VOLUME EIGHT NUMBER ONE JANUARY- JUNE 2010 ISSN 1511 …press.utp.edu.my/wp-content/uploads/2018/12/platformv8n1-new.pdf · Dr. Shahrina Mohd. Nordin Subarna Sivapalan Sub-Editor:

VOLUME EIGHT NUMBER ONE JANUARY- JUNE 2010 ISSN 1511-6794

1 VOLUME Eight NUMBER ONE jaNUaRy - jUNE 2010 PLATFORM

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

Contents

Copyright © 2009Universiti teknologi PEtRONaS

PLATFORMJanuary-June 2010

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

Mission-Oriented Research: CARBON DIOXIDE MANAGEMENT

Separation Of Carbon Dioxide From CO2/CH4 Binary Mixture Using Silicone Rubber Membrane Farooq Ahmad, Hilmi Mukhtar, Zakaria Man and Binay K Dutta

2

Mission-Oriented Research: GREEN TECHNOLOGY

Development Of Polymeric Concrete For Sustainable FuturesM.F. Nuruddin, A. Kusbiantoro and S. Qazi

10

Natural Convection Flow In Vertical Channel Due To Ramped Wall Temperature At One BoundaryM. Narahari and Vijay R Raghavan

17

Mission-Oriented Research: SUSTAINABILITY SCIENCE

Mix Design of Foamed Concrete With MIRHA Using Taguchi MethodMuhd Fadhil Nuruddin and Ridho Bayuanji

26

Technology Platform: APPLICATION OF INTELLIGENT IT SYSTEM

V2S: Voice to Sign Language Translation System for Malaysian Deaf PeopleOi Mean Foong, Tan Jung Low and Wai Wan La

35

Evaluating Pairs Analysis Threshold using Receiver Operating Characteristic (ROC) GraphEmelia Akashah P. A., Anthony T. S. Ho and Savita K. Sugathan

42

Technology Platform: RESERVOIR ENGINEERING

Transient Well Performance Modeling for Reservoir Pressure DeterminationHon Vai Yee, Suzalina Zainal and Ismail M. Saaid

48

Technology Platform: SYSTEM OPTIMISATION

Integrated Scheduling and RTO of RGP with MPC and PI ControllersNooryusmiza Yusoff and M. Ramasamy

57

A Review of Risks And Mitigation Measures in Build-Operate-Transfer ProjectsKalaikumar Vallyutham, Syed Kamarul Bakri Syed Ahmad Bokharey, Narayanan Sambu Potty and Nabilah Abu Bakar

66

2 PLATFORM VOLUME Eight NUMBER ONE jaNUaRy - jUNE 2010


INTRODucTION

Membrane based separation processes are attractive for several reasons, namely; the process is simple; there are diverse applications, which can be studied by the same basic formulations; there is no phase change involved, which is measured in commercial applications as energy savings; the process is generally carried out at atmospheric conditions which, besides being energy efficient, can be important for sensitive applications encountered in pharmaceutical and food industry and finally

modules can be added and optimized in a process design to achieve the desired separation [1]. The penetration mechanism of gas in polymer can also be applied to silicone rubber to help understand gas permeability. Free volume or “holes” exists in the rubber matrix. These “Holes” are thermally formed and disappear with the movement of polymer chains. Gases are soluble in rubber like substance. When rubber is exposed to a gas, solution occurs at the surface and the dissolved gas molecules diffuse into the interior. The permeation of gas through a membrane involves solution on one side, diffusion

SEPARATION OF cARBON DIOXIDE FROM cO2/cH4 BINARY MIXTuRE uSING SILIcONE RuBBER MEMBRANE

Farooq Ahmad*, Hilmi Mukhtar, Zakaria Man and Binay K Dutta#

Universiti Teknologi PETRONAS, 31750 Tronoh, Perak Darul Ridzuan, Malaysia #Dept of Chemical Engineering Petroleum Institute Abu Dhabi U.A.E

* [email protected]

ABSTRAcT

The permeability and selectivity of pure carbon dioxide, methane and its binary mixture at various feed pressure, temperature and composition through non-porous silicone rubber membranes was studied. It was observed that the permeability of pure methane and carbon dioxide increases with pressure, whereas such plots for carbon dioxide become convex towards pressure axis at higher pressures (17 bars) due to reduction in free volume of the polymer. The experimental results showed that for binary mixtures of carbon dioxide and methane, permeability depended not only on the feed gas pressure but also on the molar composition of the feed gas. The permeation flux was found to increase with pressure differences, and the enhancement of the proportion of the carbon dioxide in the feed gas. Experimental results showed that the selectivities estimated from pure gas varied slightly with increase in the feed pressure but reached to a maximum value of 11.4. With 20% CO2 in the feed stream, the selectivity was found to be lower by a factor of two to three times than that of pure components over the whole pressure range. High selectivities were obtained at 80% CO2 in the feed stream. An analytical model expressed in terms of pressure and feed composition was derived from permeability behavior of pure carbon dioxide and methane to predict quantitatively the permeability of binary mixtures. It was indicated that the model could be used to evaluate the separation properties and to choose the optimal feed compositions for the membrane separation systems of carbon dioxide and methane.

Keywords: silicone rubber membrane, natural gas, carbon dioxide, analytical model

This paper was presented at the 7th International Conference on Membrane Science and Technology (MST2009), Kuala Lumpur 12-15 May 2009.



through the membrane to the other side, and finally evaporation out of membrane. Polymeric membranes have been successfully used in many gas separation applications. The success has been largely based on their mechanical and thermal stability, along with good gas separation properties [2]. Understanding the transport behaviour of the target gases through membranes is the foundation of realizing effective separation of mixed gases and selecting the appropriate feed conditions [3]. Generally, the permeation behaviour of pure gas through membrane depended mainly on the properties of the gas and membrane as well as on the feed conditions. As for gas mixture, the transport behaviour of one component through membrane is affected by the presence of other penetrants so that it deviated from that of pure. Therefore, using the permeation data of pure gas to estimate the separation properties of gas mixture could lead to wrong results [3]. There have been extensive studies on the comparison of the differences between pure and mixed gas permeation behaviour [4-9]. The coupling effects (solubility coupling and diffusion coupling is an important factor that was observed in the transport behaviour of mixed gas deviating from pure gas [3]. This deviation in permeability of pure and mixed gases is more significant in glassy polymer than rubbery polymer. Plasticization also affects the transport phenomena through membrane, particularly in mixtures containing carbon dioxide and organic vapours. In plasticization of polymers, the penetrant’s interaction with the polymer caused swelling of the polymer matrix and thus increase in permeability [8].

In oder to optimize and design a real industrial gas membrane plant, it is necessary to establish a mathematical model based on the available experimental data. The model can then be used as a powerful tool to evaluate or predict the performance of a membrane at various feed conditions for a specific gas pair-membrane system. There are a few models in use to predict the practical performance of a membrane as a function of experimental parameters. Ettourney and Majeed [4] developed permeability functions to describe the permeation

behaviour of pure gas and a mixture of N2, O2, CH4 and CO2 through cellulose acetate and silicone rubber membranes. Permeability functions were developed for the N2, O2, CH4 and CO2 species in polysulfone and silicone rubber membranes. For each species all data collected for the pure and mixture gases were used to obtain the permeability functions. The functions were expressed in terms of linear dependence of permeability on species partial pressures on the feed side.

Prabhkar et al [10] developed a self-consistent model to describe the dependence of vapour and gas permeability on the concentration and temperature in rubbery polymers. The variation of the propane permeability with the permeate pressure was accurately predicted in their models. Cornesa et al [11] investigated H2-N2 binary gas mixture transported across ceramic membranes, and derived a mathematical model based on mass balance to successfully calculate the composition of the penetrants as a function of the different experimental parameters. Wu et al. [2] developed an analytical model to predict transport of pure O2, N2 and CO2 and its mixture through a PDMS membrane in which the permeability of a species was expressed in a linear relationship with the partial pressure of the component.

This study is based on the permeation behaviour of pure CO2, CH4 and its binary mixtures through silicone rubber membranes at various feed conditions. A simple and practical mathematical model expressed in terms of pressure and composition was derived to predict quantitatively the permeability of the permeated stream. There are two feed variables (pressure and feed composition). The intrinsical transport parameters such as diffusion coefficient and solubility coefficient were not introduced into the model, due to the difficulty in obtaining accurate values of them in practical applications. The variation of intrinsical parameters was considered as embodied in the change of feed variables such as temperature and pressure



METHODS/THEORY

Silicone rubber membranes have been used to study the permeability behaviour of carbon dioxide and methane. Carbon dioxide, (99.8% pure) and methane, (99.5% pure) were provided by Malaysian Oxygen Company (MOX). Silicone rubber was chosen as a material to study carbon dioxide and methane permeability in this study. It is a commercially available membrane and was provided by the manufacturer in the form of flat sheets. Silicone rubber membrane can be used in a high pressure environment; permeability studies of up to 24 bars was carried out to evaluate the performance of silicone rubber membranes at high pressure permeabilities. Silicone rubber membrane was characterized by SEM, EDX, TGA, and DSC. From the SEM image, the silicone rubber membrane used in this study was a homogenous non-porous symmetric membrane. EDX analysis showed that the silicone rubber membrane contained 38.37% elemental (Si). TGA analysis indicateed that the thermal degradation temperature of the silicone rubber membrane was 474.761°C. Glass transition temperature was determined by DSc and was found to be -43.147°C.

An existing membrane separation unit available in the Unit operation laboratory Universiti Teknologi PETRONAS was used in the experiment. The unit was capable of investigating the permeability of pure CO2 and CH4, and also a binary mixture of CO2/CH4 in various compositions. The schematic diagram is shown in Figure 2. In order to determine the permeability, the volumetric flow rate Q was then corrected to STP conditions (0°C and 1 atm) using the following equation

(1)

in which TSTP and QSTP referred to temperature (K) and volumetric of permeate gas (cm3/s) at STP condition. After conversion into STP condition, gas permeability P was then calculated using the following formula

(2)

where l is length of membrane and ∆p and A are trans-membrane pressure and effective membrane area, respectively. The CO2/CH4 ideal selectivity (unitless), αCO2/CH4

, of silicone rubber membrane can be determined in the following way.

(3)

For mixed gases the permeate flow rate was multiplied by vol percent of the component gas in the mixture. Let x be the volume percent of the carbon dioxide in the permeate stream and 1-x is the volume percent of methane in the permeate stream. Then the permeate flow rate of CO2, Qx and permeate flow rate of CH4 is Q(1-x). After correcting permeate flow rate for both gases to STP conditions (0°C and 1 atm), the permeability for CO2 and CH4 was calculated using the following formula.

(4)

(5)

Q

T

TQ STP

STP ×=

pA

lQP stp

∆×′=

4

2

42 /CH

COCHCO P

P=α

pA

lxQP stp

CO ∆×′=′

2

pA

lxQP stp

CH ∆×′−

=′)1(

4

Figure 1. SEM image of silicone rubber membrane



The selectivity based on mixed gas permeability is given by

(6)

RESuLTS AND DIScuSSION

Figure 3 shows the effect of feed pressure on the permeability of pure CO2 and CH4 across silicone rubber membrane at 25°C. Results showed that CO2

permeability increases with the increase of pressure difference across the silicone rubber membrane, but

4

2

42 /CH

CO

CHCO P

P

′′

=α

Table 1. Selectivity of CO2 over CH4 through silicone rubber membrane M1 at different pressure.

Pressure (Bar)Ideal Selectivity

PCO2/PCH4

80% CO2 - 20%CH4 Selectivity PCO2/PCH4




2 6.88 4.11 3.41 2.68 1.92

4 9.75 5.85 4.87 3.75 2.66

6 10.55 6.30 5.27 4.03 2.89

8 11.49 6.87 5.77 4.40 3.13

10 10.24 6.14 5.12 3.92 2.81

12 9.76 5.85 4.88 3.75 2.66

14 9.74 5.84 4.87 3.74 2.66

Figure 2. Schematic diagram of the membrane separation unit



when the pressure reached 17 bars, the permeability of carbon dioxide convexed towards the pressure axis due to hydrostatic compression effects. For CH4 the permeability first slightly decreased with pressure and then increased. Because of the compression effects, the free volume of the polymer decreased [8]. It appeared that the free volume reduction was strong enough to restrict passage of carbon dioxide to a limited extent. However, methane, having a smaller size, can however move through the free volume without appreciable hindrance.

The permeability data agreed reasonably well with the data reported by Jordan and Koros [8]. The slight difference between their data and the data of this study could be attributed to the fact that their silicone rubber membrane contains 4.9% silica. In general the gas permeation through a dense polymeric membrane is typically described by the solution diffusion model, that is the permeability of gas is determined by the solubility and diffusivity of gas in the membrane [12]. Silicone rubber has weak inter-molecular sieve ability due to its weak inter-molecular forces, resulting in a broad distribution of intersegmental gap sizes responsible for gas diffusion. The diffusion coefficient of penetrants often change less than the solubility coefficient so that more soluble penetrants are more permeable. In 1994 Stern [6] investigated the solubility and diffusivity coefficient of carbon dioxide in PDMS and found that CO2 solubility coefficient increased with pressure and diffusivity slightly decreased with pressure due to hydrostatic compression effects of silicone rubber membranes.

Figures 4 shows the permeability of pure CO2 and CH4 against temperature through a silicone rubber membrane at a constant pressure of 8 bars. It was observed that with increasing temperatures the permeability of carbon dioxide and methane increased. But this increase in permeability with temperature was more for carbon dioxide than methane. The effect of temperature on permeability of carbon dioxide and methane through silicone rubber membrane could be due to two aspects. The amount of free volume in silicone rubber depended on temperature. Lower temperatures resulted in less free volume and consequently lowered the permeability of carbon dioxide and methane. Similar results was observed by other researchers [13]. The mobility of the polymer chain of silicone rubber was high at higher temperatures and this high mobility of the polymer chains could have enhanced the diffusion of carbon dioxide and methane molecules. Consequently, the permeability of CO2 and CH4 increased with increasing temperatures. Diffusivity greatly depends on the size and geometry of the

Figure 3. Permeability of pure CO2 and pure CH4 permeability through silicone rubber membrane against different pressure at 25°C.

