PLATFORM - press.utp.edu.mypress.utp.edu.my/wp-content/uploads/2018/12/PlatformV6N2_page1.pdf · Enhancement Of Heat Transfer Of A Liquid Refrigerant In Transition Flow In The Annulus

P L A T F O R M

Volume 6 Number 2 Jul - Dec 2008

VOLUME SIX NUMBER TWO JULY - DECEMBER 2008 ISSN 1511-6794

PL

AT

FO

RM

VO

LUM

ES

IXN

UM

BE

RT

WO

JUL

Y-

DE

CE

MB

ER

20

08

Separation Of Nitrogen From Natural Gas By Nano- Porous Membrane Using Capillary CondensationFarooq Ahmad, Hilmi Mukhtar, Zakaria Man, Binay. K. Dutta

2

Recent Developments In Autonomous Underwater Vehicle (AUV) Control SystemsKamarudin Shehabuddeen, Fakhruldin Mohd Hashim

6

Enhancement Of Heat Transfer Of A Liquid Refrigerant In Transition Flow In The Annulus Of A Double-Tube Condenser R. Tiruselvam, Chin Wai Meng, Vijay R Raghavan

13

Fenton And Photo-Fenton Oxidation Of Diisopropanolamine Abdul Aziz Omar, Putri Nadzrul Faizura Megat Khamaruddin, Raihan Mahirah Ramli

21

Synthesis Of Well-De� ned Iron Nanoparticles On A Spherical Model Support Noor Asmawati Mohd Zabidi, P. Moodley, P. C. Thüne, J. W. Niemantsverdriet

27

Performance And Emission Comparison Of A Direct-Injection (DI) Internal Combustion Engine Using Hydrogen And Compressed Natural Gas As FuelsAbdul Rashid Abdul Aziz, M. Adlan A., M. Faisal A. Mutalib

31

The E� ect Of Droplets On Buoyancy In Very Rich Iso-Octane-Air Flames Shaharin Anwar Sulaiman, Malcolm Lawes

38

Anaerobic Co-Digestion Of Kitchen Waste And Sewage Sludge For Producing Biogas Amirhossein Malakahmad, Noor Ezlin Ahmad Basri, Sharom Md Zain

47

On-Line At-Risk Behaviour Analysis And Improvement System (E-ARBAIS) Azmi Mohd Shari� , Tan Sew Keng

52

Bayesian Inversion Of Proof Pile Test: Monte Carlo Simulation ApproachIndra Sati Hamonangan Harahap, Wong Chun Wah

65

Element Optimisation Techniques In Multiple DB Bridge Projects Narayanan Sambu Potty, C. T. Ramanathan

77

A Simulation Study On Dynamics And Control Of A Refrigerated Gas PlantNooryusmiza Yuso� , M. Ramasamy, Suzana Yusup

85

An Interactive Approach To Curve Framing Abas Md Said 91Student Industrial Internship Web Portal Aliza Sarlan, Wan Fatimah Wan Ahmad, Dismas Bismo 96Hand Gesture Recognition: Sign To Voice System (S2V) Foong Oi Mean, Tan Jung Low, Satrio Wibowo 105Parallelization Of Prime Number Generation Using Message Passing Interface Izzatdin A Aziz, Nazleeni Haron, Low Tan Jung, Wan Rahaya Wan Dagang

111

Evaluation Of Lossless Image Compression For Ultrasound Images Boshara M. Arshin, P. A. Venkatachalam, Ahmad Fadzil Mohd Hani

116

Learning Style Inventory System: A Study To Improve Learning Programming Subject Saipudnizam Mahamad, Syarifah Bahiyah Rahayu Syed Mansor, Hasiah Mohamed

122

Performance Measurement – A Balanced Score Card ApproachP. D. D. Dominic, M. Punniyamoorthy, Savita K Sugathan, Noreen I. A.

129

A Conceptual Framework For Teaching Technical Writing Using 3D Virtual Reality Technology Shahrina Md Nordin, Suziah Sulaiman, Dayang Rohaya Awang Rambli, Wan Fatimah Wan Ahmad, Ahmad Kamil Mahmood

137

Multi-Scale Color Image Enhancement Using Contourlet Transform Melkamu H. Asmare, Vijanth Sagayan Asirvadam, Lila Iznita

145

Automated Personality Inventory System Wan Fatimah Wan Ahmad, Aliza Sarlan, Mohd Azizie Sidek 152A Fuzzy Neural Based Data Classi� cation System Yong Suet Peng, Luong Trung Tuan 158

Research In Education: Taking Subjective Based Research Seriously Sumathi Renganathan, Satirenjit Kaur

166

Mission-Oriented Research: CARBON DIOXIDE MANAGEMENT

Mission-Oriented Research: DEEPW TER TECHNOLOGY

Mission-Oriented Research: GREEN TECHNOLOGY

Mission-Oriented Research: PETROCHEMICAL CATALYSIS TECHNOLOGY

Technology Platform: FUEL COMBUSTION

Technology Platform SYSTEM OPTIMISATION

Technology Platform: APPLICATION OF INTELLIGENT IT SYSTEM

Other Areas

A

:

1 VOLUME Six NUMBER twO jULy - dEcEMBER 2008 PLATFORM

I S S N 1 5 1 1 - 6 7 9 4

contents

copyright © 2008Universiti teknologi PEtRONAS

PLATFORMJuly-December 2008

Advisor: datuk dr. Zainal Abidin Haji Kasim

PLATFORM Editorial

Editor-in-Chief:Prof. ir. dr. Ahmad Fadzil Mohd. Hani

Co-Editors:Assoc. Prof. dr. isa Mohd tan

Assoc. Prof. dr. Victor Macam jr.

Assoc. Prof. dr. Patthi Hussin

dr. Baharum Baharuddin

dr. Nor Hisham Hamid

dr. Shahrina Mohd. Nordin

Subarna Sivapalan

Sub-Editor:Haslina Noor Hasni

UTP Publication Committee

Chairman: dr. Puteri Sri Melor

Members: Prof. ir. dr. Ahmad Fadzil Mohamad Hani

Assoc. Prof. dr. Madzlan Napiah

Assoc. Prof. dr. M. Azmi Bustam

dr. Nidal Kamel

dr. ismail M. Saaid

dr. M. Fadzil Hassan

dr. Rohani Salleh

Rahmat iskandar Khairul Shazi Shaarani

Shamsina Shaharun

Anas M. yusof

Haslina Noor Hasni

Roslina Nordin Ali

Secretary:Mohd. Zairee Shah Mohd. Shah

[email protected]

Address:PLATFORM Editor-in-chief

Universiti teknologi PEtRONAS

Bandar Seri iskandar, 31750 tronoh

Perak darul Ridzuan, Malaysia

http://www.utp.edu.my

[email protected]@petronas.com.my

Telephone +(60)5 368 8239

Facsimile +(60)5 365 4088


Separation Of Nitrogen From Natural Gas By Nano-Porous Membrane Using Capillary CondensationFarooq Ahmad, Hilmi Mukhtar, Zakaria Man, Binay. K. Dutta

2

Mission-Oriented Research: DEEPWATER TECHNOLOGY

Recent Developments In Autonomous Underwater Vehicle (AUV) Control Systems Kamarudin Shehabuddeen, Fakhruldin Mohd Hashim

6


Enhancement Of Heat Transfer Of A Liquid Refrigerant In Transition Flow In The Annulus Of A Double-Tube Condenser R. Tiruselvam, Chin Wai Meng, Vijay R Raghavan

13


Fenton And Photo-Fenton Oxidation Of Diisopropanolamine Abdul Aziz Omar, Putri Nadzrul Faizura Megat Khamaruddin, Raihan Mahirah Ramli

21

Synthesis Of Well-Defined Iron Nanoparticles On A Spherical Model Support Noor Asmawati Mohd Zabidi, P. Moodley, P. C. Thüne, J. W. Niemantsverdriet

27


Performance And Emission Comparison Of A Direct-Injection (DI) Internal Combustion Engine Using Hydrogen And Compressed Natural Gas As Fuels Abdul Rashid Abdul Aziz, M. Adlan A., M. Faisal A. Mutalib

31

The Effect Of Droplets On Buoyancy In Very Rich Iso-Octane-Air Flames Shaharin Anwar Sulaiman, Malcolm Lawes

38

Technology Platform: SYSTEM OPTIMISATION

Anaerobic Co-Digestion Of Kitchen Waste And Sewage Sludge For Producing Biogas Amirhossein Malakahmad, Noor Ezlin Ahmad Basri, Sharom Md Zain

47

On-Line At-Risk Behaviour Analysis And Improvement System (E-ARBAIS) Azmi Mohd Shariff, Tan Sew Keng

52

Bayesian Inversion Of Proof Pile Test: Monte Carlo Simulation Approach Indra Sati Hamonangan Harahap, Wong Chun Wah

65

Element Optimisation Techniques In Multiple DB Bridge Projects Narayanan Sambu Potty, C. T. Ramanathan

77

A Simulation Study On Dynamics And Control Of A Refrigerated Gas Plant Nooryusmiza Yusoff, M. Ramasamy, Suzana Yusup

85

Technology Platform: APPLICATION OF INTELLIGENT IT SYSTEM

An Interactive Approach To Curve Framing Abas Md Said

91

Student Industrial Internship Web Portal Aliza Sarlan, Wan Fatimah Wan Ahmad, Dismas Bismo

96

Hand Gesture Recognition: Sign To Voice System (S2V) Foong Oi Mean, Tan Jung Low, Satrio Wibowo

105

Parallelization Of Prime Number Generation Using Message Passing Interface Izzatdin A Aziz, Nazleeni Haron, Low Tan Jung, Wan Rahaya Wan Dagang

111

Evaluation Of Lossless Image Compression For Ultrasound Images Boshara M. Arshin, P. A. Venkatachalam, Ahmad Fadzil Mohd Hani

116

Learning Style Inventory System: A Study To Improve Learning Programming Subject Saipudnizam Mahamad, Syarifah Bahiyah Rahayu Syed Mansor, Hasiah Mohamed

122

Performance Measurement – A Balanced Score Card Approach P. D. D. Dominic, M. Punniyamoorthy, Savita K Sugathan, Noreen I. A.

129

A Conceptual Framework For Teaching Technical Writing Using 3D Virtual Reality Technology Shahrina Md Nordin, Suziah Sulaiman, Dayang Rohaya Awang Rambli, Wan Fatimah Wan Ahmad, Ahmad Kamil Mahmood

137

Multi-Scale Color Image Enhancement Using Contourlet Transform Melkamu H. Asmare, Vijanth Sagayan Asirvadam, Lila Iznita

145

Automated Personality Inventory System Wan Fatimah Wan Ahmad, Aliza Sarlan, Mohd Azizie Sidek

152

A Fuzzy Neural Based Data Classification System Yong Suet Peng, Luong Trung Tuan

158

Other Areas

Research In Education: Taking Subjective Based Research Seriously Sumathi Renganathan, Satirenjit Kaur

166

2


PLATFORM VOLUME Six NUMBER twO jULy - dEcEMBER 2008

This paper was presented at the 15th Regional Symposium on Chemical Engineering

In Conjunction With 22nd Symposium of Malaysian Chemical Engineers, Kuala Lumpur, 2 - 3 December 2008

INTRODUCTION

Raw natural gas contains many impurities such as, acid gases (carbon dioxide and hydrogen sulfide), lower hydrocarbons (propane and butane) and nitrogen. Studies performed by the Gas research institute reveal that 14% of known reserves in the United States are sub-quality due to high nitrogen content [Huggman et al (1993)]. The conventional cryogenic route is not favoured as it requires a lot of energy. Gas permeation through a nano-porous membrane occurs primarily by Knudsen diffusion although the interaction between the permeating molecules and the pore wall may cause other mechanisms to prevail such as surface diffusion [Jaguste et al. (1995); Uholhorn et al. (1998); Wijaman, S et al. (1995)]. Multi-layer adsorption occurs and is followed by capillary condensation. In an earlier paper [Ahmad et al. (2005)] we have reported an analysis of separation of lower hydrocarbon from natural gas by capillary condensation. It was established that

practically an acceptable flux and selectivity could be achieved by this technique.

RESULTS AND DISCUSSIONS

The technique presented by Lee and Hwang are widely used for the description of condensable gases through small pores of the membranes [Lee and Hwang (1986)]. They investigated the transport of condensable vapours through micro-porous membrane and predict six flow regimes depending on the pressure distribution and the thickness of the adsorbed layer. Here we consider the case of complete filling of a pore, with condensate in both upstream and downstream. For the condensation to occur in pore at both upstream and downstream face of membrane the condensation pressure (Pcon) should be lesser than both upstream pressure (feed pressure P1 ) and downstream pressure (Permeate pressure P2) across the membrane at certain feed temperature greater

SEPARATION OF NITROgEN FROM NATURAL gAS BY NANO-POROUS MEMBRANE

USINg CAPILLARY CONDENSATION

Farooq Ahmad*, Hilmi Mukhtar, Zakaria Man and Binay. K. Dutta,Universiti Teknologi Petronas, 31750 Tronoh, Perak Darul Ridzuan, Malaysia

*[email protected]

ABSTRACT

In the present work we have explored the potential of a nano-porous membrane to perform the separation job of binary mixture of methane/nitrogen by capillary condensation. In case of methane/nitrogen permeation rate up to 700 gmol/m2.s.bar has been achieved at a temperature lower than the critical temperature of the permeating species and higher than the critical temperature of the non-permeating species. The results have the potential to be used for further refining and optimising the process conditions to exploit this strategy for large scale removal of nitrogen from methane at a low cost.

Keywords: capillary condensation, nano-porous membrane, natural gas, nitrogen, permeability

3


VOLUME Six NUMBER twO jULy - dEcEMBER 2008 PLATFORM

than the critical temperature of the methane and lesser than the critical temperature of carbon dioxide. This situation is depicted in Figure 1. For this case, the entire pore is filled with the capillary condensate and apparent permeability is given by the following equation [Lee and Hwang (1986)]

]ln)(

ln)(

[)(

TR

0

22

2

0

12

221 P

P

r

tr

P

P

r

tr

PPM

KP d

t

−−×−

=µ

ρ

(1)

where, Kd = ρπr4/8M, t1 and t2 are the thicknesses of the adsorbed layer at upstream and downstream face pore respectively. The thickness of the absorbed layer at upstream and downstream is assumed as 10 times the molecular diameter of methane. Since methane is the more permeating component, the selectivity of methane over nitrogen is given by

2

4

2

4

N

CH

N

CH

y

y

x

x=α (2)

where x is the mol fraction in the pore, and y is the mol fraction in the bulk

The permeability of condensed methane of methane/nitrogen binary mixture has been calculated using equation (1) at different temperatures for various pore

sizes. Since the selected pore diameters are small, condensation occurs at temperature well above the normal condensation temperature at the prevailing pressure. This makes the pore flow separation more attractive than the cryogenic separation process. A wide range of pore sizes and temperatures were selected for computation of permeability and separation factors. The computed results are presented below. Figure 2 gives the permeability of methane with temperature for different pore sizes and pore lengths equal to ten times the molecular diameter of methane. With increasing temperature permeation rate is increased, because at a higher temperature more pressure is required to cause capillary condensation inside the pore. Figure 2 shows that even at moderate pressure and temperature slightly below °C, an appreciable permeability can be achieved. The permeation rate is reduced at low pore size, but at low pore size the condensation pressure is reduced and we require lesser pressure at the feed side to cause condensation inside the pore. Based on solubility of nitrogen in condensed methane using Peng-Robinson equation of state, the separation factor binary mixtures methane/nitrogen has been calculated by using equation (2) and is shown in Figure 3. From Figure 3 it can be seen that separation factor decreases with increasing temperature, because as stated earlier, at a higher temperature, more feed pressure is required to cause capillary condensation and thus solubility of nitrogen in liquid methane increases and thus separation factor decreases with increasing temperatures. From Figure 3 it is concluded that a reasonable separation factor

Figure 1. Schematic of condensation flow through a nano pore.

100 110 120 130 140 150 160 170 180 1900

100

200

300

400

500

600

700

800

900

T (K)

P (

gm

ol /

s.m

2.b

ar)

5 nm

10 nm

20 nm30 nm

40 nm

50 nm

Figure 2. Effect of pore size on permeability of methane at different temperatures.

Figure 3. Separation factor of N2/CH4 binary mixtures at various temperatures.

100 110 120 130 140 150 160 170 1800

100

200

300

400

500

T(K)

Separa

tion facto

r

4% N2- 96% CH4

8% N2- 92% CH4

12% N2- 88% CH4

16% N2- 84% CH4

4



can be achieved using a nano-porous membrane by capillary condensation which justifies that the nano-porous membrane using capillary condensation has a potential to be used. The model compares reasonably well with the experimental results which justify the validity of the model. The comparison is shown in Figure 4. Figure 4 shows that the separation factor obtained theoretically is less than the experimentally calculated separation factor. The reason for this is that the separation factor is calculated on the basis of equilibrium solubility of hydrogen in the condensed phase of methanol. In reality such a system operating at steady state will be away from equilibrium and a higher separation factor will be achieved. The permeability increases with temperature although the separation factor decreases. A balance should be struck between the two to decide upon the optimum operating temperature. For tortuous pores the increased path length will cause the permeability to decrease. The comparison between experimental and predicted values of capillary condensation pressure by the Kelvin equation given by Sperry et al. is been shown in Figure 5. The computed results and experimental data compare well at least up to a temperature of 420 K (147°C). This establishes the validity of the model up to a reasonably high temperature for gas separation applications.

CONCLUSIONS

The Kelvin equation was used to calculate the condition for capillary condensation for predicting the separation of nitrogen from methane. The separation factor of methane/nitrogen was analysed based on the principle that methane will condense preferentially. High separation factor of up to 439 was achieved, suggesting that the removal of nitrogen from natural gas by nano-porous membrane is promising. Permeation rates were also calculated which are in agreement for other condensed gas system. Also, condensation occurred at a temperature much lower than the normal saturation temperature.

REFERENCES

R. H. Hugman, E. H. Vidas, P. S. Springer, (1993), ”Chemical [1] Composition of Discovered and Undiscovered Natural Gas in the United States”, (Update, GRI-93/0456)

J. G. Wijmans, R. W. Baker, (1995), “The Solution Diffusion [2] Model, A review” J. Membr. Sci., (107) 1-21

R. J. R. Uhlhorn, K. Keizer and A. J. Burggraaf, (1998) “Gas [3] Separation Mechanism in Micro-porous modified γ-alumina Membrane”, J. Membr. Sci., (39) 285-300

D. N. Jaguste and S. K. Bhatia, (1995) ”Combined Surface [4] and Viscous Flow of Condensable Vapor in Porous Media”, Chemical Engineering Science, (50) 167-182

Figure 5. Comparison of Experimental data for methanol hydrogen separation by [Speery et al. (1991)] with the data predicted using Kelvin equation [Ahmad et al. (2008)]

Figure 4. Comparison of Experimental data for methanol hydrogen separation by [Sperry et al. (1991)] with Model predicted data [Ahmad et al. (2008)].

360 380 400 420 440 460 4800

100

200

300

400

500

600

700

T (K)

Se

pa

ra

tio

n fa

cto

r

Experimental data

Model Predicted data

360 380 400 420 440 460 4800

5

10

15

20

25

30

35

40

T(K)

Pco

n (

Ba

r)

Data Predicted by Kelvin equation

Experimental data

5



F. Ahmad, H. Mukhtar, Z. Man, Binay. K. Dutta, (2007) [5] “Separation of lower hydrocarbon from Natural gas through Nano-porous Membrane using Capillary Condensation”, Chem. Eng. Techno. 30 (9) 1266-1273

K. H. Lee and Hwang, (1986).The Transport of Condensable [6] Vapors through a Micro-porous Vycor Glass Membrane J. Colloid Interface Sci., 110 (2).544-554

F. Ahmad, H. Mukhtar, Z. Man, Binay. K. Dutta, (2008) [7] “A theoretical analysis for non-chemical separation of hydrogen sulfide from Natural gas through Nano-porous Membrane using Capillary Condensation”, Chem. Eng.Processing, (47) 2203-2208

Hilmi Mukhtar is currently the Director of Undergraduate Studies, Universiti Teknologi PETRONAS (UTP). Before joining UTP, he served as a faculty member of Universiti Sains Malaysia (USM) for about 6 years. He was a former Deputy Dean of the School of Chemical Engineering at USM. He obtained his BEng in Chemical Engineering from the University of

Swansea, Wales, United Kingdom in 1990. He completed his MSc in 1991 and later on his PhD in 1995 from the same university. His doctoral research focused on the “Characterisation and Performance of Nanofiltration Membrane”.

He has deep research interests in the area of natural gas purification using membrane processes. Currently, he is leading a research project under the Separation & Utilisation of Carbon Dioxide research group. In this project, the removal of impurities from natural gas, in particular, carbon dioxide, is the key focus of the study. In addition, he has research interests in environmental issues particularly wastewater treatment and carbon trading.

6



This paper was presented at the 5th PetroMin Deepwater, Subsea & Underwater Technology, Conference and Exhibition 2007, Kuala Lumpur

29 - 30 October 2007

INTRODUCTION

Oceans cover about two-thirds of the whole earth surface and the living and non-living resources in the oceans undoubtedly take an important role in human life. The deep oceans are hazardous particularly due to the high pressure environment. However, offshore oil industry is now forced to deal with increasing depths of offshore oil well. In recent years, oil well

depths (surface of the sea to sea bed) have increased far beyond the limit of a human diver. This has resulted in increasing deployment of unmanned underwater vehicle (UUV) for operations and maintenance of deepwater oil facilities.

UUV covers both remotely operated vehicles (ROVs) and autonomous underwater vehicle (AUVs). ROVs have tethered umbilical cable to enable remote

RECENT DEVELOPMENTS IN AUTONOMOUS UNDERWATER VEHICLE (AUV) CONTROL SYSTEMS

Kamarudin Shehabuddeen* and Fakhruldin Mohd HashimUniversiti Teknologi PETRONAS, 31750 Tronoh, Perak Darul Ridzuan, Malaysia

*[email protected]

ABSTRACT

Autonomous Underwater Vehicles (AUVs) are tethered free and unmanned. AUVs are powered by onboard energy sources such as fuel cells and batteries. They are also equipped with devices such as electronic compass, GPS, sonar sensor, laser ranger, pressure sensor, inclination sensor, roll sensor and controlled by onboard computers to execute complex preprogrammed missions. In oil and gas sector, two separate categories of AUVs are classified for application in Exploration and Production (E&P). A survey class vehicle is for inspection of offshore structures and data acquisition, and a work class vehicle for underwater manipulation and installation. AUV dynamics involve six degrees of freedom (DOF). Most of the AUV application requires very stringent positioning precision. However, AUVs’ dynamics are highly nonlinear and the hydrodynamic coefficients of vehicles are difficult to be accurately estimated due to the variables associated with different operating conditions. Experimental class AUVs provide an excellent platform for development and testing of various new control methodology and algorithms to be implemented in developing advanced AUVs. Control performance requirements of an AUV are most likely to be achieved from control concepts based on nonlinear theory. Recently developed advanced control methodologies focused on improving the capability of tracking predetermined reference position and trajectories. Due to increasing depth of operation expected from future AUVs and onboard limited power supply, the area of future research in AUV control is most likely to expand into incorporating intelligent control to propulsion system in order to improve power consumption efficiency. This paper presents a survey on some of the experimental AUVs and the past, recent and future directions of the AUV control methodologies and technologies. Keywords: Autonomous Underwater Vehicles (AUV), Experimental AUVs, Past, Recent and Future Control Methodologies

7



operator to control the operation of the vehicle. Tether influences the dynamics of vehicle, greatly reducing maneuverability.

AUVs are tethered free, unmanned, powered by onboard energy sources such as fuel cells and batteries. AUVs are also equipped with devices such as electronic compass, GPS, sonar sensor, laser ranger, pressure sensor, inclination sensor, roll sensor and controlled by onboard computers to execute complex preprogrammed missions.

With the technology advancement in battery, fuel cells, material, computer, artificial intelligence and communication, AUVs become more popular in exploring oceanic resources. In the oil and gas sectors, two separate classes of AUVs are available for application in Exploration and Production (E&P). A survey class is for inspection of offshore structures and data acquisition, and a work class vehicle for underwater manipulation required for installation of underwater facilities and equipments.

AUVs performing survey, manipulation or inspection tasks needed to be controlled in six degrees of freedom [1]. Although the control problem is kinematically similar to the control of a rigid body in a six-dimensional space, the effects of hydrodynamic forces and uncertain hydrodynamic coefficients give rise to greater challenges in developing a control system for an AUV. In order for an AUV to achieve autonomy under the ocean environment, the control system must have the adaptability and robustness to the non-linearity and time-variance of AUV dynamics, unpredictable environmental uncertainties such as sea current fluctuation and reaction force from the collision with sea water creatures, modeling difficulty of hydrodynamic forces.

The well proven linear controller based on linear theory may fail to satisfy the control performance requirements of an AUV. Therefore, interest of many AUV researchers centred around non-linear control schemes based on non-linear theory.

Several recently developed advanced control methodologies focused on improving the capability of tracking, given reference position and attitude trajectories. The objective of the paper is to address the recent and future directions of AUV control methodologies.

MARKET DRIVERS

The offshore oil industry is currently pursuing offshore oil production with well depths (surface of the sea to sea bed) that previously would have been considered technically unfeasible or uneconomical [2]. A study [3], Figure 1, shows the maximum well depth versus past years. In 1949, the offshore industry was producing in about 5 m of water depth and it took 20 years to reach about 100 m. However, in recent years, the maximum well depth increased dramatically. It indicates that the maximum well depth will continue to increase in the future. The maximum acceptable depth limit for a human diver is about 300 m. At deeper than these depths, ocean floor mapping, inspection and repair operations of facilities must be executed by either UUVs or inhabited submersibles.

Two classes of UUVs are normally used in the E&P sector. A survey class is for ocean floor mappings, inspection of offshore structures and data acquisition, and a work class vehicle for manipulation, installation and repair operations. Recently, in the UUV market, ROVs has been gradually replaced by AUVs.

