Product-Service Systemsbased on Functional Hierarchy Modeling. Journal of Cleaner Production, 51, 261-276

Analysis of the business potential of product-service systems for investment goods

General structure of Functional Hierarchy Modeling

Typology of PSS revenue mechanisms

Benjamin Kam


PhD defence 28 February 2014

Supervisor Prof. dr. ir. Paul Heremans

Co-supervisor Dr. ir. Gerwin Gelinck

FundingFunded by IWT & in cooperation with imec and Holst Centre


Introduction / ObjectiveThin-film electronics are technologies to fabricate transistors and circuits, different than silicon technology. In thelatter, transistors are fabricated by creating structures on the surface of a mono-crystalline silicon wafer. In contrast,thin-film transistors (TFTs) are fabricated by depositing and patterning material layers on a variety of possiblesubstrates such as glass or even flexible plastic sheets.

Research MethodologyExperimental, using a systematic approach:

Fabricate and evaluate device architectures of FeFETs

Provide insights in device operation

Integrate with transistor circuitry on a flexible substrate

Realize a memory array

Results & Conclusions Data can be stored for at least one year in FeFETs. ON current

decays in two phases. However, this memory performance differs for different device architectures.

A model is proposed for the device operation of FeFETs, which clarifies peculiarities found in literature.

FeFETs are integrated in a NAND-like memory array on a flexible substrate. A disturb-free read- and write operation is demonstrated.

Major publicationB. Kam, T-H. Ke, A. Chasin, M. Tyagi, C. Cristoferi, K. Tempelaars, A. J. J. M. van Breemen, K. Myny, S. Schols, J.Genoe, G. H. Gelinck, P. Heremans (2014). Flexible NAND-like organic ferroelectric memory array. IEEE Electron DeviceLetters, accepted.

Device Architecture and Materials for Organic Ferroelectric Memory Arrays

1

0

Organic electronics is an emerging thin-film technology which offersa route to ubiquitous, low-cost and flexible electronics. Due to thematerial properties of organic (as in organic chemistry) materials,organic thin-film technology could enable printing and roll-to-roll mass-fabrication methods on flexible substrates. The low-cost and flexibilitywould make new applications possible, e.g. low-cost radio frequencyidentification (RFID) tags for item-level smart labeling.

This doctoral research focuses on organic ferroelectric field-effect transistors (FeFET) to realize a non-volatile memory component in such RFID tags.

Grzegorz Pyka

Department Department of Metallurgy and Materials Engineering (MTM)


Supervisor Prof. dr. ir. Martine Wevers

Co-supervisor Dr. ir. Jan Schrooten

Co-supervisor Prof. dr. ir. Frederik Maes


Introduction / ObjectiveThe request for porous structures with highly controlled properties, coming from different application areas within theindustrial and scientific market, forced researchers to find ways to optimise additive manufacturing (AM) productiontechniques in order to control the morphological and mechanical properties of the manufactured porous structures onboth meso- and microscale. This PhD study focused on the surface modification and the subsequent characterisation ofthe porous structures having customized morphological properties. Micro- and nano-CT in-situ loading tests were to becarried out to unravel of the mechanical behaviour and failure of those porous structures.

Research MethodologyTopology changes of the Ti6Al4V porous scaffolds produced by selective laser melting (SLM) were introduced by severalsurface treatments consisting of chemical etching followed by electrochemical polishing. High resolution micro-CT imageshave been used for roughness measurements of the complex porous structures subjected to the surface treatments.Finally, an automated non-rigid image registration of images acquired prior-to and after compressive loading was appliedto assess the strains within the mechanically loaded Ti6Al4V open porous structures.

Results & ConclusionsThis work resulted in a novel protocol for a controlled post-production surface treatment of the open porous structures thatallows to eliminate the surface irregularities in a robust manner(Fig. 1), but also to produce customized structures with desiredglobal morphological properties.

In this study for the first time, micro-CT has been applied toquantify the materials’ surface texture and a novel tool for 3Dsurface roughness measurements has been developed (Fig. 2).

The local strain analysis, performed using the non-rigidregistration of the micro-CT images, revealed larger strainconcentrations at the beam geometry imperfections of thecompressed Ti6Al4V open porous structures produced by SLM(Fig. 3).

Major publication

Pyka, G., Kerckhofs, G., Papantoniou P., Speirs M., Schrooten, J., Wevers, M. (2013). Surface roughness and morphologycustomization of additive manufactured open porous Ti6Al4V structures. Materials 6: 4737-4757.

Micro-CT based morphological and mechanical characterisation of open porous metallic materials

Figure 1. SEM micrographs of a typical strut of a Ti6Al4V open porous structure a) as-produced, b) after chemical etching and c)

after additional electrochemical polishing.

a bc

a b

Figure 2. a) A typical high resolution 2D μCT cross-sectional image of a single beam of a porous Ti6Al4V structure and b) a binarized section

of a) with the corresponding profile lines. Scale bars = 200 µm.

Figure 3. a) 3D images of a single beam of Ti6Al4V porous structure generated before compression and at the 50% of ultimate

compressive strain, at the UCS and at the sample failure and b) 2D maps of the volumetric strain generated at the 50% of ultimate

compressive strain and at the UCS.

Wouter Dekeyser



Supervisor Prof. dr. ir. Martine Baelmans

Co-supervisor Prof. dr. Detlev Reiter (FZ Jülich)

Funding FWO-Vlaanderen


Introduction / ObjectiveDivertors play a key role in the development of next-step nuclear fusion reactors. Responsible for power and particleexhaust, they have to be designed such that they can safely handle the large power loads. Specifically, their designneeds to prevent from exceeding limits imposed by the materials in order to avoid excessive material erosion and melting.At the same time, sufficient particle throughput — in particular Helium pumping capacity — has to be ensured. The designof present-day divertors is heavily assisted by numerical simulations of the plasma edge. However, due to the largenumber of design variables and complex edge plasma flows, this is computationally extremely demanding. Therefore, thiswork aims at developing an automated approach based on shape optimization.