Figure 4. Permeability of pure CO2 and CH4 through silicone rubber membrane against temperature at 8 bar.

Figure 5. Permeability of CO2 though silicone rubber membrane in binary mixture of CO2/CH4 against pressure at 25°C.



gas molecule. Since the geometry of CO2 molecule is linear whereas, CH4 molecule has a three dimensional tetrahedral geometry, diffusivity of CO2 through silicone rubber membrane was faster compared to CH4, resulting in a higher increase in permeability of CO2 compared to CH4.

Figures 5 shows permeability of carbon dioxide in the binary mixture of CO2/CH4 through silicone rubber membrane, against pressure. Carbon dioxide permeability increased with the increase in carbon dioxide concentration in the binary mixtures of carbon dioxide and methane. This result is consistent with the previous results [2].

For a binary gas mixture permeating through a polymeric membrane, the selectivity of a polymer membrane towards two different penetrant gases, CO2 and CH4, is commonly expressed in terms of the ideal selectivity or ideal permselectivity, α when the downstream pressure is negligible relative to the upstream pressure.

Table 1 presents the calculated selectivities of pure CO2 over pure CH4 in CO2/CH4 binary mixtures in relationship with pressure changes through a silicone rubber membrane. An increasing trend in ideal selectivity with pressure was observed up to 8 bars, because of the compressive nature of CH4 in silicone rubber membranes. Above 8 bars, there was a slight decrease in selectivity because of the increasing permeation rate of methane. Table 1 shows that selectivities estimated from pure gas

varied slightly with an increase in the feed pressure but reached a maximum value of 11.494. With 20% CO2 in the feed stream, the selelctivity is 2 to 3 times lower than that of pure components over the whole pressure range. High selectivities were obtained at 80% CO2 in the feed stream.

Table 2 shows effect of temperature on the ideal selectivity for pure CO2 and CH4 through the silicone rubber membrane. Table 2 also shows the effect of feed composition on the selectivity of a binary mixture of CO2/CH4 at different temperatures through the silicone rubber membrane. With 20% CO2 in the feed stream, the selectivities were two to three times lower than that of pure components over the whole temperature range.

Gas diffusion coefficients typically increase appreciably with increasing temperatures [8]. The temperature changes also affects the solubility selectivity, which is governed primarily by the chemical nature of the penetrant and polymer-penetrant interactions. For both CO2 and CH4 gases, as temperature increases, the solubilities increase. Since carbon dioxide is a more condensable component to methane, its solubility increase with temperature is higher than that of methane. The solubility selectivity, therefore, will vary depending on the extent of the temperature effect on each component in the gas mixture [8].

From the permeation behavior of pure gases and carbon dioxide-methane, a simple and practical mathematical model expressed in terms of pressure and feed composition was derived to predict quantitatively the permeability at different feed composition. A similar correlation has been previously reported [2]. The pure CO2 and CH4 permeability is correlated to the binary mixture of carbon dioxide and methane by the following empirical correlation.

(7)

(8)

purepurepure 4222

)1( CHCOCOmixCO pylpynpmP −++=

purepurepuremix 2444)1( COCHCHCH pylpynpmP −++=

Table 2 Selectivity of CO2 over CH4 through silicone rubber membrane at different temperatures.

Temperature (°C) Ideal Selectivity PCO2/PCH4

25 11.49

50 10.18

75 9.45

100 9.46

125 9.17

150 9.21



where mixCO2

P and mix4CHP are the permeability

of CO2 and CH4 in binary mixture of CO2 and CH4, y is the volumetric composition of CO2 in binary mixture of carbon dioxide and methane,

pureCO2p is the permeability of CO2 at its partial pressure pureCH4

p , is the permeability of CH4 at its partial pressure and m, n and l are the balancing coefficient. The comparison of carbon dioxide experimental permeability through silicone rubber membrane with the calculated permeability at different pressures is shown in Figure 6. A regression coefficient (R2 = 0.95) was obtained showing that the correlation is best fitted to the experimental data. Similarly, the comparison of methane experimental permeability with the calculated permeability at different pressure is shown in Figure 7. A regression coefficient of R2 = 0.79 was obtained showing that the correlation is fitted well with the experimental data.

cONcLuSION

The separation characteristics of pure carbon dioxide, methane and a binary mixture of carbon dioxide and methane through silicone rubber membranes was investigated systematically. The analyses was presented in terms of variations in the permeabilities of both gases, as well as their selectivities. Experimental results showed that the permeability of CO2 was found to depend strongly on operating pressure and temperature as compared to the permeability of CH4. It was also found that the

permeability of CO2 was higher than the permeability of CH4, and CO2 permeability decreased at pressures above 17 bars due to hydrostatic compression effects.

Generally, the permeability of a binary mixture of these gases followed the same trend as the permeability of pure gases. However, the magnitude of the permeability was determined by the amount of CO2 composition present in the feed. This is consistent with earlier reports in literatures. Analysis of the selectivities estimated from pure gas permeabilities showed that silicone rubber membrane has the highest selectivity of 11.4 over the pressure range of 2 to 24 bars. For gas mixture, lower values of the selectivities were obtained at low CO2 vol% composition.

An analytical model expressed in terms of feed pressure and feed composition was developed to quantitatively predict the permeability of binary mixture of gases from pure gas permeability. The model was expressed in terms of two controllable feed parameters, compared to a common permeation model which depends on some intrinsic factors such as the diffusion coefficient or solubility coefficent. Comparison of the experimental data with the calculated values showed an excellent agreement. This model is practical in choosing the optimal separation conditions of gas mixture.

Figure 6. Comparison of experimental and calculated permeabilities of CO2 through silicone rubber membrane in binary mixtures of methane and carbon dioxide at different pressure

Figure 7. Comparison of experimental and calculated permeabilities of CH4 through silicone rubber membrane in binary mixtures of methane and carbon dioxide at different pressure



AcKNOWLEDGEMENT

The author would like to thank Universiti Teknologi PETRONAS, Malaysia for sponsoring this study

REFERENcES

[1] D. Dortmundt and K. Doshi, “Recent developments in CO2 removal membrane technology,” UOP LLC, USA, 1999.

[2] F. Wu, Lei Li, Z. Xu, S. Tan, Z. Zhang, “Transport study of pure and mixed gases through PDMS membrane,” Chem. Eng. J. vol. 117 (2006), p. 51-59.

[3] S.S. Dhingra and E. Marand, “Mixed gas transport through polymeric membranes,” J. Membr. Sci. vol. 141 (1998), p. 45–63.

[4] H. Ettouney and U. Majeed, “Permeability functions for pure and mixture gases in silicone rubber and polysulfone membranes: dependence on pressure and composition,” J. Member. Sci., vol. 135 (1997), p. 251–261.

[5] C.K. Yeom, S.H. Lee and J.M. Lee, “Study of transport of pure and mixed CO2/N2 gases through polymeric membranes,” J. Appl. Polym. Sci., vol. 78 (2000), p. 179–189.

[6] S.A. Stern, “Polymers for gas separation: the next decade,” J. Member. Sci., vol. 94 (1994), p. 1–65.

[7] R. Hughes and B.Q. Jiang, “The permeabilities of carbon dioxide, nitrous oxide and oxygen and their mixtures through silicone rubber and cellulose acetate membranes,” Gas Sep. Purif., vol. 9 (1995), p. 27–30.

[8] S.M. Jordan and W.J. Koros, “Permeability of pure and mixed gases in silicone rubber at elevated pressure,” J. Polym. Sci, Part B: Polym. Phys., vol. 28 (1990), p. 795–809.

[9] T.C. Merkel, V.I. Bondar, K. Nagai, B.D. Freeman and I. Pinnau, “Gas sorption, diffusion, and permeation in poly(dimethylsiloxane),” J. Polym. Sci. Part B: Polym. Phys., vol. 38 (2000), p. 415–434.

[10] R.S. Prabhakar, R. Roharjo, L.G. Toy, H.Q. Li, B.D. Freeman, “Self-consistent model of concentration and temperature dependence of permeability in rubbery polymer,” Ind. Eng. Chem. Res., vol. 44 (2005), p. 1547-1556.

[11] A. Cornesa, A. Fernando, J.A. Pitarch, I. Vicente-Mingaro and M.A. Rodriguez, “Separation of binary gas mixtures by means of sol-gel modified ceramic membranes, prediction of membrane performance,” J. Membr. Sci., vol. 107(1995), p. 1-21.

[12] S.D. Burnside and E.P. Giannelis, “Synthesis and properties of new poly(dimethylsiloxane) nanocomposites,” Chemistry of materials, vol. 7 (1995), p. 1597.

[13] J.G. Wijmans and R.W. Baker, “The Solution-diffusion model: A review,” J. Membr. Sci., vol. 107 (1995), p. 1-21.

Dr Farooq Ahmad received his undergraduate degree from the University of Engineering and Technology, Peshawar Pakistan in 2000 and later, his Masters degree from the University of Engineering and Technology, Lahore, Pakistan. He later obtained his PhD in Chemical Engineering from Universiti Teknologi PETRONAS Malaysia in 2008. He is currently a lecturer

at the Department of Chemical Engineering, Universiti Teknologi PETRONAS and his area of research include gas separation membrane and environmental engineering.

Hilmi Mukhtar is currently the Director of Undergraduate Studies at Universiti Teknologi PETRONAS. Before joining UTP, he served as a faculty member of Universiti Sains Malaysia for six years and was the former Deputy Dean of the School of Chemical Engineering. He obtained his BEng in Chemical Engineering from the University of Swansea, Wales in 1990 and

later completed his MSc in 1991 and PhD in 1995 from the same university. He has deep research interests in the area of natural gas purification using membrane process and environmental issues particularly wastewater treatment and carbon trading.

Associate Professor Dr Zakaria Man is a lecturer of polymer and petrochemical related subjects at the Chemical Engineering Department of Universiti Teknologi PETRONAS. His current research interest is in polymer blend, formation and properties of epoxy resins and polymer composite. He started his career in 1985 with PETRONAS Research Institute after

graduating from Universiti Sains Malaysia in Applied Science (Polymer). Initially he was responsible for petroleum products quality control and special investigation works. Later he was involved in polymer and petroleum related research works. He continued his studies at Universiti Kebangsaan Malaysia in 1991 and obtained his MSc in Polymer chemistry. In 1994, after seven years working with PETRONAS Research Institute he was transferred to Polyethylene Malaysia Sdn. Berhad (PEMSB) and was appointed as Polymer Technologist, working along side with his counterpart from British Petroleum. His new tasks were product troubleshooting, plant experimental trials, development of new catalyst formulation and product quality optimization. In 1997 he joined Universiti Teknologi PETRONAS (UTP) as a lecturer. He later pursued his PhD in Polymer Science and Technology at the Polymer Science Centre, University of Manchester Institute of Science and Technology (UMIST), United Kingdom from 2000-2003.



INTRODucTION

In 1978, J. Davidovits initiated inorganic polymeric material that can be used to react with another source material to form a binder. Recent applications of this binder is focused on the substitution of the ordinary Portland cement (OPC) portion in concrete [3]. The environmental issues caused by OPC production has further encouraged progress in the research on polymers. Depletion of raw materials and CO2

emission from fuel combustion and decomposition of limestone has placed the cement industry as one of the main culprits of environment pollution [6].

The objective to produce environment friendly concrete can be achieved by limiting the utilization of raw materials and decreasing pollutant rates from the respective OPC productions, and diminishing the cement portion in concrete [5]. Substitutions using waste materials like fly ash, rice husk ash, and other cement replacement materials (CRM) can only account for a limited percentage of the cement portion. Geopolymer, named after the reaction between polymer and the geological origin source material, has been proposed to replace all cement portions in concrete as the main binder [2].

DEVELOPMENT OF POLYMERIc cONcRETE FOR SuSTAINABLE FuTuRES

M.F. Nuruddin*, A. Kusbiantoro and S. QaziUniversiti Teknologi PETRONAS, 31750 Tronoh, Perak Darul Ridzuan, Malaysia

*[email protected]

ABSTRAcT

Environmental issues resulted from cement production have become a major concern today. World’s plan to develop sustainable futures encourages people to limit the usage of this construction material that can damage environment quality. Cement replacement material was proposed to partially replace cement portion in concrete. However another material is needed to fully replace cement function; hence can reduce cement demand throughout the world. Geopolymer is a part of inorganic polymer material that has similar bonding function like cement in concrete. It consists of alkaline solutions and geological source material. Alkaline liquids used in this research are 8M sodium hydroxide (NaOH) solution and sodium silicate (NaSiO2) solutions, while source materials are fly ash and microwave incinerated rice husk ash (MIRHA). Three different curing regimes, namely hot gunny curing, ambient curing, and external exposure curing, were applied to obtain suitable method that was suitable with cast in situ application. Geopolymer concrete samples were tested on their compressive strength and microstructure properties. It was found that external exposure curing had the highest compressive strength compared to other two curing methods. Scanning electron microscopy analysis also showed better improvement in interfacial transition zone for concrete sample with external exposure curing.

Keywords: geopolymer, sodium hydroxide, sodium silicate, fly ash, MIRHA

This paper was presented at the 13th International Conference on Structural and Geotechnical Engineering (ICSGE), Cairo, 27 - 29 December 2009.