Figure 1. Offshore Oil Fields – Maximum well depth versus past years. Data and figure source [3]

We

ll D

ep

th (

m)

0

500

1000

1500

2000

2500

1949

1954

1959

1964

1969

1974

1979

1984

1989

1994

1999

2004

8



FUTURE OF ROVS AND AUVS

Due to increasing offshore oil well depth, research is currently undertaken to enhance the capabilities of ROVs so that they can become autonomous, and hence the emergence of AUVs. AUVs are free from constraints of an umbilical cable and are fully autonomous underwater robots designed to carry out specific pre-programmed tasks such as ocean floor mappings, manipulation, installation and repair operations of underwater facilities.

AUV CONTROL

AUVs performing survey, manipulation or inspection tasks needed to be controlled in six degrees of freedom. In order for an AUV to achieve autonomy under the ocean environment, the control system must have the adaptability and robustness of non-linearity and time-variance of AUV dynamics, unpredictable environmental uncertainties such as sea current fluctuation and reaction force from the collision with sea water creatures. Experimental AUVs provide an excellent platform for the development and testing of various new control methodologies and algorithms to be implemented in developing advanced AUVs.

Highlights on Some of the Experimental AUVs

In 1995, with the intention to contribute to AUV development, the Autonomous Systems Laboratory (ASL) of the University Of Hawaii has designed and developed the Omni-directional Intelligent Navigator (ODIN-I). In 1995, ODIN-I was refurbished and ODIN-II was born. ODIN-I and ODIN-II have made precious contribution in the development and testing of various new control methodology and algorithms.

In 2003, ODIN-III was developed. It has the same external configuration and major sensors as ODIN-II, which is a closed-framed spherical shaped vehicle with eight refurbished thruster assemblies and a one DOF manipulator [4]. ODIN-III represents Experimental robotic class AUV. Eight thrusters provide instantaneous, omni-directional (six DOF) capabilities. The on-board computer system used in ODIN-III is a

PC104+ (Pentium computer CPU 300 MHZ, 128 MB RAM). The vehicle can be controlled by either an on-board computer in the autonomous mode or by operator command using ground station computer with or without orientation control via tether.

The main challenges in developing high performance experimental AUVs are the calibration of sensors and design of appropriate candidate control system, formulation of control algorithms and implementation of appropriate actuation amplitudes based on inputs from sensors.

A miniature cylindrical shaped AUV called REMUS (Remote Environmental Monitoring Units) is designed to conduct underwater scientific experiments and oceanographic surveys in shallow water [5]. REMUS is equipped with physical, biological and optical sensors. A standard REMUS is 19 cm in diameter and 160 cm long. REMUS represents experimental survey class AUV. The vehicle is controlled by PC104, IBM compatible computer. REMUS has three motors for propulsion, yaw control and pitch control. 68HC11 micro-controller is assigned to control the motors. The communication between the micro-controller and the CPU is achieved through an RS232 interface.

An even smaller experimental class AUV was developed to demonstrate that the basic components of an AUV can be packaged in a 3-inch diameter tube [6]. The hull was made from standard 3-inch Schedule 40 PVC pipe. The nose cone was machined from a solid piece of PVC. The tail was made from high density foam bonded to a piece of aluminum tubing. The AUV control system is governed by Rabbit 3000 8-bit microcontroller (RCM3200 module). The processor runs at 44.2 MHz and is programmed in C language.

AUV Navigation and Sensor Deployments

Inertial navigation system (INS) is primarily used in AUVs. INS is usually used to measure the angular displacements of yaw, pitch and roll. A pressure sensor is used to measure depth. For an experimental AUV, sonar sensors are used to measure the horizontal translation displacements. For the commercial scale

9



AUVs, the navigation suite usually consists of an INS, a velocity sensor and a Global Positioning System (GPS) receiver. The GPS is used to initialise the navigation system, and for position determination when the AUV surfaces at intervals.

Some of the Past and Recent AUV Control Methodologies

Sliding Mode Control (SMC), Neural Net Control (NNC), Proportional Integral Derivative (PID) and Adaptive Controls were among the major control methodologies explored for the position and trajectory control of an AUV.

Sliding Mode Control (SMC)

SMC is a non-linear feedback control scheme. Yoerger 1985 [7] 1991 [8] developed SMC methodology for the trajectory control of AUV. The method can control non-linear dynamics directly, without need for linearisation. This attribute is crucial for an AUV, which exhibits highly non-linear dynamics. The control methodology only require a single design for the whole operating range of the vehicle. Therefore, it is easier to design and implement than a stack of liberalised controllers.

The method was examined in simulation using a planar model of the Experimental Autonomous Vehicle (EAVE) developed by the University of New Hampshire (UNH).

Yoerger in 1991 [9] developed an adaptive SMC for the control of experimental underwater vehicle.

Song in 2000 [10] proposed a Sliding Mode Fuzzy Controller (SMFC). Song determined the parameters of the SMFC using the method based on Pontryagin’s maximum principle. Sliding mode control is robust to the external disturbance and system modeling error. SMFC takes advantage of the robustness property of the SMC and interpolating property of the fuzzy in such a way that the non-linear switching curve could be estimated and the robustness could be sustained. The effectiveness of the control philosophy was

tested on Ocean Explorer series AUVs developed by Florida Atlantic University. The problem with SMC is the chattering.

Neural Net Control (NNC)

NNC is an adaptive control scheme. NNC poses parallel information processing features of human brain with highly interconnected processing elements. The main attractive features of neural networks include self-learning and distributed memory capabilities [11].

J. Yuh in 1990 [12] proposed a multilayered neural network for the AUV control, in which the error-back propagation method is used. The development of this scheme was motivated by [13] which exhibit that the teaching error signal, the discrepancy between the actual output signal and teaching signal can be approximated by the output error of the control system.

Tamaki URA in 1990 [14] proposed a Self Organizing Neural Net Controller System (SONCS) for AUVs. SONCS had a controller called forward model and an evaluation mechanism, which were linked with initiation and modification tools. The dynamics of the vehicle was represented by a forward model. The difference between the actual motion of the vehicle and pre-programmed mission is calculated by an evaluation mechanism. The fundamental concept of SONCS uses the backward propagated signals to adjust the controller. Backward propagated signals were obtained by the evaluation of the controlled vehicle’s motion.

K. Ishii in 1995 [15] proposed an on-line adaptation method “Imaginary Training” to improve the time consuming adaptation process of the original SONCS. In this method, SONCS tunes the controller network through an on-line process in parallel with the actual operation.

J. S. Wang in 2001 [16] proposed the Neuro-Fuzzy control systems. Wang investigated the strengths and weaknesses of the rule formulation algorithms using the static adaptation and dynamic adaptation

10



methods based on clustering techniques to create the internal structure for the generic types of fuzzy models.

Proportional Integral Derivative (PID)

PID is for control over steady state and transient errors. PID has been widely implemented in process industries. It is also used as a benchmark against which any other control scheme is compared [17].

B. Jalving in 1994 [18] proposed three separate PID technique-based controllers for steering, diving, and speed control. The roll dynamic was neglected. Jalving defended it by designing the vertical distance between the centre of gravity and centre of buoyancy sufficiently long enough to suppress the moment from the propeller. The concept was tested on the Norwegian Defense Research Establishment (NDRE) AUV. The performance of the flight control system was reported to be satisfactory during extensive sea testing.

Adaptive Controls

J. Yuh in 2001 [19] proposed a non-regressor based adaptive controller. The adaptive law approximated and revised the parameters defined by the unknown bounds of the system parameter matrices and then tunes the controller gains based on performance of

the system instead of the knowledge of the dynamic model. The controller was tested on ODIN in the pool of the University of Hawaii.

S. Zhao in 2004 [20] proposed an adaptive plus disturbance observer (DOB) controller. Zhao used a non-regressor based adaptive controller as an outer-loop controller and a DOB as an inner-loop compensator. Zhao carried out experimental work on the proposed adaptive DOB controller using ODIN. The experimental work involved determining the tracking errors in associated predetermined trajectories. The performance of the proposed adaptive plus DOB controller was compared with other controllers such as the PID controller, PID plus DOB controller and the adaptive controller. The proposed adaptive DOB controller was reported to be effective in compensating the errors arising from external disturbances.

ENERgY STORAgE, POWER CONSUMPTION AND EFFICIENCY

Limited onboard power supply gives rise to the principal design challenge of an AUV system design. Limited onboard energy storage primarily restrictsthe range of a vehicle. Key design parameters are specific energy (energy storage per unit mass) and energy density (energy per unit volume). Fairly new battery technologies such as lithium ion and lithium polymer

Figure 2. AUV Range as a function of hotel load and speed. Calculations were made for a low drag 2.2m long, 0.6m diameter vehicle with high efficiency propulsion and batteries providing 3.3 kW.h of energy. Data and figure source: [21]

Hotel = 10 W Hotel = 40 W

Hotel = 160 W

0.1 1 10

10

100

1000

AUV speed (ms-1)

Range (km)

11



have higher specific energy and energy density than the conventional lead acid batteries and are now available for AUV applications.

Two types of power loads are typically identified by the AUV designers. One is propulsion load (propulsion and control surface actuator loads) and the other is hotel load (on-board computers, depth sensor and navigation sensors).

Propulsion load typically constitutes a large portion of the power required to operate an AUV. The amount of energy required to move the vehicle through the water is a function of both the drag of the vehicle and efficiency of the propulsion system [21]. Overall loss of efficiency of the propulsion system is contributed by electric motor efficiency losses, hydrodynamic inefficiency of the propeller, shaft seal frictional losses and viscous losses.

Reduction in hotel load is obviously beneficial, and is aided by continuing advances in low power consumption computers and electronic components. As the hotel power consumption decreases, the vehicle speed for best range also decreases. This was shown by a study on an existing AUV [21], in Figure 2. Currently, this research has found only one source of reference on the relationship between hotel load, range and speed.

CONCLUSION

Recently developed advanced control methodologies focused on improving the capability of tracking pre-determined reference positions and trajectories. Control performance requirements of an AUV have been achieved from control concepts based on non-linear adaptive theory rather than the linear theory based on fixed parameters. Due to the non-linear (time variant) hydrodynamic coefficients and unforeseen disturbances that the AUV control has to deal with, future research work on AUV position and trajectory control is likely to investigate the effectiveness of newly developed non-linear DOB based control concepts.

Due to the increasing oil well depth, future control methodologies of AUVs will most likely involve broadening the scope of control from the sole position and trajectory control method to that of incorporating intelligence system for power efficient orientated mission planning, and will also involve intelligent control based on actuation of control surface and thrusters to maximise efficiency in consumption of the limited onboard power supply.

Due to increasing depth of operations expected from future AUVs and the varying water pressures and densities with depth, a key area for future research and development in AUV control is likely to investigate the possibility of incorporating innovative power train technologies (between the electric motor and propeller) with intelligent control in order that the limited onboard power supply is consumed with maximum efficiency. Variable propeller pitch with intelligent control is also likely to optimise power consumption.

Due to the deep ocean environment where a human diver could not reach, future control methodologies of work class AUVs are also likely to focus on autonomous coordination based control of cooperating manipulators or humanoid end effectors between multiple AUVs performing underwater equipment installation or repair tasks. While the adaptive control concepts are better suited for position control of AUVs, the learning capability of the NNC system may be considered for coordination of manipulators. I f the newly developed coordination based control methodologies were to be represented in a virtual reality environment, it would be easier to evaluate responsiveness and behavior of cooperating manipulators.

REFERENCES

Gianluca Antonelli, Stefano Chiaverini, Nilajan Sarkar, [1] and Michael West, “Adaptive Control of an Autonomous Underwater Vehicle: Experimental Results on ODIN”, IEEE Transaction on Control Systems Technology, Vol. 9, No. 5, pp. 756–765, September 2001

Loius L. Whitcomb, “Underwater Robotics: Out of the [2] Research Laboratory and Into the Field”, International Conference on Robotics and Automation, IEEE 2000

12



D. Harbinson and J. Westwood, “Deepwater Oil & Gas – An [3] Overview Of The World Market”, Deep Ocean Technology Conference, New Orleans, 1998

H. T. Choi, A. Hanai, S. K. Choi, and J. Yuh, “Development of [4] an Underwater Robot, ODIN-III”, Proceedings of the 2003 IEEE/RSJ International Conference on Intelligent Robots and Systems, pp. 836–841, Las Vegas, Nevada, October 2003

Mike Purcell, Chris von Alt, Ben Allen, Tom Austin, Ned [5] Forrester, Rob Goldsborough and Roger Stokey, “Nee Capablities of the REMUS Autonomous Under Vehicle”, IEEE 2000

Aditya S. Gadre, Jared J. Mach, Daniel J. Stilwell, Carl E. Eick, [6] “Design of a Prototype Miniature Autonomous Underwater Vehicle”, Proceedings of the 2003 IEEE/RSJ International Conference on Intelligent Robots and Systems, pp. 842–846, Las Vegas, Nevada, October 2003

D. Yoerger, J. Newman, “Demonstration of closed-loop [7] Trajectory Control of an Underwater Vehicle”, 1985 Proceedings of OCEANS, vol 17, pp. 1028–1033, Nov. 1985

D. Yoerger, J. Slotine, “Robust Trajectory Control of [8] Underwater Vehicles”, IEEE Journal of Oceanic Engineering, Vol. OE-10, No.4, pp. 462 – 470, October 1985

D. Yoerger, J. Slotine, “Adaptive Sliding Control of and [9] Experimental Underwater Vehicles”, Proceedings of the 1991 IEEE International Conference on Robotics and Automation, pp. 2746 – 2751, Sacramento, California, April 1991

F. Song and S. Smith, “Design of Sliding Mode Fuzzy [10] Controllers for Autonomous Underwater Vehicle without System Model”, OCEANS’2000 IEEE/MTS, pp. 835-840, 2000

Arthur G. O. Mutambara, “Design And Analysis of Control [11] Systems”, CRC Press 1999

J. Yuh, “A Neural Net Controller for Underwater Robotic [12] Vehicles,” IEEE Journal of Ocean Engineering, Vol. 15, No. 3, pp. 161–166, July 1990

J. Yuh, R. Lakshmi, S. J. Lee, and J. Oh, “An Adaptive Neural-[13] Net Controller for Robotic Manipulators”, in Robotics and Manufacturing, M. Jamshidi and M. Saif, Eds. New York: ASME, 1990

Tamaki URA, Teruo FUJII, Yoshiaki Nose and Yoji Kuroda, [14] “Self-Organizing Control System for Underwater Vehicles”, IEEE 1990

Kazuo Ishii, Teruo Fujii, and Tamaki Ura, “An On-line [15] Adaptation Method in a Neural Network Based Control System for AUVs”, IEEE Journal of Ocean Engineering, Vol. 20, No. 3, pp. 221–228, July 1995

Jeen-Shing Wang and C. S. George Lee, “Efficient Neuro-[16] Fuzzy Control Systems for Autonomous Underwater Vehicle Control”, Proceedings of the 2001 IEEE International Conference on Robotics and Automation, pp. 2986–2991 Seoul, Korea, 2001

Ahmad M. Ibrahim, “Fuzzy Logic for Embedded Systems [17] Applications”, Elsevier Science 2004

Bjvrn Jalving, “The ADRE-AUV Flight Control System”, IEEE [18] Journal of Ocean Engineering, Vol. 19, No. 4, pp. 497–501, October 1994

J. Yuh, Michael E. West, P. M. Lee, “An Autonomous [19] Underwater Vehicle Control with a Non-regressor Based Algorithm +”, Proceedings of the 2001 IEEE International Conference on Robotics and Automation, pp. 2363–2368, Seoul, Korea, May 2001

S. Zhao, J. Yuh, and S. K. Choi, “Adaptive DOB Control [20] for AUVs”, Proceedings of the 2004 IEEE International Conference on Robotics and Automation, pp. 4899–4904, New Orleans, LA, April 2004

James Bellingham, MIT Sea Grant, Cambridge, MA, USA. [21] (doi:10.1006/rwos.2001.0303)

Kamarudin Shehabuddeen received the B. Eng. degree in mechanical engineering from University of Sunderland, UK, in 1996, and M.S. degree in engineering design from Loughborough University, UK, in 1998. He is registered with the Board of Engineers Malaysia (BEM) as a Professional Engineer (P.Eng.) in the mechanical engineering discipline.

He worked in Wembley I.B.A.E. Sdn. Bhd., Shah Alam, Malaysia, as a mechanical design engineer from 1996 – 1997. After receiving the M.S. degree, he worked with Sanyco Grand Sdn Bhd., Shah Alam, Malaysia, as a test rig design engineer for automotive brake master pumps. In 1999, he joined Universiti Teknologi PETRONAS (UTP), Malaysia, where he is currently a lecturer in the department of mechanical engineering. His current research interests include neuro-fuzzy based control, adaptive controls, Global Positioning System (GPS) based navigation of autonomous underwater vehicle and unmanned air vehicle. Currently he is pursuing his PhD degree in mechanical engineering at UTP.

Fakhruldin Mohd. Hashim is currently an Associate Professor at the Department of Mechanical Engineering, Universiti Teknologi PETRONAS (UTP). Formerly the head of the department and now the Deepwater Technology R&D cluster leader, he holds a BEng in Mechanical Engineering (RMIT), an MSc (Eng) in Advanced Manufacturing Systems & Technology

(Liverpool) and a PhD in Computer Aided Engineering (Leeds). Currently he is a UTP Senate and Research Strategy Committee member, and the Chairman of the Research Proposal Evaluation Committee for UTP. Dr. Fakhruldin’s areas of interest include Engineering Systems Design, Sub-sea Facilities and Underwater Robotics. He has been engaged as a consultant in over 20 engineering projects and has secured a number of major R&D grants during his 20-year career as an academician. He is one of the assessors for the Malaysian Qualifications Agency (MQA) and an external examiner of one of the local public university. He is also a trained facilitator in strategic planning and organisational management.

13



This paper was presented at the 5th European Thermal Sciences Conference, Netherlands

18 - 22 May 2008

INTRODUCTION

Condensation heat transfer, both inside and outside horizontal tubes, plays a key role in refrigeration, air-conditioning and heat pump applications. In recent years, the adoption of substitute working fluids and new enhanced surfaces for heat exchangers has thrown up new challenges in condensation heat transfer research. Well-known and widely established correlations to predict heat transfer during condensation may prove to be inaccurate in some new applications, and consequently a renewed effort is now being dedicated to the characterization of flow conditions and associated predictive procedures of heat transfer with condensing vapour. Much research effort has been directed at miniaturizing thermal systems and identifying innovative techniques for heat transfer enhancement. These techniques are classified as: passive enhancement techniques and active enhancement techniques. The double-

tube condenser is an example of the use of passive enhancement technique. For the given arrangement of double tube, the refrigerant flows in the annulus and the cooling water flows in the inner tube, shown in Figure 1.

ENHANCEMENT OF HEAT TRANSFER OF A LIQUID REFRIgERANT IN TRANSITION FLOW

IN THE ANNULUS OF A DOUBLE-TUBE CONDENSER

R. Tiruselvam1, W. M. Chin1 and Vijay R. Raghavan**Universiti Teknologi PETRONAS, 31750 Tronoh, Perak Darul Ridzuan, Malaysia

1OYL R&D Centre, Selangor, Malaysia

ABSTRACT

The basis of the present study is that augmentation can reduce the temperature difference across the condenser and evaporator of a refrigeration system and increase its energy efficiency. This research is conducted on the inner tube having a 3D corrugated outer surface. The annulus-side coefficients are determined using the Wilson Plot technique. It was found that the form of correlation by Hausen for transitional annular flow region with adjusted empirical constants predicts to a good accuracy the subcooled transitional flow of liquid. For the single phase heat transfer correlation proposed, all predicted data lie within +5% of the experimental values.

Keywords: Double-Tube Condenser, Transition Flow in Annulus, Augmented Surface, Heat Transfer Enhancement.

Figure 1. Concentric Tube Configuration (Double Tube Heat Exchanger)

14



OBJECTIVES

The purpose of the study is to obtain the single phase heat transfer coefficient for the subcooled liquid refrigerant flow in the annulus of the double tube exchanger. This research is conducted for a 3D corrugated outer surface enhancement (Turbo-C Copper Top Cross) on the inner tube used in a double-tube design. It is difficult to generate a mathematical model for such a complex geometry, and no standard correlation for the heat transfer coefficient is available for the Turbo-C copper top cross. Modification of equations from previous literature will require extensive measurements of pressure and temperature of the refrigerant and water sides, and the surface temperature of the fin and fin root surfaces. With the above mentioned motivation, the current research aims to characterize the heat transfer performance of the Turbo-C Copper Top Cross tube for subcooled liquid and the enhancement in comparison with a plain tube. Experiments are conducted to obtain the necessary data using R-22 (Chlodifluoromethane) as test fluid and plain tube and Copper Top Cross as test surfaces.

ExPERIMENTS

The commissioning of the test section was carried out using R-22 refrigerant in the annulus and water as coolant in the inner tube. The primary objective of these tests is to determine the correlation for overall heat transfer of the condensation process. The overall energy flow in the test section can be determined using three independent routes. These routes use:-• temperatureincreaseinthecoolantflow• the mass of condensate collected in the test

section • circumferentially averaged temperature drop

across the tube wall

Deans et al. (1999) have reported that the maximum difference in the calculated overall energy flows using these three routes to analyse the condensation process was less than 5%. The temperature increase in the coolant flow is chosen in the present case due to its ease in experimental measurements as well as

in the Wilson Plot. The test facility was designed and assembled so as to cater for this need. The test facility is capable of testing either plain or enhanced straight double-tube condenser at conditions typical of a vapour compression refrigeration system.

Setup

The double-tube heat exchanger configuration used in this study is a one-pass, single-track system. Single-track system means that that only one test section and one refrigerant compressor may be installed in the system for individual testing. The compressed refrigerant is initially condensed into subcooled liquid using a pre-condenser before it passes through the double-tube condenser only once as it travels from the high-side (compressor discharge line) to the low-side (compressor suction line), as shown in Figure 2. The single-track and single-pass system makes it possible to obtain one successful data point for every setting. If the operating conditions such as refrigerant mass flow rate or compressor discharge pressure are varied (non-geometric variables), it is possible to obtain additional data points without changing the test section or compressor. The use of the electronic expansion valve (EXV) permits such operation. Use of a conventional expansion device such as the capillary tube will involve repetition of work where the refrigerant circuit has to be vacuumed, leak tested and recharged for the individual length of capillary tube needed to throttle the refrigerant flow. The two main variable components in this test facility are the test section and the refrigerant compressor. The facility is designed and installed with valves and fittings for both the refrigerant medium and the cooling water. This is to allow for quick and easy replacement of the test section and/or refrigerant compressor. Each refrigerant compressor has an individual range of refrigerant flow rate, depending on the amount of refrigerant charge and compressor suction and discharge pressure. Three different refrigerant compressors were chosen (1HP, 2HP, and 3HP) to provide a sufficient range of refrigerant mass flow rates. A straight horizontal test section was used in this study, with two types of inner tube, i.e. Plain Tube (Plain Annulus) and Turbo-C (Enhanced Annulus),

15



Figure 2. Schematic Diagram of Experimental facility

16



with data as shown in Table 1. An illustration of the enhanced annulus is shown in Figure 3.

Data Reduction & Experimental Uncertainties

In the data run the subcooled liquid from the pre-cooler enters and exits in the same phase. The heat transferred to the cooling water is kept sufficiently low to allow the subcooled liquid temperature at the test section exit to be below the saturation temperature; hence no phase change occurs. This will allow us to obtain the single phase heat transfer coefficient for subcooled liquid.

( ) ( ) ( ) )()( LMTDAUTCpMhMQ oCSCCCCRRSC =∆=∆= (1)

( ) ( ) ( )iCCW

oCSCoCSC AhR

AhAU

111 ++= (2)

From equation (1) and (2) the heat transfer coefficient for single phase liquid is calculated using the Wilson Plot technique by subtracting the thermal resistance of the tube wall and cooling water from the overall thermal resistance. This gives us the heat transfer coefficient for subcooled liquid given in equation (3).

( ) ( )( ) ( )( )

1−

−−

=

CiC

oCWoC

SC

oCSC hA

ARA

Q

LMTDAh

(3)

The heat transfer coefficient hi for cooling water in the tube is got from Petukhov and Popov(1963), for the application range, 0.5 < Pr < 106 and 4000 < Re < 5x106, with a reported uncertainty of 8%.

Lk

d

d

RC u

i

O

W π2

ln

= (4)

( ) ( )( )

−

+

+

−

+

=

−3

1

Pr1Pr2

7.12Pr101

63.0Re900

07.1

PrRe2

5.0

CCCC

CCC

i

C

C

f

f

D

kh

(5)

Uncertainties in the experimental data were calculated based on the propagation of error method, described by Kline and McClintock (1953). Accuracy for various measurement devices, refrigerant properties and water properties are given in Table 3 and 4.

Table 1. Description of the Test Section Inner Tubes

Description Plain Tube Turbo-C

Inner Tube Outer Diameter, do 22.2 mm 22.2 mm

Inner Tube Inner Diameter, di 17.82 mm 17.82 mm

Length 2980 mm 2980 mm

Outer Surface Smooth Surface 3-D Integral Fin

Inner Surface Smooth Surface Smooth Surface

Other Information N.A. Pitch of Fin = 0.75 mmPitch of Corrugation = 5 mmFin height = 0.8 mm

Figure 3. Illustration of enhanced surface (Turbo-C)

17



Table 2. Uncertainties of Various Measurement Devices

Parameter (Make, Type) Uncertainties

Water Volume Flow (YOKOGAWA, Magnetic Flow Meter) + 0.08% of reading

Refrigerant Mass Flow Meter (YOKOGAWA, Coriolis) + 0.05% of reading

Refrigerant Pressure (HAWK, Pressure Transducer) + 0.18 psig

Water temperature (CHINO, Pt-100 RTD) + 0.1 °C

Refrigerant Temperature (VANCO, Type-T Thermocouple) + 0.6 °C

Data provided by OYL R&D Centre Malaysia

Table 3. Uncertainties of Properties

Predicted Properties (R-22) Uncertainties Source

Density + 0.1%Kamei et al. (1995)

Isobaric Heat Capacity + 1.0%

Viscosity + 2.1% Klein et al. (1997)

Thermal Conductivity + 3.7% McLinden et al. (2000)

Predicted Properties (Water) Uncertainties

Density + 0.02%Wagner and Pruß (2002)

Isobaric Heat Capacity + 0.3%

Viscosity + 0.5%Kestin et al. (1984)

Thermal Conductivity + 0.5%

Predicted Properties (Copper) Uncertainties

Thermal Conductivity + 0.5% Touloukian et al. (1970)

Property data obtained from ASHRAE (2001)

Table 4. Uncertainty Analysis for Experimental Data

Test Sequence Plain Tube, hsc (W/m2.K) Turbo-C, hsc (W/m2.K)

Compressor Highest Lowest Highest Lowest

1 HP + 5.40% + 4.81% + 5.32% + 5.04%

2 HP + 4.90% + 3.86% + 4.24% + 3.95%

3 HP + 4.42% +3.72% + 3.99% + 3.89%

18



Uncertainties in the single phase heat transfer coefficient (subcooled liquid) are calculated for various test runs in the smooth and enhanced annulus as a root-sum-square (RSS) method. Experimental results and the associated uncertainties are listed in Table 6. The uncertainties are dominated by the uncertainties associated with the refrigerant and water properties. Higher uncertainties were found at higher refrigerant mass flow rate.