Research MethodologyIn this thesis, an efficient adjoint approach to divertor target shape optimizationis developed. The divertor design problem is formulated as a mathematicaloptimization problem, in which the divertor target shape is the control variable,while the edge plasma model appears as a constraint. Using the velocitymethod, shape sensitivities are obtained which depend on boundary data only.Therefore, they can be evaluated at almost negligible cost. Furthermore, with aone-shot algorithm — which solves state, adjoint and design equations as onecoupled system — solutions to the entire design problem are found at anequivalent computational cost of only 4 to 10 edge plasma simulations.

Results & ConclusionsThe developed algorithms are applied to divertor target design for optimalpower load spreading. A number of test problems is considered in which theedge plasma is described with models of increasing complexity, ranging fromfluid descriptions to coupled fluid-kinetic models including radiation transport.An almost perfectly uniform target load isachieved for the test cases, significantlyreducing the peak loads, and optimallymaking use of the high-heat-flux plasma-facing components.

Major publicationW. Dekeyser, D. Reiter, M. Baelmans (2012). Optimal Shape Design for Divertors. International Journal of ComputationalScience and Engineering, in press.W, Dekeyser, D. Reiter, M. Baelmans (2012). Divertor Design through Shape Optimization. Contributions to PlasmaPhysics, 52 (5–6), 544–549.

Optimal Plasma Edge Configurations for Next-Step Fusion Reactors

Fig. 1: Velocity method for shape sensitivities

Fig. 2: Optimized divertor target geometry(left), with perfectly uniform load to thedivertor targets (right: outer target).

Niels Van Steenbergen

Department Civil Engineering


Supervisor Prof. dr. ir. Patrick Willems


Introduction / ObjectiveFloods are among the most destructive natural hazards. They affect thousands of people worldwide and cause damagefor millions of euros. It is important that local authorities and the people at risk can be warned timely. Flood forecastingsystems are being developed for that purpose. However, these systems are subject to significant uncertainties. The threeobjectives of this doctoral research were: (1) quantification of these uncertainties; (2) identification and testing of mostefficient uncertainty reduction options; (3) communication and visualization of the uncertainties.

Research Methodology Statistical analysis of flood forecast errors

Identification of most important uncertaintysources by resimulation of historical floodforecasts

Development of improvement actions

Calibration hydrological models

Data assimilation

Results & Conclusions A non-parametric statistical methodology has been developed and applied to

the historical forecasts of the operational river flood forecasting system of Flanders Hydraulics Research, to quantify the total uncertainty in flood forecast results for river water levels and discharges.

This total flood forecast uncertainty was decomposed into the contributions from the main sources of uncertainty. This provided insight and guidance on priorities for model improvement actions.

Given that improved rainfall-runoff model calibration was identified as a model improvement action with high priority, but the number of historical observations to evaluate the model under extreme flow or flood conditions is limited, a novel technique was developed to evaluate the rainfall-runoff model performance to extrapolate peak flows. The technique is based on the evaluation of peak flow changes under increasing rain storm volumes.

A practical tool was developed for enhancing flood preparedness, based on indirect soil moisture indicators and rainfall forecasts.

A technique was developed for probabilistic flood mapping.

Major publicationVan Steenbergen, N., Ronsyn, J., Willems, P. (2012). Non-parametric data-based approach for probabilistic flood forecasting in support of uncertainty communication. Environmental Modelling & Software, 33, 92-105.

Uncertainty quantification and reduction for river flood forecasting

Total uncertainty before improvement actions

Total uncertainty after improvement actions

Kameswaran Vengattaramane

Department Electrical Engineering (ESAT)


Supervisor Prof. dr. ir. Michiel Steyaert

Funding IMEC


logo fundingagency ifapplicable

Introduction / ObjectiveA sub-gate delay resolution Time-to-Digital Converter (TDC) with power-performance re-configurability for use in DigitalPLL

Research MethodologyThe work begins with a study of Digital PLL architecture andidentified the key performance metrics of the frequencysynthesizer based on a Digital PLL. An simulation platform isthen developed to accurately model the functional behavior andto predict the impact of TDC quantization on the output spectralquality. Sub-gate delay effective resolution is identified as themain requirement. A qualitative comparison of the existingtechniques is provided, highlighting the analog-intensive nature ofexistent techniques. In this work, the concept of SpatiallyOversampled TDC System is presented, where the resolutionenhancement over a gate delay is obtained by low-complexitydigital post-processing of several parallel TDC outputs. Theseparallel TDCs have slightly different resolutions and measure thesame input. Knowing the individual TDC resolutions, multiplecoarse time-interval estimates can be derived. These estimatesare then arithmetically processed to result in a final estimate withmuch better precision and accuracy than any individual TDCused in the parallel system. This system emulates a fineresolution TDC using only digital-domain techniques. Theachieved resolution enhancement depends on the number ofchannels (TDCs) used and hence resolution re-configurability isreadily incorporated.

Results & ConclusionsTwo test prototypes are realized on silicon using 90-nm CMOStechnology. In the first prototype, a system with 8 parallel currentstarved ring oscillator based TDCs is designed. The resolution ofTDCs is changed by setting the current drive through aprogrammable word. Measurements show a 3X improvementover a buffer delay [52-62]ps validating the system principle. Inthe second prototype, a fully synthesizable architecture with 8channels, described in Verilog is developed. This incorporates anonline background calibration scheme to track the individual TDCresolutions and is a fully integrated implementation with all thepost-processing done on-chip. The measured effective resolutionis [14-40]ps. These two prototypes demonstrate the viability of thefully digital alternative for resolution enhancement in TDCs.

Major publicationVengattaramane, K.; Borremans, J.; Steyaert, M.; Craninckx, J., "A standard cell based all-digital Time-to-Digital Converter with reconfigurable resolution and on-line background calibration," ESSCIRC (ESSCIRC), 2011 Proceedings of the , vol., no., pp.275,278, 12-16 Sept. 2011

Kameswaran Vengattaramane, Jonathan Borremans, Michiel Steyaert, and Jan Craninckx, Spatially oversampled TDC with digital resolution enhancement. Analog Integr. Circuits Signal Process. 70, 3 (March 2012), 311-322.

Reconfigurable All-Digital Time-to-Digital Converter for Digital PLL

Carlos Fuentes

Department Department of Materials Engineering


Supervisor Prof. dr. ir. Ignace Verpoest

Co-supervisor dr. ir. Aart Van Vuure

Funding SBA-KU Leuven


Introduction / ObjectiveUnderstanding the fibre-matrix interface in order to improve the mechanical performance is not an easy task. Although itis widely recognized that the properties of the interface are governed by physical and chemical compatibility between thefibre and the matrix, the interaction forces responsible for the work of adhesion and the strength of the interface aredifficult to estimate.