The main constituents of geopolymer are alkaline liquid and source material. Alkaline liquid is usually a combination of sodium hydroxide or potassium hydroxide with sodium silicate or potassium silicate [1]. The use of only alkaline hydroxide activator will result in a low rate of reaction compared to those containing soluble silicate (Palomo et al., 1999). The addition of sodium silicate to the sodium hydroxide solution will enhance the reaction rate between the alkaline liquid and the source material [8].

Source materials used in this research were combinations of fly ash and microwave incinerated rice husk ash (MIRHA). Both of these materials have similar specifications for calcium content, i.e. low in calcium. High calcium content in source materials is not recommended as it can obstruct the polymerization process [4]. A blended composition of fly ash and MIRHA was observed for the effect of a different silicate content in the source material (MIRHA has a higher silicate content than fly ash) [6].

Several experimental set ups on geopolymer concrete has been conducted. For example, the curing method was found to have some limitations to the geopolymer concrete applications. Heat required in the curing process can only be supplied by electrical instruments, hence it is currently only applicable in the precast concrete industry. This research focused on a curing method for geopolymer concrete suitable for cast in situ applications.

EXPERIMENTAL METHODS

Materials

Alkaline liquids in this research were obtained from a supplier in Malaysia with specific requirements. NaOH was supplied by QuickLab Sdn Bhd, Malaysia in pellet form with 99% purity. 8 molar NaOH solution was used for all mix proportions. NaSiO2 was obtained from Malay-Sino Chemical Industries Sdn Bhd, Malaysia with proportion of Na2O: 14.73%, SiO2: 29.75%, and water: 55.52%.

Fly Ash was obtained from Manjung Power Plant, Malaysia with oxide compositions as described in Table 1. Rice husk used to produce MIRHA was from Bernas Milling Plant, Malaysia. Rice husk was first dried under direct sunlight to remove its moisture content to reduce excess smoke from its combustion. Rice husk was then incinerated in a microwave incinerator at 400°C. The UTP Microwave Incinerator (UTPMI) was the Air Cooled Magnetron system with an overall dimension of 2.3(H) x 4.0(W) x 4.0(L) and a chamber capacity of 1 m3. MIRHA was ground 2000 times to fineness in a ball mill. The oxide content of MIRHA is described in Table 2. Coarse aggregates of 20 mm size were prepared under saturated surface dry (SSD) conditions. Since commercial water-reducing admixture was not suitable for mixtures in this research, glucose solution was added into the

Table 1. Fly Ash Chemical Composition

Oxide Percentages (%)

SiO2 51.19 %

Al2O3 24.00 %

Fe2O3 6.60 %

CaO 5.57 %

MgO 2.40 %

SO3 0.88 %

K2O 1.14 %

Na2O 2.12 %

Table 2. MIRHA Chemical Composition

Oxide Percentage (%)

SiO2 88.90 %

MgO 0.72 %

SO3 0.32 %

CaO 0.63 %

K2O 3.65 %

Al2O3 0.16 %

Fe2O3 0.45 %



mixture to delay the setting time during the mixing and casting process.

Experimental Setup

The mixture proportion was designed with different amounts of blended source materials to investigate the resultant geopolymer concrete properties. Constant amounts of NaOH and NaSiO2 were used throughout the mix proportions. Details of each mixture is described in Table 3. Alkaline solutions were prepared fresh 1 hour before the mixing process to avoid formation of precipitates in the NaOH solution. The mixing process was divided into two stages: dry mix and wet mix. Initially, coarse aggregate, fine aggregate and fly ash (and MIRHA), were mixed in a rotating pan mixer for 3.5 minutes. Alkaline and glucose solutions were then poured into the dry mixed material and continued as a wet mix for 1.5 minutes. Fresh geopolymer concrete was then hand-mixed to ensure homogeneity of the mixture.

Fresh concrete was then cast in 100 mm cube moulds and compacted using a poker vibrator. The curing process was divided into three curing methods: hot gunny curing, ambient curing and external exposure curing. In hot gunny curing, concrete samples were covered with a hot gunny sack for 48 hours (the hot gunny was replaced every 24 hours). To prevent heat being released immediately, the samples were covered with plastic sheet. In ambient curing, concrete samples were placed outside the room while still protected from direct sunlight and rain. In external exposure curing, concrete samples were placed in a transparent chamber that placed outside, hence heat radiation from sunlight can penetrate into the chamber while still protect the samples from rainfall.

Hardened concrete samples were tested for their compressive strength at 3, 7, 28, and 56 days for all curing regimes. Analysis of interfacial transition zones (ITZ) and microstructure properties were conducted using scanning electron microscopy (SEM) on 56-day concrete sample.

Table 3. Details of Mix Proportion

Mix code* Fly Ash (kg/m3)

MIRHA(kg/m3)

cA FA NaOH NaSiO2 Water Sugar(kg/m3)(kg/m3) (kg/m3) (kg/m3) (kg/m3) (kg/m3)

A1 350 0 1200 645 41 103 35 10.5

A2 339.5 10.5 1200 645 41 103 35 10.5

A3 332.5 17.5 1200 645 41 103 35 10.5

A4 325.5 24.5 1200 645 41 103 35 10.5

B1 350 0 1200 645 41 103 35 10.5

B2 339.5 10.5 1200 645 41 103 35 10.5

B3 332.5 17.5 1200 645 41 103 35 10.5

B4 325.5 24.5 1200 645 41 103 35 10.5

C1 350 0 1200 645 41 103 35 10.5

C2 339.5 10.5 1200 645 41 103 35 10.5

C3 332.5 17.5 1200 645 41 103 35 10.5

C4 325.5 24.5 1200 645 41 103 35 10.5

* A: hot gunny curing; B: ambient curing; C: external exposure curing




compressive Strength Test

The results obtained from compressive strength tests illustrate the polymeric reactions in geopolymer concrete with various curing treatments. Compressive strength development of geopolymer concrete for all curing regimes is described in Table 4.

In hot gunny curing, compressive strength did not develop rapidly. The presence of high humidity from the hot gunny significantly decreased concrete strength because the heat from the hot gunny could not be maintained for a long duration. In contrast with conventional OPC concrete, water in geopolymer concrete did not take part in the polymeric reaction. It helped only during the mixing and casting process to increase its workability.

However, during drying period (after 48 hours), water evaporated from the hardened concrete and left micropores inside the concrete. The presence of these pores led to the premature failure of the

concrete sample. Blended source material with 95% fly ash and 5% MIRHA provided better results than other samples in the hot gunny curing as showed in Figure 1. The compressive strength for a blended sample increased by 36% compared to a non-blended sample of the same curing regime It showed that different amounts of Al-Si materials in fly ash and MIRHA affected the properties of geopolymer concrete.

In ambient curing, heat generated from the environment was absorbed by the polymeric material to initiate the reaction. In this regime,

Table 4. Compressive Strength Development of Geopolymer Concrete

Type of curing code

% Replacement

by MIRHA

compressive Strength (MPa)

3 days 7 days 28 days 56 days

Hot gunny Curing

A1 0 4.67 10.53 15.74 16.62

A2 3 3.43 8.41 12.83 14.96

A3 5 7.11 11.31 19.01 22.66

A4 7 3.19 7.01 11.5 12.77

Ambient Curing

B1 0 7.46 14.11 19.73 21.92

B2 3 6.3 11.05 17.92 19.19

B3 5 9.38 14.74 25.3 27.28

B4 7 8.55 11.35 16.75 17.03

External Exposure Curing

C1 0 32.78 44.76 48.88 50.96

C2 3 19.93 24.96 35.2 36.35

C3 5 14.14 24.29 28.7 33.62

C4 7 16.81 30.22 41.34 44.84

0

5

10

15

20

25


Com

pres

sive

Str

engt

h (M

Pa)

Curing Days

A1

A2

A3

A4

Figure 1. Compressive Strength Development of Concrete Sample with Hot Gunny Curing



compressive strength development also showed a similar trend to that of hot gunny curing. However, the difference in the amount of heat in the first week of the maturing period affected the performance of these two methods. Concrete samples in ambient curing did not have excess water in the environment from the first day; hence the polymeric reaction could take place faster than those in hot gunny curing. Compressive strengths of up to 24% was measured compared to non-blended samples in the same curing regime and 64% compared to non-blended hot gunny concrete samples. Similar to the hot gunny mixture, the mixture of 95% fly ash and 5% MIRHA performed better than other ambient curing samples as described in Figure 2.

Different characteristics were observed for the external exposure curing. Sufficient amounts of heat during daylight accelerated the polymeric reaction inside the concrete samples. Figure 3 shows that non-blended samples provided the best development among all concrete samples. On average, concrete samples in the external exposure curing showed compressive strengths 162% higher than the hot gunny curing samples, and 102% higher than the ambient curing samples. It was also observed that the critical period for geopolymer concrete is within the first week of the mixing-casting process. The lack of polymeric reaction within this period results in the lower performance of geopolymer concrete. Like

conventional OPC concrete, strength development in geopolymer concrete stabilises only after 28 days.

Scanning Electron Microscopy (SEM) Analysis

SEM analysis was carried out to observe the microstructure properties of geopolymer concrete. Figures 4, 5 and 6 shows the inner condition for hot gunny curing, ambient curing and external exposure curing concrete samples respectively.

Figure 4 shows the presence of gaps at the interfacial transition zone between aggregate and paste. It is the result of the low bonding strength of geopolymer paste caused by low amounts of heat absorbed during polymeric reaction in the maturing period.

Figure 2. Compressive Strength Development of Concrete Sample with Ambient Curing

Figure 3. Compressive Strength Development of Concrete Sample with External Exposure Curing

Com

pres

sive

Str

engt

h (M

Pa)

Curing Days

0

5

10

15

20

25

30


B1

B2

B3

B4 Com

pres

sive

Str

engt

h (M

Pa)

Curing Days

0

10

20

30

40

50

60


C1

C2

C3

C4

Figure 4. SEM Image of Hot Gunny Curing Sample



The ITZ connects other microcracks and micropore paths in the paste matrix and if the concrete sample has loads, it could result in premature failure of the concrete sample.

In Figure 5, the presence of ITZ was covered by a paste matrix which resulted in the higher compressive strength of the concrete samples. Even though it did not completely cover the paste matrix, the increased bonding strength showed a better performance than the hot gunny curing.

The superior strength of the external exposure curing sample was verified by the SEM image as shown in Figure 6. It supports the analysis that an elevated temperature is important to accelerate the polymeric reaction in geopolymer concrete. With the strong bonding between aggregate and paste, microcrack paths will be discontinued with loads applied on the concrete and would result in higher compressive strengths of concrete samples.

cONcLuSION

In order to develop a suitable geopolymer concrete for the sustainable future, a proper curing method is required in the production of the geopolymer concrete. In high humidity environments, blended source material of 95% fly ash and 5% MIRHA could improve geopolymer concrete performance up to 24% compared to non-blended concrete. However

Figure 5. SEM Image of Ambient Curing Sample

in elevated temperatures, the blended source material did not significantly improve concrete strength compared to the non-blended concrete. The concrete compressive strength in the external exposure curing was 162% higher than the hot gunny curing and 102% higher than the ambient curing. It shows that a sufficient amount of heat is needed to increase polymeric reaction rate.

AcKNOWLEDGEMENTS

The authors would like to acknowledge Universiti Teknologi PETRONAS, Malaysia for the research financial support.

REFERENcES

[1] V.F.F. Barbosa, K.J.D. MacKenzie and C. Thaumaturgo, “Synthesis and characterisation of materials based on inorganic polymers of alumina and silica: sodium polysialate polymers,” International Journal of Inorganic Materials, vol. 2(4), 2000, pp. 309-317

[2] J. Davidovits, “High alkali cements for 21st century concretes,” Concrete Technology: Past, Present and Future, P. K. Mehta, ACI, Detroit, USA, 1994, SP 144-19, pp. 383-397.

[3] P. Duxon, J.L. Provis, G.C. Lukey and J.S.J. van Deventer, “The Role of Inorganic Polymer Technology in the Development of ‘Green Concrete’,” Cement and Concrete Research, vol. 37, 2007, pp. 1590 – 1597

[4] J.T. Gourley, “Geopolymers: opportunities for environmentally friendly construction materials,” Materials 2003 Conference: Adaptive Materials for a Modern Society, Sydney, Institute of Materials Engineering Australia, 2003.

[5] P.K. Mehta, “Greening of the Concrete Industry for Sustainable Development,” ACI Concrete International, vol. 24 (7), 2002, pp. 23 – 28

Figure 6. SEM Image of External Exposure Curing Sample



[6] M.F. Nuruddin, A. Kusbiantoro and N. Shafiq, “Microwave incinerated rice husk ash (MIRHA) and its effect on concrete properties,” IMS International Conference, 2008, American University of Sharjah.

[7] A.M.W. Palomo and M.T. Blanco, “Alkali-activated fly ashes, a cement for the future,” Cement and Concrete Research, vol. 29(8), 1999, pp. 1323-1329.

[8] H. Xu and J.S.J. van Deventer, “The Geopolymerisation of Alumino-Silicate Minerals,” International Journal of Mineral Processing, vol. 59(3), 2000, pp. 247-266.

Associate Professor Ir Dr Muhd Fadhil Nuruddin has been involved with research related to carbonation in concrete, cement replacement materials, polymeric concrete and concrete technology for more than 20 years now. He is the Past President and a Fellow Member of The Concrete Society Of Malaysia (FCSM), a Member of The Institution Of Engineers Malaysia (Miem),

and Board Of Director Of The International Association On Concrete Technology (IACT). To date he has published more than 100 technical papers at international and national levels.

Andri Kusbiantoro is a Doctoral of Philosophy student in concrete technology. He obtained his MSc from UTP and his main research areas relate to concrete technology, cement replacement materials, and polymeric concrete.