RESULTS & ANALYSIS

This experimental work was conducted to develop heat transfer correlations so that the results of the study could be directly used by the HVAC community for design and development of double tube condenser systems. The first step in this effort was comparison of the current data with available correlations. The correlations available in the literature range from purely theoretical ones to those, purely empirical in nature. All fluid properties required in the data

reduction and the correlations were evaluated using property tables in ASHRAE (2001). All subsequent analysis of correlations is given in non-dimensional form, shown in Figure 4. Comprehensive reviews of literature led to selection of a few correlations which best represent the annulus geometries and flow characteristics of the fluids in the test section. The selected correlation was compared with experimental work on the plain tube. This is done to standardise the apparatus and assess the applicability of the selected correlations.

The one that most accurately represents the range and conditions of interest was used as the starting point. These correlations give the basic representation of the average heat transfer coefficients for a given set of parametric conditions. Next, it was assumed that the presence of the fluid inlet and exit fittings (both refrigerant and cooling water) and surface enhancement did not alter the form of the heat transfer coefficient substantially and that any

Figure 4. Comparison of augmented single phase Nusselt number

19



difference present can be handled by adjusting the empirical constants. Base on the above conditions, the correlation by Hausen (1934) was found to be most suitable for the following analysis where it applies for transitional flow development until fully developed turbulent flow, given in equation (6). The use of Hausen (1934)’s correlation was reviewed and presented by Knudsen and Katz (1950) and by Knudsen et al. (1999).

( ) ( )

+=

32

3

1

3

2

1PrReO

i

D

DCNu

for 2,000 < Re < 35,000 (6)

Considering that the Reynolds Number for the test is <10,000 the flow of subcooled liquid in both the plain and enhanced annulus is taken to be in the transition region. Equation (6) was used to evaluate the Nusselt type correlation which is given by Hausen (1934). Thus, the single phase heat transfer coefficient for subcooled liquid for the plain annulus is:

( ) ( ) ( )

+=

32

31

1PrRe0055.0 8058.0

O

iRhLIQUID

TUBEPLAIN

D

DNu

(7)Equation (6) was also used to evaluate the Nusselt type correlation for enhanced annulus since applicable correlations for transition flow in the enhanced annulus are lacking. Thus, the single phase heat transfer coefficient for subcooled liquid flow for enhanced annulus is:

( ) ( ) ( )

+=−

32

31

1PrRe0086.0 8175.0

O

iRhLIQUID

CTURBO

D

DNu

(8)

Comparison of experimental Nusselt value using (7) and (8) against predicted Nusselt value is illustrated in Figure 5.

OVERVIEW REMARKS

The overall objective of the present study was to develop the single phase heat transfer coefficient for subcooled liquid in transition flow. The correlation by Hausen (1934) was used for both plain annulus

and enhanced annulus for the transition region. The new empirical constants resulted in good prediction for transitional subcooled liquid flow in the annulus for both plain annulus and enhanced annulus. By examining the accuracy of the single phase heat transfer correlation proposed, all predicted data is within the +5% of experimental value. The subcooled liquid flow of plain annulus has an absolute deviation of +2.52%. Similar results are observed for the subcooled liquid flow for enhanced annulus with +2.46%.

NOMENCLATURE

C Coefficient in heat transfer correlation (∆h)R Enthalpy change of liquid R-22 (J/kg)LMTD Log mean temperature difference Reh Reynolds Number of liquid R-22 based on annulus

hydraulic diameter

U Average Overall heat transfer coefficient (W/m2K)

di Inner tube inner diameterdo Inner tube outer diameter Di Outer tube inner diameter

Subscripts:c Cooling Waterco copper R Refrigerant R-22SC subcooled liquidw wall

ACKNOWLEDgEMENTThe authors wish to acknowledge the support provided for this research by OYL Research and Development Sdn. Bhd.

Figure 5. Experimental Nu vs. Predicted Nu for Subcooled Liquid

20



REFERENCES

ASHRAE, 2001, “Fundamentals Handbook”, Appendix E [1] Thermophysics Properties

Deans, J., Sinderman, A., Morrison, J.N.A., 1999, “Use Of [2] The Wilson Plot Method To Design and Commission A Condensation Heat Transfer Test Facility,” Two-Phase Flow Modelling and Experimentation, Edizioni ETS, pp. 351-357

Hausen, 1934, “C.H., Z. Ver. Dtsch. Ing. Beih.” Verfahrenstech., [3] Vol. 91, No. 4

Kamei, A., Beyerlein, S. W., and Jacobsen, R.T., 1995, [4] “Application of Nonlinear Regression in the Development of a Wide Range Formulation for HCFC-22,” International Journal of Thermophysics, Vol. 16, No. 5, pp. 1155-1164

Kestin, J., Sengers, J. V., Kamgar-Parsi, B., and Levelt Sengers, [5] J.M.H., 1984, “Thermo Physical Properties of Fluid H2O,” Journal of Physical and Chemical Reference Data, Vol. 13, No. 1, pp. 175-183

Klein, S. A., McLinden, M. O., 1997, “An Improved Extended [6] Corresponding States Method for Estimation of Viscosity of Pure Refrigerants and Mixtures,” International Journal of Refrigeration, Vol. 20, No. 3, pp. 208-217

Kline, S., and McClintok, F., 1953, “Describing Uncertainties [7] in Single-Sample Experiments,” Mechanical Engineering, Vol. 75, pp. 3-8

Knudsen, J. G., and Katz, D. L., 1950 “Chemical Engineering [8] Progress”, Vol. 46, pp. 490

Knudsen, J. G., Hottel, H. C., Sarofim, A. F., Wankat, P. C., [9] Knaebel, K. S., 1999, “Heat and Mass Transfer”, Ch. 5, McGraw-Hill, New York.

Mclinden, M. O., Klein, S. A., Perkins, R. A., 2000, “An [10] Improved Extended Corresponding States Model of Thermal Conductivity Refrigerants and Refrigerant Mixtures,” International Journal of Refrigeration, Vol. 23, pp. 43-63

Petukhov, B. S., and Popov, V. N., 1963 “Theoretical [11] Calculation of Heat Exchanger in Turbulent Flow in Tubes of an Incompressible Fluid with Variable Physical Properties,” High Temp., (1/1), pp. 69-83

Tiruselvam, R., 2007, “Condensation in the Annulus of [12] a Condenser with an Enhanced Inner Tube”, M.S. thesis (research), University Tun Hussein Onn Malaysia

Tiruselvam, R., Vijay R. Raghavan., Mohd. Zainal B. Md. [13] Yusof., 2007, “Refrigeration Efficiency Improvement Via Heat Transfer Enhancement”, Paper No. 030, Engineering Conference “EnCon”, Kuching, Sarawak, Malaysia, Dec. 27-29

Toulaukian, Y. S., Powell, R. W., HO, C. Y., and Klemens, P. G., [14] 1970, “Thermophysical Properties of Matter, Vol. 1, Thermal Conductivity, Metallic Elements and Alloys,” IFI/Plenum, New York

Wagner, W., and Pruß, A., 2002, “New International [15] Formulation for the Thermodynamic Properties of Ordinary Water Substance for General and Scientific Use,” Journal of Physical and Chemical Reference Data, Vol. 31, No. 2, pp. 387-535

R. Tiruselvam is currently a PhD candidate at the Faculty of Mechanical Engineering, Universiti Teknologi PETRONAS. He received his Master of Engineering (Research) in Mechanical Engineering from University Tun Hussein Onn Malaysia (UTHM) in 2007. Previously, he was conferred the B. Eng. (Hons) (Mechanical) from UTHM and Dip. Mechanical Eng. from

POLIMAS. His research interest is mainly in heat transfer enhancement in thermal systems. He has been collaborating with OYL R&D Centre for a period of 4 years. Currently he holds a position as Research Engineer in OYL R&D Centre.

Chin Wai Meng is the Research Manager of OYL Research & Development Centre Sdn Bhd, the research and design centre for OYL Group, whose primary business is in the Heating, Ventilation and Air-Conditioning (HVAC). He has been with the company for the past 19 years where his expertise is in the testing of air-conditioning units to determine performance and

reliability. He also has experience in the design and construction of psychrometric test facilities. For the past 3 years, he has established and led the Research Department which specialises in the research on Heat Transfer and Refrigeration Systems. Mr. Chin holds a Bachelor’s degree in Mechanical Engineering from Universiti Malaya and he is currently pursuing a Master of Science in Universiti Teknologi PETRONAS.

Vijay R. Raghavan is a professor of Mechanical Engineering at Universiti Teknologi PETRONAS. Earlier he was a professor of Mechanical Engineering at Universiti Teknologi Tun Hussein Onn Malaysia (UTHM) and at the Indian Institute of Technology Madras. His areas of interest are Thermofluids and Energy. He obtained his PhD in Mechanical Engineering in the

year 1980 from the Indian Institute of Technology. In addition to teaching and research, he is an active consultant for industries in Research and Development, Design and Troubleshooting.

21



This paper was presented at the UK – Malaysian Engineering Conference 2008, London

14 -16 July 2008.

INTRODUCTION

Diisopropanolamine (DIPA) is a secondary amine of aliphatic amine group. It has been widely used as an additive in cosmetic and personal care products, metalworking fluids as a corrosion inhibitor and lubricity agent, pharmaceutical industry for drug intermediates as well as solvent to remove acid gas from raw natural gas. Wastewater contaminated with DIPA has an extremely high chemical oxygen demand (COD) which exceeds the limits set by local authorities. High influent COD levels make the biological treatment of the wastewater not possible. Typical practice is to store the wastewater in a buffer tank prior to pick up by a licensed scheduled waste contractor. The cost of the wastewater disposal is huge due to the large volume of wastewater generated.

DIPA is highly water soluble. Thus, the removal of this organic pollutant is tricky and very little literature is available on this topic. Extracting the pollutant may be one of the possible ways to solve the problem, but the production of secondary waste may still be an issue. In the past several years, advanced oxidation processes (AOPs) have attracted many researchers. Numerous reports on AOPs are available with a handful of organic pollutants that are degradable through this technique. According to Huang et al. [1], Fenton’s reagent has been discovered over 100 years ago but its application as a destroying agent for organic pollutants was only explored several decades later. Among the applications of Fenton’s reagent that have been reported included the degradation of azo dye Amido black [2], textile effluents [3], cork cooking [4], and pharmaceutical waste [5]. The Fenton system is an attractive oxidant for wastewater treatment due

FENTON AND PHOTO-FENTON OxIDATION OF DIISOPROPANOLAMINE

Abdul Aziz Omar*, Putri Nadzrul Faizura Megat Khamaruddin and Raihan Mahirah RamliUniversiti Teknologi PETRONAS, 31750 Tronoh, Perak Darul Ridzuan, Malaysia.

*[email protected]

ABSTRACT

Diisopropanolamine has been widely used as an additive in cosmetic and personal care products, metalworking fluids as a corrosion inhibitor and lubricity agent, pharmaceutical industry for drug intermediates as well as solvent to remove acid gas from raw natural gas. Although it is well applied in industry, the in-situ wastewater treatment method for diisopropanolamine contaminated wastewater has not yet been developed. The applicability of Fenton’s reagent and photo-Fenton for the degradation of diisopropanolamine was investigated. The effect of H2O2 concentration towards the degradation was investigated by changing the concentration of H2O2 while keeping the initial COD, concentration of FeSO4 , pH and temperature constant at 50,000 mg/L, 4.98 g/L, 3 and 30 °C respectively. 31% and 24% of the diisopropanolamine degradation were achieved for Fenton and photo-Fenton respectively at Fe:H2O2 ratio 1:50. Further research work will be conducted to increase the degradation efficiency and determine other optimum parameters.

Keywords: Diisopropanolamine, wastewater, fenton oxidation, photo-fenton oxidation

22



to the fact that iron is a highly abundant and non-toxic element, as well as the simple handling procedure and environmentally benign of H2O2 [6].

THEORY

Glaze et al. [7] defined AOPs as “near ambient temperature and pressure water treatment processes which involve the generation of hydroxyl radicals in sufficient quantity to effect water purification”. The main feature of AOP is the hydroxyl radical, ·OH, which has a high oxidation potential and acts rapidly with most organic compounds to oxidize them into CO2 and water. Hydroxyl radicals have the second largest standard redox potential after fluorine, which is 2.8 V [8].

Hydrogen peroxide (H2O2) is a strong oxidant, but it alone is not effective in high concentration of refractory contaminant such as amines at a reasonable amount of H2O2. Thus, a relatively non-toxic catalyst like iron was introduced to increase the rate of ·OH radical production. Fenton’s reagent consists of ferrous ion (Fe2+) and H2O2 which generates the ·OH radical according to

Fe2+ + H2O2 → Fe3+ + ·OH + OH− (1)

·OH radical may also be scavenged by reaction with another Fe2+.

·OH + Fe2+ → Fe3+ + OH− (2)

The reaction between Fe3+ and H2O2 slowly regenerated Fe2+.

In their report, Walling and Goosen [12] have simplified the overall Fenton chemistry by considering the dissociation of water as in Equation (3) which suggests that the acidic environment is needed in order to dissociate H2O2 by the presence of H+.

2Fe2+ + H2O2 + 2H+ → Fe3+ + 2 H2O (3)

Another way to initiate the generation of ·OH radical is by supplying UV radiation to the system containing

H2O2. The direct photolysis of H2O2 leads to the formation of the ·OH radical.

H2O2 →uv 2·OH (4)

Combination of UV irradiation with Fenton’s system known as photo-Fenton has also been a promising technique in wastewater treatment and research has been conducted on the application of this technique to some organic compound. Based on the literature, the presence of light has increased the production rate of ·OH radical by an additional reaction as in (5).

Fe(OH)2+ →uv Fe2+ + ·OH (5)

According to Matthew [8], although reactions (4) and (5) are important, the most vital aspect of photo-Fenton is the photochemical cycling of Fe3+ back to Fe2+.

There are few main factors affecting the process efficiency. The concentration of H2O2, concentration of FeSO4, UV power dosage, temperature and pH are among the main contributing factors towards process efficiency.

Effect of H2O2 concentration

H2O2 plays an important role as the oxidising agent in the AOP. Thus, it is vital to optimise the concentration of H2O2 because the main cost of these methods is the cost of H2O2 and excessive dosed of H2O2 trigger side-effects [9] due to self-scavenging of ·OH radical by H2O2 itself (Eqn. 6). Jian et al., [2] reported in their study that the degradation of Amido black 10B was reducing with high concentration of H2O2.

·OH + H2O2 → H2O + HO2· (6)

Effect of the initial Fe2+ concentration

Optimum concentration of Fe2+ is vital for the reaction as too low a concentration will slow the generation of ·OH radical thus reducing the removal of COD. But too high a concentration could lead to self-scavenging of ·OH radical by Fe2+ (Eqn. 2) [2].

23



Effect of UV power dosage

Hung at el. [9] in their report, wrote that COD removal could be increased by increasing UV power dosage. This is due to the faster formation rate of ·OH radical. The dosage of UV could be controlled by the number of lamps inside the reactor.

Effect of Temperature

Temperature is a critical factor to reaction rate, production yield and distribution. Reaction rate is expected to increase with increased temperature. However, in these processes, no optimal temperature was detected [10, 11]. Some researchers, however, gave the opposite result. Anabela et al. [4] reported in their literature that the optimal temperature was 30 °C in the degradation of cork cooking wastewater.

Effect of pH

Previous studies have shown that the acidic level of near pH3 was usually the optimum. High values of pH (>4) decreased the generation of ·OH radical because of the formation of ferric hydroxo complexes but at too low a pH value (<2), the reaction slowed down due to the formation of [Fe(H2O2)6]2+ [11].

RESEARCH METHODOLOgY

Materials

DIPA was obtained from Malaysia Liquified Natural Gas (MLNG); H2O2 and NaOH were from Systerm; FeSO4·7H2O was from Hamburg Chemicals; and H2SO4 was from Mallinckrodt.

Experimental procedure

A stirred jacketed glass reactor was used to monitor the progress of the reaction. The simulated waste of DIPA in the desired concentration was prepared and concentrated H2SO4 and 1M NaOH were added to adjust the solution to the desired pH value. The ferrous sulfate catalyst was added into the amine solution at the beginning of the experiment and stirred to get a

homogeneous solution. This is then followed by the addition of H2O2. The reaction started immediately and the temperature was maintained by circulating cooling water through the jacket. Samples were taken at regular intervals of time for COD analysis. COD was then measured by a Hach 5000.

For photo-Fenton oxidation, the experimental procedure was similar to the Fenton process, except for the additional UV irradiation. A quartz tube containing a UV lamp 4W was inserted into the reactor.

Analysis

Samples of 3 ml in volume were taken and put into the test tube containing 4 ml of 1M NaOH at the regular interval for COD analysis. NaOH was added into the samples to increase pH to 12 so that hydrogen peroxide became unstable thus decomposing into oxygen and water. Besides, this method can precipitate iron into ferric hydroxide [Fe(OH)3] [12]. The precipitate Fe(OH)3 was then separated from the solution by using microfilter. In order to further ensure that no interference of H2O2 to the COD measurement, the test tubes containing samples were heated in the boiling water for 10 minutes to remove the remaining H2O2 [13-15] as the peroxide is also unstable at temperature higher than 40 °C. The level of reproducibility for this system is high where the COD percentage removal will only vary about 5 percent between runs for the same parameters.

RESULT AND DISCUSSION

Effect of H2O2 Concentration

The effect of H2O2 concentration on COD removals was examined by changing the H2O2 concentration while keeping the concentration of FeSO4, pH and temperature constant at 4.98 g/L, 3 and 30 °C, respectively (Fig. 1).

Figure 1(a) shows the percentage of COD removal versus time for different H2O2 concentration at initial COD of 50,000 mg/L. From the figure, the COD removal increases with increasing H2O2 concentration.

24



However, when the Fe:H2O2 ratio is more than 1:50, the percentage removal decreases as in Figure 1(b). This may be due to the scavenging effect of H2O2 as in Eqn. (6). When too much H2O2 was in the solution, it reacted with the ·OH radical and subsequently reduced the concentration of ·OH radical available to attack the organic compound.

·OH + H2O2 → H2O + HO2· (6)

Figure 2(a) shows the percentage of COD removal by using photo-Fenton oxidation method. It followed the same trend as Fenton’s system. The percentage of COD removal increased when the hydrogen peroxide concentration increased. But when the Fe:H2O2 ratio was more than 1:30, the percentage removal decreased as in Figure 2(b).

The efficiency between Fenton’s system and photo-Fenton was compared as in Figure 3. From the figure,

0 10 20 30 40 50 60 70 80 900

10

20

30

40

50

CO

D R

emov

al (

mg

/L)

Fe:H2O2 = 1:20

Fe:H2O2 = 1:30

Fe:H2O2 = 1:40

Fe:H2O2 = 1:50

Fe:H2O2 = 1:60

Time (min)

1:20 1:30 1:40 1:50 1:60

18

20

22

24

26

28

30

CO

D R

emov

al (

%)

Fe:H2O2

(a) (b)

Figure 1. Effect of H2O2 concentration on COD removal for Fenton Oxidation (Initial COD = 50,000 mg/L; FeSO4 was 4.98 g/L; temperature = 30 °C, pH = 3)

0 10 20 30 40 50 60 70 80 900

5

10

15

20

25

30

35

Time (min)

CO

D R

emov

al (

%)

Fe: H2O2 = 1:20

Fe: H2O2 = 1:30

Fe: H2O2 = 1:40

Fe: H2O2 = 1:50

Fe: H2O2 = 1:60

1:20 1:30 1:40 1:50 1:60

19

20

21

22

23

24

25

26

CO

D R

emov

al (

%)

Fe:H2O2

(a) (b)

Figure 2. Effect of H2O2 concentration on COD removal for photo-Fenton Oxidation (Initial COD = 50,000 mg/L; FeSO4 was 4.98 g/L; temperature = 30 °C, pH = 3)

25



the percentage COD removal for photo-Fenton is slightly higher compared to Fenton. This could be due to the additional UV irradiation added into the system. The addition led to the additional production of ·OH radical which then increased the concentration of ·OH radical. Besides, UV irradiation was able to regenerate ferrous catalyst by the reduction of Fe(III) to Fe(II) as in Eqn. (7).

Fe(III)OH2+ →uv Fe(II) + ·OH (7)

However, at ratio Fe:H2O2 of 1:50, the result obtained was different. The percentage COD removal for Fenton was higher. The problem could be due to the stirring method. The stirrer used was small and the volume of the solution inside the reactor was high. Besides, the presence of a quartz tube inside the reactor might have reduced the stirring efficiency. One way to overcome the problem is by using aeration to mix the solution and this is recommended for further research plans.

CONCLUSION

The applicability of Fenton and photo-Fenton for the degradation of diisopropanolamine was investigated. By keeping the initial COD, concentration of FeSO4, pH and temperature constant at 50,000 mg/L, 4.98 g/L, 3 and 30 °C respectively, the optimum ratio of Fe:H2O2

for both Fenton and photo-Fenton was 1:50. COD removal of 31% and 24% was achieved for Fenton and photo-Fenton, respectively. Initial comparison between Fenton and photo-Fenton showed that photo-Fenton gave better degradation. However, this is not conclusive and further research will have to be conducted to increase the percentage of COD removal and reduce sludge formation.

REFERENCES

C. P. Huang, C. Dong, Z. Tang, “Advanced chemical oxidation: [1] its present role and potential future in hazardous waste treatment”, Waste Mgmt, 13 (1993) 361-577

J. H. Sun, S. P. Sun, G. L. Wang, and L. P. Qiao (2006) [2] “Degradation of azo dye Amido black 10B in aqueous solution by Fenton oxidation process”, Dyes and Pigment, B136, 258-265G

M. Perez, F. Torrades, X. Domenech, J. Peral (2001) “Fenton [3] and photo-Fenton oxidation of textile effluents”, Water Research 36, 2703-2710

A. M. F. M. Guedes, L. M. P. Madeira, R. A. R. Boaventura and [4] C. A. V. Costa (2003) “ Fenton oxidation of cork cooking wastewater-overall kinetic analysis” , Water Research 37, 3061-3069

Huseyin, T., Okan, B., Selale, S. A., Tolga, H. B., I. Haluk [5] Ceribasi, F. Dilek Sanin, Filiz, B. D. and Ulku, Y. (2006) “ Use of Fenton oxidation to improve the biodegradability of a pharmaceutical wastewater” , Hazard. Mater. B136 (258-265)

Rein, M (2001) “Advanced oxidation processes – current [6] status and prospects” , Proc. Estonian Acad. Sci. Chem., 50, 2, 59-80

Glaze, W. H., Kang, J. W. and Chapin, D. H. (1987) “ The [7] chemistry of water treatment processes involving ozone, hydrogen peroxide and ultraviolet radiation”, Ozone Science Engineering

Matthew, A. T., (2003) “ Chemical degradation methods [8] for wastes and pollutants, environmental and industrial applications” , pp. 164-194, Marcel Dekker Inc

Hung, Y. S, Ming, C. C. and Wen, P. H. (2006) “ Remedy of [9] dye manufacturing process effluent by UV/H2O2 process”, Hazard. Mater. B128, 60–66

Dutta, K., Subrata, M., Sekhar B., and Basab C. (2001) @ [10] Chemical oxidation of methylene blue using a Fenton-like reaction@ , Hazard Mater. B84, 57-71

Figure 3. Comparison of degradation efficiency between Fenton and photo-Fenton oxidation of DIPA (Initial COD=50,000 mg/L; FeSO4 was 4.98 g/L; temperature = 28 °C, pH = 3)

26



Ipek, G., Gulerman, A. S. and Filiz, B. D. (2006) “Importance [11] of H2O2/Fe2+ ratio in Fenton’s treatment of a carpet dyeing wastewater”, Hazard. Mater. B136, 763-769

C. Walling and A. Goosen, (1973) “Mechanism of the ferric [12] ion catalysed decomposition of hydrogen peroxide: effects of organic substrate”, J. Am. Chem. Soc. 95 (9) 2987-2991

Kavitha, V. and Palanivelu, K. (2005) “The role of ferrous ion [13] in Fenton and photo-Fenton processes for the degradation of phenol”, Chemosphere 55, 1235-1243

Lou J. C. and Lee S. S. (1995) “Chemical Oxidation of BTX [14] using Fenton’s reagent”, Hazard Mater. 12, No 2, 185-193

Jones C. W. (1999) “Introduction to the preparation and [15] properties of hydrogen peroxide”. In: Clark, J. H. (Ed) Application of Hydrogen Peroxide and Derivatives. Royal Society of Chemistry, Cambridge, UK, pp. 30

Abdul Aziz Omar is currently Head of the Geosciences and Petroleum Engineering Department, Universiti Teknologi PETRONAS.