The main objective of this dissertation is to study the surface and the wetting behaviour of natural fibres, in order toestablish the feasibility of performing meaningful wetting measurements on them. The current methodology for measuringcontact angles on natural fibres is revised, in order to take into account equilibrium conditions during the measurement ofwetting. The work of adhesion (obtained by wetting analysis) is correlated with the practical adhesion (obtained bymicromechanical tests).

Research MethodologyAtomic force microscopy (AFM), X-ray Photoelectron Spectroscopy (XPS), Time-of-Flight Secondary Ion MassSpectrometry (ToF-SIMS), and optical profilometry techniques were used to study the fibres topography and chemistry. Anew approach, based on an autoclave treatment, to reduce the noise in performing contact angle measurements onrough natural fibre surfaces is presented.

The wetting dynamics of various test liquids on natural fibres is analysed by applying the molecular-kinetic theory ofwetting. Additionally, a novel way to measure equilibrium contact angles by using sound excitation is proposed, and Themechanical strength of the interfaces was assessed by single fibre pull-out tests.

Results & ConclusionsThe typical large fluctuation during wetting of naturalfibres is mainly a consequence of surface topographyirregularities (roughness and waviness).

A methodology for calculating the surface energycomponents of natural fibres and thermoplasticmatrices, and an acoustic vibration method to measureequilibrium contact angles on synthetic and naturalfibres were developed.

The influence of physical adhesion and roughness onthe interface strength was characterized by the localinterface shear strength and friction stress, and thenormal interface stress at the moment of crack initiation.Both interfacial parameters show a poor performancefor natural fibre composites.

SEM cross sectionimage of a technicalbamboo fibre, showingthat the fibre breaksthrough the elementaryfibre- lignin interface.

Major publicationFuentes Rojas, C.A., Tran, L., Dupont-Gillain, C., Vanderlinden, W., De Feyter, S., Van Vuure, A., Verpoest, I. (2011).Wetting behaviour and surface properties of technical bamboo fibres. Colloids and Surfaces A, Physicochemical andEngineering Aspects, 380 (1-3), 89-99. http://dx.doi.org/10.1016/j.colsurfa.2011.02.032

Interfacial adhesion in natural and synthetic fibre composites: a physical-chemical-mechanical approach

The surface chemicalcomposition of naturalfibres is assed by XPSanalysis.

Introduction / ObjectiveWe aim to develop and implement both nonlinear and linear distributedoptimization methods that are applicable, but not restricted to the optimalcontrol of distributed systems. Such systems are typically large scale, thus thewell-established centralized solution strategies may be computationally overlyexpensive or impossible and the application of alternative control algorithmsbecomes necessary. Moreover, it is often desired that the information on thecoupled subsystems inner dynamics is kept local, making the deployment of acentralized optimal control scheme impossible in practice. In such a case,optimization approaches based on some limited exchange of informationremain the only possible option.

Attila Kozma

Department Electrical Engineering


Supervisor Prof. dr. Moritz Diehl

Funding EU FP7


Research MethodologyWe consider nonlinear optimal control problems with distributed structure, forwhich we design an efficient distributed shooting method.

We investigate linear optimal control problems with quadratic cost functions, i.e.quadratic programs (QP). An overview of dual decomposition based approaches iscarried out, followed by the development of a novel dual decomposition scheme.The proposed method employs second-order derivatives.

Results & Conclusions

Distributed Multiple Shooting: Results in up to 19 times fasterintegration time when compared to the conventional approaches.

Dual Newton-CG method for Quadratic Programming:Experimental results suggest that it requires up to 152 times lesslocal optimization steps than classical first-order schemes.

Major publicationA. Kozma, J. Frasch, and M. Diehl. A Distributed Method for Convex Quadratic Programming Problems Arising in OptimalControl of Distributed Systems, Proceedings of the 52nd Conference on Decision and Control (CDC), 2013

Distributed optimization methods for large scale optimal control

Tommaso TamarozziDepartment Department of Mechanics, Division PMA



Co-supervisor Prof. dr. ir. Paul Sas

FundingMarie Curie ITN VECOM FP7-213543,KUL Industrial Research Fund (IOF)IWT O&O on HEV-NVH project


Introduction and ObjectivesCurrent numerical strategies do not allow to efficiently capture the local behavior of mechanical systems with variabletopology (e.g. bearings and gears). The main goal of this thesis can be summarized as follows: Modify existingcomponent-level model order reduction (MOR) techniques to develop novel MOR strategies allowing to solve flexiblemultibody systems in which small moving interfaces are present.

Results & ConclusionsTwo methods named Static Modes Switching (SMS) andStatic Modes Sliding (SMS2) have beed developed andtested on different test cases ranging from academical toindustrially relevant The SMS method proved to be efficient and reliable for simulation of flexible multibody systems with variable topology

Numerical strategies have been developed to deal with the discontinuous character of the method

In order to improve the usability of the SMS technique, the novel parametric MOR method name SMS2 is introduced

The method shows similar benefits as compared to the SMS approach with improved usability and removing the discontinuities introduced by SMS

Future work will focus on extensions of SMS2 to non linear contact problems

Major publicationTamarozzi, T., Heirman, G.H.K., Desmet, W. (2014), An on-line time dependent parametric model order reductionscheme with focus on dynamic stress recovery, Computer Methods in Applied Mechanics and Engineering. 268,336–358, Elsevier

Efficient Numerical Simulation Strategies for Flexible Multibody Systems with Variable Topology

Research MethodologySeveral numerical tools has been used and extended. Mainly the fields of (parametric) model order reduction, flexiblemultibody and contact mechanics have been explored:

Component level MOR has been extended to develop 2 novel methodologies to deal with sliding and contactproblems.