Sobia Anwar Qazi graduated in Civil Engineering from NED University of science and technology, Karachi, Pakistan in 2007. From there, she joined UTP to pursue her MSC. Her research topic is ‘Mix Design of polymeric Concrete Incorporating Fly ash, Rice husk ash and Silica fume’



INTRODucTION

Natural convection in vertical channels is of interest in many technological applications such as cooling of electronic equipment, solar collectors, passive heating of buildings, channel nuclear reactors, etc. The study of fully developed free convection flow and steady state heat transfer between two parallel plates at constant temperature was initiated by Ostrach [1]. Also, Ostarch [2] studied the combined effects of the steady free and forced convective laminar flow and heat transfer with linear wall temperature profile in a vertical channel. The first exact solution for free convection in a vertical parallel plate channel with asymmetric heating for a fluid of constant properties was presented by Aung [3]. Some of the early studies in this field were those by Bodoia and Osterle [4], Aung et al. [5], Miyatake and Fujii [6], Miyatake et al. [7]. Recent contributions on this problem include the numerical investigation of natural convection flow by Lee and Yan [8] in a vertical channel with unheated entry and unheated exit sections. The study revealed interesting results

on the induced volumetric flow rate of the fluid and the Nusselt number. Many of the early work have been reviewed by Manaca et al. [9]. Olsson [10] derived relations for buoyancy driven flow in vertical channels for Nusselt number and flow rate, for cases where the Rayleigh number range spans from fully developed duct flow to isolated plate boundary layer flow. Campo et al. [11] considered natural convection for heated iso-flux boundaries of the channel containing a low-Prandtl number fluid. Pantokratoras [12] studied the fully developed free convection flow between two isothermal vertical parallel plates with large temperature differences for a fluid of varying thermophysical properties. All these studies are restricted to fully developed steady state free convection flows for different physical situations. The unsteady laminar free convection flow between infinite vertical parallel surfaces has its applications in a multitude of technological processes such as the early stages of melting, transient heating of insulating air gaps by heat input at the start-up of furnaces, some types of passive solar heating and ventilating systems. Singh et al. [13] studied the

NATuRAL cONVEcTION FLOW IN VERTIcAL cHANNEL DuE TO RAMPED WALL TEMPERATuRE AT ONE BOuNDARY

M. Narahari and Vijay R RaghavanUniversiti Teknologi PETRONAS, 31750 Tronoh, Perak Darul Ridzuan, Malaysia

ABSTRAcT

An exact solution of unsteady natural convection of a viscous incompressible fluid in a vertical channel due to ramp heating at one boundary is presented. The temperature at one channel plates increases linearly over a certain time period, then remains constant. That at the other plate is maintained at the initial fluid temperature. The influence of the physical parameters on the velocity field, the temperature field, rate of heat transfer, skin-friction and volumetric flux of the fluid are analyzed systematically. Natural convection due to ramp heating is compared with the baseline case of flow with constant temperature.

Keywords: free convection, ramped temperature, heat transfer, vertical parallel plate channel

This paper was presented at the 2009 ASME Summer Heat Transfer Conference, California, 19 – 23 July 2009



transient free convection flow between two long vertical parallel plates maintained at constant but unequal temperatures. The same problem was extended by Jha et al. [14] to consider symmetric heating of the channel walls. Narahari et al. [15] analyzed the transient free convection flow between two long vertical parallel plates with constant heat flux at one boundary and the other being maintained at a constant initial fluid temperature. Singh and Paul [16] presented an analysis of the transient free convective flow of a viscous and incompressible fluid between two parallel vertical walls occurring as a result of asymmetric heating/cooling of the walls. Recently, an exact analytical solution to the problem of unsteady free convective flow and heat transfer of a viscous incompressible fluid between two long vertical parallel plates with the plate temperature linearly varying temperature with time at one boundary and the other being held at a constant temperature was presented in [17]. All these were performed by assuming the plates at two different constant temperatures or temperature of the plates varying linearly along the plates.

However, several practical problems may necessitate considering wall conditions which are non-uniform or arbitrary. To understand such situations it is useful to investigate problems subjected to step discontinuities in the wall temperature. An approximate analytical model for the free convection flow from a vertical plate with discontinuous wall temperature conditions was developed by Schetz [18]. A new analytical model for the laminar natural convection from a vertical plate with a step change in wall temperature was presented by Lee and Yovanovich [19]. The validity and accuracy of the model was demonstrated by comparing with the existing results. The unsteady natural convection flow of an incompressible viscous fluid near a vertical plate with ramped wall temperature was studied by Chandran et al. [20] and they compared the natural convection flow near a ramped temperature plate with the flow near a plate with constant temperature. Recently, the natural convection boundary layer adjacent to an inclined semi-infinite flat plate subjected to ramp heating was investigated by Saha

et al. [21]. The flow development from the start-up to a steady state has been described based on scaling analysis and verified by numerical simulations.

In the present investigation, we have considered the unsteady natural convection flow between two infinite vertical parallel plates subject to a temperature boundary condition which follows a ramp function up until some specified time and then remains constant at one boundary and the other being held at a constant initial fluid temperature. Closed form analytical solutions to velocity distribution, viscous drag, temperature distribution and heat flux have been derived with the help of Laplace transform technique. The natural convection due to ramp heating has also been compared with the baseline case of flow due to constant temperature.

GOVERNING EQuATIONS AND THEIR SOLuTIONS

Here, an unsteady natural convection flow of a viscous incompressible flow through an infinite vertical parallel plate channel is considered. The x'-axis is taken along the vertical direction and the y’-axis normal to it. Initially both the plates and the enclosed fluid are at a temperature, dT′ and the fluid is stagnant. At time 0>′t , the surface temperature of the plate at y′=0 is raised or lowered to 0/)( ttTTT dwd ′′−′+′ when 0tt ≤′ , and thereafter, for 0tt >′ , is maintained at the constant temperature

wT ′ while that of the other plate at dy =′ continues to remain at the initial temperature dT′ , causing the flow of free convection currents. Then the flow can be shown to be governed by the following equations under usual Boussinesq’s approximation [13-17]:

(1)

and

(2)

with following initial and boundary conditions

2

2

y

Tk

t

TCp ′∂

′∂=′∂′∂ρ

2

2

y

Tk

t

TCp ′∂

′∂=′∂′∂ρ



(3)

All the quantities are defined in Nomenclature. It is assumed that the upper and lower ends of the channel between two parallel surfaces are open to the ambient medium at temperature dT′ .

The non-dimensional quantities are defined as follows:

(4)

Then in view of equations (4), equations (1) – (3) reduce to the following non-dimensional forms.

(5)

(6)

According to the above non-dimensional variables, the characteristic time can be defined as

(7)

The initial and boundary conditions given in Eq. (3) now becomes

(8)

>′=′′=′=′

>′=′′=′

≤′<=′′′−′+′=′

>′=′=′

≤′≤′≤′=′=′

0forat,0

for0at

0for0at)(

0for0at0

0and0for,0

0

00

tdyTTu

ttyTT

ttyt

tTTTT

tyu

tdyTTu

d

w

dwd

d

=′−′′−′

=′−′

=

′=

′−′′

=′

=′

=

.Pr,,)(

,)(

,,

2

3

20

k

C

TT

TTdTTgGr

Gr

du

TTgd

uu

t

tt

d

yy

p

dw

ddw

dw

µθ

νβ

νβν

2

2

y

u

t

u

∂∂+=

∂∂ θ

2

2

Pryt ∂

∂=∂∂ θθ

ν

2

0d

t =

>===

>==

≤<==

>==

≤≤≤==

0for1at0,0

1for0at1

10for0at

0for0at0

0and10for0,0

tyu

ty

tyt

tyu

tyu

θ

θ

θ

θ

These Eqs. (5) and (6), subject to the conditions (8) has been solved by the Laplace transform technique and the solutions are derived under various conditions.

Case 1: 1Pr ≠

(9)

(10)

Case 2: 1Pr =

(11)

where yncynbyna ++=−+=+= 22,22,2

and H is the unit step function which, in general, defined as

Here k is a constant, z is a dummy variable and f1, f2, f3 are functions of dummy variable.

{[

}{

} ])1()1,Pr(

)1,Pr()1,()1,(

),Pr(),Pr(

),(),()1(Pr

1),(

1

111

11

011

−−+

−−−−−−

+−

−−

= ∑∞

=

tHtbf

taftbftaf

tbftaf

tbftaftyun

{ }[{ } ])1()1,Pr()1,Pr(

),Pr(),Pr(),(

22

022

−−−−−

−= ∑∞

=

tHtbftaf

tbftaftyn

θ

{ }[

{ }

{ }])1()1,(),(2

)1()1,(),(2

)1()1,(),()1(),(

33

33

033

−−−+

−−−−

−−−+= ∑∞

=

tHtaftafy

tHtbftbfb

tHtcftcfntyun

( )

+++

−+−=

t

zttzz

t

zttz

ztzf

2erfc1212

24

1

4exp)/()10(

12),(

224

22

1 π

++

−−=

t

zt

z

t

ztztzf

2erfc

24exp)/(),(

22

2 π

( )

+−

−+=

t2

zerfc

64exp)/(4

3

1),(

322

3z

ztt

zttztzf π

≥<≤=− .1

,00)( ktktktH



NuSSELT NuMBER AND SKIN-FRIcTION

The other two practically important measures of transport of energy and momentum are the heat flux and the skin-friction at the plates. These quantities in dimensionless form can be derived from the solutions for the distribution of temperature [Eq. (10)] and velocity [Eqs. (9) and (11)]. The dimensionless heat flux which is nothing but the Nusselt number at the two plates can be derived as

(12)

and

(13)

The non-dimensional skin-friction at the plates are given as

Case 1: 1Pr ≠

(14)

{ }[

{ } ])1()1,1()1,(

),1(),(

)(Nu

44

044

0

00

−−++−−

++=

∂∂−=

′−′′

=

∑∞

=

=

tHtnftnf

tnftnf

yTTk

dq

n

ydw

θ

∑∞

=

=

−

−+−

+=

∂∂−=

′−′′

=

044

1

11

)1(1,2

12,

2

122

)(Nu

n

ydw

tHtn

ftn

f

yTTk

dq θ

0

00 )( =∂

∂=′−′

′=

ydw y

u

TTdgβρτ

τ

{[

}{

} ])1()1,1(

)1,()1,1()1,(

),1(),(

),1(),()1(Pr3

4

6

655

66

055

−−+−

−−−++−−

+−−

++−

= ∑∞

=

tHtnf

tnftnftnf

tnftnf

tnftnfn

and

(15)

Case 2: 1Pr =

(16)

and

(17)

where

{ }[

{ } ])1()1,12()1,12(

),12(),12()1(Pr3

1

)(

87

087

1

11

−−+−−+−

+−+−

−=

∂∂−=

′−′′

=

∑∞

=

=

tHtnftnf

tnftnf

y

u

TTdg

n

ydwβρτ

τ

{[

}{

{ } }])1()1,1()1,(

),1(),()3/1(

)1()1,22(

),22()1(2

55

55

2

02

00

−−++−−

+++

−−+−

++−=

∂∂=

∑∞

=

=

tHtnftnf

tnftnf

tHtnf

tnfn

y

u

n

y

τ

{[

}{

{ } }])1()1,1()1,(

),1(),()3/1(

)1()1,22(

),22()1(2

55

55

2

02

00

−−++−−

+++

−−+−

++−=

∂∂=

∑∞

=

=

tHtnftnf

tnftnf

tHtnf

tnfn

y

u

n

y

τ

−

−=

t

zz

t

zttzf

PrerfcPr2

Prexp)/Pr(2),(

2

4 π

( )

−+−

+=

t

ztzt

t

ztzztzf

222

5 exp)/(erfc2

3),( π

( )

−+−

+=

t

ztzt

tzztzf

PrexpPr)/Pr(

t

Przerfc

2

Pr3Pr),(

22

226

π

( ) ( )

−+−

+=

t

ztzttzztzf

4exp4)/(2

t2

zerfc6),(

222

7 π



and 87654 ,,,, fffff are functions of dummy variable.

( )

( )

−+−

+=

t

ztzt

t

ztzztzf

4

Prexp4Pr)/Pr(2

2

PrerfcPr6Pr),(

22

228

π


This study investigated the effects of Prandtl number and time on the unsteady natural convection flow and heat transfer in a vertical parallel plate channel subject to a ramped wall temperature at the

Figure 1. Velocity profiles for Pr=0.71 (air)

Ramped: Solid curvesIsothermal: Dashed curves

Figure 2. Velocity profiles for Pr=7.0 (water)


Figure 3. Temperature profiles for Pr=0.71 (air)

Figure 4. Temperature profiles for Pr=7.0 (water)





boundary y=0 and the other at y=1 was maintained at the initial fluid temperature. In order to highlight the influence of the ramped temperature, the present study has also been compared with the study of free convection flow between two infinite vertical parallel plates with constant temperature at one boundary (at y=0) and the other (at y=1) being held at initial fluid temperature [13]. The expressions for the velocity and temperature fields have been derived by the usual Laplace transform technique in series form. The series in Eqs. (9), (10), and (11) can be shown to be absolutely convergent because of the presence of standard mathematical functions. The numerical values of the velocity, temperature, Nusselt number, skin-friction and volume flow rate are computed for different Prandtl number and time. The results are limited to two most commonly

occurring fluids, atmospheric air (Pr=0.71) and water (Pr=7.0).