Associate Professor Aziz completed his tertiary education in the United States, where he obtained his Master’s and Bachelor’s degrees from Ohio University

in 1982. While his undergraduate training was in Chemistry and Chemical Engineering, his graduate specialisation was in Environmental Studies. He has over 15 years of experience as an academician, 6 years as a process/project engineer and 4 years as a senior manager. He has also worked on many projects related to EIA (Environmental Impact Assessment), safety studies and process engineering design. Among his many experiences, one significant one would be the setting up of the School of Chemical Engineering at Universiti Sains Malaysia. He was appointed the founding Dean of the School. In March 2001, he joined Universiti Teknologi PETRONAS (UTP) as a lecturer, where he continues to teach and pursue his research interests. Assoc. Prof. Aziz is a Professional Engineer in Chemical Engineering, registered with the Malaysian Board of Engineers since 1989, and a Chartered Engineer in United Kingdom from 2006. He is a Fellow of the Institution of Chemical Engineers (IChemE), UK and is a Board member of IChemE in Malaysia.

27



This paper was presented at the International Conference on Nanoscience and Nanotechnology 2008, Shah Alam,

18 - 21 November 2008

INTRODUCTION

Iron has been the traditional catalyst of choice for the Fischer-Tropsch synthesis due to its favorable economics. However, knowledge on the relation between the rate of the reaction to the composition and morphology of the catalyst is still lacking [1]. The use of spherical model catalyst system enables investigation on the fundamental aspects of the catalyst, such as influence of particle size, phase and composition on the catalytic activity [2]. The objective of the present work is to prepare and characterize spherical model SiO2-supported iron catalysts. The catalysts were prepared using the colloidal synthesis approach [3], the reverse microemulsion method [4] and the ammonia deposition method [5]. The colloidal synthesis approach was adapted from a method described by Sun and Zeng [3] which involved

homogeneous nucleation process. However, our aim is to stabilise the iron nanoparticles on the SiO2 spheres through a heterogeneous nucleation process.

The usage of spherical model silica support allows for viewing of iron nano-particles in profile with transmission electron microscopy. Supported iron nano-particles in combination with electron microscopy are well suited to study morphological changes that occur during the Fischer-Tropsch synthesis. The spherical model catalyst enables investigation on the fundamental aspects of the catalyst, such as influence of particle size, phase and composition on the catalytic activities. This paper presents the results of three synthesis approaches for the spherical Fe/SiO2 model catalysts as well as their morphological changes upon exposure to syngas.

SYNTHESIS OF WELL-DEFINED IRON NANOPARTICLES ON A SPHERICAL MODEL SUPPORT

Noor Asmawati Mohd Zabidi*, P. Moodley1, P. C. Thüne1, J. W. Niemantsverdriet1

*Universiti Teknologi PETRONAS, 31750 Tronoh, Perak Darul Ridzuan, Malaysia 1Schuit Institute of Catalysis, Eindhoven University of Technology

*noorasmawati_mzabidi @petronas.com.my

ABSTRACT

Spherical model catalysts consisting of SiO2-sphere-supported iron nanoparticles were prepared using the colloidal synthesis approach, the reverse microemulsion and the ammonia deposition methods. Amongst these synthesis methods, the colloidal synthesis approach was found to be the most promising synthesis route for the Fe/SiO2 model catalysts. The modified colloidal synthesis method produced nearly monodisperse spherical-shaped iron oxide nanoparticles with average diameters of 6.2 ± 0.9 nm on the SiO2 spheres. X-ray photoelectron spectroscopy (XPS) analyses revealed that the catalyst contained Fe2O3 (hematite). Morphological changes were observed on the spherical Fe/SiO2 model catalysts during the Fischer-Tropsch synthesis (FTS) reaction.

Keywords. Nanoparticles, iron, spherical model catalyst, Fischer-Tropsch reaction

28



ExPERIMENTAL

For the colloidal synthesis method [3], non-porous silica spheres (BET surface area = 23 m2g-1, pore volume = 0.1 cm3g-1 and diameter = 100 - 150 nm) were sonicated in a mixture of olelylamine, oleic acid and cyclohexane for 1 hour and then heated and stirred in a multi-neck quartz reaction vessel. A liquid mixture of iron(III) acetyl acetonate, oleylamine, oleic acid, 1,2-hexadecanediol, and phenyl ether was slowly added to the stirred SiO2 suspension once the reaction temperature reached 150 °C. The reaction mixture was refluxed under nitrogen atmosphere at 265 °C for 30 minutes. The reverse microemulsion method [4] involved preparing two reverse microemulsions. The first reverse microemulsion consisted of Fe(NO3)3.9H2O (aq) and sodium bis-(2-ethylhexyl) sulfosuccinate (AOT, ionic surfactant) in hexanol. The second reverse microemulsion was prepared by mixing an aqueous hydrazine solution (reducing agent) with the AOT solution. SiO2 spheres were added to the mixture and the slurry was stirred for 3 hours under nitrogen environment. The ammonia deposition method utilised Fe(NO3)3.9H2O and 25 wt% aqueous ammonia [5].

The calcined catalyst samples were placed on carbon-coated Cu grids for characterisation by transmission electron microscopy. TEM studies were carried

out on a Tecnai 20 (FEI Co) transmission electron microscope operated at 200 kV. XPS was measured with a Kratos AXIS Ultra spectrometer, equipped with a monochromatic Al Kα X-ray source and a delay-line detector (DLD). Spectra were obtained using the aluminium anode (Al Kα = 1486.6 eV) operating at 150 W. Spectra were recorded at background pressure, 2 x 10-9 mbar. Binding energies were calibrated to Si2s peak at 154.25 eV.

The activities of the spherical model catalysts for Fischer-Tropsch synthesis were evaluated in a fixed-bed quartz tubular microreactor equipped with an on-line mass spectrometer (Balzers). Catalyst samples were pre-reduced in H2 at 450 °C for 2 h and then exposed to syngas (H2:CO = 5:1) at 270 °C. The samples were regenerated via heating in a flow of 20% oxygen in helium up to 600 °C.

RESULTS AND DISCUSSION

Figures 1(a), (b) and (c) show the TEM images of spherical model catalysts comprising iron oxide nano-particles anchored on SiO2 spheres prepared via the modified colloidal synthesis approach, the reverse microemulsion method and the ammonia deposition method, respectively. Spherical-shaped iron oxide nano-particles with average diameters of 6.2 ± 0.9 nm were formed via the modified colloidal synthesis method and the nano-particles were almost evenly dispersed on the SiO2 surfaces. An equimolar mixture

Figure 1(a) Figure 1(b) Figure 1(c)

Figure 1. TEM images of calcined catalysts of iron oxide nanoparticles on SiO2 spheres prepared via the (a) colloidal synthesis approach (b) reverse microemulsion method (c) ammonia deposition method.

29



of oleylamine and oleic acid was used in the colloidal synthesis approach and these surfactants were able to prevent the agglomeration of the iron oxide nano-particles. Iron oxide nano-particles were anchored on the SiO2 surfaces and did not lie in between the SiO2 spheres, as shown in Figure 1(a), thus suggesting that nucleation occurred heterogeneously. The iron loading was kept at 6 wt% as we have discovered that increasing the iron loading resulted in highly agglomerated nano-particles. The size of the nano-particle is influenced by temperature, time, surfactant, amounts of metal precursor as well as the ratio of the metal precursor to the surfactant [6]. The reverse microemulsion method also produced spherical-shaped iron oxide nano-particles with average diameters of 6.3 ± 1.7 nm, however, the coverage of the SiO2 surfaces was found to be less than that obtained using the colloidal synthesis approach. The result of the synthesis via the ammonia deposition method showed extensive agglomeration of the iron nanoparticles, as depicted in Figure 1(c).

The spherical model catalysts synthesised by the colloidal synthesis method and the reverse microemulsion method were tested in a Fischer Tropsch reaction. However, only the catalyst synthesised via the colloidal method showed some activities in the Fischer Tropsch reaction. Changes on the morphology were investigated upon exposure to the syngas. Figure 2 shows the TEM image of the catalyst after two hours exposure to the syngas. It shows ~ 50% increase in the size of the nano-particles and formation of an outer rim of thickness 3.2 ± 0.6 nm, following exposure to the syngas. A darker color at the centre of the used catalyst nano-particles suggests that the iron oxide remained in the core whereas the outer rim consists of amorphous carbon (EB = 284.5 eV), as confirmed by the XPS analyses (Figure 3). Table 1 shows the atomic ratios obtained from XPS analyses. Figure 4 shows the presence of Fe3p peak at EB= 56.0 eV for the fresh and the used catalyst, thus suggesting that the catalyst remained as Fe2O3 (hematite). Upon contact with H2/CO, oxygen-deficient Fe2O3 was observed, however, the oxygen vacancies did not reach a critical value that can lead to nucleation of Fe3O4.

Table 1. Atomic ratios based on XPS analyses

Sample O1s Fe / O1s Si Fe 3p / O Fe2O3 C1s / Fe3p

Fresh 0.248 0.186 2.02

Spent 0.199 0.066 7.64

Regenerated 0.231 0.076 6.24

275 280 285 290 295 300

Binding energy (eV)

Inte

nsity

(AU

)

(a)

(b)

(c)

Figure 2. TEM image of iron oxide nanoparticles on SiO2 spheres after the Fischer Tropsch reaction at 270 °C for 2 hours at H2:CO ratio of 5:1.

Figure 3. XPS showing carbon region for (a) fresh (b) regenerated (c) spent Fe/SiO2

30



CONCLUSIONS

Spherical model catalysts consisting of SiO2-supported iron nano-particles have been prepared and characterized using TEM and XPS. The modified colloidal synthesis approach resulted in spherical-shaped iron oxide nano-particles with average diameters of 6.2 ± 0.9 nm. The modified colloidal synthesis method produced better dispersion of the iron oxide nano-particles compared to that obtained from the reverse microemulsion method. The spherical Fe/SiO2 model catalyst prepared via the modified colloidal synthesis method exhibited some activity towards Fischer-Tropsch synthesis whereas the one synthesized via the reverse microemulsion method showed negligible activity for FTS. Morphological changes were observed on the spherical Fe/SiO2 model catalysts upon exposure to syngas and during the re-oxidation step. TEM results show ~ 50% increase in the size of the iron oxide nano-particles and formation of the carbon rim around the iron oxide nano-particles (confirmed by XPS) upon a 2-hour exposure to the syngas. XPS measurements confirmed the presence of Fe2O3 nano-particles in the fresh and the used catalyst samples. The results of our investigation show that the spherical Fe/SiO2 model nano-catalyst of well-defined size can be prepared and characterised. This can facilitate the size-dependent studies of the iron-based catalyst in FTS.

ACKNOWLEDgEMENTSThe authors thank Mr. Tiny Verhoeven for the TEM and XPS measurements. The authors would also like to thank Mr. Denzil Moodley for his assistance in the activity studies. We acknowledge the financial support for this project from Sasol South Africa. Noor Asmawati Mohd Zabidi acknowledges the support given by Universiti Teknologi PETRONAS under the sabbatical leave programme.

REFERENCES

A. Sarkar, D. Seth, A.K. Dozier, J.K. Neathery, H. H. Hamdeh, [1] and B. H. Davis, “Fischer-Tropsch synthesis: morphology, phase transformation and particle size growth of nano-scale particles”, Catal. Lett. 117 (2007) 1

A. M. Saib, A. Borgna, J. van de Loosdrecht, P. J. van Berge, [2] J. W. Geus, and J. W. Niemantsverdriet, “Preparation and characterization of spherical Co/SiO2 model catalysts with well-defined nano-sized cobalt crystallites and a comparison of their stability against oxidation with water”, J. Catal. 239 (2006) 326

S. Sun and H. Zeng, “Size-controlled synthesis of magnetite [3] nanoparticles”, J. Am. Chem. Soc. 124 (2002) 8204

A. Martinez and G. Prieto, “The key role of support surface [4] tuning during the preparation of catalysts from reverse micellar-synthesized metal nanoparticles”,Catal. Comm. 8 (2007) 1479

A. Barbier, A. Hanif, J.A. Dalmon, G.A. Martin, “Preparation [5] and characterization of well-dispersed and stable Co/SiO2 catalysts using the ammonia method”, Appl. Catal. A. 168 (1998) 333

T. Hyeon, “Chemical synthesis of magnetic nanoparticles, [6] ”Chem. Commun. (2003) 927

Noor Asmawati Mohd Zabidi obtained her PhD in 1995 from University of Missouri-Kansas City, USA. She joined Universiti Teknologi PETRONAS in 2001 and was promoted to an Associate Professor in 2005. She was on a sabbatical leave at Eindhoven University of Technology, The Netherlands from March–December 2007. During the sabbatical leave, she carried out

a research project on the synthesis of SiO2-supported iron nanocatalysts for the Fischer Tropsch reaction. Her research interests are photocatalysis and catalytic conversion of gas to liquid (GTL) fuels and chemicals.

Fe 3p

45 50 55 60 65 70

Binding energy (eV)

AU

Fresh

Regenerated

Spent

Figure 4. XPS showing Fe3p region for (a) fresh (b) spent (c) regenerated Fe/SiO2

Fe 3p

45 50 55 60 65 70

Binding energy (eV)

AU

Fresh

Regenerated

Spent

31



This paper was presented at the International Gas Research Union Conference 2008, Paris

8 - 10 October 2008

PERFORMANCE AND EMISSION COMPARISON OF A DIRECT-INJECTION (DI) INTERNAL COMBUSTION ENgINE USINg HYDROgEN AND COMPRESSED NATURAL gAS AS FUELS

A. Rashid A. Aziz*, M. Adlan A., M. Faisal A. MuthalibUniversiti Teknologi PETRONAS, 31750 Tronoh, Perak Darul Ridzuan, Malaysia

*[email protected]

ABSTRACT

Hydrogen internal combustion engine is considered as a suitable pathway to hydrogen economy before fuel cell technologies became more mature and cost effective. In this study, combustion of hydrogen and compressed natural gas (CNG) in a direct-injection single cylinder research engine was investigated. Engine performance parameters such as the power, torque, BMEO and COV of hydrogen DI operation were measured and compared to CNG-DI operation. Stoichiometric combustion of CNG at part load (50% throttle opening) is presented and compared with hydrogen at 0.2 and 0.8 equivalent ratio. The slightly-lean hydrogen (0.8 equivalent ratio) resulted in a better overall performance and emission of the engine.

INTRODUCTION

Global warming and air pollution issues have brought international efforts to scale down the use of hydrocarbon fuel, which is one of the biggest contributors to a number of greenhouse gases. On the other hand, fossil fuel reserves, especially petroleum, is depleting and will be at its peak in just a couple of decades while the demand from industry and transportation is increasing [1].

Hydrogen has been introduced as an alternative energy carrier. It can be made from both renewable and fossil energy. By using renewable energy or nuclear power plant to produce hydrogen, greenhouse gases can be totally eliminated. It is better than electricity in terms of distribution efficiency, refuelling speed and energy capacity. On the other hand, hydrogen vehicle performance is comparable to hydrocarbon-fuelled vehicles. In addition, the only emission of hydrogen is water vapour.

However, the current hydrogen engines still face practical problems that mask the actual capability of hydrogen as fuel for transportation. Hydrogen engine’s power and speed are limited due to knock, backfire [2], low volumetric efficiency [3] and a number of other problems. Some researchers suggested the use of a direct-ignition system and a number of experiments have been conducted that provided proof that using hydrogen as fuel produces better output than its gasoline counterparts [2].

In this study, the main objective was to examine the performance of a hydrogen-fuelled direct-injection spark ignition internal combustion engine. Aspects that were observed were power and torque output, emissions, fuel consumption and operating characteristics.

ExPERIMENTS

Experiments were done using a four-stroke Hydra Single Cylinder Optical Research Engine at the

32



Centre for Automotive Research, Universiti Teknologi PETRONAS, Malaysia. The specification of the engine is listed in Table 1. Engine control parameters such as injection timing, ignition timing and air-fuel ratio were controlled by an ECU that was connected to a computer. All output parameters of the engine was obtained from a high-speed data acquisition as well as a low-speed data acquisition from the engine dynamometer control interface.

Two experiments were conducted in this study – “ultra-lean combustion” (equivalence ratio of 0.2) and “slightly-lean combustion” (0.8 which is near stoichiometric). The first experiment used low-pressure injector (7.5 bar). However, because the flow-rate of that injector was too slow, a high-pressure injector had to be used for the second experiment. Both experiments used the stratification method, also known as stratified-lean combustion. Figure 1 illustrates a piston with 35 mm bowl that was used in the experiments and Figure 2 shows the position of the injector relative to the spark plug in the direct injection system. Table 2 lists the main operating parameters.

Table 1. Engine details

Engine type 4-stroke spark ignition

No. of cylinders One

Displacement volume 399.25 cm3

Cylinder bore 76 mm

Cylinder stroke 88 mm

Compression ratio 14:1

Exhaust valve open 10° ATDC

Exhaust valve closed 45° BBDC

Inlet valve open 12° BTDC

Inlet valve closed 48° ABDC

Fuel induction Direct-injection

Rail Pressure 7.5 and 18 bar

Injector position Centre

Injector nozzle type Wide angle

Table 2. Main parameters of the two experiments

Parameter

Value

Ultra-lean Slightly-lean

Equivalence ratio 0.2 0.8

Injector Rail Pressure 7.5 bar 18 bar

Injection Timing 130 deg BTDC 130 deg BTDC

Load Part throttle Part throttle

Ignition Timing MBT MBT

Table 3. Main parameters of the CNG-DI experiment

Parameter Value

Equivalence ratio Stoichiometric

Injector Rail Pressure 18 bar

Injection Timing 130 deg BTDC

Load Part throttle

Ignition Timing MBT

Besides comparing results from both experiments, another result from an experiment on natural gas fuel was also made (Table 3).

Figure 1. Stratified-lean piston

Figure 2. Location of injector and spark-plug relative to the piston at TDC

33



RESULTS AND DISCUSSIONS

Engine Stability

The engine was unstable in ultra-lean mode. It alternated between producing positive torque and negative torque – showing that the engine was not producing positive work. Peak pressure varied as much as 13 percent (Figure 3).

Possible cause of the variation could be from misfiring. The injector that was used for this experiment was a low pressure injector with low mass flow rate – about 2.7 ms to fully inject the fuel. At 3 500 rpm, this corresponded to 56.7 CA degree which was a long duration for hydrogen injection. A study showed that, for equivalence ratio between 0.7 to 1.4, the coefficient of variation (CoV) is low but increases significantly when the ratio moves far from the range. The study also concluded that cycle variation is caused by variation in the early combustion period [4]. Running the engine at slightly-lean resulted in lower CoV. The combustion produced very high peak pressure, common for hydrogen engine because of its high flame temperature and high flame speed. The peak pressure reached up to 60 bars (Figure 4) but lower than what actually the engine could produce. It was seen that the ignition timing was set near TDC (Figure 3) to achieve MBT. Advancing the ignition resulted in a peak pressure reaching up to 90 bars with knocks occurring in the cylinder.

A study showed that MBT ignition advance increases when equivalence ratio decreases [5]. For a mixture of 0.2 equivalence ratio, the MBT ignition timing range is between 30 to 50 degrees BTDC. This is consistent with the current results.

Peak pressures of natural gas combustion nearly doubled the peak pressure for slightly-lean hydrogen combustion (Figure 5). Despite the high peak pressure, natural gas was still running without knock as opposed to hydrogen.

Performance

Combustion of ultra-lean mixture faced consistent misfiring which resulted in very low BMEP (Figure 6). Judging from the CoV, when the speed was increased,

10

20

30

40

50

60

-20 -10 0 10 20 30 40

Crank Angle (deg ATDC)

Cylinder Pressure (bar)

2500 rpm

3000 rpm

3500 rpm

4000 rpm

4500 rpm

Figure 3. Pressure developed during compression and expansion stroke at various engine speeds for 0.8 equivalence ratio

Figure 4. Coefficient of variation at different engine speed with its corresponding spark advance

Figure 5. BMEP comparison of ultra-lean H2, slightly-lean H2 and stoichiometric natural gas

0

2

4

6

8

10

12

14

16

1500

2000

2500

3000

3500

4000

4500

5000

Engine Speed (rpm)

CoV (%)

-5

0

5

10

15

20

25

30

35

Ignition (deg BTDC)

0.2 CoV

0.8 CoV

0.2 Ign.

0.8 Ign.

0

2

4

6

8

1500 2000 2500 3000 3500 4000 4500 5000

Engine Speed (rpm)

BMEP (bar)

Ultra-lean H2

Slightly-lean H2

CNG

34



misfiring occurred frequently (Figure 3). BMEP decreased until it could no longer produce any work after 4 500 rpm.

Usage of a wide angle injector caused only small amounts of fuel to be concentrated inside the piston bowl, which was less than ignitable air-fuel ratio. A possible solution is to utilise a narrow angle injector with stoichiometric-lean piston. A stoichiometric lean piston has a smaller bowl and could concentrate the mixture to be around the stoichiometric ratio. [6]

For the slightly-lean mixture, the usage of late ignition hampered its potential performance. At TDC, when the pressure is around 30 to 40 bars, the pressure diagram (Figure 4) clearly shows that rather than a smooth line, the pressure line dropped and ended up with lower peak pressure. In Figure 7, it was seen that near TDC, the pressure drop slightly after ignition, which was caused by the ignition delay.

In comparison with natural gas, hydrogen produced more torque and power. For torque, as well as BMEP curve, the values decreased with increase in engine speed (Figures 6 & 8). However, the power output of slightly-lean hydrogen did not show a drop with increase speed (Figure 9).

Figure 7: Brake-torque comparison of ultra-lean H2, slightly-lean H2 and stoichiometric natural gas

Figure 6. Pressure map comparison of ultra-lean H2, slightly-lean H2 and stoichiometric natural gas at 3500 rpm

Figure 8: P-V diagram of a cycle at 4500 rpm and 0.8 equivalence ratio.

Comparison at 3500 rpm

-10

0

10

20

30

40

50

60

0 0.0001 0.0002 0.0003 0.0004 0.0005

Volume (m3)

Pre

ssu

re (

bar

)

H2 - 0.8

CNG

H2 - 0.2

0

5

10

15

20

25

30

1500 2000 2500 3000 3500 4000 4500 5000

Engine Speed (rpm)

Brake-Torque (Nm)

Ultra-lean H2

Slightly-lean H2

CNG

Comparison at 4500 rpm

-10

0

10

20

30

40

50

60

70

0 0.0001 0.0002 0.0003 0.0004 0.0005

Volume (m3)

Pre

ssu

re (

bar

)

H2 - 0.8

CNG

H2 - 0.2

0

2

4

6

8

10

1500 2000 2500 3000 3500 4000 4500 5000

Engine Speed (rpm)

Brake-Power (kW)

Ultra-lean H2

Slightly-lean H2

CNG

Figure 9. Brake-power comparison between ultra-lean and slightly-lean hydrogen with stoichiometric natural gas.

35



Fuel Economy

The engine’s fuel consumption at ultra-lean mixture is fairly constant before 4 000 rpm (Figure 10). The sharp increase after that point could be linked to the misfiring that became worse at higher speeds. As the speed increases, the fuel had less time to mix with the air, making a heterogeneous mixture. At slightly-lean mixture, the consumption of fuel was also low – primarily caused by a stable combustion without misfire.

Figure 11 shows the fuel consumption comparison between hydrogen with natural gas. The graph shows the BSFC in equivalent gasoline consumption. Conversion was based on the fuels’ lower heating value. From the graph, it was seen that slightly-lean hydrogen had lower BSFC than natural gas. The ultra-lean hydrogen performed poorly at higher engine speeds.

Emissions

The only significant emission for hydrogen engine is nitrogen oxides (NOx). However, the comparison shows that natural gas emits more NOx than hydrogen (Figure 12). At ultra-lean, the amount of emitted NOx is considerably high especially at higher engine speeds. The existence of NOx is usually related to high combustion temperature in the cylinder. As speed increases, there is less time for heat to transfer from cylinder to the atmosphere, which increases temperature (Figure 13).

The test-bed did not have an in-cylinder thermocouple which could be used to determine temperatures inside the cylinder. However, the system did have an exhaust gas temperature. The exhaust temperature value was used to show a rough estimation of the combustion cylinder temperature.

Comparison in Figure 14 shows that the amount of carbon dioxide emitted by hydrogen engine is really insignificant compared to hydrocarbon fuel. The large percentage of carbon dioxide in natural gas exhaust is mainly caused by the carbon element in the structure

Figure 10. Brake specific fuel consumption (BSFC) at various engine speeds.

Figure 11. NOx comparison between ultra-lean and slightly-lean hydrogen with stoichiometric natural gas.

Figure 12. Comparison of BSFC in term of gasoline usage.

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1500 2000 2500 3000 3500 4000 4500 5000

Engine Speed (rpm)

BSFC (kg/kW-h) Ultra-lean H2

Slightly-leanH2

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1500 2000 2500 3000 3500 4000 4500 5000

Engine Speed (rpm)

NOx (g/m3)

Ultra-lean H2

Slightly-lean H2

CNG

0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

4.0

1500 2000 2500 3000 3500 4000 4500 5000

Engine Speed (rpm)

Gasoline Equivalent BSFC

(kg/kW-h)

Ultra-lean H2

Slightly-lean H2

CNG

36



of methane – a major constituent of natural gas. On the other hand, the existence of carbon dioxides in hydrogen fuelled engine is mainly contributed by combustion of engine oil on the cylinder wall [7].

The negative value of CO2 on the graph occurred because the emitted CO2 was less in relative to the calibration value. The system calculated CO2 in percentage of total exhaust gas – not in ppm unit. Supposedly, there must be no CO2 in the exhaust line while calibrating the gas analyser but a very little amount of CO2 could possibly exist. This also shows the low amount of CO2 emitted by a hydrogen-fuelled engine.

CONCLUSIONS

Based on the above results and discussions, the following conclusions were derived:• Direct-injection could avoid the backfire

phenomenon and reduce the likelihood of pre-ignition.

• Atpartload,poweroutputofhydrogenwasbetterthan natural gas. This proved that the actual power output of hydrogen is higher than current commercial fuels.

• Hydrogenatslightly-leanmixturehasbetterfueleconomy than ultra-lean mixture.

• Theonlysignificantemissionofhydrogenengineis NOx but it is still lower than the amount emitted by natural gas.

• TheamountofCO2 by hydrogen engine is much less than natural gas.