The techniques are readily implementable in flexible multibody software

Gear contact mechanics has been deeply studied to apply the proposed methods

Kavin Karunratanakul


PhD defence 31 January 2014

Supervisor Prof. dr. ir. Hans Van Oosterwyck

FundingNational Metal and Materials Technology Center, Thailand


Introduction / ObjectiveThe treatment of large segmental defects in long bones remains a clinical challenge. Bone tissue engineering (TE) is amultidisciplinary research field that has the potential to revolutionise the treatment of non-healing bone defects. However,TE products for large defects do not exist yet in the clinic. One of the reasons for this lack of successful TE products is aninsufficient understanding of the role of the mechanical environment to which TE products are exposed. Finite element(FE) modelling is a powerful tool to quantify the mechanical environment inside porous scaffolds used for bonereconstruction. Therefore, the general goal of this study is to develop realistic finite element models that can be used toquantify the mechanical environment inside porous scaffolds.

Research MethodologyA 3D FE model of an external fixator is developed and validated by comparing the calculated results to the experimentaltesting results. Several aspects of the model are studied, in particular, the contact settings between fixator parts andbetween the fixator and the bone. Then, the validated FE model is used to quantify mechanical stimuli inside the scaffoldunder in vivo loading conditions. Finally, predicted values of mechanoregulation stimuli are compared to in vivo results ontissue ingrowth in porous scaffolds.

Results & ConclusionsBy carefully adjusting the contact settings between each part ofthe model, a good correspondence between predicted axialstiffness and measured stiffness can be obtained. The FE modelcan quantify mechanical stimuli inside the scaffold underrealistic in vivo loading condition. This study demonstrates theimportance of in vivo loading conditions as well as mechanicalproperties of scaffold and fixator on mechanical stimuli insidethe scaffold.

Top: Experimental set-up for axial loading ofreconstructed tibia (A). A close-up view of the scaffold-bone interface is shown in a separate box. Therepresentative finite element model shows elementsize for each part of the model (B).Left: histological section (a) and mechanoregulationstimuli in porous scaffold (b)Major publication

K. Karunratanakul, G. Kerckhofs, J. Lammens et al. (2013). Validation of a finite element model of a unilateral externalfixator in a rabbit tibia defect model. Medical Engineering & Physics, 35 (7), 1037-1043.

Quantification of the mechanical environment for tissue engineering in load-bearing bone

(a) (b)

Illustration: UNSW

Loic Tous

Department Elektrotechniek (ESAT)


Supervisor Prof. dr. ir. Robert Mertens

Co-supervisor Prof. dr. ir. Jozef Poortmans

Funding KU Leuven


Introduction / ObjectiveAs most of a photovoltaic (PV) system cost is area related, the highest impact on cost can be achieved by increasing theefficiency of the solar cells in the PV modules while reducing manufacturing costs. This is followed in this thesis byreplacing conventional silver (Ag) screen printed front side contacts by nickel/copper (Ni/Cu) plated contacts since theypresent several advantageous properties (reduced optical shading, high conductivity, reduced contact resistance tooptimum emitters) including being much cheaper than Ag.

Research MethodologyDespite the potential advantages of Ni/Cu plated contacts their commercialization has so far been limited due to a numberof challenges (increased process complexity, lack of suitable low-cost production techniques/tools, doubts over cost-advantage and long-term reliability) as well as recent progress made with screen printing of Ag. Thus, the approachfollowed in this thesis is to reduce process complexity by introducing suitable low-cost production techniques/tools in anincremental manner since many processes are inter-linked and need to be co-optimized. Reliability and cost-of-ownershipissues are addressed in parallel.

Results & ConclusionsEfficiency ~20.5% using inline plating/sintering tools: high-efficiency process that is robust and repeatable

Ni/Cu/Ag contacts ~4.4c€/cell cheaper than Ag contacts: return-on-investment <3years

First evaluation of environmental module testing andthermal ageing: long-term (>25 years) reliability appears feasible

Major publicationL. Tous, M. Aleman, R. Russell, E. Cornagliotti, P. Choulat, A. Uruena, S. Singh, J. John, F. Duerinckx, J. Poortmans,and R. Mertens. Evaluation of advanced p-PERL and n-PERT large area silicon solar cells with 20.5% energy conversionefficiencies, accepted in Progress in Photovoltaics: Research and Applications

Nickel/Copper Plated Contacts as an Alternative to Silver Screen Printing for the Front Side Metallization of Industrial high efficiency Silicon Solar Cells

60 cells PV module Silicon solar cell schematicPrinted Ag contact (top view) Plated Ni/Cu contact (top view) Plated Ni/Cu contact (cross-section)

0

10

20

30

40

50

19,2 19,4 19,6 19,8 20,0 20,2 20,4 20,6 20,8

nu

mb

er o

f ce

lls

solar cell efficiency [%]

p-type PERC Ag p-type PERC Ni/Cu

+0.5%abs. gain

Shoaib Iqbal


PhD defence 30th January 2014

Supervisor Prof. dr. ir. Farid Al-Bender

Co-supervisors Prof. dr. ir. Wim Desmet, Dr. Ir. Bert Pluymers

Funding EMECW Lot 15


Introduction / ObjectiveIn recent years, there has been a growing concern regarding the increasing prices and requirements of energy around theglobe. One of the largest and cost effective ways to tackle this issue is by improving the energy efficiency ofcomponents/systems. This leads to a reduced total cost of ownership of machines, lower fossil fuel dependency andreduced greenhouse gas emissions. In this thesis two critical high-dynamic components, namely bearings and wetclutches are subjected to detailed energy analysis through modeling and experimental characterization. Bearings arecommonly used component in machines/systems, whereas the wet clutches find application in most of the transmissiondevices.

Research MethodologyThe approach adopted in the present thesis in order totackle the issue of energy analysis of bearings and wetclutches that could enable the system level energyefficiency improvement is shown schematically in adjacentfigure.

Results & ConclusionsThe main contributions are divided into five parts as follows: Improvement in the state-of-the-art bearing power loss models.

Development of an extended Reset-Integrator friction for wet clutch applications.

Investigation of degradation and driving energy level effects on wet clutches performance and energetic behavior.

Experimental characterization of drag torque in disengaged wet clutches

Development of a drag torque model for disengaged wet clutches

Major publication S. Iqbal, F. Al-Bender, J. Croes, B. Pluymers, and W. Desmet, ‘‘Frictional power loss in solid-grease-lubricated needle

roller bearing,’’ Lubrication Science, vol. 25, no. 5, 2012. S. Iqbal, F. Al-Bender, B. Pluymers, and W. Desmet, ‘‘Model for predicting drag torque in open multi-disks wet

clutches,’’ ASME Journal of Fluids Engineering, art.no. FE-13-1241, 2013.