The velocity and temperature profiles of air and water have been shown in Figures 1 to 4 for different values of time . It was also observed from Figures 1 and 2 that an increase in the Prandtl number has a diminishing effect on the velocity but the velocity increases with increasing t. Hence the flow is accelerated upwards in the channel with greater values of t. It was also noted that the velocity profiles are slightly skewed towards the plate having an increasing temperature and correspondingly lower the density of the fluid in its vicinity. Also, seen was that the transient velocity is much greater in the isothermal plate case than in the case of ramped temperature of the plate at y=0 for small values of

(Nu 0

)

Pr = 7.0 = 0.71

Figure 5. Variation of the Nusselt number (Nu0) for different Pr


(Nu 1

)

Pr = 7.0

Pr = 0.71


Figure 6. Variation of the Nusselt number (Nu1) for different Pr

(τ0)

Pr = 0.71

Pr = 7.0

Figure 7. Variation of the skin-friction (τ0) for different Pr

(τ1)

Pr = 0.71

Pr = 7.0

Figure 8. Variation of the skin-friction (τ1) for different Pr





t and this difference is not significant at large values of t.

Close inspection of the velocity profiles also reveals that the transient velocity reaches the steady state velocity earlier in the case of isothermal plate than in the case of ramped temperature of the plate at y=0 for low Prandtl number fluids and this time gap is not significant for high Prandtl number fluids.

Observed in Figures 3 and 4 is that the temperature increases with increasing time but it falls owing to an increase in the Prandtl number. The fall in temperature with increasing y is sharper for water than air because of the lesser contribution of transport of energy (heat) as compared to that of momentum. The fluid temperature is greater in the isothermal plate case than in the case of ramped temperature of the plate at y=0. This is because in the latter case, the heating of the fluid takes place more gradually than in the former case. For small values of t, the transient temperature due to isothermal plate is much higher than the transient temperature due to ramped temperature of the plate at y=0 and this difference is not significant at large values of t. Also, the transient temperature reaches the steady state temperature earlier in the case of isothermal plate than that of the plate with ramped temperature at y=0 for low Prandtl number fluids and this time gap is not significant for higher Prandtl number fluids.

The variation of the Nusselt number against time t is plotted in Figures 5 and 6 for different values of Pr. The ramped temperature plate in Figure 5 shows that the Nusselt number Nu0 increases for 0<t <1, and decreases for t > 1 for both air and water. Whereas for the case of the isothermal plate, Nu0 decreases with increasing t. It was observed from Figure 6 that the Nusselt number Nu1 increases with increasing t in both ramped and isothermal cases. From these figures, it was also clear that an increase in Prandtl number leads to an increase in Nu0 but a decrease in Nu1. The Nusselt number Nu0 is much higher than Nu1 indicating that much of the energy released from the plate at y=0 because of its higher temperature is convected out by the fluid before it reaches the plate at y=1. Close observation of the profiles also revealed that both Nu0 and Nu1 reaches the steady state earlier in the case of isothermal plate than in the case of ramped temperature of the plate at y=0 but this time difference is not significant for higher Prandtl number fluids.

Computed vales of the dimensionless shear stress against the time t were plotted in Figures 7 and 8 for different values of Pr. It was observed from these figures that the skin-friction increases with increasing time but decreases with a rise in the value of Prandtl number. Physically, this is possible because fluids with high Prandtl number move slowly and hence there is less friction at the plate. The variation of the skin-friction is significant at small values of t. The plate at y=0 is subject to a high shear stress than the one at y=1 because of steeper velocity profile in the vicinity. It was also noticed that the shear stress reaches steady state earlier in the isothermal case than in the case of ramped temperature for low Prandtl number fluids but this time gap is not significant for fluids with high Prandtl numbers.

The non-dimensional volumetric flux, Q, of the fluid can be calculated from the integral of the axial velocity distribution as

(18)∫=′−′

′=

1

03 )(

dyuTTgd

QQ

dwβν

Pr = 0.71

Pr = 7.0

Figure 9. Variation of the volume flow rate (Q) for different Pr




Computed time-varying values of Q obtained by numerical integration of the velocity using Simpson’s 1/3 rule were plotted in Figure 9 for different values of the Prandtl number. It was observed that the volume flux increases with increasing time but decreases with increasing Prandtl number. It was also noticed that the volume flux reaches steady state earlier in the case of isothermal plate than that of the plate with ramped temperature but it is not significant for high Prandtl number fluids.

cONcLuSION

An exact solution to the problem of unsteady natural convection flow in a vertical channel subjected to a ramped wall temperature at one boundary was determined with the help of Laplace transform technique. The solutions were obtained in the form of rapidly converging infinite series. The effects of the Prandtl number and time on the flow fields were discussed. It was found that the velocity decreases with increasing the Prandtl number but it increases with increasing time. The temperature increases with increasing time but it falls owing to an increase in the Prandtl number. The Nusselt number at the ramped temperature plate increases for 10 << t and decreases for t>1 for both air and water. An increase in the Prandtl number leads to an increase in Nu0 but a decrease in Nu1. The Nusselt number is consistently much higher at the plate with ramped temperature. The skin-friction increases with increasing time but decreases with a rise in the value of the Prandtl number. The plate with ramped temperature was subject to a high shear stress because of steeper velocity profile in the vicinity. The volume flux increases with increasing time but decreases with increasing Prandtl number. It was also found that the velocity, temperature, Nusselt number, skin-friction, and volume flux in the case of ramped temperature plate is always less than that of the case of an isothermal plate. The transient state reaches the steady state earlier in the case of isothermal plate than that of the plate with ramped temperature for low Prandtl number fluids and this time gap is not significant for fluids with higher Prandtl number.

NOMENcLATuRE

Cp = specific heat at constant pressure

d = distance between two plates

g = acceleration due to gravity

Gr = Grashof number

k = Thermal conductivity

u′ = velocity of the fluid

u = dimensionless velocity

T′ = fluid temperature

wT ′ = temperature of the plate at y′=0

dT′ = temperature of the plate at y′=dt ' = time

t0 = characteristic time

t = non-dimensional time

y = non-dimensional coordinate in -direction

x′, y′ = coordinate system

Q′ = mass flux

Pr = Prandtl number

0q′ = rate of heat transfer at the plate y′=0

1q′ = rate of heat transfer at the plate y′=dNu0 = the Nusselt number at the plate y=0Nu1 = the Nusselt number at the plate y=1

Greek symbols

ß = thermal expansion coefficient

v = kinematic viscosity

ρ = density

μ = viscosity

θ = dimensionless temperature

1τ ′ = skin-friction at the plate y′=0

1τ ′ = skin-friction at the plate y′=dτ0 = dimensionless skin-friction at the plate y=0τ1 = dimensionless skin-friction at the plate y=1

REFERENcES

[1] S. Ostrach, “Laminar Natural-Convection Flow and Heat Transfer of Fluids with and without Heat Sources in Channels with Constant Wall Temperatures,” Report No. NACA TN 2863, 1952.



[2] S. Ostrach, “Combined Natural and Forced Convection Laminar Flow and Heat Transfer of Fluids with and without Heat Sources in Channels with Linearly Varying Wall Temperature,” Report No. NACA TN 3141, 1954.

[3] W. Aung, “Fully Developed Laminar Free Convection Between Vertical Plates Heated Asymmetrically,” Int. J. Heat and Mass Transfer, vol.15(8), 1972, pp. 1577-1580.

[4] J.R. Bodoia and J.F. Osterle, “The Development of the Convection Between Heated Vertical Plates,” ASME J. Heat Transfer, vol. 84(1), 1962, pp. 40-44.

[5] W. Aung, L.S. Fletcher and V. Sernas, “Developed Laminar Free Convection Between Vertical Flat Plates with Assymmetric heating,” Int. J. Heat and Mass Transfer, vol. 15, 1972, pp. 2293-2308.

[6] O. Miyatake and T. Fujii, “Free Convection Heat Transfer Between Vertical Parallel Plates – One Plate Isothermally Heated and the Other Thermally Insulated,” Heat Transfer – Japanese Research, vol. 1(3), 1972, pp. 30-38.

[7] O. Miyatake, T. Fujii and H. Tanaka, “Natural Convection Heat Transfer Between Vertical Parallel Plates – One Plate with a Uniform Heat Flux and the Other Thermally Insulated,” Heat Transfer – Japanese Research, vol. 2(1), 1973, p. 25-33.

[8] K.T. Lee and W.M. Yan “Laminar Natural Convection Between Partially Heated Vertical Parallel Plates,” Wärme-und Stoffübertragung (Heat and Mass Transfer), vol. 29, 1994, pp. 145-151.

[9] O. Manca, B. Morrone, S. Nardini and V. Naso, Computational Analysis of Convection Heat Transfer, WIT Press, Southampton, GB, 2000, Chap. 7.

[10] C.O. Olsson, “Prediction of Nusselt Number and Flow Rate of Buoyancy Driven Flow Between Vertical Parallel Plates,” ASME J. Heat Transfer, vol. 126(1), 2004, pp. 97-104.

[11] A. Campo, O. Manca and B. Morrone, “Numerical Investigation of the Natural Convection Flows for Low-Prandtl Fluids in Vertical Parallel-Plates Channels,” ASME J. Applied Mechanics, vol. 73, 2006, pp. 96-107.

[12] A. Pantokratoras, “Fully Developed Laminar Free Convection with Variable Thermophysical Properties Between Two Open-Ended Vertical Parallel Plates Heated Asymmetrically with Large Temperature Differences,” ASME J. Heat Transfer, vol. 128, 2006, pp. 405-408.

[13] A.K. Singh, H.R. Gholami and V.M. Soundalgekar, “Transient Free Convection Flow Between Two Vertical Parallel Plates,” Wärme-und Stoffübertragung (Heat and Mass Transfer), vol. 31, 1996, p. 329-331.

[14] B.K. Jha, A.K. Singh and H.S. Takhar, “Transient Free Convection Flow in a Vertical Channel due to Symmetric Heating,” Int. J. Appl. Mech. and Eng., vol. 8(3), 2003, pp. 497-502.

[15] M. Narahari, S. Sreenadh and V.M. Soundalgekar, “Transient Free Convection Flow Between Long Vertical Parallel Plates with Constant Heat Flux at One Boundary,” J. Thermophysics and Aeromechanics, vol. 9(2), 2002, pp. 287-293.

[16] A.K. Singh and T. Paul, “Transient Natural Convection Between Two Vertical Walls Heated / Cooled Asymmetrically,” Int. J. Appl. Mech. and Eng, vol. 11(1), 2006, pp. 143-154.

[17] M. Narahari, “Free Convection Flow Between Two Long Vertical Parallel Plates with Variable Temperature at One Boundary,” Proc. International Conference on Mecahanical & Manufacturing Engineering (ICME2008), Johor Bahru, Malaysia, 2008.

[18] J.A. Schetz, “On the Approximate Solution of Viscous-Flow Problems,” ASME J. Appl. Mech., vol. 30, 1963, pp. 263-268.

[19] S. Lee and M. M. Yovanovich, “Laminar Natural Convection From a Vertical Plate With a Step Change in Wall Temperature,” ASME J. Heat Transfer, vol. 113, 1991, pp. 501-504.

[20] P. Chandran, N.C. Sacheti and A.K. Singh, “Natural Convection Near a Vertical Plate With Ramped Wall Temperature,” Heat Mass Transfer, vol. 41, 2005, pp. 459-464.

[21] S.C. Saha, C. Lei and J.C. Patterson, “On the Natural Convection Boundary Layer Adjacent to an Inclined Flat Plate Subject to Ramp Heating,” Proc. 16th Australian Fluid Mechanics Conference, Crown Plaza, Gold Coast, Australia, 2007, pp. 121-124.

Dr. Vijay R. Raghavan is a professor of Mechanical Engineering at the 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 Trouble-shooting.

Narahari Marneni graduated in 1993 with a first class distinction BSc (Mathematics, Physics and Chemistry) from Sri Venkateswar University, India. He earned his MSc degree in Applied Mathematics with first rank from Sri Krishnadevaraya University, India in 1995. He completed his MPhil in Mathematics at Sri Venkateswara University in 1997 and followed by PhD in

2001. Currently he is a Senior Lecturer in the Fundamental and Applied Sciences Department at Universiti Teknologi PETRONAS (UTP). He has published several research papers in refereed national and international Journals. He has presented research papers in peer reviewed international conferences. His research interests are Fluid Dynamics, Porous Media, Magneto-hydrodynamics, Heat and Mass Transfer and Computational Fluid Dynamics.



ABSTRAcT

This paper deliberates on the performance of foamed concrete containing cement replacement material namely Microwave Incinerated Rice Husk Ash (MIRHA). Mixes incorporating 0%, 5%, 10% and 15% MIRHA were prepared. The optimal mix design based on compressive strength was established. Taguchi method with L16 (45) orthogonal array was adopted to determine the best possible mix design and the number of experiment and trial mixes to be conducted was minimized. The optimal MIRHA replacement level was foud to be at 5%. It was also found that the target strength of MIRHA foamed concrete achieved met the load bearing strength requirement of 17 N/mm2.

Keywords: foamed concrete, Microwave Incinerated Rice Husk Ash (MIRHA), Taguchi’s method

MIX DESIGN OF FOAMED cONcRETE WITH MIRHA uSING TAGucHI METHOD

Muhd. Fadhil Nuruddin*, Ridho Bayuaji#

*Universiti Teknologi PETRONAS, 31750 Tronoh, Perak Darul Ridzuan, Malaysia [email protected]

# Institut Teknologi Sepuluh Nopember (ITS), Surabaya, Indonesia

INTRODucTION

Lightweight concrete has been introduced broadly all over the world as a new approach in replacing the normal concrete. However, in producing lightweight concrete, the low compressive strength compared to Ordinary Portland Cement (OPC) concrete is a concern. OPC is widely used as one of the concrete ingredients. However, a problem that the construction industries have to overcome when using the OPC is the carbon dioxide (CO2) emission. The largest emissions from cement manufacture is CO2, amounting to nearly 1 metric ton of gas per metric ton of cement.

Therefore, it is a good practice if the usage of OPC to be reduced let alone if it is being replaced by waste material as disposing wastes materials can affect the environment. The purpose of this research was to produce high strength lightweight concrete that

incorporates waste rice husk as replacement of a certain percentage of OPC content in the concrete.