REFERENCES

Rifkin, J. (2002). “When There is No More Oil…: The [1] Hydrogen Economy”. Cambridge, UK: Polity Press

Das, L. M. (1990). “Fuel induction techniques for a hydrogen [2] operated engine”. International Journal of Hydrogen Energy, 15, 833-842

Yi, H. S., Lee S. J., & Kim, E. S. (1996). “Performance evaluation [3] and emission characteristics of in-cylinder injection type hydrogen fueled engine”. International Journal of Hydrogen Energy, 21, 617-624

Kim, Y. Y., Lee J. T. & Choi, G. H. (2005). “An investigation on [4] the causes of cycle variation in direct injection hydrogen fueled engines”. International Journal of Hydrogen Energy, 30, 69-76.

Mohammadi, A., Shioji, M., Nakai, Y., Ishikura, W. & Tabo, E. [5] (2007). “Performance and combustion characteristics of a direct injection SI hydrogen engine”. International Journal of Hydrogen Energy, 32, 296-304

Zhao, F. F., Harrington, D. L. & Lai, M. C. (2002). “Automotive [6] Gasoline Direct-Injection Engines”. Warrendale, PA: Society of Automotive Engineers

Norbeck, J. M., [7] et al. (1996). “Hydrogen Fuel for Surface Transportation”. Warrendale, PA: Society of Automotive Engineers

-1

0

1

2

3

4

5

6

1500 2000 2500 3000 3500 4000 4500 5000Engine Speed (rpm)

CO2 (vol %)

Ultra-lean H2

Slightly-lean H2

CNG

0

20

40

60

80

100

120

140

1500 2000 2500 3000 3500 4000 4500 5000

Engine Speed (rpm)

NOx (mg/m3)

775

800

825

850

875

900

925

950Temperature (K)NOx

Temp.

Figure 13. CO2 comparison between ultra-lean and slightly-lean hydrogen with stoichiometric natural gas.

Figure 14. Temperature of exhaust gas and the amount of emitted NOx at various speeds for slightly-lean combustion.

37



ACKNOWLEDgEMENTThe authors would like to thank the Ministry of Science, Technology and Innovation for the initial grant under the IRPA project on CNG-DI Engine as well as UTP for general support.

Abd Rashid Abd Aziz graduated with a PhD in Mechanical Engineering (Thermo-fluid) from the University of Miami in 1995. He is currently an Associate Professor and the Research Head for Green Technology (Solar, Hydrogen and Bio-fuels). He is also the group leader for the Hybrid Vehicle Cluster. He leads the Centre for Automotive Research (CAR), which carried out several

research projects with funds from the Ministry of Science, Technology and Innovation (MOSTI) and PETRONAS. His areas of interest are in internal combustion engines, laser diagnostics, flow visualisation, CFD, alternative fuels and hybrid powertrain.

38



13th International Conference on Applied Mechanics and Mechanical Engineering, Egypt,

27-29 May 2008

INTRODUCTION

The combustion of clouds of fuel droplets is of practical importance in engines, furnaces and also for prevention of explosion and fire in the storage and use of fuels. Theoretical [1] and experimental [2-4] evidence suggests that flame propagation through clouds of droplets, under certain circumstances, is higher than that in a fully vapourised homogeneous mixture. Even though this may be advantageous in giving more rapid burning, its effects on emissions are uncertain.

Most combustion in engineering applications takes place under turbulent conditions. Nevertheless,

it is well established that the laminar burning rate plays an important role in turbulent combustion [5]. Information on laminar burning velocity is sparse, even for gaseous mixtures at conditions pertaining to engines, which range from sub-atmospheric to high pressure and temperature. Such data for fuel sprays and for gas-liquid co-burning [6-8] are even sparser than for gases. As a consequence, there is little experimental data of a fundamental nature that clearly demonstrates the similarities and differences in burning rate, either laminar or turbulent, between single and two-phase combustion.

In the present work, the influence of droplets in iso-octane-air mixtures within the upper flammability

THE EFFECT OF DROPLETS ON BUOYANCY IN VERY RICH ISO-OCTANE-AIR FLAMES

Shaharin A. Sulaiman and Malcolm Lawes1

Universiti Teknologi PETRONAS, 31750 Tronoh, Perak, Malaysia 1School of Mechanical Engineering, University of Leeds, UK

[email protected]

ABSTRACT

An experimental study is performed with the aim of investigating the effect of the presence of droplets in flames of very rich iso-octane-air mixture under normal gravity. Experiments are conducted for initial pressures in the range 100-160 kPa and initial temperatures 287-303 K at an equivalence ratio of 2.0. Iso-octane-air aerosols are generated by expansion of the gaseous pre-mixture (condensation technique) to produce a homogeneously distributed suspension of near mono-disperse fuel droplets. The droplet size varies with time during expansion; hence the effect of droplet size in relation to the cellular structure of the flame was investigated by varying the ignition timing. Flame propagation behavior was observed in a cylindrical vessel equipped with optical windows by using schlieren photography. Local flame speeds were measured to assess the effect of buoyancy in gaseous and aerosol flames. It was found that the presence of droplets resulted in a much earlier onset of instabilities, at a rate faster than that taken for the buoyancy effect to take place. Flame instabilities, characterised by wrinkling and cellular surface structure, increase the burning rate due to the associated increase in surface area. Consequently, the presence of droplets resulted in a faster flame propagation rate than that displayed by a gaseous flame. The mechanism of flame instabilities that caused a significant reduction of the buoyancy effect is discussed.

Keywords: buoyancy, combustion, droplets, flame, instabilities

39



limit (rich) was investigated. Such gaseous mixtures are well known, from a number of previous works, for example in [9-11], to experience the effect of buoyancy or natural convection during combustion. Some of the previous studies were performed in tubes. However, Andrews and Bradley [12] suggested that the limit of flammability obtained by the tube method would be subject to the same sources of error [13] as would be the burning velocity measurements using tubes, mainly due to wall quenching. Thus studies in larger tube diameters or in large explosion vessels have been recommended.

Figure 1 shows an illustration of two centrally ignited spherical flames to describe the effect of buoyancy [14]. The open arrow represents the local velocity which resulted from gas expansion during flame propagation. The solid arrow, which points upward, represents the local velocity caused by the buoyancy or natural convection effect. The dashed lines show the resulting flame front accounting for the net velocity. With the absence of buoyancy effect, the resulting flame would be spherical or circular when viewed from the side. However, with the presence of buoyancy effect, the resulting flame has a flatter bottom surface as that illustrated in Figure 1.

ExPERIMENTAL APPARATUS

Figure 2 shows the photograph and schematic of the aerosol combustion apparatus. The combustion vessel, which essentially resembled a Wilson cloud chamber [15], was a cylindrical vessel of 305 mm diameter by 305 mm long (internal dimensions), with

a working volume of 23.2 litres. On both end plates of the combustion vessel circular optical access windows of 150 mm diameter were provided for characterisation of aerosol and photography of flame propagation. To initially mix the reactants four fans, driven by electric motors, were mounted adjacent to the wall of the

FLAME FLAME

Buoyancy driven flow velocity

Velocity resulted from gaseous expansion

without buoyancy effect with buoyancy effect

the resulting flame front

Gravity

Figure 1. Illustration of the buoyancy effect [14] on spherical flames. The arrows indicate the local velocities resulted by gas expansion and buoyancy driven convection

CV : Combustion Vessel EV : Expansion Vessel SL : Supply Line DL: Discharge Line VP: Vacuum Pump Orifice Pipe

Valve

DL

VP

CV

EV

SL

28 litres 23 litres

(a)

(b)

Figure 2. Aerosol combustion apparatus: (a) photograph (b) schematic

40



vessel. Two electrical heaters were attached to the wall of the vessel to preheat the vessel and mixture to 303 K. The expansion vessel, which has a volume of 28 litres, was connected to the combustion vessel by an interconnecting pipe through a port. The vacuum pump, indicated in Figure 2 (a), was used to evacuate the system and to remove particulates prior to preparation of the mixture.

The aerosol mixtures were prepared by a condensation technique, which generated near mono-dispersed droplet suspensions. This was achieved by controlled expansion of a gaseous fuel-air mixture from the combustion vessel into the expansion vessel that was pre-evacuated to less than 1 kPa. The expansion caused a reduction in the pressure and temperature of the mixture, which took it into the condensation regime and caused droplets to be formed.

The characteristics of the generated aerosol were calibrated by in-situ measurements of the temporal distribution of pressure, temperature, and droplet size and number, without combustion, with reference to the time from start of expansion. The diameters of individual droplets were measured using a Phase Doppler Anemometer (PDA) system, from which the droplet mean diameter, D10, was obtained. Since the expansion took place over a period of several seconds while combustion took place over less than 100 ms, the far field values of D10 were assumed to be constant during combustion.

The mean droplet diameter varied with time during expansion; hence the effect of droplet size in relation to the cellular structure of the flame was investigated by varying the ignition timing. The iso-octane-air mixture was ignited at the centre of the combustion vessel by an electric spark of approximately 500 mJ. The flame front was monitored through the vessel’s windows by schlieren movies, which were recorded using a high-speed digital camera at a rate of 1000 frames per second and with a resolution of 512 × 512 pixels. The flame image was processed digitally by using image-processing software to obtain the flame radius. The velocity of the flame front, also known as the stretched flame speed, Sn, was obtained directly

from the measured flame front radius, r, by

dt

drSn = (1)

Similarly, the local flame speed is given by

dt

dLS L = (2)

where L is the distance between the local flame front and the spark electrode as measured from the schlieren image of the flame.


Figure 3 shows the schlieren photographs of flames at maximum viewable radius and the corresponding contour plots at 2 ms intervals for gaseous and also aerosol mixtures. The mixtures were initially at equivalence ratio, φov, of 2.0, temperatures between 287 and 303 K, and pressures between 100 and 159 kPa. For the aerosol mixtures, the droplet mean diameters, D10, were 5 µm and 13 µm. It must be noted that the relatively small differences in pressure and temperature between the conditions in the three images have been shown, for gaseous flames, to have little effect on the flame structure [16]. Hence, it was assumed that the difference in the flame structure is entirely due to the effects of droplets. In Figure 3(a), the image is slightly darker than the others due to the low exposure setting of the camera. The circular black areas at the corners of the photographs represent the region beyond the window of the combustion vessel. The electrode holder and thermocouple are clearly shown on the middle-right section in each photograph.

It is shown in the schlieren image in Figure 3(a) that for a gaseous flame, the upward distance of the flame propagation is greater than the downward one, which is a sign of the buoyancy effect as described for Figure 2. The upper surface of the gaseous flame is relatively smoother and has fewer cells as compared to the lower surface. In the contour plot for the gaseous flame in Figure 3(a), larger spacing between flame contours is shown for the upper propagation as compared to the lower one. This suggests faster upward flame

41



propagation than the downward propagation. Conversely, the difference between the leftward and rightward propagations is shown to be small, the rightward propagation being very close to the flame radius. Hence it is shown that significant difference in propagation rate occurs only in the vertical direction.

With the presence of droplets of 5 µm, it is shown in Figure 3(b) that trend of vertical flame propagation is almost similar to that for the gaseous flame in Figure 3(a). However the aerosol flame has more cells on its surface than the gaseous one. Interestingly, with bigger droplets (D10 = 13 µm), it is shown in Figure 3(c) that the resulting flame is highly cellular. The contour plot of the flame shows faster burning rate (large spacing between contours) and also smaller difference between the upward and downward propagation rate (more uniform contour spacing in all directions) as compared to those in Figures 3 (a) and (b). Hence it is suggested that with the presence

of large diameter droplets, the buoyancy effect demonstrated in gaseous flames of rich mixtures is overcome by instabilities and consequently faster burning rate, such that there was less time available for natural convection to be significant.

Figure 4 shows graphs displaying the temporal variations of the flame radius and local vertical and horizontal propagation distances (from the electrode) for the gaseous iso-octane-air flame depicted in Figure 3(a) and for the aerosol flame in Figure 3(c), both at φov = 2.0. Here the effect of the presence of large droplets (D10 = 13 µm) is presented. The upward, downward, leftward and rightward propagation distances between the spark gap and the corresponding flame edge were measured at 1 ms interval.

It is shown in the graph in Figure 4(a) that for the gaseous flame, the upward propagation distance of the flame propagation is always greater than

(a) D10 = 0 µm (gaseous) (b) D10 = 5 µm (c) D10 = 13 µm

Figure 3. Schlieren images and contour plots (superimposition of edges) throughout propagation for iso-octane-air flames at φov = 2.0 and various droplet sizes. The time interval between each contour is 2 ms

42



that of the downward one. In addition, the upward distance of the flame propagation increases at a steady rate, whereas the downward one decelerates; these indicate the effect of buoyancy force acting on the hot flame kernel. The difference between the leftward and rightward propagations is shown to be small, although the rightward propagation distance seems to be slightly different from the flame radius. Obviously, the flame radius and horizontal propagation distances are shown to be at approximately midway between the upward and downward components. The deceleration in the downward propagation is only experienced by the gaseous flame, as shown in Figure 4(a). Hence, the cellularity on the bottom half of the gaseous flame in Figure 3(a) is probably caused by hydrodynamic instabilities, as a result of an upward flow of unburned gas which velocity exceeded the flame speed at the base of the flame, as illustrated in Figure 2. Conversely, the smoother upper surface of the flame is probably due to flame stabilisation as a result of high stretch rate. This occurs when the expanding flame front propagates through a velocity gradient in the unburned gas that is induced by the upward, buoyant acceleration of hot products as explained by Andrews and Bradley [12].

With the presence of large enough droplets (D10 = 13 µm), the aerosol flame burned faster than the gaseous

flame. This is shown in Figure 4(b), in which the aerosol flame took approximately 60 ms to reach a radius of 50 mm, as compared to about 90 ms for the gaseous flame to reach the same radius. This is very likely caused by earlier instabilities in the aerosol flames, as depicted in Figure 3, which promotes a faster burning rate due to increase in the flame surface area. In relation to the buoyancy, it is shown in Figure 4(b) that such effect is absent, as implied by the gap between the upward and downward flame propagation that is significantly and consistently smaller as compared to that for the gaseous flame. Furthermore, the aerosol flame exhibits acceleration in the downward propagation as compared to deceleration in the gaseous flame.

Figure 5 shows the vertical components of the flame propagation distance from the spark electrode for the gaseous flame and also the aerosol flames (D10

values of 5, 13, and 16 µm) at initial conditions similar to those described for Figure 3. The negative values indicate the downward flame propagation distances. The propagation rates for the fine aerosol (D10 = 5 µm) flames are shown in Figure 5 to be similar to those for the gaseous flames, as indicated by their nearly identical curve plots, particularly for the downward propagation of the flame. In addition, the buoyancy effect is evident by the greater values of the positive distance as compared to the negative distance. It is

0

10

20

30

40

50

60

70

0 20 40 60 80 100

Time (ms)

Dis

tanc

e (m

m)

U

L

R

D r

Propagation direction:

U: upward L: leftward R: rightward D: downward r: schlieren radius

0

10

20

30

40

50

60

70

0 20 40 60 80 100

Time (ms)

Dis

tanc

e (m

m)

D

U

L

R r

(a) gaseous, D10 = 0 µm (b) aerosol, D10 = 13 µm

Figure 4. Flame propagation distance from spark electrode as a function of time for iso-octane-air mixtures at φov = 2.0. Also shown is the corresponding direction of the flame propagation with respect to the spark electrode

43



clear in Figure 5 that the flames within aerosols of large droplets (13 and 16 µm) propagate at a faster rate than those of fine droplets (5 µm) and gaseous. For these aerosol flames, the effect of buoyancy is not obvious. Interestingly, it is shown that the curves for upward propagation for all values of D10 are coincident for approximately the first 25 ms of propagation; a similar trend is observed for downward propagation up to about 16 ms. This was probably because the buoyancy is not yet in effect during those periods of initial flame kernel growth.

Figure 6 shows graphs of variation in the local flame speed (deduced by time derivative of the graphs in Figure 4) with time from the start of ignition. The speed for the upward propagating flame is represented by the diamond markers, and the bottom one by the square markers. The circle and triangle markers represent the rightward and leftward flames respectively. The gaseous flame is shown in Figure 6(a) to propagate faster in the upward component by

about 0.4 m/s, as compared to that of the downward component, which also decreases towards a near zero value throughout propagation. A negative downward component, if were to occur, would implicate an upward flow of unburned gas at the central base of the flame; such cases were reported elsewhere; e.g. in [9], but this is beyond the scope of the present work. The sideway components of the flame speed are shown in Figure 6(a) to be similar, which suggest the independencies of these components from the natural convection effect in the gaseous flame. With droplets (D10 = 13 µm), it is shown in Figure 6(b) that all the components of flame speed nearly coincide with each other, indicating a more uniform distribution of flame speed throughout the flame surface, and hence evident the absence of the natural convection effect. However, after about 35 ms from the start of ignition, the curve for the upward component of the flame started to burn at a significantly faster rate than the other components. The reason for this is not clear and thus further investigation is required.

-6 0

-4 0

-2 0

0

2 0

4 0

6 0

0 2 0 4 0 6 0 8 0 10 0

Time(ms)

Dis

tanc

e fr

om s

park

ele

ctro

de (

mm

)

13 µm 16 µm 5 µm 0 µm

0 µm

5 µm

13 µm

16 µm

Downward

Upward

Figure 5. Comparison of vertical flame propagation distance from spark electrode as a function of time for iso-octane-air mixtures at φov = 2.0. Negative distances indicate downward flame propagation

44



0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

1.8

0 20 40 60 80 100

Time (ms)

Loc

al F

lam

e S

peed

(m

/s)

.

Upward

Downward

Leftward

Rightward

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

1.8

0 20 40 60 80 100

Time (ms)

Loc

al F

lam

e Sp

eed

(m/s

) .

UpwardDownwardLeftwardRightward

Figure 6. Comparison of vertical flame propagation distance from spark electrode as a function of time air for iso-octane-air mixtures at φov = 2.0 for (a) D10 = 0 µm (gaseous) and (b) D10 = 13 µm

a) gaseous, D10 = 0 µm

(b) aerosol, D10 = 13 µm

45



The mechanism of flame instabilities, which caused increased cellularity and insignificance of the buoyancy effect, in aerosol flames is probably related to the heat loss from the flame and local rapid expansion through droplet evaporation. Although droplet evaporation can also cause high gradients in the mixture strength (variations in local equivalence ratio), which might have an effect on the flame, this was negated experimentally [17] using water aerosol in propane-air mixtures. In another study [18] using a rig similar to that of the present work, it was shown that the presence of 30 µm diameter hollow spherical glass beads (no evaporation) in a gaseous iso-octane-air mixture did not alter the smooth characteristics of the flame structure as well as the burning rate. Thus it is evident that the presence of droplets probably plays an important role in the introduction of instabilities due to evaporation.

CONCLUSION

The effects of the presence of near mono-dispersed droplets in flames of very rich iso-octane-air mixture (φov = 2.0) were investigated experimentally in a spherical explosion vessel at near atmospheric conditions. The fuel droplets, which were in the form of aerosols/vapour, were generated by condensation of the gaseous pre-mixture through expansion and this resulted in a homogeneously distributed suspension of near mono-disperse fuel droplets. The effects of droplet size in relation to the structure of the flame surface and to the burning rate were investigated by varying the ignition timing, as the droplet size varied with time during expansion. Observations of the gaseous flame using schlieren photography through the vessel’s windows revealed the buoyancy effect, with distinct differences in flame surface structure and local burning rates between the upper and lower halves of the flame, similar to those described in previous studies. The presence of fine droplets (5 µm) did not cause significant change with respect to the gaseous flame in terms of the buoyancy effect, flame structure and burning rate. However, with larger droplets (13 µm) the flame became fully cellular at a faster rate and more importantly the effect of buoyancy was significantly reduced. The increased

propensity to instability results in the burning rate of aerosol mixtures being faster than those in the gaseous phase at similar conditions. This is so, although the fundamental unstretched laminar burning velocity is probably unchanged by the presence of droplets.

REFERENCES

J. B. Greenberg, “Propagation and Extinction of an Unsteady [1] Spherical Spray Flame Front,” Combust. Theory Modelling, vol. 7, pp. 163-174, 2003

D. R. Ballal and A. H. Lefebvre, “Flame Propagation in [2] Heterogeneous Mixtures of Fuel Droplets, Fuel Vapor and Air,” Proc. Combust. Inst., 1981

G. D. Myers and A. H. Lefebvre, “Propagation in [3] Heterogeneous Mixtures of Fuel Drops and Air,” Combustion and Flame, vol. 66, pp. 193-210, 1986

G. A. Richards and A. H. Lefebvre, “Turbulent Flame Speeds [4] of Hydrocarbon Fuel Droplets in Air,” Combustion and Flame, vol. 78, pp. 299-307, 1989

D. Bradley, A. K. C. Lau, and M. Lawes, “Flame Stretch Rate as [5] a Determinant of Turbulent Burning Velocity,” Phil. Trans. R. Soc. Series A, vol. 338, pp. 359, 1992.

Y. Mizutani and A. Nakajima, “Combustion of Fuel Vapour-[6] Drop-Air Systems: Part I, Open Burner Flames”, Combustion and Flame, vol. 20, pp. 343-350, 1973

Y. Mizutani and A. Nakajima, “Combustion of Fuel Vapour-[7] Drop-Air Systems: Part II, Spherical Flames in a Vessel”, Combustion and Flame, vol. 21, pp. 351-357, 1973

F. Akamatsu, K. Nakabe, Y. Mizutani, M. Katsuki, and T. Tabata, [8] “Structure of Spark-Ignited Spherical Flames Propagating in a Droplet Cloud”, in Developments in Laser Techniques and Applications to Fluid Mechanics, R. J. Adrian, Ed. Berlin: Springer-Verlag, 1996, pp. 212-223

H. F. Coward and F. Brinsley, “Dilution Limits of Inflammability [9] of Gaseous Mixtures”, Journal of Chemical Society Transaction (London), vol. 105, pp. 1859-1885, 1914

O. C. d. C. Ellis, “Flame Movements in Gaseous Mixtures”, [10] Fuel, vol. 7, pp. 195-205, 1928

I. Liebman, J. Corry, and H. E. Perlee, “Dynamics of Flame [11] Propagation through Layered Methane-Air Mixtures”, Combustion Science and Technology, vol. 2, pp. 365, 1971

G. E. Andrews and D. Bradley, “Limits of Flammability [12] and Natural Convection for Methane-Air Mixtures”, 14th Symposium (International) on Combustion, 1973

G. E. Andrews and D. Bradley, “Determination of Burning [13] Velocities, A Critical Review”, Combustion and Flame, vol. 18, pp. 133-153, 1972

D. Bradley, Personal Communications, 2006[14]

46



C. T. R. Wilson, “Condensation of water vapour in the [15] presence of dust-free air and other gases”, in Proceedings of the Royal Society of London, 1897

D. Bradley, P. H. Gaskell, and X. J. Gu, “Burning Velocities, [16] Markstein Lengths and Flame Quenching for Spherical Methane-Air Flames: A Computational Study”, Combustion and Flame, vol. 104, pp. 176-198, 1996

F. Atzler, M. Lawes, S. A. Sulaiman, and R. Woolley, “Effects of [17] Droplets on the Flame speeds of Laminar Iso-Octane and Air Aerosols”, ICLASS 2006, Kyoto, Japan, 2006

F. Atzler, “Fundamental Studies of Aerosol Combustion”, [18] Department of Mechanical Engineering, University of Leeds, 1999

Shaharin Anwar Sulaiman graduated in 1993 with a BSc in Mechnical Engineering from Iowa State University. He earned his MSc in Thermal Power and Fluids Engineering from UMIST in 2000, and PhD in Combustion from the University of Leeds in 2006. During his early years as a graduate, he worked as a Mechanical and Electrical (M&E) Project Engineer in Syarikat

Pembenaan Yeoh Tiong Lay (YTL) for five years. His research interests include combustion, sprays and atomisation, air-conditioning and ventilation, and biomass energy. He joined UTP in 1998 as a tutor. At present he is a Senior Lecturer in the Mechanical Engineering programme and also the Programme Manager for MSc in Asset Management and Maintenance. Certified as a Professional Engineer with the Board of Engineers, Malaysia. He is also a Corporate member of the Institution of Engineers Malaysia.

47



This paper was presented at the 2nd International Conference on Environmental Management, Bangi,

13 -14 September 2004

INTRODUCTION

Malaysia, with a population of over 25 million, generates 16 000 tones of domestic waste daily. At present, the per capita generation of solid waste in Malaysia varies from 0.45 to 1.44 kg/day depending on the economic status of an area. There are now 168 disposal sites but only 7 are sanitary landfills. The rest are open dumps and about 80% of these dumps have filled up to the brim and have to be closed in the near future. The Malaysian government is introducing a new law on solid waste management and also drafting a Strategic Plan for Solid Waste Management in Peninsular Malaysia. The principal processes options available and being recognised as hierarchy for integrated waste management are: waste minimisation, reuse,

material recycling, energy recovery and landfill [1].

Municipal solid waste (MSW) contains an easily biodegradable organic fraction (OF) of up to 40%. Conventional MSW management has been primarily disposal by land filling. Sewage sludge is characterised by high content of organic compounds and this is the cause of its putrescibility. Therefore, sludge before landfill disposal or agricultural application should undergo chemical and hygienic stabilisation. One possible method of stabilisation and hygienisation involves methane fermentation [2].

The anaerobic co-digestion of sewage sludge with organic fraction of municipal solid waste (OFMSW) seems to be especially attractive [3]. The feasibility of

ANAEROBIC CO-DIgESTION OF KITCHEN WASTE AND SEWAgE SLUDgE FOR PRODUCINg BIOgAS

Amirhossein Malakahmad*, Noor Ezlin Ahmad Basri1, Sharom Md Zain1

*Universiti Teknologi PETRONAS, 31750 Tronoh, Perak Darul Ridzuan, Malaysia. 1Universiti Kebangsaan Malaysia, 43600 UKM, Bangi, Selangor, Malaysia.