Energy Analysis of High-Dynamic components: Model Development and Experimental Evaluation for Needle Roller Bearings and Wet Clutches

02

46

8

0

200

400

6000

0.2

0.4

0.6

0.8

Load [kN]Speed [rpm]

Bea

ring

pow

er lo

ss [

kW]

02

46

8

0

200

400

6000

0.2

0.4

0.6

0.8

Load [kN]Speed [rpm]

Bea

ring

pow

er lo

ss [

kW]

0 500 1000 1500 2000 25000

2

4

6

8

10

12

Clutch Speed (rpm)

Dra

g T

orq

ue (

Nm

)

ExperimentProposed ModelTraditional ModelKato ModelYuan Model

Research Methodolgy

Bearing

Existing Model Improved Model

Wet Clutch

Film Shape

Model Results Drag Torque Curve

Marlene Brito MartínezDepartment Department of Chemical Engineering


SupervisorsProf. dr. ir. Bart Van der BruggenProf. dr. ir. Patricia L. AlconeroProf. dr. Zenaida Rodríguez Negrín

FundingCenter of Bioactive Chemicals.UniversidadCentral “Marta Abreu” de Las Villas. Cuba


Introduction / ObjectiveThe Center of Bioactive Chemicals Cuba is a research center focused on products with high biological activity. In thiswork, the recovery of the active pharmaceutical ingredient 1-(5-bromo-fur-2-il)-2-bromo-2- nitroethane, (referred G-1), isperformed by means of the use of nanofiltration and membrane crystallization from the waste streams produced duringthe purification process of G1 and Vitrofural®. Thus, experimental work was to be performed for specific applications inorder to evaluate the technical viability of this technology.Research MethodologyThe potential of nanofiltration(NF) for the separation of G-1 is evaluated by using a cross-flow configuration. Bothcommercial and self-made polyethersulfone NF membranes are studied in synthetic mixtures of G-1 in ethanol and theretention of G-1 and flux of ethanol through the membranes are determined in order to evaluate the applicability of thestudied membranes at production scale.

In addition, during the manufacture of APIs, residual streams with variable content of APIs can be produced and theirfurther recovery is as attractive as challenging due to the complex mixture of compounds that can be present. Membranecrystallization can be used in this case to crystallize selectively the target compound and recover it from the mixture as anintegrated process solution.

Results & ConclusionsThe application of NF for the recovery of G-1 from ethanolsolutions is feasible; the best results were found with themembranes PES/NMP25%, PES/NMP27% and thecommercial membrane NF 270 in terms of percent retentionand permeance.The application of membrane crystallization for G-1 recoverycan be considered as a novel technique with high potential torecover pharmaceutical compounds from waste streams.

Major publicationBrito, M., Jullok, N., Rodriguez, Z., Van der Bruggen, B., Luis, P., (2014). Membrane crystallization for the recovery of apharmaceutical compound from waste streams. Journal Chemical Engineering Research and Design, 92, 264–272..

Application of nanofiltration and membrane cristallization for the recovery of the active pharmaceutical ingredient G-1

1,21x10-5

(0,025%)3,22x10-9

(96,97%)4.53x10-11

(3%)1.36 x10-12

1,8

1,85

1,9

1,95

0 1 2 3 4 5 6 7 8

Tra

nsm

embr

ane

flux

(ml/m

2 .m

in)

Concentration G-1 (mg/L )CBQ

The mass transfer coefficients inside the fiber(kf), shell side(kP),the membrane (kmp) and the total coefficient (ktotal).

Maenhoudt Marijn

Department Electrical Engineering


Supervisor Prof. dr. ir. Geert Deconinck

Co-supervisor Prof. dr. ir. Ronnie Belmans.

FundingEU FP7 OPTIMATE (FP7-239456-optimate)


Introduction / ObjectiveTo understand the evolution of the electricity market, dynamic market modelling tools can be applied.Using such models, stakeholders gain insights on the sensitivity of the market design against potentialdisturbances or market imperfections, and take necessary actions to pro-actively address them. How thestate of an interconnected electrical system evolves after clearing the day-ahead market as organisedunder the European power exchange model, subject to strategic gaming behaviour has been studied.

Research MethodologyFirstly, a novel profit risk hedging offering strategy is presented. It submits the coordinated dispatchschedule of thermal, hydropower and renewable power plants to the market operator. The generatorpursues a total profit-maximising objective by simultaneously exercising physical and economicwithholding while explicitly taking into account underlying technical constraints, plant economics andstep-wise, price-responsive demand.

Secondly, the offering strategy is integrated in a newly designed dynamic electricity market model. Usingmulti-agent systems, each generator updates its perception of the market environment, includingcompetitor behaviour and the inherent complexities of the electricity market, by evaluating theperformance of historic decisions on its profit.

Results & ConclusionsThe model is commercially available on www.optimate-platform.eu.It is unique compared to competing models becauseAny generation portfolio can be fully modelledInherent market complexities are explicitly taken into account,(e.g. limited transmission capacity, price-responsive demand…)The offering strategy is proven to significantly outperformalternative ones assumed to support policy decisionsResults are robust and easily tractableImportant market design parameters can be altered to assesstheir impact on the efficient operation of the electricity market,subject to suboptimal behaviour of market participantsThe model is currently being used by Transmission SystemOperators, Regulators and Energy Administrations in Europe

Strategic Behaviour in Power Wholesale Electricity Markets

Major publicationMaenhoudt, M.; Deconinck, G., "Strategic Offering to MaximizeDay-Ahead Profit by Hedging Against an Infeasible Market ClearingResult," Power Systems, IEEE Transactions ondoi: 10.1109/TPWRS.2013.22860052718-2818.

Rodrigo B. Moraes



Supervisor Prof. dr. ir. Marc Moonen

Co-supervisor Dr. ir. Paschalis Tsiaflakis

Funding


IntroductionThe ability to easily exchange and access data has transformed the way we work, study, inform and entertain ourselves. In particular, the Internet has had an effect on people’s lives in the past two decades that is profound. Profound as this effect may be, people seem not to grow tired of it, and the broadband access market grows at a fast pace. Of the technologies currently available for broadband access, digital subscriber lines (DSL) is by far the market leader. One of the major problems with this technology is severe levels of multi-user interference, commonly known as crosstalk.