Foamed concretes, produced by introducing preformed foam, are lightweight concretes consisting of a system of macroscopic air voids of approximately 0.1 to 1 mm size [1] uniformly distributed in either a matrix of aggregate and cement paste or cement paste alone.

Rice Husk Ash (RHA) is a by-product of paddy that is obtained through combustion process of rice husk. The high content of amorphous silica and very large surface area of the RHA make it a highly reactive pozzolanic material that can be used to improve the strength and durability of concrete [2].

In this paper, an investigation was carried out to better understand the effect of MIRHA on strength development of the foamed concrete by using

This paper was presented 3rd International Engineering Convention, Damascus, Syria, 11 - 14 May 2009.



Taguchi method [3-5]. The experimental work was designed to give the optimum working conditions of the parameters that affect some mechanical properties of foamed concrete.

One of the advantages of the Taguchi method over conventional experimental design methods, in addition to keeping the experimental cost at the minimum level [6], is that it minimizes the variability around the investigated parameters when bringing the performance value to the target value. Its other advantage is that the optimum working conditions determined from the laboratory work can also be reproduced in a real production environment [3].

EXPERIMENTAL STuDY MATERIALS

The constituent materials used in the laboratory to produce foamed concrete are shown in Table 1 and Table 2 shows the chemical properties of MIRHA and OPC used.

PREPARATION OF MIXTuRE

The mix proportion used in this project was generated using the Taguchi method. It was done by arranging the parameters (refer Table 3) which were microwave incinerated rice husk ash content (MIRHA), water to cementitious materials ratio (w/c), sand cement ratio (s/c), super plasticizer content (SP) and foam content (FC) using an orthogonal array (refer Table 4). The

Table 1. Constituent materials of foamed concrete used and standard adopted

Materials Standard Remarks

Cement BSEN 197-1, 2000 Ordinary Portland Cement

Sand BS EN 12620, 2002 natural sand, with 100 % passing 425 mm sieve

MIRHA BS EN 450, 2000 In order to produce MIRHA with high reactive silica content, controlled combustion of rice husk is adopted [7, 8]

Foam ASTM C 869-91 (reapproved 1999), ASTM C 796-97

Preformed foam by aerating Palm oil base, hydrolyzed natural protein a ratio of 1:30 (by volume), aerated to a density of 110 kg/m3

Table 2. Chemical properties of MIRHA and OPC

Oxide composition

Weight (%)

Na2O MgO Al2O3 SiO2 P2O5 K2O CaO TiO2 Fe2O3 SO3 MnO

MIRHA 0.02 0.63 0.75 90.8 2.5 3.77 0.87 0.02 0.28 0.33 0.08

OPc 0.02 1.43 2.84 20.4 0.1 0.26 67.7 0.17 4.64 2.2 0.16

Table 3. Parameters and their variation level

Variable unit Level 1 Level 2 Level 3 Level 4

MIRHA (%) 0 5 10 15

w/c ratio 0.3 0.35 0.4 0.45

s/c ratio 0.25 0.5 0.75 1

SP (%) 1 1.5 2 2.5

Foam content (%) 20 25 30 35



Table 4. Standard L16 (45) orthogonal array

Exp. no

Independent variables Dependent Variable (s)

Performance parameter valueVariable 1 Variable 2 Variable 3 Variable 4 Variable 5

1 1 1 1 1 1 P1

2 1 2 2 2 2 P2

3 1 3 3 3 3 P3

4 1 4 4 4 4 P4

5 2 1 2 3 4 P5

6 2 2 1 4 3 P6

7 2 3 4 1 2 P7

8 2 4 3 2 1 P8

9 3 1 3 4 2 P9

10 3 2 4 3 1 P10

11 3 3 1 2 4 P11

12 3 4 2 1 3 P12

13 4 1 4 2 3 P13

14 4 2 3 1 4 P14

15 4 3 2 4 1 P15

16 4 4 1 3 2 P16

mix proportion adopted was given in Table 5. MIRHA was incorporated at 0%, 5%, 10% and 15% replacing Portland cement by weight. Super plasticizer was used to improve the workability. The cube samples of 50 mm size were prepared to determine the effect of these parameters on compressive strength.

The Orthogonal Array (OA) experimental design method was chosen to determine the experimental plan, L16(45), which is one of the standard experimental plans improved by Taguchi and it is the most appropriate approach for the condition being investigated (five parameters with four levels). In Table 4, it should be noted that the parameter Variable 1 (i.e. MIRHA) has four levels. Similarly Variable 2 (i.e. w/c), Variable 3 (i.e. s/c), Variable 4 (i.e. SP) and variable 5 (i.e. FC) have four levels.

SAMPLE PREPARATION, cuRING AND TESTING

The LWFC samples were cast using 50 mm cube moulds in accordance with BS 4500-2002. There were three stages for the preparation of foamed concrete. The first was the preparation of mortar, second was the preparation of foam from a pre-mixed foaming agent, and finally, the generation and introduction of foam using compressed air. The mortar procedure of machine mixing used in this research was according to BS 1881-125.1986. The mould surfaces were greased with mineral oil in order to prevent the development of bond between the mould and the concrete. After casting, the moulded specimens were covered with plastic sheet and left in the casting room for 24 hours. They were then de-moulded and moved into the curing tank with the temperature at 20±2°C. The curing was in accordance with BS EN 12390-2:2000. For the compressive test, 6 samples as recommended by ASTM standards, of each mix were tested. The strength development of LWFC



Table 5. Mixture proportion of foamed concrete

code cement (kg/m³) Sand (kg/m³) Water (kg/m³) MIRHA (kg/m³) SP (kg/m³)

LWFC-1 930 233 419 0 23

LWFC-2 950 475 285 0 19

LWFC-3 770 578 270 0 12

LWFC-4 620 620 248 0 6

LWFC-5 893 223 282 47 13

LWFC-6 703 352 333 37 7

LWFC-7 732 549 308 39 18

LWFC-8 703 703 259 37 18

LWFC-9 900 225 350 100 9

LWFC-10 810 405 360 90 12

LWFC-11 572 429 286 64 11

LWFC-12 648 648 216 72 16

LWFC-13 748 187 352 132 15

LWFC-14 663 332 273 117 17

LWFC-15 774 580 273 137 8

LWFC-16 565 565 299 100 8

was monitored for 3, 7 and 28 days. The test was in accordance with BS EN 12390-3.


Table 6 shows the dry density and strength development of the various mix proportion. The compressive strengths of 3-day, 7-day and 28-day specimens were in the range of 6.9 to 45 N/mm2, 7.5 to 56.3 N/mm2 and 10.2 to 68.9 N/mm2 respectively. The highest compressive strengths of 3-day, 7-day and 28-day specimens were obtained from mixes LWFC-8 LWFC-8, and LWFC-2 respectively.

The investigation of the best possible levels of mix proportions were conducted for the maximization of compressive strength and for the minimization of dry density values using the Taguchi method. The performance statistics for ‘‘the larger the better” situations were evaluated for maximization properties of LWFC and ‘‘the smaller the better” situations were evaluated for minimization properties

of LWFC. Since all parameters are interconnected in the LWFC mix, the best possible testing conditions of the LWFC properties could be determined from the main effect plot graphs for compressive strength and dry density (Figures 1–4). From Figures 1–4, the best mix proportions of the target properties were obtained.

Figure 1 shows the variation of the performance statistics as a function of compressive strength. In general, Figures 1, 2 and 3 have similar trends. Increasing of MIRHA, w/c and foam parameters decreased the compressive strength but increasing the content of SP increased the compressive strength. MIRHA is the most influential factor on the 3-day compressive strength of the LWFC with 34.4% contribution. The second most influential factor is superplastizier (SP) with 25.1% contribution. The contribution rank of the parameters on the compressive strength can be seen in Table 6. An optimal condition for maximum 3-day compressive strength was obtained at 5% MIRHA, 0.35 w/c,



Figure 1. Main plot for 3-day compressive strength

Table 6. Test result on hardened foamed concrete

codeDry Density

(kg/m3)

compressive strength (N/mm2)

3rd day 7th day 28th day

LWFC-1 1,856 31.2 28.6 61.3

LWFC-2 1,899 54.9 43.0 76.5

LWFC-3 1,541 19.6 21.6 25.2

LWFC-4 1,340 12.4 11.0 13.8

LWFC-5 1,400 25.5 26.1 29.7

LWFC-6 1,504 22.8 24.5 26.6

LWFC-7 1,918 36.6 39.2 54.8

LWFC-8 1,886 45.0 56.3 63.9

LWFC-9 1,409 21.0 24.4 30.0

LWFC-10 1,517 24.6 30.3 37.5

LWFC-11 1,208 8.6 13.6 18.2

LWFC-12 1,668 23.3 30.2 40.1

LWFC-13 1,249 16.1 20.9 27.8

LWFC-14 1,412 26.6 25.2 30.8

LWFC-15 1,703 6.9 7.5 10.2

LWFC-16 1,374 15.7 20.1 25.5

12.5

17.5

22.5

27.5

32.5

0 5 10 15 0.3 0.35 0.4 0.45 0.25 0.5 0.75 1 1 1.5 2 2.5 20 25 30 35

3-da

y C

ompr

essi

ve s

tren

gth

(N/m

m2 )

MIRHA (%) w/c s/c SP (%) Foam (%)



0.5 s/c, 2% SP and 25% foam. However, most of the concretes with MIRHA achieved the design strength which was higher than 17 N/mm2. It was evident that, the foamed concrete which incorporated MIRHA could be used as load bearing structure.

As seen from Figure 4, increasing MIRHA, w/c and foam parameters decreased the dry density but increasing the content of SP increased the dry density. Foam content was the most influencing factor on dry density of LWFC with 44.4% contribution. The second most influencing factor was MIRHA with 24.1% contribution. The contribution rank of the parameters on the compressive strength is shown in Table 6. An optimal condition for minimum was obtained at 15% MIRHA, 0.45 w/c, 0.25 s/c, 1.5% SP, and 35% foam.

The response data given in Table 7 were analyzed using analysis of variance (ANOVA). A statistical analysis was performed to determine the statistical significant factors and the data analysis was

presented in Table 7. Finally, degree of contribution of each significant factor was obtained so as to determine the level of its statistical importance in the model. The contribution percentage in Table 7 gave an idea about the degree of contribution of the factors to the measured response. If the contribution percentage was high, the contribution of the factors to that particular response was more. Likewise, lower contribution percentage lowered the contribution factors on the measured response. According to Figures 2–4, the best mix proportions of the target properties were tabulated in Table 8.

cONcLuSION

LWFC consists of many components; it was critical to use a systematic approach for identifying optimal mixes and investigate the most effective factors under a set of constraints. Due to this reason, the Taguchi method with L16 (45) orthogonal array was used in this study to investigate ranking of the effective parameters and best possible mix

7-da

y C

ompr

essi

ve s

tren

gth

(N/m

m2 )

12.5

17.5

22.5

27.5

32.5

37.5

42.5

0 5 10 15 0.3 0.35 0.4 0.45 0.25 0.5 0.75 1 1 1.5 2 2.5 20 25 30 35MIRHA (%) w/c s/c SP (%) Foam (%)




Figure 4. Main effect plot for dry density

1325

1375

1425

1475

1525

1575

1625

1675

1725

0 5 10 15 0.3 0.35 0.4 0.45 0.25 0.5 0.75 1 1 1.5 2 2.5 20 25 30 35


Dry

Den

sity

(Kg/

mm

3 )


17.5

22.5

27.5

32.5

37.5

42.5

47.5

0 5 10 15 0.3 0.35 0.4 0.45 0.25 0.5 0.75 1 1 1.5 2 2.5 20 25 30 35


28-d

ay C

ompr

essi

ve s

tren

gth

(N/m

m2 )



Table 7. Analysis of variance results of LWFC properties

Parameterstatistical

parameters3-day compr.

strength7-day compr.

strength28-day compr.

strength Dry Density

MIRHA DFa 3 3 3 3

SSSb 633.2 690.6 1068.8 204558.3

ASSc 633.2 690.6 1068.8 204558.3

MSd 211.1 230.2 356.3 68186.1

Contribution (%) 34.4% 25.4% 23.3% 24.1%

sand/cement DF 3 3 3 3

SSS 243.2 404.7 391.2 32772.6

ASS 243.2 404.7 391.2 32772.6

MS 81.1 134.9 130.4 10924.2

Contribution 13.2% 14.9% 8.5% 3.9%

water/binder DF 3 3 3 3

SSS 168.0 228.3 125.7 79573.3

ASS 168.0 228.3 125.7 79573.3

MS 56.0 76.1 41.9 26524.4

Contribution (%) 9.1% 8.4% 2.7% 9.4%

superplasticizer DF 3 3 3 3

SSS 462.3 851.5 1829.7 156036.0

ASS 462.3 851.5 1829.7 156036.0

MS 154.1 283.8 609.9 52012.0

Contribution (%) 25.1% 31.3% 39.8% 18.4%

foamed DF 3 3 3 3

SSS 332.0 548.9 1180.8 374984.9

ASS 332.0 548.9 1180.8 374984.9

MS 110.7 183.0 393.6 124995.0

Contribution (%) 18.1% 20.2% 25.7% 44.2%a Degree of freedom c Adjusted sum of squareb Sequential sum of square d Mean square (variance)

Table 8. Optimal mix design properties for properties of Lightweight Foamed Concrete

Optimal Mix Proportional MIRHA/

cement (%)water/binder

sand/cement

SP/cement (%)

Foam/Total Vol (%)

3-day compr. strength 5 0.35 0.5 2 25



Dry Density 15 0.45 0.25 1.5 35



proportions of fresh and hardened properties of LWFC. At the end of this research, it was seen that the Taguchi method is a promising approach for optimizing mix proportions of LWFC to meet hardened concrete properties. The Taguchi method could simplify the test protocol required to optimize mix proportions of LWFC by reducing the number of trial batches. This study showed that it was possible to design foamed concrete that satisfed the criteria of high strength lightweight concrete. As was seen, results of properties of fresh and hardened concrete of the samples produced satisfied the expected properties of LWFC.