*[email protected]

ABSTRACT

In this paper, an attempt was made to present the results of some experiments conducted on anaerobic digesters to make comparative study of the biogas generation capacity of the mixture of organic fractions of municipal solid waste from kitchen waste and sewage sludge in different compositions. Batch digestion of samples with various percentage of kitchen waste and sewage sludge was carried out under controlled temperature 35 °C and pH7 conditions for 15 days for each experiment. In all experiments the content of total solid and volatile solid, pH, Kjeldahl nitrogen, chemical oxygen demand, biogas productivity and the content of biogas were measured. The results obtained showed that biogas productivity varied between 4.6 and 59.7 ml depending on the composition of each component in the sample which were added to the digesters. The bioprocess efficiency was observed to be 7.5% - 70.5% for total solid, 51.2% - 81.0% for volatile solid and 8.3% - 43.8% for COD. The overall effluent chemical oxygen demand concentration indicated that it should be treated before using for other applications. From the study results, it is evident that the second bioreactor with 75% of kitchen waste and 25% of sewage sludge produced the maximum quantity of methane gas as compared to other bioreactors.

Keywords: kitchen waste, sewage sludge, biogas production

48



anaerobic co-digestion of waste activated sludge and a simulated OFMSW was examined by Poggi-Varaldo and Olesz-kiewicz [4]. The benefits of co-digestion include: dilution of potential toxic compounds, improved balance of nutrients, synergistic effects of microorganisms, increased load of biodegradable organic matter and better biogas yield. Additional advantages include hygienic stabilisation, increased digestion rate, etc. during methane fermentation the two main processes that occur are:i. Acidogenic digestion with the production of

volatile fatty acid; andii. The volatile fatty acids are converted into CH4 and

CO2.

In batch operation the digester is filled completely with organic matter and seed inoculums, sealed, and the process of decomposition is allowed to proceed for a long time until gas production is decreased to a low rate (duration of process varies based on regional variation of temperature, type of substrate, etc.). Then it is unloaded, leaving 10-20 percent as seed, then reloaded and the operation continues. In this type of operation the gas production is expected to be unsteady and the production rate is expected to vary from high to low. Digestion failures due to shock load are not uncommon. This mode of operation, however, is suitable for handling large quantities of organic matter in remote areas. It may need separate gasholders if a steady supply of gas is desired [5]. Callaghan et al. worked on co-digestion of waste organic solids which gave high cumulative production of methane [6].

ExPERIMENTAL

Batch digestion of samples was carried out under controlled temperature 35 °C and pH7 conditions for 15 days for each experiment. All five samples were fed into a 1 L bioreactor operated under mesophilic condition.

In the first experiment, 100% kitchen waste was used while the second experiment was conducted with the mixture of kitchen waste (75%) and sewage sludge (25%). The third experiment was conducted with 50%

kitchen waste and 50% sewage sludge, in the fourth run, mixture of 25% kitchen waste and 75% sewage sludge was used. The fifth experiment was conducted with only sewage sludge.

In all experiments total solid, volatile solid, pH, Kjeldahl nitrogen and chemical oxygen demand for initial and final properties of samples were determined. Biogas productivity and the content of biogas were also measured. All analytical procedures were performed in accordance with Standard Methods [7].


i. Variation in pH value throughout the experiments

As shown in Figure 1, from the graph plotted, the pH variation could be categorised into 3 main zones. The first zone started from first day till fourth day, which showed a drastic drop of the pH. This is due to the high development rate of volatile fatty acids by microorganisms. The pH is maintained at neutral with the addition of sodium hydroxide solution. The second zone started from the fifth till the twelfth day of experiment. In the second zone, the pH was in the range of 6.9 to 7.3. This is due to the development of CO3HNH4 from CO2 and NH3, which were produced during the anaerobic process.

The percentage of CO3HNH4 had caused the increase of alkalinity of the samples. Due to this, any differences

2

3

4

5

6

7

8

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

Sample 1

Sample 2

Sample 3

Sample 4

Sample 5

pH

Time (day)

Figure 1

49



in the volatile fatty acid content did not affect the pH value. The third zone started on the thirteenth till the last day of the experiment. In this zone, it was found that the pH value of the samples started to increase. This is due to the development of CO3HNH4 that still continues, but no more volatile fatty acid was produced.

ii. Biogas production

• The production of cumulative biogas

Figure 2 shows the production of cumulative biogas for all the samples. It was found that the second sample with the composition of 75% kitchen waste and 25% activated sludge produced the highest quantity of biogas, which was 59.7 ml. This was followed by the first sample (100% kitchen waste), then fourth sample (25% kitchen waste and 75% activated sludge), then the third sample (50% kitchen waste and 50% activated sludge), and lastly the fifth sample (100% activated sludge). The productions of the biogas of the respective samples were 47.1 ml, 22.3 ml, 8.4 ml and 4.6 ml. The fifth sample produced the least biogas; this is consistent with the literature data by Cecchi et al., (1998) and Hamzawi et al., (1998) that showed the production of the cumulative biogas is high when organic components that are easily biodegradable in the sample are higher. According to Schmidell (1986), the anaerobic digestion process for MSW alone is possible but will produce less biogas as compared to a mixture of MSW and activated sludge. This is due to the production of volatile fatty acids by

microorganisms is more likely to accumulate rather than to release biogas.

An increment of 5% of activated sludge is enough to reduce the accumulation of volatile fatty acid and release more biogas. From Figure 2 for the first and second samples, the results comply with the Schmdell theory. Samples three and four that had different composition of kitchen waste and activated sludge produced less biogas due to the composition of activated sludge that has unsuitable C:N ratio for the anaerobic digestion process.

• Thebiogasproductionrate

The rate of biogas production for every sample is shown in Figure 3. It was found that the production of biogas for the first sample was the slowest that started from the fourth day of experiment and reached the highest quantity on the eighth day. On the other hand the fifth sample started to produce biogas the earliest on the second day and reached the highest amount on the fifth day. The production rate of biogas for samples two and three occurred on the seventh day of the experiment while the fourth sample occurred on the sixth day. Therefore, it can be concluded that the results obtained are consistent with the research done by Cecchi et al. [8], which stated that the production of biogas is slower for high organic loading as compared to a lower organic loading.

Table 1 summarises the amount of total suspended solid (SS), volatile suspended solid (VSS), alkalinity,

0

10

20

30

40

50

60

70

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

Time (day)

Pro

duct

ion

of c

umul

ativ

e bi

ogas

(ml) Sample 1

Sample 2

Sample 3

Sample 4

Sample 5

0

5

10

15

20

25

30

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

Time (day)

Bio

gas

prod

ucti

on r

ate

(ml/d

ay)

Sample 1

Sample 2

Sample 3

Sample 4

Sample 5

Figure 2 Figure 3

50



kejldahl nitrogen, pH, and COD before and after treatment in all five experiments. According to the results, the bioprocess efficiency was observed to be 7.5% - 70.5% for total solid, 51.2% - 81% for volatile solid and 8.3% - 43.8% for COD. The first bioreactor is most efficient in treating the volatile solid component, which achieved efficiency of 81.0%.

CONCLUSIONS

Five experiments were conducted under mesophilic conditions in batch bioreactor for 75 days. Five different kinds of feedstock were loaded into the reactor. It was found that the cumulative biogas production increased, when the mixture kitchen waste and activated sludge was used.

However, the highest value of methane production was for sample 2 (75% kitchen waste and 25% activated sludge), which produced 59.7 ml. For the rate of biogas production the situation was the same and the best result was for sample 2, after the samples 1, 4, and 3 were settled respectively. The 5th sample produced the least biogas. The anaerobic co-digestion of kitchen waste and activated sludge were demonstrated to be an attractive method for environmental protection and energy savings, but it is clear that applying better equipment and adjustment of conditions could give more reasonable results.

REFERENCE

Mageswari, S., “GIAI global meeting”. Penang, Malaysia. 17-[1] 21 March 2003

Sosnowski, P., Wieczorek, A., Ledakowicz, S., 2003. “Anaerobic [2] co-digestion of sewage sludge and organic fraction of municipal solid wastes”. Advance in environmental researches 7, 609-616

Hamzawi, N., Kennedy, K.J., Mc Lean, D.D., 1998. “Technical [3] feasibility of anaerobic co-digestion of sewage sludge and municipal solid waste”. Environ. Technol. 19, 993-1003

Poggi-Varaldo, H. M., Olesz-kiewicz, J. A., 1992. “Anaerobic [4] co-composting of municipal solid waste and waste sludge at high total solid level”. Environ technol. 13, 409-421

Table 1. The value of measured parameters before and after treatment

ParameterBefore treatment

After treatment

Sample I

SS (%) 5.00 3.29

VSS (%) 42.14 8.00

N (mg/L N-NH3) 0.40 0.55

pH 7.00 7.36

COD (mg/L) 4864 3520

Sample II

SS (%) 5.10 4.72

VSS (%) 38.30 7.50

N (mg/L N-NH3) 0.29 0.47

pH 7.00 7.31

COD (mg/L) 4000 3600

Sample III

SS (%) 6.44 1.90

VSS (%) 32.55 8.80

N (mg/L N-NH3) 0.24 0.45

pH 7.01 7.27

COD (mg/L) 2800 2560

Sample IV

SS (%) 2.45 1.40

VSS (%) 27.52 6.69

N (mg/L N-NH3) 0.23 0.30

pH 7.00 7.50

COD (mg/L) 2400 2200

Sample V

SS (%) 0.27 0.19

VSS (%) 0.41 0.20

N (mg/L N-NH3) 0.02 0.10

pH 7.00 7.59

COD (mg/L) 320 180

51



Polprasert, C., 1996. “Organic waste recycling, technology [5] and management”. Second edition. Chichester: John Wiley and sons

Callaghan F. J., Wase, D. A. J., Thayanithy, K., Forster, C. F., [6] 1998. “Co-digestion of waste organic solids: batch study”. Bioresearch technology 67, 117-122

Standard methods for water and wastewater treatment [7] 18th edition (1992)

Cecchi, F., Pavan, P., Musacco, A., Mata-Alvarez, J., Sans, C., [8] Ballin, E., 1992. “Comparison between thermophilic and mesophilic digestion of sewage sludge coming from urban wastewater plants”. Ingegneria Sanitaria Ambientale 40, 25-32

Amirhossein Malakahmad is an academic staff at Universiti Teknologi PETRONAS. He graduated with BEng in Chemical Engineering in 1999 from Islamic Azad University, Tehran, Iran. He completed his MSc in 2002 in Environmental Engineering from the same university. In 2006, he received his PhD from the National University of Malaysia, UKM for his research

on an application of zero-waste anaerobic baffled reactor to produce biogas from solid waste. He joined Universiti Teknologi PETRONAS in August 2007. His research interests are in water and wastewater treatment and solid waste engineering.

52



This paper was published in the Journal of Loss Prevention in the Process Industries,

doi:10.1016/j.jlp.2007.11.007 /(2008)

ON-LINE AT-RISK BEHAVIOUR ANALYSIS AND IMPROVEMENT SYSTEM (E-ARBAIS)

Azmi Mohd Shariff* and Tan Sew KengUniversiti Teknologi PETRONAS, 31750 Tronoh, Perak Darul Ridzuan, Malaysia.

*[email protected]

ABSTRACT

Behaviour Based Safety (BBS) is a programme that has been implemented in many organisations to identify at-risk behaviour and to reduce injury rate of their workers. The effectiveness of the BBS was proven with many companies recorded high percentage of reduction of injury rate especially during the first year of implementation. However, the BBS process could be very labour intensive. It requires many observers to make the process effective. Very much effort was required to train the employees to become the observers. Many organisations which attempted to obtain the benefits of BBS did not sustain comprehensive participation required in BBS related activities. With this drawback, it calls for a simplified process that could achieve the same result as BBS. This study was intended to establish an alternative to the BBS, termed as On-line At-Risk Behaviour Analysis and Improvement System (e-ARBAIS). The e-ARBAIS utilises computer technology to play a role in making the routine observation process more sustainable and hence instilling the habitual awareness through the cognitive psychology effect. A database was set up with the pre-programmed questions regarding at-risk behaviours in the organisation. The employees then utilised the database to feedback their observations. Through this process, the traditionally tedious observations by trained observers as required in BBS were now done naturally by all respondents. From the collective feedback, at-risk behaviours can be easily identified in the organisation. The HSE committee within the organisation can thus, take the appropriate action by reinforcing the safety regulations or safe practices to correct all the unsafe behaviours either by changing the design (“Hard-ware”), system (“Soft-ware”) or training the employee (“Human-ware”). This paper introduces the concept, framework and methodology of e-ARBAIS. A case study was conducted in X Company (not a true name as the permission to use their real name was not given). A prototype computer program based on e-ARBAIS was developed using Microsoft Excel named as “1-min Observation” programme. A preliminary result based on one-month data collection is highlighted in this paper. Based on this preliminary result, “1-min Observation” programme has received positive feedback from the management and employees of Company X. It was done with very small resources and thus saved time and money compared to traditional BBS technique. The e-ARBAIS concept is workable and practical since it is easy to implement, collect data and correct unsafe behaviour. Some recommendations by the employees of Company X were presented in this paper to further improve the “1-min Observation” programme. The project at Company X is still in progress in order to see a long term impact of this programme. The e-ARBAIS has shown its potential to reduce injury in the organisation if implemented with a thorough plan and strong commitment from all levels.

Keywords: behaviour based safety, at-risk behaviour, human factor, safety programme, injury rate reduction

53



INTRODUCTION

Behaviour based safety (BBS) was first established by B. F. Skinner in the 1930s (Skinner, 1938). He was a psychologist who developed a systematic approach called behaviour analysis to increase safe behaviours, reduce risky behaviours and prevent accidental injury at work and on the road. This approach was later known as applied behaviour analysis (Hayes, 2000). Behaviour study was important because H. W. Heinrich, a workplace safety pioneer, reported that out of 550 000 accidents, only 10% was caused by unsafe working conditions, another 88% was caused by worker’s unsafe actions (SCF, 2004). A “Workplace Attitude Study” conducted by Missouri Employers Mutual Insurance (MEM) which was published in Occupational Hazards (September 2003) revealed that 64.1% of Americans thought that a workplace accident would never happen to them. 53.4% believed that the probability was very low for a work injury that could cause them to become permanently disabled (SCF, 2004). This showed that people generally perceived that there was a low risk of injury possibility in a workplace. This showed that accidents could happen if workers continue to work with at-risk behaviour and perceived it was safe to do so. The human toll of unsafe behaviour was high. According to the U.S. Bureau of Labour Statistics, unintentional injury was the leading cause of death to people aged 44 and under. In 2001, private industry had more than 5.2 million non-fatal accidents and injuries, with more than 5 000 fatal injuries. Behaviour-based safety programmes that target and document behaviour changes indeed save lives, money and productivity (APA, 2003).

Behaviour is an “upstream” approach to safety. It focuses on the “at-risk behaviour” that might produce an accident or near miss rather than trying to correct a problem after an accident or occurrence. The behaviour-based aim then, is to change the mindset of an employee by hopefully making safety a priority in the employee’s mind (Schatz, 2003).

However, it was noted to many that not all organisations had successful experience in implementing the BBS as the others did (Geller, 2002). Over years, some safety

professionals had started to develop alternatives to the BBS programme, i.e. people-based safety, ProAct Safety and Value-based Safety. It was desirable to develop another alternative to the BBS programme via the help of computer technology.

THE E-ARBAIS CONCEPT

The e-ARBAIS programme is meant to provide alternative solutions to certain limitations of BBS as mentioned below.

a. Prevent coyness in direct feedback with computer interface

Problem arises when employees dare not approach the peers to give feedback directly (Gilmore et al., 2002). The e-ARBAIS programme provided another channel for the peers to communicate. The peers may now give feedback on their observations to the database and publish the feedback via the computer. This helps to reduce the problem of coyness through direct feedback with peers.

b. Inculcate safety culture with e-ARBAIS

The e-ARBAIS utilises computer software to prompt the employees if they observe any unsafe behaviour relating to the topic asked. For instance, the e-ARBAIS database could have a question like, “Did you see anybody drive faster than 20 km/h today?” The employees would be reminded to observe occurrence around them naturally without bringing the checklist. The e-ARBAIS questions that would be prompted regularly may also serve as the checklist in the ordinary BBS process. However, instead of focusing on many items during one observation, the questions would require the employees to respond to certain particular areas in a day. Different topics would be asked everyday. This will eventually instil a psychological effect in which people are “reminded” on the safety rules and regulations. The cultivating of habitual awareness is always the heart of designing the e-ARBAIS. Only when it becomes habitual, the safety culture could be inculcated.

54



As employees perform observations, they come to recognise any discrepancies between their own behaviour and what is considered safe, and they begin to adopt safe practices more consistently. McSween said “We have created a process where they raise their personal standards” (Minter, 2004). This is the objective of this whole e-ARBAIS – to inculcate the safety culture in organisations.

c. Tackling slow data collecting and analysis with IT

Questions on observations are then being repeated randomly. All the feedbacks are collected in the database and would be analysed by the HSE committee on a regular basis. When the HSE committee analyse the data, they would be able to identify the high risk issues faced by the employees. For example, if the speeding behaviour of the employees remains

high in the statistics after a few times being asked through the software, it implies that many people are speeding in the plant and refuse to change their behaviour. From there, the HSE committee could provide some recommendations such as building a few road humps in the plant, doing spot checks and issuing warnings to those who speed. The action item could also be derived from an established method such as ABC analysis or ABC model (Minshall, 1977). The data analysed would highlight the areas in which the HSE committee needs to focus on and to provide the solution for improvement. This would also help the committee to identify if the unsafe behaviours are due to• "Hard-ware" problem like inadequate structural

safety design,• "Soft-ware" problem like poor system

implementation or obsolete operating procedure, or

Figure 1. Framework of e-ARBAIS

ABC Analysis (optional)

Set up of e-ARBAIS Database

Observations of At-Risk Behavior at Workplace

Review by HSE Committee

Action Plan to Correct At-Risk Behavior

Analysis of At-Risk Behavior

On-line Feedback to Database

Interview (optional)

Feedback to peers (optional)

Briefing to End Users on e-ARBAIS

55



Table 1. The comparison of ordinary BBS and e-ARBAIS

Element Ordinary BBS e-ARBAIS

Training Training given to observers on how to define unsafe behaviour and how to provide feedback.

No training needed. All will participate in the observations. Only briefing on what is e-ARBAIS and how it works.

Checklist Checklist must be used to go through all items and see which does not comply.

No checklist is used. Checklist is built in the database as pre-programmed questions.

Observation frequency

Observers are required to make certain observations in a certain period, i.e. 1 observation a week.

Observation is done on daily basis or flexible adjustment to frequency can be made.

Cost Additional cost for training and printing checklist.

Minimum cost since training and checklist are not required.

Feedback Feedback is given directly when observation completes, whether it is positive or negative. Results of observation need to be reported to HSE committee for further analysis.

Feedback can be given either by face to face or through the database to prevent “sick feeling” with peers. Feedback is displayed directly in the database. HSE committee uses the same set of data for further action.

Communication The analysed results normally can only be accessed by the HSE committee. Employees generally not being communicated on the overall performance of the observation.

Feedback is recorded in database which everyone could access to see what unsafe behaviour is observed.

Involvement Only those who are trained will be involved in the observations. To involve all, much training is required.

All will be involved in the observation since the observation is done naturally without focusing on only one activity.

Management commitment

Management commitment determines if the process will be successful. Most BBS fail due to poor management commitment.

Database displays the management participation and thus motivates management to further commit and improve the programme.

• "Human-ware" problem like employees’ riskybehaviour.

d. Reduce intensive labour and high cost with e-ARBAIS 24 hour functionality

The advantage of e-ARBAIS is that the observation can be done 24 hours a day and 7 days a week with minimum man-hour needed. It is all operated by the computer. Thus, resulting in higher efficiency and cost saving.

The differences between e-ARBAIS to the ordinary BBS programme are given in Table 1.

FRAMEWORK OF E-ARBAIS

The framework of e-ARBAIS is given in Figure 1.

To start the e-ARBAIS program, a database first needs to be set up. A systematic approach can be established to develop the questions in the database. The database consists of pre-programmed questions that can be based on the past incident records in the organization to focus on the problematic area, or it can be the near miss cases. It can also be the questions that purely derived from the BBS checklist alone. Basically, to effectively use the e-ARBAIS program, the questions must be custom designed for each company. The discussion needs to be done with HSE committee on

56



the development and selection of the questions and agreed by the management of the company.

After that, the end users needed to be briefed about the e-ARBAIS and how the program will be implemented. The end users were not required to observe any specific activities to ensure that e-ARBAIS is naturally done according to the pre-programmed questions in the database. The end users only needed

to be more aware of at-risk behaviour observation by their colleagues when they are doing daily work. At any occasion, they could also give their feedback directly to the peers. To those who are afraid that their feedback could cause ill-feeling, they have the option of giving their feedback to the database. The database would calculate the analysis automatically and publish the result online based on the feedback received. The employees are indirectly reminded by the online analysis and feedback on the unsafe behaviours from this exercise and this could change their behaviour eventually to avoid similar observations in the future. This could be achieved through the cognitive psychology effect.

With the analysis done by the database, the HSE committee could use those data and discuss them in their regular meeting directly to correct the at-risk behaviours that frequently occur. The committee could apply the ABC analysis (Minshall, 1977) to understand the behaviours. Optionally, the HSE committee could also conduct some interviews with the identified employees to further understand why they take risk when performing their tasks. With that, the necessary action plan could be established to rectify the at-risk behaviours that are contributed by factors such as “hard-ware”, “soft-ware” or “human-ware” mentioned above.

The questions in the pre-programme database need to re-visit time to time and should be updated with any new potential at-risk behaviour as necessary.

CASE STUDY

A case study using e-ARBAIS concept was implemented in Company X for a month. The program was termed as “1-min Observation”. One of the important tools of this study was to use IT (information technology) to help on the observation processes. The database was developed on Microsoft Excel spreadsheet. One master list that consists of 16 questions was generated in the database based on the previous frequent at-risk behaviour identified and reported as shown in Table 2. The case study was run for a month and each question was repeated four times to observe any

Table 2. The questions in “1-min Observation” database

1. Did you see people NOT using handrail when travelling up and down stairs?

2. Did you see employee driving faster than 20km/h in the plant area?

3. Did you experience back pain after work?

4. Did you see any reckless forklift driver?

5. Did you see people wearing glasses instead of safety glasses in the plant?

6. Did you see people lifting thing in improper position?

7. Did you see people NOT wearing ear plug in noisy area?

8. Did you see anyone working at height with falling hazard due to improper PPE/position?

9. Did you see people working (in the plant/lab/packaging) NOT wearing safety glass?

10. Did you see any leak in the plant but NOT barricaded?

11. Did you see any area (office or plant) unclean and expose to tripping hazard?

12. Did you see people using hand phone at restricted area?

13. Did you see people NOT looking in the direction that they are walking (eyes NOT on path)?

14. Did you see people NOT following 100% of the procedure when doing work?

15. Did you see people working without hand glove in the plant area?

16. Did you see people walking/working inside the maintenance workshop yellow line area without minimum PPE required?

57



Figure 2. The flowchart for “1-min Observation” programme

No

Yes

Yes

Database would show today's date and search for questions which matched with today's date.

Database would prompt the 2 questions which matched with today's date into the front page

User to select "yes" or "no" for both questions and department name

User clicked "submit".

Data was captured in database. Data collected consist of date, user ID, answer for 1st question,

answer for 2nd question, and department.

Database would count the

participation in a day based on

departments. The data was

transferred as chart and would be

shown in the page of "participation".

Database would count the percentage of total unsafe

behavior observed in a day. The data was transferred to

chart and would be shown in the page of "unsafe

behavior".

ionsparticipat

behaviour unsafebehaviour unsafe total

∑∑

=

Database would display yesterday responses for both

questions and would show the percentage of

unsafe behavior. The statistics was shown in the page

of "statistics".

New data was submitted?

Start

User open database

End

58



Figure 3. The main page of the “1-min Observation” program

Figure 4. The data sharing page with all employees of Company X

59



possible trend in the data collected. At the front page of the on-line database, only two questions would be prompted. Two questions per day were adopted after the consideration of the human factor. The program was intended to let people feel that it was so simple to participate and “why not take part?” The HSE Committee felt that three questions may cause irritation and one question was just too little and not efficient in data collection. After much consideration, two questions were considered the most appropriate. The questions consisted of areas like Personal Protection Equipment (PPE) usage, ergonomics, safety rules, safe practice and housekeeping. The questions were custom designed to suit Company X interests to reduce at-risk behaviour of their employees. The flowchart of the “1-min Observation” program is shown in the Figure 2. The main page of the “1-min Observation” program and the data sharing page with all the employees are given in Figure 3 and Figure 4 respectively.

Participation

The participation from the employees at the early stage of launching was not good. This was primarily due to the unfamiliarity of the program and the routine of going into the web page for the “1-min Observation” file everyday. After much explanation and encouragement by the safety manager to

the employees, the response started to increase. The responses of the employees from respective departments are shown in Figure 5.

The responses were expected to be low during weekends and public holidays. Figure 6 shows the responses received from all the employees and the responses were on the lower side during weekends. The trend showed that there was improvement in their participation over time. However, there were some occasions when feedback were lower due to some visitors’ plant visit or corrupted master file. The master file which was compiled using Microsoft Excel was easily corrupted due to multiple sharing with many people and the huge size of the file. The problem was then fixed by using a standby master file, consistently backed up with double password protection. Also, the file size was then reduced by removing some unnecessary decorative pictures in the file.

Based on the record from the Human Resource Department, the daily attendance of the employees was used to compare against the participation rate. Figure 7 shows the percent of participation relative to the attendance. The highest participation received was 86% whereas sometimes it went below 10%. This depended heavily on the plant activities. If the plant was experiencing some problems then the response

Figure 5. The responses received from all the respective departments.

60



would be lower as most of the employees were tied up on the rectification of plant problems.

Data Analysis

The feedback from the employees were analysed by the database. Each question was prompted four times. The calculation is shown below.

%100Responses Total

Observed Behaviours UnsafeTotalBehaviours UnsafeofPercent ×=

where

4 Time Observed Behaviours Unsafe

3 Time Observed Behaviours Unsafe2 Time Observed Behaviours Unsafe

1 Time Observed Behaviours UnsafeObserved Behaviours UnsafeTotal

+++=

4 Time responses ofNumber 3 Time responses ofNumber

2 Time responses of Number 1 Time responses ofNumber Responses Total

+++=

From this, the percentages of unsafe behaviours were sorted accordingly. However, it was noted that even the topmost unsafe behaviour was only 35% of the total response as shown in Figure 8. Generally, most of the respondents were practising the safe behaviour.