Results & Conclusions Proposed algorithm for optimal spectrum coordination is up to 60 times faster than the benchmark algorithm.

Low complexity proposed algorithm for sub-optimal spectrum coordination is up to 15 times faster than the benchmark algorithm.

Combined signal and spectrum coordination is seen to provide gains of several gigabits per second in realistic DSL scenarios (see figure on the right).

Algorithm for combined signal and spectrum coordination is very robust and flexible, and can be used in several scenarios with different types of coordination on the transmitter and receiver sides or with per-user of per-transceiver power constraints or with asynchronous transmission.

A typical DSL network, where lines depart the centraloffice (CO) of the network operator to reach businessesand houses.

Signal and Spectrum Coordination for Next Generation DSL Networks

Research MethodologyIn this thesis, we develop signal and spectrum coordination techniques that aim at avoiding, minimizing or even profiting from crosstalk. Among other things, we develop:

A system model that takes into account multi-user and multitone transmission. We also consider multiple input, multiple output (MIMO) transmission and possible asynchronism between the different users.

Competitive algorithms using spectrum and signal coordination so as to maximize the data rate of the users.

Major publicationR.B. Moraes, P. Tsiaflakis, J. Maes and M. Moonen, “DMT MIMO IC Rate Maximization in DSL With Combined Signal and Spectrum Coordination,” IEEE Transactions on Signal Processing, vol. 61, no. 7, pp. 1756-1769, April 2013

Pradip Mainali

Department Electrical Engineering - ESAT


Supervisor Prof. dr. ir. Rudy Lauwereins

Co-supervisorsProf. dr. ir. Luc Van GoolProf. dr. ir. Gauthier Lafruit

Funding IMEC


Introduction / ObjectiveHyperspectral Imaging (HSI) combines spectroscopy with imaging, enabling the identification of the chemical signature ofscene materials and objects without physical contact. Hyperspectral data is collected in a 3D hyperspectral imagecube/data structure with two spatial (horizontal) and one spectral (vertical) dimension (cfr. Fig. 1). The process of stackingspectral images acquired from a 2D hyperspectral camera to form a 3D hyperspectral image cube (cfr. Fig. 2 and 3),generally suffers from misalignment/misregistration, preventing the accurate reading of the scene’s chemical-spectralsignatures. To achieve a perfect hyperspectral cube, images have to be aligned/registered with sub-pixel accuracy.

Research MethodologyTo meet the above research objective, this thesis proposes a novel image feature-based spectral image registrationalgorithm (cfr. Fig. 4) to achieve immunity against the severe hyperspectral image artifacts such as blur, noise andspectral reflectance variations. The main contributions of this research work are:

o proposed two novel feature detection algorithms using the cosine modulated Gaussian and the tenth order Gaussianderivative filters.

o proposed a low complexity implementation approach with a fixed-length approximation filter, providing constant-time,low-cost image filtering implementations.

o uniquely combines the proposed advanced feature detection algorithms with RANdom SAmple Consensus’s(RANSAC) to robustly estimate the registration model parameters.

Results & ConclusionsThe proposed feature detection algorithms show a two times higher immunity against the image artifacts as compared tostate-of-the-art methods. The proposed registration algorithm reduces the registration error from hundreds of pixels inthe raw hyperspectral image cube down to half-pixel registration inaccuracy, without the need of costly calibration,deblurring and/or denoising techniques in hardware or software. This eventually paves the way to a low-cost and reliablehardware/software co-designed hyperspectral imaging system.

Major publications1. P. Mainali, G. Lafruit, Q. Yang, Geleen B., L. V. Gool, and R. Lauwereins, “SIFER: Scale-Invariant Feature Detectionwith Error Resilience”, International Journal of Computer Vision, Springer, 2013, vol. 21, pp. 435–445.2. P. Mainali and et. al., “Feature Detection in Numeric Data”, US patent(US 2013/0308860) and European patent (EP13168131).

Robust Registration in Integrated Hyperspectral Imaging

Fig. 1: Hyperspectral data Fig. 2: Hyperspectral sensor Fig. 3: Hyperspectral imagingsystem: scene is scanned by the sensor

Fig. 4: Hyperspectral imageregistration in software

Noël Deferm

Department Electrical Engineering - ESAT


Supervisor Prof. dr. ir. ing. Patrick Reynaert

Funding IWT-Vlaanderen


Introduction / ObjectiveThis work focuses on the development of circuit and system design techniques for millimeter wave wirelesscommunication systems above 90GHz and fabricated in nanometer scale CMOS technologies. Thanks to the CMOStechnology scaling, the speed of the transistors has increased towards a maximum frequency of oscillation above300GHz for the latest technology nodes. So CMOS has become a millimeter wave technology, but with the greatadvantage of high integration capability. Although the speed of the CMOS transistors has increased, the target operationfrequency of the circuits in this work (above 90GHz) is still close to the technology’s maximum frequency of oscillation. Socoping with a low power gain is one of the major design challenges in CMOS. Also, the transistors tend to show potentialunstable behavior, which makes multistage design a real challenge. Furthermore, the metal stack of CMOS technologiesis not optimized for the design of high frequency passives, which results in a decreased performance of traditionalmatching circuits. At system level, problems like the implementation of the millimeter-wave chip interface and bandwidthand linearity requirements of the ADC’s, DAC’s and up- and downconversion circuits emerge.

Research MethodologyTo solve the above mentioned problems, six different chips are designed in which circuit and system level solutions areproposed and implemented. Three different multistage differential W-band amplifiers are designed in which capacitiveneutralization is extensively applied to improve the gain and stability properties of the transistors. The adoption oftransformers and slow-wave transmission lines in the impedance matching networks resulted in small chip footprints,while improving the performance of the amplifiers even more. At system level, direct digital modulation schemes wereapplied in three different F-band and D-band transmitters which has lead to new digital modulation system topologies.These direct carrier modulator topologies have resulted in a relaxation of the requirements of the upconverter and made itpossible to omit high speed DAC’s. Also the design complexity of these systems could be considerably reduced.