REFERENcES

[1] T.H. Wee, D.S. Babu, T. Tamilselvan and H.S. Lin, “Air-Void System of Foamed Concrete and its Effect on Mechanical Properties”. ACI Mater J, vol. 103 (1) (2006), pp. 45–52.

[2] D. Bui, Rice Husk Ash as Mineral Admixture For High Performance Concrete. 2001.

[3] Ibrahim Türkmen, Rüstem Gul and Cafer Çelik. “A Taguchi approach for investigation of some physical properties of concrete produced from mineral admixtures”. Building and Environment vol. 43, 2008, p. 1127-1137.

[4] Erdoğan Ozbay, Ahmet Oztas, Adil Baykasoglu and Hakan Ozbebek, “Investigating mix proportions of high strength self compacting concrete by using Taguchi method”. Construction and Building Materials, vol. 23, 2009, p. 694-702.

[5] Harun Tanyildizi and Ahmet Coskun. “Performance of lightweight concrete with silica fume after high temperature”. Construction and Building Materials, vol. 22, 2008, p. 2124-2129.American Society for Testing and Materials. ASTM E-122. Standard recommended practice for choice of sample size to estimate the average quality of a lot or process. Am Soc Test Mater, Philadelphia, 1972.

Muhd Fadhil Nuruddin is an Assocaite Professor at the Civil Engineering Department of University Technology PETRONAS. His areas of interest are carbonation in concrete and concrete technology. He obtained his PhD in Civil Engineering from USM.

Ridho Bayuaji is a PhD candidate at the Civil Engineering Department of Universiti Teknologi PETRONAS. He received his Master in Civil Engineering, Universiti Gadjah Mada Indonesia in 2005. He was conferred the BEng in Civil Engineering at Institut Teknologi Sepuluh Nopember Surabaya Indonesia. His research interest is mainly in hydration process in lightweight-

foamed concrete. He is also a lecturer in Civil Engineering Department at Institut Teknologi Sepuluh Nopember Surabaya Indonesia.



INTRODucTION

There are at least 70 million people around the globe who suffer from speech and hearing disabilities, either at birth or by accident [1]. It is somehow difficult for us to interact with them because of the unfamiliar communication means used. Sign Language (SL) is a common form of communication which is widely used by the speech and hearing impaired. Thus, probably the only easier way of communication/interaction with them is by learning their language – the sign language [2].

We may have friends or family members who have hearing or speech disabilities. Such disabilities may be from birth, or by accident. Surely it is difficult for us to communicate with them if we do not know their language – the Sign Language (SL). It is also difficult for them as well to communicate with us since they have such disabilities.

One may be interested to learn up this language; however it may be costly to attend tuition classes to learn this language. Furthermore, tuition classes exhibits time constraint, where one does not have the flexibility in time on whether or not to attend the tutorial. He/she may prefer to have a self-tutorial, however there is no such inexpensive software. These may contribute to the reluctance of the public to learn SL to better communicate with those with hearing or speech impairments.

There are campaigns of speeches and talks given to the public. However, these talks usually are not able to reach those with hearing disabilities. So far in Malaysia, only the news on RTM 1 uses SL extensively to present the daily news. The lack of programs using this technique may be the reason for the extra cost incurred in hiring a translator from speech.

V2S: VOIcE TO SIGN LANGuAGE TRANSLATION SYSTEM FOR MALAYSIAN DEAF PEOPLE

Oi Mean Foong*, Tan Jung Low and Wai Wan LaUniversiti Teknologi PETRONAS, 31750 Tronoh, Perak Darul Ridzuan, Malaysia

*[email protected]

ABSTRAcT

The process of learning and understanding the sign language may be cumbersome to some, and therefore, this paper proposes a solution to this problem by providing a voice (English Language) to sign language translation system using Speech and Image processing technique. Speech processing which includes Speech Recognition is the study of recognising the words being spoken, regardless of whom the speaker is. This project uses template-based recognition as the main approach in which the V2S system first needs to be trained with speech patterns based on some generic spectral parameter set. These spectral parameter set will then be stored as a template in a database. The system will perform the recognition process through matching the parameter set of the input speech with the stored templates to finally display the sign language in video format. Empirical results show that the system has 80.3% recognition rate.

Keywords: image processing, sign language, speech recognition, spectral parameter.

This paper was presented at the First International Visual Informatics Conference, IVIC 2009, Kuala Lumpur, 11-13 November, 2009.



Furthermore, there is a lack of trained personnel in Malaysia who are able to translate these speeches into SL. The result is that those with hearing disabilities know less about the ongoing news around them. The objectives of this research are three-fold:

1. To ease the communication between normal people and the hearing/speech impaired.

2. To eliminate the need of attending costly SL classes – it can be done at home.

3. To be able to reach a wider audience (by including the hearing impaired) during speeches and campaigns.

Research Motivation

According to The Star Online on the 20th December 2006, “Radio Television Malaysia (RTM 1) will be incorporating more SL in their news segments and dramas for the benefit of the hearing-impaired” and on 22nd July 2007, “there is an acute shortage of SL interpreters because at present there are only 10 qualified interpreters catering to 24,000 registered deaf people nationwide, according to the Malaysian Federation of Deaf”.

This simply says that media agencies such as RTM is offering opportunities for more SL interpreters to join its organization. But according to The Star report, the opportunity is not taken up due to the shortage of people with such qualification in this country.

The V2S system is a proposed solution to reduce the cost of employing special skilled employees for media agencies such as RTM. With the V2S, the media agencies need not hire more SL interpreters in their news segments or talk shows. Hiring extra interpreters increase cost in salaries. The proposal helps not only in reducing cost incurred in hiring SL interpreters, the media agencies would be directly providing community services to the less fortunate audience.

The V2S system can be taken as an alternative means to an SL interpreter, i.e. to replace the old-fashioned way (a real person doing interpretation) of translating into SL. The interpretation process is made available by taking the advantages of modern ICT technology via easily affordable gadgets such as computers, hand phones or PDAs as a mediator (translator). This system may also solve the problem of shortage of SL interpreters.

RELATED WORKS

Voice recognition can be generally classified into speaker recognition and speech recognition categories. Speaker recognition is a way of recognising people from their voices. Such systems extract features from speeches, then modeled and used them to recognise the person from his/her voice. There is a difference between speaker recognition (recognising who is speaking) and speech recognition (recognising what is being said). Voice recognition is a synonym for speaker and thus, not speech recognition. Speaker recognition has a history dating back some four decades, where the output of several analog filters were averaged over time for matching. Speaker recognition uses the acoustic features of speech that was found to be different between individuals. These acoustic patterns reflect both anatomy (e.g. size and shape of the throat and mouth) and learned behavioural patterns (e.g., voice pitch, speaking style). The incorporation of learned patterns into the voice templates (the latter, called “voiceprints”) has earned speaker recognition its classification as a “behavioural biometric” [3].

The fundamental task of speech recognition is the derivation of a sequence of words from a stream of acoustic information. A more general task is automatic speech understanding, which includes the extraction of meaning (for instance, a query to a database) or producing actions in response to speech. For many applications, interaction between system components devoted to semantics, dialog generation, etc. and the speech recognition subsystem can be critical [4].



Feature extraction is a critical element in speech recognition since it is the first step of the recognition process and generates the parameters of which the recognition is based. It is well known that the Mel Frequency Cepstral Coefficients (MFCC) are the most widely used feature parameter. One of the steps of MFCC is the Mel-scaled filter bank processing. This step may result in some loss of information from the original signal but it is widely accepted that such a step is helpful in the component for the extraction of information from the signals [5, 6].

A vector quantiser is a system for mapping a sequence of continuous or discrete vectors into a digital sequence suitable for communication over or storage in a digital channel. The goal of such a system is data compression to reduce the bit rate so as to minimise communication channel capacity or digital storage memory requirements whilst maintaining the necessary fidelity of the data. The mapping for each vector may or may not have memory which depended on past actions of the coder, just as in well established scalar techniques such as PCM. Even though information theory implies that one can always obtain better performance by coding vectors instead of scalars, scalar quantisers have remained by far the most common of data compression systems because of their simplicity and good performance when the communication rate is sufficiently large. In addition, relatively few design techniques have existed for vector quantisers [7]. Even though there are other techniques for pattern matching, vector quantisation is considered as one of the best approaches for its flexibility in training as well as recognising.

SL is used primarily by deaf people throughout the world. SL differs from spoken language in that it is visual rather than auditory, and is composed of precise hand shapes and movements. This language has evolved in a completely different medium, using the hands and face rather than the vocal and is perceived by the eyes rather than the ears.

SL is not a universal language shared by deaf people of the world because there are many sign languages

that have evolved independently of each other. Just as spoken languages differ in grammatical structure, rules and historical relationships, sign languages also differ along these parameters.

An important property of human sign language is that the form of words is generally arbitrary, and there are no indigenous sign languages that are simply a transformation of a spoken language to the hands. Sign language is also equipped with the same expressive power that is inherent in spoken languages and it can express complicated, intricate concepts with the same degree of explicitness and eloquence as spoken language. Sign language portrays the image, identity and culture of the country in which the deaf community belongs to. In Malaysia, we have the Malaysian Sign Language (Bahasa Isyarat Malaysia, BIM).

BIM has many dialects, differing from state to state. American Sign Language (ASL) has had a strong influence on BIM but the two are different enough to be considered separate languages. Other sign languages in use in Malaysia are Penang Sign Language (PSL), Selangor Sign Language (SSL or KLSL), Kod Tangan Bahasa Malaysia (KTBM) and Chinese Sign Languages [8].

PROPOSED SYSTEM

The proposed V2S system architecture is shown in Figure 1. As illustrated in the diagram, the main components are the sound recording component (with its supporting sound/voice training algorithm), digital signal processing component (with its supporting MFCC – Mel Frequency Cepstral Coefficients algorithm counterparts) and the vector quantisation component (supported by its matching sub-component).

Main System components Description

Sound Recording

The sound recording process is responsible for capturing and recording sound using a microphone.



Output of this process is the recorded sound which can be in .wav or .midi format. The quality of the recorded sound is highly dependent on the sound recording software used. However the quality of the recorded sound can be enhanced by applying proper noise filtering. The sound training algorithm allows the user to “train” the system to capture the same sound/voice an appropriate number of times so as to produce a good quality recorded signal. The quality of the recorded sound/voice is an important factor in determining the accuracy of the voice-to-signal translation of the system.

Digital Signal Processing

The “trained” sound from the sound recording is then fed into the Digital Signal Processing (DSP) part. The DSP is the most important and a difficult implementation of the S2V system. The main task of the DSP is to convert the recorded sound from a time domain to a frequency domain. This is a necessary for extracting the features (formant) of the recorded sound so that the system can recognise words spoken into it. The MFCC (Mel Frequency Cepstral Coefficients) algorithm is used in this formant extraction process.

Vector Quantisation

Vector quantisation is used to perform speech recognition. In fact, vector quantisation is one of the most effective matching techniques popularly used for speech recognition. The basic concept of vector quantisation is to compress any vector of a speech/voice feature into one scalar vector. By compressing the feature vector, a lot of space for feature storage

can be saved and this helps to increase the efficiency of the matching process i.e. to compare a new feature with one value instead of many. In vector quantisation, the system is to be trained first. Then the trained sound will be stored in a codebook in the database. Each trained sound will have its own codebook. During the recognition stage, the new input signal will be used to compare with all the stored codebooks and the codebook which has the closest distance will be chosen as the recognised word.

In general, the input (sound/voice) is compared with the existing codebooks in the database for video retrieval. The values of each codebook and the voice input are represented using matrices. The respective average value of each codebook and the voice input will be computed. Then, each codebook will be compared with the input voice (the trained recorded signal) value. The confirmation of which video to be played is based on the closest input voice value (distance) to the codebooks stored in the database.

Figures 2(a) and (b) show samples of sound/voice for a single word-matching process. That is, the matching of calculated codebook (from the input voice) with the stored codebook. The samples of these sound/voice matching were processed using one of the utilities available in the MatLab® application.

EMPIRIcAL RESuLTS AND FINDINGS

The interface of this project was designed to have a display panel and few buttons for simplicity purposes. By clicking on the “V2S” button on the screen the user is allowed to input raw voice i.e. recording of spoken voice into the system. There is a display panel of the appropriate video output (Sign Language) which is the relevant translation of the spoken word(s).

Figure 3 shows the main interface of the V2S prototype application. This prototype includes the S2V (Sign-to-Voice) system [9]. The S2V system was presented in WASET 2008 Congress in Singapore.

Figure 1. The proposed V2S system architecture

TrainingAlgorithm

Sound Recording

Voice Input

MFCC

Digital Signal Processing

Vector Quan�za�on

VideoDatabase

MatchingPlay Video

Exis�ng codebook loaded from database

Trained sound



Figures 4(a) to (d) show some samples of the video output for the relevant translated words.

SYSTEM EVALuATION

100 people comprising elderly women and men, young male and female children, and middle aged male and female were sampled for the accuracy of the V2S system. The test was monitored and conducted during the PECIPTA 2007 (Expositions of Research and Inventions of International Institution of Higher Learning) in Kuala Lumpur and CDC (Career Development Carnival) exhibitions in UTP. It was found that at least 80 out of the 100 people were able to callback the desired sign language video. This means the system was at least 80% accurate in its display of the correct sign language. It should be mentioned here that the words uttered by the test samples were “you”, “us”, “come here”, and “turn

Figure 2(b). Matching of the “us” voice with the stored codebookFigure 2(a). Matching of the “me” voice with the stored codebook

Figure 3. The main user interface of the V2S system. (1) The V2S button – clicking it shall prompt the user for training the voice input. The system will then matches the voice to the corresponding video in the database. (2) The display panel – displays the relevant video for the translated voice. (3) Close button – exit the V2S system.