Figure 6. The responses received from the employees.

Figure 7. The percentage of participation based on daily attendance.

61



From Figure 9, the highest unsafe behaviour observed was the usage of handphones at restricted areas, which contributed to 35% of the responses. There was no occasion whereby anyone was observed lifting goods with improper position.

DISCUSSION

The employees of Company X gave good support and responses to the “1-min Observation” program. Using the e-ARBAIS program, it was rather easy to identify at-risk behaviours that needed improvement. It did not involve many additional resources to gather the useful data. As this is a preliminary result, the program was considered quite successful. A longer time is needed to see a long–term impact of e-ARBAIS.

Figure 8. The percentage of unsafe behaviours observed based on the total feedback

Figure 9. The top five unsafe behaviours observed

62



Some challenges in implementation of e-ARBAIS in Company X and its limitations are shared below. The Challenges to Implement the e-ARBAIS in an Organisation

a. Ensure clear communication

As e-ARBAIS was a new concept, it was very important to communicate clearly to the employees about the implementation. During the case study, unclear communication and explanation from safety department were some of the feedback quoted from the employees.

b. Ensure management commitment

Management commitment was undeniably an important role. Management commitment on the participation would lead others to follow. Consistent management commitment from each level was imperative.

c. Ensure follow up action

The e-ARBAIS program could be more effective if the employees could see the follow-up action by the Safety Department or HSE Committee. Employees who participated in the program would be eager to report their observations and wanted to see the changes. Thus, if the HSE committee was not able to take appropriate action to make the changes, employees would begin to feel disappointed with the management as there was no follow up action. Eventually, the program may cease. There was no motivation that could continue to thrill the employee to participate in the program.

d. Honest Participation

The data collected would be very useful if everybody participated and responded honestly. However, there was a possibility that people did not answer based on the observations. Generally, this should not happen and the overall analysis would be useful.

The Limitation of “1-min Observation” Program

As the case study was rather short, it was difficult to measure if there was any improvement in the safety behaviour in the long term.

Also, the “1-min Observation” programme was an IT based programme. It was thus vital that the database worked appropriately. During the case study, the database was created using Microsoft Excel and the file was corrupted several times and interrupted the program. Additionally, the file was shared among 79 employees and only one was accessible at one time. A lot of time was wasted while waiting. Some of them gave up when they could not open the file on a particular day. The file malfunctions were occasionally due to the huge size of the database. Improvement on this was required to make the programme more successful.

The questions in the database were developed based on the recommendations by the company and were focused on the observation of unsafe act. However, one of the questions, “Did you experience back pain after work?” is the result of unsafe acts and might not be suitable to be included in the database for the observation purposes. However, it shows the flexibility of the e-ABAIS that can also be extended to cover general safety issues resulting from an unsafe act.

Some of the questions may not be directly understood by the employees. For instance, “Did you see people lift things in improper positions?” The employees might not fully understand what “improper position” means and therefore inaccurate responses might be given. The refinement of these questions is required to ensure accurate feedback from the observers.

Additionally, there was a possibility that one unsafe act was observed by multiple observers which created a false statistic in the database. In this case, a “feedback” column that enables the observer to write and describe the observation would definitely help to minimise the problem.

63



CONCLUSION

In conclusion, the concept of e-ARBAIS was to serve as another alternative to the current BBS programme. It required fewer resources, more involvement, and low cost which all added on to the sustainability of the program. The e-ARBAIS was easy to implement and also in collecting data. It also emphasised and reminded employees on conducting task with the correct behaviour. This was intended to provide a psychological effect to the employees on the safe behaviour and inculcate the habitual awareness. Thus, this can encourage a safety culture in an organization.

The case study of implementing e-ARBAIS in Company X, which was named as “1-min Observation” had received positive support and feedback. There were some constraints in fully implementing the “1-min Observation” such as, a well designed database, effective communication between safety department with the employees and efficient follow up action on the data analysed. All these would add up to the success of the case study. It can be further fine tuned and used in many organisations. Some recommendations were given below. The programme

could be more effective if the recommendations were considered and adopted.

Overall, the e-ARBAIS concept was feasible and practical. Given a longer time and with the implementation stage improved, the e-ARBAIS would definitely benefit the organisation.

RECOMMENDATIONS

a. Appropriate planning for the e-ARBAIS programme

Most of the employees would like to be informed about the analysis after their participation. They wanted to know more about the findings and what were the unsafe behaviours that were observed most frequently. Therefore, more proper planning from the safety department was required. The HSE committee should take immediate action once the employees had completed the feedback. The timely analysis should be shared with everybody. Also, appropriate action must be taken to show to the employees that their feedback was valuable. No one would like to waste their time if they knew nothing was going to happen with their feedback. It was very important to note that the sustainability of the program depended on the confidence level of the employees on the program.

The flow chart in Figure 10 shows the appropriate cycle in the e-ARBAIS program.

b. Improvement on database and feedback column

As mentioned earlier, one of the weaknesses during the case study was the frequent corruptions and malfunctions of the database.

To overcome this problem, a more stable programme should be used. A web-based database is much more user-friendly in this case.

Additionally, it would be value enhancing if there were some feedback columns on top of the questions

Figure 10. The flow chart of how the e-ARBAIS programme should be implemented

Feedback

Execute Action Item

Share data & action

Team Review

Update e-ARBAIS

Get ready new questions

64



posted so that the respondents could more accurately describe the problems and give feedback to the safety department.

c. Sustainability of the program

It is important to ensure that the program is sustainable. One of the recommendations was to give rewards to the employees for their feedback given to the program. Rewards may help to encourage participation and continuous feedback. In the long term, the safety program would sustain and the unsafe behaviour of the employees could be improved.

ACKNOWLEDgEMENTThe authors would like to thank Tuan Haji Mashal Ahmad and his staff on their valuable contribution in this work.

REFERENCES

American Psychological Association (APA) (2003). [1] “Behaviour analyses help people work safer”, Washington. www.apa.org.

Geller, E. S. (2002). “How to get people involved in Behaviour-[2] Based Safety – selling an effective process”, Cambridge, MA: Cambridge Center for Behavioural Studies.

Gilmore, Michael R., Perdue, Sherry R., Wu, Peter (2002). [3] “Behaviour Based Safety: The next step in injury prevention”. SPE International on Conference on Health, Safety & Environment in Oil and Gas Exploration and Production, Kuala Lumpur, MAL, 20-22 Mar 2002.

Hayes, S. C. (2000). “The greatest dangers facing behaviour [4] analysis today”. The Behaviour Analyst Today, Volume 2, Issue Number 2. Cambridge Center for Behavioural Studies.

Minshall, S (1997). “An opportunity to get ahead of the [5] accident curve”. Mine safety and Health News, Vol 4, No. 9.

Minter, S. G. (2004). “Love the process, hate the name, [6] Occupational Hazards”, 3rd June 2004.

SCF Arizona Loss Control (2004). “Behavioural Safety: The [7] right choice – Listen to your conscience and eliminate dangerous behaviours”, SCF Arizona, 2nd July 2004, http://www.scfaz.com/publish/printer_549.shtml

Schatz, J. R. (2003). “Behaviour-Based Safety: An introductory [8] look at Behaviour-Based Safety”, Air Mobility Command’s Magazine, Jan/Feb 2003.

Skinner, B. F. (1938). “The behaviour of organisms: An [9] experimental analysis”. Acton, Mass.: Copley Publishing Group.

Azmi Mohd Shariff received his MSc in Process Integration from UMIST, United Kingdom in 1992. He furthered his studies at University of Leeds, United Kingdom and received his PhD in Chemical Engineering in 1995. He joined Universiti Kebangsaan Malaysia in 1989 as a tutor upon his return from Leeds and was appointed as lecturer in 1996.

He joined Universiti Teknologi PETRONAS (UTP) in 1997 and was appointed as the Head of Industrial Internship in 1998. He was later appointed as the Head of Chemical Engineering Programme 1999–2003. He is currently an Associate Professor in the Department of Chemical Engineering and a Leader of the Process Safety Research Group.

He teaches Process Safety and Loss Prevention at undergraduate level and Chemical Process Safety at graduate level. He started research work in the area of Process Safety in 2002. He has successfully supervised and is currently supervising a few PhD, MSc and final year undergraduate students in the area of Quantitative Risk Assessment (QRA), Inherent Safety and Behaviour Based Safety. He has presented and published more than 15 articles relating to process safety in conferences and journals. He currently leads an E-Science research project under Ministry of Science and Technology entitled ‘Development of Inherent Safety Software for Process Plant Design’. He has also experience in conducting QRA for a few local companies in Malaysia. Recently, he was successful in conducting a short-course on ‘Applied QRA in Process Industry’ to OPU and Non-OPU staff.

65



This paper was presented at the International Conference on Science & Technology: Application in Industry & Education 2008 (ICSTIE 2008), Penang,

12 - 13 December 2008

INTRODUCTION

During construction, proof tests to verify pile design are conducted. Specification usually calls for the amount of absolute and permanent pile displacement during a proof test to be less than a specified amount. The number of piles subjected to proof tests is proportional to the total number of piles being constructed and the number of piles that ‘failed’ relative to the number of proof tests should not exceed a certain prescribed number. Load applied for proof test is usually twice the design load at constant loading rate using one or two load cycles.

This paper attempts to interpret pile proof test for obtaining soil parameters (soil and rock unit skin resistance as well as base resistance). The probabilistic inverse analysis method as given in [1] was used to

obtain actual soil parameters in the form of its joint probability density. Parameters obtained were then utilised to generate histograms of ultimate pile load capacity using Monte Carlo simulation technique. The arrangement in this paper are as follows: Section 2 outlines geotechnical aspects of the drilled shaft particularly its design methodology and interpretation of pile-load-test results at project site near Kuala Lumpur. The soil condition and pile-load-test results at project site near Kuala Lumpur are explained in Section 3. In Section 4, the salient features of the probabilistic inverse method are given, followed by its application to interpret pile-load-test results in Section 5. Section 6 concludes results from this work.

BAYESIAN INVERSION OF PROOF PILE TEST: MONTE CARLO SIMULATION APPROACH

I. S. H. Harahap*, C. W. Wong1

*Universiti Teknologi PETRONAS, 31750 Tronoh, Perak Darul Ridzuan, Malaysia. 1Malaysia LNG Sdn. Bhd., Malaysia.

ABSTRACT

Pile load test is commonly conducted during both design and construction stages. The objective of pile load test during design stage is to obtain the actual soil parameters and ultimate load in-situ. On the other hand, for the test conducted during construction stage the objective is to prove that the actual pile capacity conforms to the design loads. This paper presents probabilistic interpretation of proof pile test to obtain the ultimate pile capacity. As the first step, the “actual” field parameters are back calculated using ultimate pile capacity from proof pile load tests. The probabilistic inverse method is used for back calculation of parameters. Soil parameters obtained from back calculation are then sampled using Monte Carlo simulation technique to generate histogram of ultimate pile capacity. From the synthetic histogram, other statistical metrics such as mean, standard deviation and cumulative probability density of ultimate pile capacity can be obtained.

Keywords: Socketed drilled shaft, Monte Carlo simulation, probabilistic inverse analysis, proof pile load test

66



gEOTECHNICAL ASPECTS

Design of Socketed Drilled Shaft

The ultimate capacity of socketed drilled shaft can be determined using the following equation:

Qu + Qfu + Qbu

Qu = (fSCS + FRCR) + qbAb (1)

where Qu is ultimate pile capacity, Qfu is ultimate shaft capacity, Qbu is ultimate base capacity. The ultimate skin resistance consist of contribution from soil part (fS) and rock part (fR) where fS and fR are unit shaft resistance, Cs and CR are circumferential area of pile embedded in each layer (soil and rock). qb is unit base resistance for the bearing layer (rock) and Ab pile base area. Figure 1 shows components of ultimate capacity of drilled shaft socketed into rock.

The unit skin resistance of cohesionless material usually has the form of fR = KSσO tanø where KS is coefficient of lateral pressure, σO is vertical overburden pressure and ø is friction angle [2]. For cohesive material the unit skin resistance is commonly taken as proportional to the undrained shear strength as fR = αSu where α is proportional coefficient and Su is undrained shear strength [3]. For rock, the unit skin resistance is empirically determined from rock unconfined compressive strength, qu [4], or Rock Quality Designation, RQD. Table 1 shows available empirical correlation to determine fR from qu and

Table 2 to determine fR from RQD. However, the unit skin resistance has a limiting value depends on the unconfined compressive strength of the rock [5]. The empirical correlations proposed is either has linear relation with qu or power-curve relation to qu. Evaluation by [6] indicated that the SPT N-value may not a good indicator of fR due to its sampling rate because it is too infrequent and suffers too much variability. From evaluation of pile load test results [7], there is a significant difference of load-settlement behaviour among sedimentary, granitic and decomposed rock, and hence the range of qu for these rocks. Generally granitic rock has a softer response compared to sedimentary rock. The ultimate unit resistance range between 6 to 50 MPa for granitic rock compared to between 1 to 16 MPa for sedimentary rock. Empirical correlation by [8] in Table 2 is the lower bound for sedimentary rock [7]. The rock unit skin resistance, fR and RQD relationship as in Table 2 is commonly used in Malaysia to design drilled shaft socketed into rock [9]. Other note, the unit resistance for uplift load should be adjusted due to contraction of pile under uplift load, and hence reducing confinement stress [10].

Design approach for socketed drilled shaft varies from place to place [11]. For example the ultimate capacity could be determined by considering all resistances (soil and rock skin resistances, and base resistance) or totally omitting the base resistance. It is due to the fact that less displacement is required to mobilise skin resistance compared to the displacement that required mobilising base resistance. On the practical side, the length of rock socket and hence total pile length is determined in situ during construction based on observed rock condition, i.e. RQD at that particular location.

Discussions on various construction method and constructability issues of drilled shaft can be found in Ref [12], and the effect of construction method on skin and base resistance in Ref [13,14]. However, from pull out test results [6], drilled shaft construction method has no significant effect on ultimate resistance.

qb

fS

fR LR

LS

D

Figure 1. Ultimate capacity of socketed drilled shaft

67



Pile Load Test

Pile-load-test generally serves two purposes. When conducted during design stage, it helps to establish parameters to be used in design, and when conducted during construction stage, it proves working assumptions during design. To obtain the “actual soil parameters” for design purpose, the pile test is instrumented [4] and parameters are back calculated from data obtained during testing. The number of this type of test is limited due to its cost; therefore site variability of pile ultimate capacity cannot be established.

The load settlement curves do not always show a sign of failure as specified by various methods, for example Davisson’s, Terzaghi’s, Chin’s methods and others. As such, it is difficult to ascertain the validity of design assumption, i.e. the ultimate skin resistance, based on information’s obtained from this test. Other elaborate method to interpret pile load test in sand and clay can be found in [15, 16]. A new and novel approach that utilizes a data base of pile load tests is recently proposed by [17]. In their method design parameters are extracted from the data base using Bayesian neural network that intelligently update its knowledge when new information is added to the data base.

Other than instrumented pile as previously cited, [18] proposed method to derive soil parameters from load settlement curve. The approach uses “projected load settlement curve” to obtain the ultimate capacity. The projected load settlement curve is an analytical function for load settlement relations with parameters of the function are obtained from regression of the actual load settlement curve. For this purpose, failure

Table 1. Empirical value of unit skin resistance for socketed drilled shaft determine from rock unconfined compressive strength and SPT N-value. Complete references are given in [6]

No. Empirical Correlation Reference

1 fR(tsf) = 1.842 qu 0.367 Williams et al. (1980)

2 fR(tsf) = 1.45 uq for clean sockets, and fR(tsf) = 1.94 uq for rough sockets

Rowe and Armitage (1987)

3 fR(tsf) = 0.67 uq Horvath and Kenney (1979):

4 fR(tsf) = 0.63 uq Carter and Kulhawy (1988)

5 fR(tsf) = 0.3 (qu) Reynolds and Kaderabek (1980):

6 fR(tsf) = 0.2 (qu) Gupton and Logan (1984)

7 fR(tsf) = 0.15 (qu) Reese and O’Neill (1987)

8 fR = 0.017V (tsf), orfR = – 5.54 + 0.4 N(tsf)

Crapps (1986)

9 N-value = {10, 15, 20, 25, 30, >30}fR(tsf) = {0.36, 0.77, 1.1, 1.8, 2.6, 2.6}

Hobbs and Healy (1979)

10 N-range = 10 - 20, 20 - 50, 50 - 50/3 in., >50/3 in.fR(tsf) = 1.5, 2.5, 3.8, 5

McMahan (1988)

Table 2. Empirical value of unit skin resistance for socketed drilled shaft determine from RQD Ratio

RQD Ratio %Working Rock Socket Resistance fR (kPa)

Below 25 300

25 - 70 600

Above 70 1000

68



is defined as correspond to a settlement of 10% of pile diameter.

As a proof test, pile-load-test results are interpreted based on criteria establishes to achieve the design objectives and elucidated in the technical specification. As an example, the settlement of the pile tested should not exceed specified settlement at working load and permanent settlement should not exceed settlement at twice of the working load. The pile is considered “pass” if both criteria are satisfied. The number of proof pile test is prescribed based on total length of pile constructed; as such more than one proof pile-load-test is conducted within one project. Furthermore, it is common that proof pile-load-test is “fail to reach failure”, in other words the applied load is less than ultimate capacity of the pile.

While the interpretation of proof test based on settlement criteria lay out in the technical specification is sufficient for practical purposes, there are also attempts to further exploit information’s from proof pile-load-test. For example, from pile-load-test that reaches failure, information’s can be obtained to update the reliability of pile [19-22]. For pile-load-test that fails to reach failure, information’s can be obtained to update the probability distribution of pile capacity [23]. These approaches follow trend of migration of geotechnical analysis from factor of safety based to reliability based [24]. It is worth to note that, besides ultimate load limit state approach

previously cited, serviceability limit state for drilled shaft to establish probability of failure and reliability index from load settlement curve of pile load test also have been attempted in [25,26]. In their approach, the pile load settlement curves are calculated using “t-z” approach and finite difference method. Probabilistic load-settlement curves are developed using Monte Carlo simulation. From the histogram generated, the probability of failure and reliability index can be determined.

For the work reported herein, the probability density of soil parameters are calculated from ultimate pile capacity, deduced from pile-load-test, using probabilistic inverse method. Monte Carlo simulation technique is then used to generate the histogram of pile capacity. The probability of ultimate pile capacity, or the reliability index, can be obtained from cumulative probability density of pile capacity.

DESCRIPTION OF THE PROJECT

Site Condition

A total of 12 bore holes were carried out for soil investigation during design stage. The soil investigations were mainly carried out by using Standard Penetration Test (SPT) as well as standard soil index and physical properties tests. The site condition was mainly formed by 3 types of soil, which were silt, clay and sand. Silt was found on the top of the soil layer

Figure 2. Load settlement curve for proof pile-load-test. Davisson’s ultimate load can be obtained only in two out of nineteen tests.

69



while very stiff or hard sandy silt were encountered on the next layer which range from Reference Level (RL) 70 m to RL 50 m. Generally, high organic content was observed for the upper layer materials. The RQD for the site ranged from 10% to 30%, with average of 25%.

Pile Loading Test

There were three types of pile being used at the site consisting of a 450 mm, 600 mm and 900 mm diameter bored pile with design load of 1500 kN, 4000 kN and 9000 kN respectively. Out of a total of 19 proof pile-load-tests conducted, only two gave ultimate pile capacity based on Davisson’s criteria. The calculated

ultimate pile capacity is 8900 kN and 3900 kN for 900 mm and 600 mm diameter piles, respectively. Figure 2 shows load deflection curves and ultimate load determination procedure using Davisson’s method for 600 m and 900 mm piles, and Table 3 shows the schedule of all tests. It should be noted that for both the 600 mm and 900 mm piles one out of seven test piles being tested was fail.

PROBABILISTIC INVERSE METHOD

Suppose that we have function f that map parameters into theoretical quantity such that d = f(m) where d = {di ,…, dND and m = {mi, ,…, mNM}, the objective of inverse analysis is to determine m given m. In the

Table 3. Attributes of pile proof test result

Pile LocationEstimate of Qu

(kN)Pile Length

(m)Pile Diameter

(mm)Socket Length

(m)

FF1- P28 Not fail 6.000 900 4.5

FF1- P67 8900 10.245 900 4.5

FF2 - P184 Not fail 6.000 900 4.5

FF2 - P472 Not fail 11.000 900 4.5

FF3 - P234 Not fail 11.075 900 4.5

FF4 - P236 Not fail 12.325 450 1.5

FF4 - P275 Not fail 14.625 450 1.5

FF5 - P33 Not fail 12.425 600 3.0

FF5 - P85 Not fail 22.125 600 3.0

FSK 1,5 - P47 Not fail 14.000 900 4.5

FSK 1,5 - P44 Not fail 8.200 450 1.5

FSK 2,3,4 - P200 Not fail 8.800 900 4.5

FSK 2,3,4 -P387 Not fail 18.000 900 4.5

FSK 6 - P 74 Not fail 2.600 900 2.6

FSK 6 - P 307 Not fail 5.700 600 3.0

FSK 7 - P40 Not fail 21.025 600 3.0

FSK 7 - P370 Not fail 7.330 600 3.0

PSB - P138 Not fail 19.500 600 3.0

PSB- P183 3900 10.775 600 3.0

70



context of pile-load-test, to determine fS, fR and qbu knowing Qu obtained from pile-load-test and f is the relationship in Eq. (1).

Data Space

Suppose that we have observed data values dobs, the probability density model to describe experimental uncertainty, such as Gaussian model, can be written as follows:

= )dd(C)dd(2

1expk)d( obs

1D

TobsD (2)

where CDis the covariance matrix. If the uncertainties are uncorrelated and follow Gaussian distribution, it can be written as

=

2

i

iobs

i

Ddd

2

1expk)d( (3)

Model Space

In a typical problem we have model parameters that have a complex probability distribution over the model space. The probability density is denoted as ρM(m). Suppose that we know joint probability density function ρ(m,d) and d = f(m), then the conditional probability density function, σM(m) = ρM|d(m)(m|d = f(m))can be obtained as follows [1]:

)m(fdDMDM

DT

MM .

g.ggdetgdet

)FgFgdet ())m( )f,m(k)m(

=

=FDgTFMg

2/1

2/12/1

(4)

For constant gM(m) and gD(d), and linear or weak linearity problem [1], Eq. (4) reduces to

)m(fdD

DMM )d(

)d()m(k)m(=

= (5)

where k is the normalizing factor, µD(d) is homogenous probability density function, which upon integration over the data space become unity.

EVALUATION OF POSTERIOR DISTRIBUTION

The analytical form of posterior distribution, i.e. Eq. (5), is difficult to obtain, or, even if obtainable, is difficult to interpret. One has to resort to simulation approach such as Monte Carlo simulation to obtain parameter pairs over the model space and used such data for any application. In Monte Carlo simulation, after sufficient number on sampling of random variables X0, X1, …, Xn the expectation µ = E{g(Xi} is approximated as:

(6)

Other approach is to use Markov Chain Monte Carlo (MCMC) simulation that generate sampling points over the model space by “controlled random walk”, the Markov Chain, that eventually converged to the conditional probability distribution of the posterior (or parameters). In Markov Chain approach the sequence of random variables X0, X1, X2, … at each time t ≥ 0the next state Xt+1 is sampled from a distribution P(Xt+1 | Xt) that depends on the state at time t. Similar to Monte Carlo simulation, if sufficient numbers of sampling points are obtained, then the approximation to the expected value is evaluated through Eq. (6).

In general the MCMC has three basic rules:

(i) a proposal Markov Chain rule expressed by a transition kernel q(x,y),

(ii) an accept reject rule which accept or reject a newly proposed Yk = q(Xk,.)where Xk is recently accepted random variable, and

(iii) a stopping rule.

A typical MCMC has the basic algorithm shown as the following algorithm

71



Basic MCMC Algorithm

1. Draw initial state X0 from some initial distribution

2. For I = 0 to N do

3. Modifying Xt according to some proposal distribution to obtain a proposed sample Yt+1 where Yt+1 = q (Xt, Y)

4. With some probability A(Xt, Yt+1) accept Yt+1

= +++

t

1tt1t1t Xotherwise

)Y,X(AyprobabilitwithYX

Some acceptance rules are given as follow(a) Metropolis sampling:

)X(

)Y()Y,X(A

t

1t1tt = +

+

(b) Metropolis-Hasting sampling

)Y,X(q)X(

)X,Y(q)Y()Y,X(A

1ttt

t1t1t1tt

+

+++

+=

(c) Boltzman sampling:

)Y()X(

)Y()Y,X(A

1tt

1t1tt

+

++ +

=

BAYESIAN INTERPRETATION OF PROOF PILE LOAD TEST

The simplistic model of bearing capacity of socketed drilled shaft is given by Eq. (1). Assuming known pile geometry, the model space is then m = (fz, fR, qb). The probability density model to describe experimental uncertainty (Eq. 3) is formed using the theoretical model d = f(m) as in Eq. (1), and observed pile ultimate capacity as dobs. The joint probability density is then σM(m) = σM(fS, fR, qb). Prior knowledge can be incorporated in ρM(m) = ρM(fS, FR, qb) particularly knowledge on those parameters specific for the rock type and its locality.

The simplistic model of bearing capacity of socketed drilled shaft is given by Eq. (1). Assuming known pile geometry, the model space is then m = (fS, fR, qb). The

probability density model to describe experimental uncertainty (Eq. 3) is formed using the theoretical model d = f(m) as in Eq. (1), and observed pile ultimate capacity as dobs. The joint probability density is then σM(m) = σM(fS, fR, qb). Prior knowledge can be incorporated in ρM(m) = ρM(fS, fR, qb) particularly knowledge on those parameters specific for the rock type and its locality.