Results & ConclusionsThe applied design techniques have resulted in state-of-the-artperformance. Measurements of the amplifiers show high gainsranging from 11dB up to 18dB and high output powers up to8dBm above 90GHz. Measurements of the fully integratedtransmitters show high speed wireless data communicationcapabilities. The F-band ASK transmitter and the 120GHz PSKtransmitter respectively support data rates up to 5Gb/s and10Gb/s. The fully integrated 120GHz Star-QAM transmitter, whichcombines both modulation techniques of the previouslymentioned transmitters is capable of supporting a 10Gb/s wirelessdata link. The integration of a frequency generator, modulator,power amplifier, baseband circuits and bondwire antenna haslead to a fully integrated solution which solves the problem ofmillimeter wave interfacing and hereby closes the gap between alaboratory chip solution and a real-life application.

Chip photograph of the 120GHz 10Gb/s Star-QAMtransmitter in 45nm CMOS.

Major publicationN. Deferm and P. Reynaert, “Design, implementation and measurement of a 120 GHz 10 Gb/s phase-modulating transmitter in 65 nm LP CMOS,” Springer Analog Integrated Circuits and Signal Processing, vol. 75, no. 1, pp. 1-19, 2013.

CMOS Front Ends for Millimeter Wave Wireless Communication Systems

Christophe Van der Kelen




Supervisor Prof. dr. ir. Peter Göransson (KTH)

FundingMRTN-CT-2006-035559IWT-Vlaanderen


Introduction / ObjectivePoroelastic materials for acoustic applications are oftenanisotropic due to the production processes involved, andanelastic due to the constituent material used. An accuratecharacterisation of these materials must therefore take intoaccount the possible anisotropy and anelasticity. Theobjective of this thesis is to enable the characterisation ofa constitutive material model which is as general aspossible, and which includes the inherent materialanisotropy and anelasticity. For this purpose, a set ofadvanced characterisation techniques needs to beengineered to characterise the anisotropic flow resistivitytensor and the anisotropic dynamic Hooke’s tensor.

Research MethodologyThe developed characterisation techniques are based onan inverse estimation procedure, and includes bothexperiments and numerical predictions. The property tocharacterise is isolated in a specially designed set-up suchthat it can be modelled by physics solely involving thisproperty. The obtained experimental and numerical datathen serve as the input to an optimisation, in which thedifference between both is minimised. The obtained modelparameters are then considered to give a best possiblerepresentation of the material parameters.

Results & ConclusionsThe relevance of including anisotropy in the material model is demonstrated by the characterisation of the flow resistivityand the elastic material properties of a melamine foam, clearly showing that both properties are non-isotropic. Moreover,it is found that the production process has a significant influence on the material properties, and that the anelasticity of thematerial needs to be included for a correct representation of the material behaviour.

Procedure of inverse estimation.

Major publication C. Van der Kelen, P. Göransson (2013). Identification of the full anisotropic flow resistivity tensor for multiple glass wool and melamine foam samples. Journal of The Acoustical Society of America , 134 (6), 4659 – 4669.

J. Cuenca C. Van der Kelen, P. Göransson (2014). A general methodology for inverse estimation of the elastic and anelastic properties of anisotropic open-cell porous materials with application to a melamine foam. Journal of Applied Physics , 115, 084904.

Vibro-acoustic modelling of anisotropic poroelastic materials –characterisation of the anisotropic properties

Joris Maervoet

Department Department of Computer Science


Supervisor Prof. dr. Patrick De Causmaecker

Co-supervisor Prof. dr. ir. Greet Vanden Berghe

Funding IWT-Baekelandmandaat


Introduction / ObjectiveThe present thesis describes various applications based on concrete questions originating from GIS practice. Thefollowing three themes are treated throughout this work.1.Quality maintenance in geographic information systems (GIS). It is pursued through the identification of patterns,describing relational regularities, and corresponding outliers indicating probable inconsistencies in the data. Thisapproach is different from classical approaches to quality analysis in GIS, traditionally relying on techniques prospectingfor statistical and positional deviation in geo-data. The quality of this type of data is of high influence on the servicequality offered by navigation systems, online route planners and apps providing location-based services.2.Efficiency enhancement of shortest path (cost) approximation components for spatial networks. Shortest pathalgorithms are the basic component of point-to-point route planners and navigation systems and operate on spatialnetworks, extracted from geographical data. An approximation component returns a path of which the lengthapproximate the length of the shortest path between two locations in a network. This work focusses on suchcomponents in resource-constrained environments where exact approaches cannot be applied.3.The fast and automated discovery of attractive closed paths. This process operates in spatial networks where theedges have both a length and an attractiveness score. An algorithm of low computational impact generating heuristicsolutions is presented, and gives rise to a web component for tour suggestion for outdoor activities.

Research MethodologyThe first theme implies the modeling and design of a tool for data engineers, tracing relation regularities and outliers inlarge and dynamic geographical databases. It is validated during an extensive sanity check. Theme 2 involves thedesign of approximation components based on auxiliary structures in spatial weighted graphs. The presentedtechniques are supported by large-scale experimental analyses of the approximation accuracy, the calculation timerequired to produce an approximation and the structures' compactness. The last theme implies the introduction of thetour suggestion problem for outdoor activities and a fast algorithm generating heuristic solutions to this problem.

Results & ConclusionsThe underlying concepts and approaches introduced in the present thesis are applicable in more general domains. Foursoftware components resulting from this research have been adopted by industry.

Major publicationMaervoet, J., Vens, C., Vanden Berghe, G., Blockeel, H., De Causmaecker, P. (2012). Outlier detection in relationaldata: a case study in geographical information systems. Expert Systems with Applications, 39 (5), 4718-4728.

Quality Maintenance in Geographical Data and Services for Spatial Networks

POI

POI

POI

POI

Thanh Hieu LEDepartment Computer science


Supervisor Prof. dr. ir. Marc Van Barel

Co-supervisor Prof. dr. Raf Vandebril

FundingKU Leuven Center of Excellence: Optimization in Engineering (OPTEC)


Introduction / Objective

The purpose of this research is to study of the Pythagoras number of polynomials, together with theircorresponding low-rank representations, from theoretical and practical points of view. The results are used tosolve several optimization problems that can be reformulated over non-negative polynomials, such as low-pass filter design problems.