The effects of prior knowledge on ultimate pile capacity can be investigated using various forms of density distribution. In this work, only the effect of unit skin and base resistances of rock are considered, and the joint probability density is obtained from Eq (5) as

= SbRSMbRM df)q,f,f()q,f( (7)

In this work two aspects are being investigated: (a) the effect for prior knowledge on predicted pile bearing capacity and (b) comparison between the “brute force” Monte Carlo and Markov Chain Monte Carlo simulations. For the first objective, four cases are considered. In Cases 1 to 3 lognormal prior distributions of fR are assumed with mean values range between 300 to 800 kPa. For the fourth case, normal distribution is assumed for fR with mean value of 300 kPa. These values conform to an empirical value for unit skin resistance for rock at low RQD (Table 2). This number is somewhat lower than back calculated from pile load test in limestone which range from 900 kPa to 2 300 kPa [29]

Results from Case 1 to 4 are compared in term of: (a) plot of posterior probability density, (b) Monte Carlo sampling points and (c) histogram of predicted ultimate pile capacity and shown in Figure 3 to 6. Figure 7 shows relative comparison of prior density distribution used in Case 1 to 3. A sample of 3D plot is shown in Figure 8 for Case 1. In Case 5 MCMC is used to draw sampling points with prior density distribution as used in Case 4. All cases use 15 000 trials. Results for all cases are shown in Table 2 for a 600 mm diameter pile.

72



(a) (b) (c)

Figure 3. (a) Plot of posterior probability density for Case 1, (b) Sampling points superimposed to probability density plot and (c) Histogram of ultimate pile capacity

(a) (b) (c)


(a) (b) (c)


73



(a) (b) (c)

Figure 6. (a) Plot of posterior probability density for Case4, (b) Sampling points superimposed to probability density plot and (c) Histogram of ultimate pile capacity

Figure 7. Prior distribution of unit skin resistance of rock

Figure 9. Case 5: (a) Plot sampling points generated by Markov chain and (b) Histogram of ultimate pile capacity

Figure 7. A 3D plot of posterior joint probability distribution for Case 1

74



DISCUSSIONS

From Table 4 and Figures 3, 4, 5, 6 and 9 the following observations can be advanced:

For all practical purposes the prior density distributions have minimal effect on calculated ultimate pile capacity. The mean values ranged between 4 103 to 4 145 kN.

Markov Chain Monte Carlo is more efficient compared to the brute force Monte Carlo method. Out of 15 000 trials, 6 042 points have been generated using MCMC method compare to 2 066 points using brute force Monte Carlo method. The sampling points for MCMC method are more concentrated around the maximum probability density (Figure 9a) compare to MC method (Figure 6b) resulting in a more accurate ultimate pile capacity.

Markov Chain Monte Carlo yield more accurate results (3 902 kN to 3 900 kN from pile test) compare to brute force Monte Carlo method (4 145 kN to 3 900 kN from pile test).

CONCLUDINg REMARK

Based on previous discussions the following conclusions can be put forward:• Method to interpret pile test to obtain

probabilistic characteristics of ultimate load has been presented. The first step is to obtain joint probability distribution of soil parameters in the material (or parameter) space and the second step is to generate histogram of ultimate pile capacity using Monte Carlo technique. Statistical characteristics of the ultimate pile capacity are then obtained from the synthetic histogram using standard discrete method.

• From this study the effect of prior probabilitydistribution, for all practical purposes, is negligible.

• Markov ChainMonte Carlomethod yieldsmoreaccurate results compare to brute force Monte Carlo method and more efficient in term of ratio of generated sampling points to number of trial.

REFERENCES

Mosegaard, K. & Tarantola, A. “Probabilistic Approach to [1] Inverse Problem”. In International Handbook of Earthquake & Engineering Seismology (Part A), Academic Press. 2002. pp. 237-265

Rollins, K. M., Clayton, R. J., Mikesell, R. C. & Blaise, B. C. [2] “Drilled Shaft Side Friction in Gravely Soils”. Journal of Geotechnical and Geoenvironmental Engineering, 2005. 131(8):987-1003

Table 4. Comparison of mean, median and standard deviation

CaseMean(kN)

Median(kN)

Standard Deviation

(kN) Remark

Case 1 4103 4147 521 Lognormal Distribution, MC



Case 4 4145 4169 590 Normal Distribution, MC

Case 53840 3905 351 (a) Normal Distribution, MCMC (burned point 100)

3902 3931 270 (b) Normal Distribution, MCMC (burned point 500)

75



O’Neil, M. W. “Side Resistance in Piles and Drilled Shafts”. [3] In The Thirty-Fourth Karl Terzaghi Lecture. Journal of Geotechnical and Geoenvironmental Engineering, 2001. 127(1):1-16

Zhang, L. & Einstein, H. H. “End Bearing Capacity of [4] Drilled Shafts in Rock”. Journal of Geotechnical and Geoenvironmental Engineering. 1998. 124(7):574-584

Amir, J. M. “Design of Socketed Drilled Shafts in Limestone, [5] a Discussion”, Journal of Geotechnical Engineering. 1994. 120(2):460-461

McVay, M. C., Townsend, F. C. & Williams, R. C. “Design of [6] socketed drilled shafts in limestone”. Journal of Geotechnical Engineering. 1992. 118(10): 1626-1637

Ng, C. W. W, Yaw, T. L. Y, Li, J. H. M. & Tang, W. H. “Side Resistance [7] of Large Diameter Bored Piles Socketed into Decomposed Rocks”. Journal of Geotechnical and Geoenvironmental Engineering. 2001.127(8):642-657

Horvath, R. G. & Kenney, T. C. “Shaft Resistance of Rock-[8] socketed Drilled Piers”. Proceedings Symposium on Deep Foundation. 1979

Tan, Y. C. & Chow, C. M. “Design and Construction of Bore Pile [9] Foundation”. Geotechnical Course for Foundation Design & Construction. 2003

Fellenius, B. H. Discussion of ‘‘Side Resistance in Piles [10] and Drilled Shafts’. Journal of Geotechnical and Geoenvironmental Engineering. 2001. 127(1): 3–16

Hejleh, N. A., O’Neill, M. W, Hanneman, D. & Atwooll, [11] W. J. “Improvement of the Geotechnical Axial Design Methodology for Colorado’s Drilled Shafts Socketed in Weak Rocks”. Colorado Department of Transportation. 2004.

Turner, J. P. “Constructability for Drilled Shafts”. Journal [12] of Construction Engineering and Management. 1992. 118(1):77-93

Majano, R. E., O’Neill, M. W. & Hassan, K. M. “Perimeter Load [13] Transfer in Model Drilled Shafts Formed Under Slurry”. Journal of Geotechnical Engineering. 1994. 120(12.):2136-2154

Chang, M.F. & Zhu, H. Construction Effect on Load [14] Transfer along Bored Pile. Journal of Geotechnical and Geoenvironmental Engineering. 2004. 130(4):426-437

Cherubini, C., Giasi, C.I. & Lupo, M. Interpretation of Load [15] Tests on Bored Piles in the City of Matera. Geotechnical and Geological Engineering. 2004. 23:239-264

Pizzi, J.F. Case history: Capacity of a Drilled Shaft in [16] the Atlantic Coastal Plain. Journal of Geotechnical and Geoenvironmental Engineering, 2007. 133(5):522-530.

Goh, A. T. C., Kulhawy, F. H. & Chua, C. G. “Bayesian Neural [17] Network Analysis of Undrained Side Resistance of Drilled Shafts”. Journal of Geotechnical and Geoenvironmental Engineering, 2005. 131(1):84-93

Boufia, A. “Load-settlement Behaviour of Socketed Piles [18] in Sandstone”. Geotechnical and Geological Engineering. 2003. 21:389-398

Kay, J. N. “Safety Factor Evaluation for Single Piles in Sand”. [19] Journal of Geotechnical Engineering Division. 1976. 102(10):1093-1108

Lacasse, S. & Goulois. A. “Uncertainty in API Parameters [20] for Predictions of Axial Capacity of Driven Piles in Sand”. Proceeding of the 21st Offshore Technology Conference, Society of Petroleum Engineers, Richardson, Texas. 1989. 353-358

Baecher, G. R. & Rackwitz, R. “Factor of Safety and Pile Load [21] Tests”. International Journal of Numerical and Analytical Methods in Geomechanics. 1982. 6(4):409-424

Zhang, L. M. & Tang, W. H. “Use of Load Tests for Reducing Pile [22] Length”. Proceeding of the International Deep Foundations Congress. Geotechnical Special Publication No. 116, M. W. O’Neill and F. C. Townsend, eds., ASCE, Reston, Va., 2002. 993–1005

Zhang, L. M. “Reliability Using Proof Pile Load Tests”. Journal [23] of Geotechnical and Geoenvironmental Engineering. 2004. 130(2): 1203-1213

Duncan, J. M. “Factors of Safety and Reliability in [24] Geotechnical Engineering”. Journal of Geotechnical and Geoenvironmental Engineering. 2000. 126(4):307-316

Misra, A. & Roberts, L. A. “Axial Service Limit State Analysis of [25] Drilled Shafts using Probabilistic Approach”. Geotechnical and Geological Engineering, 2006. 24:1561–1580

Misra, A., Roberts, L. A. & Levorson, S. M “Reliability Analysis [26] of Drilled Shaft Behaviour Using Finite Difference Method and Monte Carlo Simulation”. Geotechnical and Geological Engineering. 2007. 25:65–77

Hassan, K. M. & O’Neill, M. W. “Side Load-Transfer Mechanisms [27] in Drilled Shafts in Soft Argillaceous Rock”, Journal of Geotechnical and Geoenvironmental Engineering. 1997. 123(2):145-152

Hassan, K. M., O’Neill, M. W., Sheikh, S. A. & Ealy, C. D. “Design [28] Method for Drilled Shafts in Soft Argillaceous Rock”. Journal of Geotechnical and Geoenvironmental Engineering. 1997. 123(3):272-280

Gunnink, B. & Kiehne, C. “Capacity of Drilled Shafts in [29] Burlington Limestone. Journal of Geotechnical and Geoenvironmental Engineering”. 2002. 128(7):539-545

76



I. S. H. Harahap holds a Bachelor’s Degree (Sarjana Muda, SM) and Professional Degree (Insinyur, Ir.) from Universitas Sumatera Utara, Medan, Indonesia. After a short stint with the industry, he continued his tertiary education in the United States, obtained his Master’s of Science in Civil Engineering (MSCE) degree from Ohio University, Athens, Ohio and Doctor of

Philosophy (PhD) degree from Northwestern University, Evanston, Illinois. He was with Universitas Sumatera Utara (USU) before joining Universiti Teknologi PETRONAS (UTP) in August 2005.

His research interests include: (1) application of expert system in geotechnical engineering, (2) implementation of geotechnical observational method, (3) subsurface exploration and site characterization, (4) landslide hazard identification and mitigation, and (5) robust and reliability based structural optimization. He is the recipient of 1993 Thomas A. Middlebrook Award from American Society of Civil Engineer (ASCE) for his research on simulation of the performances of braced excavation in soft clay.

Wong Chun Wah graduated with a Bachelor of Science degree in Civil Engineering from Universiti Teknologi PETRONAS (2008). Currently, he is working with PETRONAS’s Malaysia LNG Sdn. Bhd. in Bintulu as a Civil and Structural Engineer.

77



This paper was presented at the International Conference on Construction and Building Technology 2008, Kuala Lumpur,

16 - 20 June 2008

INTRODUCTION

Bridge structures are designed with high quality and safety standards but sometimes with not enough attention to construction methods, site conditions and details. Construction problems encountered during execution are complex and costly. Many construction problems can be avoided with proper attention and consideration of the construction process during the design phase [1]. Factors of simplicity, flexibility, sequencing, substitutions and labour skill and availability should be the part of design. The appropriate use of standardisation can have several benefits [1]. These include increased productivity and quality from the realization of repetitive field operations, reduction in design time, savings from volume discounts in purchasing, and simplified materials management. This method of standardising bridge elements may be suitable for selective projects

of same nature but are less significant and more complex when constructing multiple bridge projects situated at different site conditions.

The construction process and success in management of multiple bridge projects directly relies on the selection and optimisation of their elements/components. A systematic optimization process is adopted during the conceptual design stage to overcome the resource constraints during the construction phase. The knowledge of construction experience is also utilised during the element optimisation process.

D&B combines the design and construction functions to vests its responsibility with one entity: the design-builder. The D&B process changes some fundamental relationships between the client and the contractor. The client employs a Project Director as the

ELEMENT OPTIMISATION TECHNIQUES IN MULTIPLE DB BRIDgE PROJECTS

Narayanan Sambu Potty*, C. T. Ramanathan1, *Universiti Teknologi PETRONAS, 31750 Tronoh, Perak Darul Ridzuan, Malaysia

1Kumpulan Liziz Sdn Bhd, Kota Kinabalu, Sabah, Malaysia. *[email protected]

ABSTRACT

Management of multiple bridge projects relies on the selection and optimisation of elements. Problems are avoided if construction knowledge and experience are utilised. A systematic selection process is required during design stage to maximise the use of available resources. Multiple bridge designs incorporate practical suggestions from the field personals. The case study in East Malaysia considers problems of material, labour and equipment availability. The average ratio presence of each element is calculated to show its impact. The need of element optimisation techniques in bridges is emphasised. The database is presented and also the process undertaken during the design is discussed.

Keywords: Design and Build process, Multiple project management, Bridge elements, Average ratio presence.

78



representative, whereas the contractor has to engage design consultant and manage the construction of the whole project. Owners employ only one contractor who is solely responsibility for delivering the assigned project with defined requirements, standards and conditions. Both parties are expected to aim for a successful project outcome.

BACK gROUND AND PROBLEM STATEMENT

Sabah situated on the northern tip of the island of Borneo is the second largest state in Malaysia. Over 70% of its population lives in rural area as majority are depending directly or indirectly on agriculture. The state has several development projects procured by D&B Method for Upgrading rural roads and Bridges replacement for the benefit of rural sectors contributing to the national economy. Five contract packages comprising 45 bridges located in 12 districts in the state having 76 115 square kilometer coverage area [2] and two road projects in one district was assigned to the first author’s company. As summarised in Figure 1, the Bridge projects and the two road projects were handled simultaneously.

This study examines the use of element optimisation technique through a case study for managing the above multiple DB bridge projects in Sabah, East Malaysia. The data of bridge elements of all the 45 bridges were collected and compiled. The ratio of

each element in individual bridges and their average ratio presence in each projects were compared for study.

RESEARCH METHODOLOgY

The element ratio comparison and analysis were made in the following steps.1. Review of all the five projects and compilation of

the summary.2. Prepare the element ratio table and Pi chart for

each bridge of all the projects and analyse the ratio of their impact on the individual bridges.

3. Compress and derive a single and common average ratio table and pi chart showing the overall impact of each elements for the entire multiple project.

4. Identify the critical and crucial elements that need attention.

5. Discussion on the element of major contributions which is to analyse the element with maximum element ratio.

ANALYSIS OF COMPONENTS OF MULTIPLE DB BRIDgE PROJECTS

Schedules of the multiple projects

The projects started at different times having simultaneous construction periods at various stages

Figure 1. Duration of Multiple D&B projects in Sabah

79



of completion as shown in Table 1. The five projects involving 45 Bridges have been completed successfully and delivered to client for public usage.

Bridge element ratio

The element ratio for bridges for each project was calculated as given in the following section.

Project 1 – Package A

Table 2 and Figure 2 show the percentage presence of each element in constructing the bridges in Package A. From the weight breakdown it is clear that critical elements “Piling work (Foundation)” has greater than 20% presence and production of beams and erection has nearly 40% presence. The influence of these elements in the project is high in spite of their low quantum in physical work (20% and 40% respectively) because of their specialty nature. Hence, care should be taken while choosing these specialties works to suite the local availability and eventually use few specialised designs for those particular elements. In this manner these crucial elements in Package A were optimised in the design as follows:

Critical element No. 1: Piling works– Steel H piles of driven type for 6 Bridges– Micro pile of drilling type for 3 Bridges

Critical element No. 2: Beams – Cast in situ Post tensioned beams of two

shapes I-16 and I-20.

Table 2. Weight of each Element in bridges of Package A

Table 1. Schedule of Multiple contracts handled in Sabah

Project Number

No. of Bridges

No. of districts Project Duration

Locations Months Period

1 12 3 18 Jul 03 – Jan 05

2 5 3 18 Jan 05 – Jul 06

3 8 3 18 Jul 05 – Jan 07

4 13 3 20 Oct 05– Jun 07

5 7 3 20 Oct 05– Jun 07

Total 7 45 12 48 Jul 03–Jun 07

Figure 2. Weight of elements expressed as a percentage

22.64%

16.33%

1.57%

2.87%

13.93%2.77%

39.88%

Description Ratio

Foundation (Piling) 22.64%

Abutment and Wing Wall 16.33%

Piers, Crossheads and Pilecaps 1.57%

Precast, Prestressed Beam and Ancillary 39.88%

Diaphragms 2.87%

Bridge Deck and Run-On-Slab 13.93%

Parapet 2.77%

TOTAL 100.00%

80



Project 2 – Package B

Table 3 and Figure 3 show the percentage presence of each element in constructing the bridges in Package B. From the weight breakdown it is clear that critical elements are “Piling work (Foundation)” which has greater than 20% presence and “production of beams and erection” with 30% presence. The reasons as explained above for Project 1 have eventually resulted due to the use of specialised designs for those particular elements. In this manner these crucial elements in Package B were optimised in the design as follows:

Critical element No. 1: Piling works– Micro pile for 3 Bridges– Spun pile for 1 Bridges

Critical element No. 2: Beams – Cast in situ Post-tensioned beams of two

shapes I-16 and I-20 for 2 Bridges– Prestressed precast beams (in factory) for 3

Bridges.

Project 3 – Package C

Table 4 and Figure 4 show the percentage presence of each element in constructing the bridges in Package C. From the weight breakdown it is clear that critical elements are again “Piling work (Foundation)” with greater than 20% presence and the “production of beams and erection” with greater than 30% presence. The reasons as explained above has eventually resulted due to the use of specialised designs for those particular elements. In this manner these crucial elements in Package C were optimised in the design as follows:

Critical element No. 1: Piling works– Bored pile for 3 Bridges– Micro pile for 2 Bridges– Spun pile for 1 Bridges

Critical element No. 2: Beams– Cast in situ Post tensioned beams of two

shapes I-16 and I-20 for 4 Bridges.– Prestressed precast beams (in factory) for two

bridges.Figure 4. Weight of each element expressed in percentage

Table 3. Weight of each element in bridges of Package B

Figure 3. Weight of elements expressed in percentage

29.41%

12.68%

8.79%

32.50%

1.63%

11.88%

3.11%

Table 4. Weight of each element in bridges of Package C

22.44%

12.13%

13.56%

31.73%

2.64%

13.84%

3.66%

Description Ratio





Diaphragms 1.63%


Parapet 3.11%

TOTAL 100.00%

Description Ratio

Foundation 22.44%




Diaphragms 2.64%


Parapet 3.66%

TOTAL 100.00%

81



Project 4 – Package D

Table 5 and Figure 5 show the percentage presence of each element in constructing the bridges in Package D.

Critical element No. 1: Piling works – Micro pile for 9 Bridges– Spun pile for 2 Bridges

Critical element No. 2: Beams – Cast in situ Post tensioned beams of two

shapes I-16 and I-20 for 6 Bridges.– Prestressed precast beams (in factory) for 6

bridges.

From the weight breakdown of the elements of bridge, it is clear that critical elements are “Piling work (Foundation)” has greater than 30% presence and the “production of beams and erection” with 30% presence. The reasons as explained above has eventually resulted due to the use of specialised designs for those particular elements. In this manner these crucial elements in Package D were optimised in the design as follows:

Project 5 – Package E

Table 6 and Figure 6 show the percentage presence of each element in the construction of the bridges in Package D. From the weight breakdown it is clear that critical elements are again “Piling work (Foundation)” with greater than 30% presence and the “production of beams and erection” with presence of greater than 55%. The reason is due to the design using Steel girder to suite the site condition. In this manner these crucial elements in Package E were optimised in the design as follows:

Critical element No. 1: Piling works– Micro pile for 5 Bridges– Spun pile for 1 Bridge

Critical element No. 2: Beams – Cast in situ Post tensioned beams I-16 for 1

Bridge.– Steel girders 5 Bridges– Steel trusses 1 Bridge

Table 6. Weight of each element in bridges of Package E

Figure 5. Weight of each element expressed in percentage

Table 5. Weight of each element in bridges of Package D

33.82%

16.24%

1.42%

31.36%

12.15% 2.27%2.73%

21.05%

8.67%

1.78%

56.84%

.2 16%

7.77% 1.72%

Figure 6. Weight of each element expressed as a percentage of total

Description Ratio





Diaphragms 2.27%


Parapet 2.73%

TOTAL 100.00%

Description Ratio





Diaphragms 2.16%


Parapet 1.72%

TOTAL 100.00%

82



Overall average ratio

The Table 7 and Figure 7 show the overall influence of these elements in the multiple projects.

Table 7 Overall Weight of Each Element for bridges in Package A to E

It was observed that the critical elements that needed more attention were “the foundation (piling) works” and “Superstructure Beam works”.

Discussion of Results

The elements having high ratios (or high presence) were high impact causers. In this multiple project the overall influence of the elements “Piling” and “Beams” in bridge completion are critical. Even though the quantum of these elements were less compared to other elements of the bridges, the ratio of their influence in the construction was more due to their level of special technology, specialist availability, method of construction, risk involved and limited usage/availability of resources to produce.

Piling has 25.87% and beams have 38.46% – they have the maximum presence. The presence of these two items were more with less volume of work because of its speciality and use of uncommon materials.

Hence, extra care was given while deciding the design for these critical elements. Then, few design optimisations were adopted to reduce the complexity and to ease implementation in the field as shown in Table 8. The techniques adopted in element optimisation for the multiple bridge construction were successful and resulted in the projects being executed in time and within budget.

Table 7. Overall weight of each element for bridges in Package A to E

25.87%

13.21%

5.43%

38.46%

2.32%

11.91%

2.80%

Figure 7. Overall Weight of each element expressed in percentage

Ratio

Description Package A Package B Package C Package D Package E Average

Foundation (Piling) 22.64% 29.41% 22.44% 33.82% 21.05% 25.87%

Abutment and Wing Wall 16.33% 12.68% 12.13% 16.24% 8.67% 13.21%

Piers, Crossheads and Pilecaps 1.57% 8.79% 13.56% 1.42% 1.78% 5.43%

Precast, Prestressed Beam and Ancillary 39.88% 32.50% 31.73% 31.36% 56.84% 38.46%

Diaphragms 2.87% 1.63% 2.64% 2.27% 2.16% 2.32%

Bridge Deck and Run-On-Slab 13.93% 11.88% 13.84% 12.15% 7.77% 11.91%

Parapet 2.77% 3.11% 3.66% 2.73% 1.72% 2.80%

TOTAL 100.00% 100.00% 100.00% 100.00% 100.00% 100.00%

83



FINDINgS AND LESSONS LEARNED

The following findings were obtained and lessons learnt from the case study:1. The natural tendency for more cautiousness/

attention for Girders selection was advantages inthe handling and construction of all the beams(about 355 beams of five different varieties)without major problems. This enhanced thefinishing of those beams as scheduled in thebridge programme.

2. Conversely, piling works on the foundation partwere taken lightly at the design stage. Therewas no attempt to rationalise and the decisionwas left to the individual designers. This resultedin usage of the same micropile method in themajority of the bridges. Difficulties which arosein implementing the micropiles for many bridgeswere:-(i) Availability of specialists to perform the works

was very limited in Sabah(ii) The risk of loosing equipment in the drilled

holes requires skilled operators for drilling the

Table 8. Overall “Piling” and “Beam” element optimisation summary for Project 1 to 5.

84



pile. There were not enough skilled operators in this field.

(iii) The component materials like API pipe G80 and permanent casings were very scarce to obtain/procure.

(iv) The method had various stages to complete one point, which is time consuming for each bridge.

Hence remedial measures were taken to catch up with the progress at the cost of spending extra resources, time and money.

CONCLUSIONS

1. In general, the Element Optimisation Techniquewas needed to be adopted for all the elementsin compliance with the required standards. Extraimportance is required for elements that havemore influence in the execution and completionof the project.

2. In multiple DB projects the element optimisationshave to be started well ahead during theconceptual design stage. But this optimisationshould not interfere in the functional quality andthe integrity of the structure which are designedfor a stipulated life period.

3. In this process it is also important to consider andreview the feedback from field personnel afterconducting a proper study on site conditions.

4. In spite of the lag in piling works as mentionedin “lessons learned” from the case study, thecompany was able to make up and completein time with recognition and credits due to theproper and timely planning in rest of the elementoptimisations.

ACKNOWLEDgEMENTSThe authors would like to thank the Directors of Kumpulan Liziz Sdn Bhd., Hj. Ghazali Abd Halim, Mr. Simon Liew and Mr. Liew Ah Yong for making available data regarding the multiple projects being executed by the company in Sabah.

REFERENCES

Rowings, J. E., Harmelink, D. J., & Buttler, L. D., 1991, [1] “Constructability in the Bridge Design Process”, Research project 3193, Engineering Research Institute, Iowa State University. 1-11

Wikipedia, 2007, Area of Sabah State[2]

Narayanan Sambu Potty received his Bachelor of Technology in Civil Engineering from Kerala University and Master of Technology degree from National Institute of Technology in Kerala. His PhD work “Improving Cyclone Resistant Characteristics of Roof Cladding of Industrial Sheds” was done at Indian Institute of Technolgy Madras India.

Currently Associate Professor at UTP, he has earlier worked in Nagarjuna Steels Ltd., TKM College of Engineering, Kerala, India and Universiti Malaysia Sabah. His main research areas are steel and concrete structures, offshore structures and construction management.

C.T. Ramanathan has more than 15 years of experience in various construction projects in India and Malaysia. As Project Manager for Sabah, East Malaysia, he leads a team of professionals in managing the Design and Build Infrastructural projects for this region. He has full responsibility throughout the D&B project life cycle (initial project definition to close out) involving

Highways, Roads and Bridges in the state. He also has extensive experience in the entire process of D&B operation methods. His Master’s research was on “Complexity Management of Multiple D&B Projects in Sabah”. He has published and presented ten papers for international conferences.

Documents

PLATFORM - press.utp.edu.mypress.utp.edu.my/wp-content/uploads/2018/12/PlatformV6N2_page1.pdf · Enhancement Of Heat Transfer Of A Liquid Refrigerant In Transition Flow In The Annulus