Results & Conclusions Gave and proved the lower and upper bounds on the Pythagorasnumber of sosm-polynomials; gave a new proof for the upper bound onthe Pythagoras number of sos-polynomials which is known in theliterature

Proposed algorithms for approximating the low-rank representationsof sos(m)-polynomials and for computing their Pythagoras number

Proposed an algorithm for decomposing a non-negative polynomialas a sum of squares of rational functions

Proposed an optimization model that can be used to solve several(IIR, FIR) filter design problems

Figure: Specification for discrete-time low-pass filter and itsmathematical formulation.

Major publication T. H. Le, M. Van Barel. A convex optimization method to solve a filter design problem. J. Comput. Appl. Math,255:183192, 2014.

T. H. Le, L. Sorber, and M. Van Barel. The Pythagoras number of real sum of squares polynomials and sum of squaremagnitudes of polynomials. Calcolo, 50(4): 283303, 2013.

M. Ferranti, T. H. Le, R. Vandebril. A comparison between the complex symmetric based and classical computation ofthe singular value decomposition of normal matrices. (Accepted to Numer. Algor.)

Low-Rank Representations for Sum of Squares Polynomials

Research Methodology Recast IIR or FIR filter problems as optimizationproblems over sos-polynomials.

With the knowledge on the Pythagoras number ofsos-polynomials, the resulting problems can be solvedby combining a bisection rule on the design parameterand a conic LP program solving the feasibility problems.

Implemented in MATLAB with CVX toolbox.

Figure: FIR filter of degree 30

Anke Van Campen



Supervisor Prof. dr. ir. Joris De Schutter, Prof. dr. I. Jonkers

Co-supervisor dr. ir. Friedl De Groote

Funding FWO


Introduction / ObjectiveBiomechanical analysis for the evaluation of human motion is a valuable tool. The potential applications are very diverse:treatments of mobility impairments in order to enhance the individual's quality of life, optimization of sports performance,and design of prosthetics. These analyses rely on musculo-skeletal models from which the muscle-tendon model is asubmodel. The better these models reflect the patient, the more reliable the analyses are. However, in general genericmodels are used. Therefore, this thesis aimed at subject-specifically describing the force generating capacity of themuscles and tendons by identification of the muscle-tendon parameters based on experimental force data.

Research Methodology• Sensitivity analysis: which parameters are crucial for the accuracy of the simulations?

• Design experimental set-up: enhance regular dynamometer to allow full 3D inverse dynamic analysis.

• Experimental design: find minimal set of experiments for maximum amount of information on parameters.

• Parameter identification: non-linear optimization, validation in simulation environment.

• Proof of concept: experimental data, different measurement conditions.

• Focus on knee joint.

Results & Conclusions• Hierarchy in muscle-tendon parameters: reduction

parameter space.

• 3D experimental joint moment registration and proper modelling of joint: more accurate data.

• A priori intelligence results in definition of feasible set and hot start for non-linear estimation algorithm.

• Estimation algorithm outperformes benchmark bymore than one order of magnitude when noisy dataare involved.

• Subject-specific muscle-tendon parameterestimation: added value for motion simulation.However, development of other modelling aspectsmight show its full potential.

Major publicationA. Van Campen, G. Pipeleers, F. De Groote, I. Jonkers, J. De Schutter. A new method for estimating subject-specificmuscle-tendon parameters of the knee joint actuators: a simulation study. Int j numer method biomed eng, accepted forpublication (2014).

Identification of subject-specific parameters of a Hill-type muscle-tendon model for simulations of human motion

Joyce Van den Broeck



Supervisor Prof. dr. ir. J. Vander Sloten

Co-supervisor Prof. dr. J. Bellemans

Funding IWT – Baekeland scheme


Introduction / ObjectiveA recent trend in orthopaedic surgery is the use of preoperative plans based on 3D imaging data of the patient. Onemethod to transfer this plan to the actual surgery is the use of patient-specific instrumentation. To allow an accuratetransfer of the plan, three important criteria are identified: stability of the fit of the guide on the anatomy, uniqueness of thefit position and correct use of the guide by the surgeon. This work focuses on the requirement of stability.

Research MethodologyA mathematical model is introduced to predict the fit stability during the virtual design process, by providing thedesigner with a stability score. This model originates from robotics theory and analyses the resistance of the contactsurface towards a potential translational or rotational movement. Validation experiments have been performed by usingboth fit experiments and quantitative measurements relating applied forces and resulting displacements. The validationexperiments confirmed that the stability score resulting from the model provides a reliable indication of the hapticfeeling of stability when fitting the guide on the respective bone model. As both the preoperative plan and guide designare based on 3D models of the patient’s anatomy reconstructed from medical imaging data, one chapter of this thesisdiscusses the accuracy of image segmentation and 3D model reconstruction.

Results & ConclusionsThe results presented in this work suggest that the currently used process ofsurgical guide design could benefit from the introduction of mathematicalalgorithms to reduce design uncertainties such as the stability of fit, therebyimproving the quality of the instrumentation sent to the surgeon for patienttreatment. The methods currently developed provide a good starting point foroptimizing this design process, creating an opportunity for additionalimprovement in automatic design of the optimal guide for each patient.

Major publicationVan den Broeck, J, Wirix-Speetjens, R, Vander Sloten, J (2013). Preoperative analysis of the stability of fit of a patient-specific surgical guide. Computer Methods in Biomechanics and Biomedical Engineering, April 29, doi: 10.1080/10255842.2013.774383.

Patient-specific instrumentation in orthopaedic surgeryA computational model for design optimization

Example of a patient-specific surgical instrument designed to fit on the distal femur, to guide the drilling of four pins.

In addition, the knowledge obtained during the development of the stability modelwas used to create tools to optimize the guide design process. Methods weredeveloped to indicate how an existing guide design can be improved with respectto the predicted stability, either by extending the current guide with functionalfeatures or by enlarging the contact surface. In addition, a method to create theoptimal contact surface for a specific anatomy was investigated, maximizing bothstability and accuracy.

AcknowledgementThis PhD research was funded by both the Baekeland scheme of the Agency for Innovation by Science and Technology(IWT) and Materialise NV.

Documents

PhD Defence Topics