Figure 4(a). Video for “come here”

Figure 4(b). Video for “turn left Figure 4(c). Video for “you” Figure 4(d). Video for “us”



left”. Figure 5 shows the overall statistics of the test conducted.

The graph shows the accuracy of each category by averaging the accuracy percentage of the 4 spoken words. The “middle-aged man” category shows the highest accuracy. This may be caused by training of the system via ONLY the “middle-age man”. However, other categories do show high accurate responses. This implies that the more training the system gets for each word, the more accurate it would be.

cONcLuSION

Natural Language to SL translation is the main scope of this research. The fundamental idea of this system is to translate the human voice to SL. The system is to match the captured voice with the pre-stored SL videos in the database and to display the appropriate sign/gesture. Thus, this provides an alternative interactive way of communication between a normal person and a hearing-impaired person.

The prototype allowed translation of spoken English to SL in the Malaysian context. The accuracy of the system depended on how much system training was conducted. The system was first tested with an accuracy of 80.3%. With sufficient training, it should be able to recognise all the trained commands or words and execute the corresponding translation.

This system would be useful to those who wish to learn SL and to help those who wish to communicate with the hearing-impaired more effectively. For future work, the system could be implemented in mobile devices with animated hand gestures [10] for the deaf.

REFERENcES

[1] The World Federation of the Deaf, http://www.hearinglossweb.com/res/hlorg/wfd.htm

[2] Cornucopia of Disability Information: Disability Statistics, http://codi.buffalo.edu/graph_based/demographics/.statistics.htm

[3] E. Zetterholm, Voice Imitation. A Phonetic Study of Perceptual Illusions and Acoustic Success, PhD thesis, Lund University (2003).

[4] V. Zue, R. Cole and W. Ward, Speech Recognition, Cambridge University Press, New York (1997).

[5] J.W. Hung, “Optimization of Filter-Bank to Improve Extraction of MFCC Features in Speech Recognition,” IEEE Transactions on Intelligent Multimedia, Video and Speech Processing, 2004, pp. 675-678.

[6] H. Md. Rashidul, J. Mustafa and Md. Golam Rabbani Md. Saifur Rahman, “Speaker Recognition Using Mel Frequency Cepstral Coefficient,” 3rd International Conference on Electrical and Computer Engineering, 2004, pp. 565-568.

[7] R.M. Gray, Vector Quantization. Morgan Kaufmann, CA, 1990.

[8] Wikipedia, Malaysia Sign Language, http://en.wikipedia.org/wiki/ MalaysianSignLanguage

[9] O.I. Foong, T.J. Low and W. Satrio, “Hand Gesture Recognition: Signs to Voice System (S2V),” WASET conference proceedings, ISSN 2070-3740, vol. 33, article 6, 2008, pp. 32-36.

[10] R.S. Sequndo, J.M. Montero, J.M. Guarasa, R. Cordoba, J. Ferreiros, J.M. Pardo, “Proposing a Speech to Gesture Translation Architecture for Spanish Deaf People,” Journal of Visual Languages and Computing, vol. 19, 2008, pp. 523-538.

Elde

rly m

en –

84%

Elde

rly w

omen

–74

%

Mid

dle

age

men

–90

%

Mid

dle

age

wom

en –

78%

Youn

g m

ale

child

ren

–80

%

Youn

g fe

mal

e ch

ildre

n –

76%

Test category

Perc

enta

ge a

ccur

acy

30

60

90

Figure 5. Overall system accuracy test



O. M. Foong received her BSc Computer Science from Louisiana State University and MSc Applied Mathematics from North Carolina A&T State University in the USA. She worked as Computer System Manager at Consolidated Directories at Burlington, USA and also as Lecturer at Singapore Polytechnic prior to joining the CIS Department at Universiti Teknologi

PETRONAS. Her research interests include data mining, decision support systems and artificial intelligence.

T. J. Low received his BEng. (Hons) in Computer Technology from Teesside University, UK in 1989 and MSc IT from National University of Malaysia in 2001. Low has been in the academic line for the past 20 years as lecturer in various public and private institutes of higher learning. His research interests include wireless technology, embedded systems, and grid/

HPC computing. Some of his current R&D projects include Biologically Inspired Self-healing Software, VANET, and Noise Removal in Seismograph using High Performance Computer. His other completed projects have been recognised at national as well as international levels. For instance, his Free Space Points Detector was presented at INPEX 2008 Pittsburgh, USA in August 2008. Low has published various papers in systems survivability, HPC/Grid, and wireless/mobile technology.



INTRODucTION

Steganalysis or the detection of the message in an image is one of the methods to attack the secret communication between two parties. Many researchers have conducted studies to break the steganography algorithm.

This research was carried out for Pairs Analysis detection technique developed by Fridrich [1]. The focus was on greyscale images and Least Significant Bit (LSB) embedding.

Since there was no threshold value to distinguish between stegogramme and non-stegogramme classes, it would lead to an incorrect classification.

This paper addresses the limitation of the incorrect classification and proposes a solution to get the most appropriate threshold value.

RESEARcH WORK

Steganalysis software are used to hide messages the carrier images. Pairs Analysis, Chi-squared Attack, F5,

RS Steganalysis and Outguess are famous algorithms which attack the image carrier.

An attack developed by Provos based on Chi-squared is the detection algorithm which was developed before pairs algorithm and can be applied to any steganographic software [1]. The detection algorithm works for a fixed set of Pairs of Values (PoVs), or other fixed group of values and are flipped into each other to embed message bits [1]. If the secret message were embedded sequentially in the cover image pixels or indices, an abrupt change would be observed in the statistical evidence encountered at the end of the message [1]. This detection algorithm was developed and used for sequential embedding, i.e. an embedded message in the sequence order of pixels or indices or coefficients. Chi-squared technique could also be used for random message embedding, but would be less effective unless 97% of pixels or coefficients or indices were used for embedding. Westfeld developed an idea to group colours from one pixel or neighbouring pixels and fusing their values using a special hash function. Westfeld claimed that messages as small as 33% of the maximal image capacity could be detected [1].

EVALuATING PAIRS ANALYSIS THRESHOLD uSING REcEIVER OPERATING cHARAcTERISTIc (ROc) GRAPH

Emelia Akashah P. A., Anthony T. S. Ho and Savita K. Sugathan

ABSTRAcT

This paper explains the implementation of Receiver Operating Characterictic (ROC) graph addressing the incorrect classification of images for stegogramme and non-stegogramme classes using Pairs Analysis detection technique. The threshold value to discriminate between the two classes is identified, to reduce the rate of False Negative (FN) .

Keywords: non-stegogramme, pairs analysis, ROC, stegogramme, threshold

This paper was presented at the World Congress Of Engineering And Computer Science (WCECS 2008), San Francisco, 22 – 24 October 2008.



An improved Chi-squared Attack, known as Generalized Chi-squared developed by Provos, allowed for random embedding [1]. The use of a sliding window of a fixed size that could be moved along the image, rather than by increasing the window size, provided the capability of random message detection [1]. However, Provos did not elaborate or perform the analysis any further on this new proposed technique.

RS Steganalysis is another detection technique which was introduced before Pairs Analysis. The estimation of the number of flipped pixels during LSB embedding could be identified, thus it could estimate the length of a message [1].

Pairs Analysis was developed as an improvement of RS Steganalysis. It is able to detect the presence or absence of a hidden message in an image, either a greyscale image or true colour image. The detection algorithm could be applied separately to each colour channel for true colour images [1]. This could increase the value of the discrimination function [1].

Fridrich tested the algorithm on the EzStego steganography software, focusing on gif images [1].

Andrew Ker stated that Pairs and RS Steganalysis Attack are threshold-free statistics, which meant that the algorithm could detect the presence or absence of a message and try to estimate its length, without having to set the threshold to discriminate between the two classes. He said that the output would only be yes for stegogramme or no for non-stegogramme [2].

Pairs Analysis algorithm would detect randomly the spread of messages in 8-bit images, embedded using LSB flipping of palette indices to a pre-ordered palette [1]. It could detect the message length in the cover image [1]. It was said that this method is reliable and can accurately estimate the secret message length. The advantages of this detection algorithm are that it can be used for many different steganographic systems and also for different image formats (jpg, bmp, gif etc).

METHODOLOGY OF STuDY

For this research, a bmp format of natural images was chosen to be tested. 0, 10, 20, 40, 60, 80 and 100 percent message lengths were generated and compared to the image capacity. For the 100 percent message length, each bit of the message corresponded to one pixel of the image.

For the embedding technique, the secret message was hidden in a random location using the key generated from the randperm function in Matlab. This function re-arranged the location of the matrices and the new random location was used as a key for embedding.

Once the result for this image database was obtained, the algorithm was tested on other steganographic systems. All the results from the test were gathered and compared to see how this technique worked for different steganography software. The results were analysed to see whether the detection algorithm was reliable to be used for the chosen steganography software.

In the first step, colours from the image were extracted and split to make colour pairs. The colour cuts were concatenated and put into a single stream. The sequence of colour was converted to a binary vector. It was assumed that the image had up to 256 palette colours, P <=256. The set of colour pairs that was exchanged during embedding was:

Z = {(c0,c1),(c2,c3),…..,(cP-2,cP-1)} (1)

For example, let (c1,c2) be a colour pair and associate c1 with a ‘0’ and c2 with a ‘1’. The same thing was applied to the rest of the sequence. The next stage was to make shifted pairs and this was applied to the same concept as before. The set of shifted colour pairs was as follows:

Z’ = {(c1,c2),(c3,c4),…..,(cP-1,c 0)} (2)

After that, the homogenous bit-pairs were counted (e.g. the sequence of 11 or 00) for both sequences,



E and E’. In this stage, R(p) denotes the expected number of homogeneous bit-pairs in Z after flipping the LSB indices, divided by n – the length of Z. The same process was done for Z’ where R’(p) is the relative number of homogeneous bit-pairs in Z’ [1]. The value that was obtained from these expressions was used in the quadratic equation to get the result of the unknown message length, q. To get the approximation value of q, the smaller root was chosen from the quadratic equation. It would be used to prove the relative number of the pairs after embedding a relative message length. In Pairs Analysis theorem, one additional assumption is required where the structure of homogeneous bit-pairs in Z and Z’ should be the same if there is no message embedded in the image. There is no reason why they should have different structures [1].

EXPERIMENTAL RESuLT

An overlapped distribution graph produced from the experiment performed on the image sources could lead to incorrect classifications. Figure 1 below is the result of a distribution graph for 100 cover images and stegogrammes from the image database. The overlapped area might be or might not be a stegogramme.

The implementation of the FN and FP concept was applied in distinguishing the image class. A threshold

value to discriminate between the two classes could help in reducing the mistake of classifying the images.

Besides the use of FN and FP in the image classification, the experiment results for the test images could be true-positive (TP) or true-negative (TN). Such results could be obtained from the non-overlapped area of the graph distribution. Once the threshold was set and if the output were above the threshold, the test was considered as positive.

Cover image < Threshold value <= Stegogramme

Classifying the images to their classes was dependent on the accuracy of the threshold value. FP and FN in Table 1 were also referred as Type I and Type II errors.

Between the two types of errors, Type II or FN would be more dangerous if it were to occurr. False positives were tolerated to reduce the number of false negatives [7]. Type II errors means not to accept something when the condition were true. The observer, in detecting any secret communication would see it as a normal communication. Type I, for example, is just like accusing someone doing something he actually did not commit. FP would happen when a cover image is assumed as a stegogramme [7].

Figure 1. Distribution graph for 100 images (stego & non-stego)

-0.05 0 0.05 0. 0.15 0.

2

4

6

8

10

12

14

q

De

nsi

ty

PDF of Stego and Non-Stego

StegoNon-Stego

Table 2. Type of error

Test result Actual condition Error Type

Stegogramme Cover image Type I

Cover image Stegogramme Type II

Table 1. Classification of cover image and stegogramme

True positive(TP) Stegogramme which we detect as stegogramme

True negative (TN) Cover image which we detect as cover image

False positive (FP) Cover image which we detect as stegogramme

False negative (FN) Stegogramme which we detect as cover image



The distribution graph is divided into fractions TP, TN, FP and FN. Axis Y shows the density value of the distributions while axis X shows the value of q, which represents the bitflips. The value of a cut-off point was identified on the distribution graph before a threshold value was selected. Figure 2 shows the fraction for TP, TN, FP and FN with cut-off value 0.02. All the cut-off values chosen belonged to the overlapping area.

THRESHOLD VALuE uSING ROc

ROC can help us in making decisions, for example in the choice of threshold value to reduce an incorrect classification. To plot the ROC curve, only true-positive rates and false-positive rates are needed. True-positive rate is also known as sensitivity and false-positive rate is known as specificity. To plot the ROC, the sensitivity value for axis-Y and 1-specificity value for axis-X was chosen. ROC curve is sometimes known as sensitivity vs. 1-specificity graph. Each threshold chosen represented one point on the ROC graph. Sensitivity (Se) is a statistical measure of how well a classification test correctly identifies a condition and it is a proportion of true-positives [4]. To perform the test, a high sensitivity rate is required. Specificity (Sp) is used to identify negative cases, where the test would correctly indicate ‘negative’ if the image does not contain any hidden message. It would represent the proportion of true-negatives of all negative cases [4]. To perform the

test, a high specificity rate is required. Se and Sp are commonly used to measure test performance. Errors would occur if sensitivity and specificity were not considered. Before the values of Se and Sp were calculated, it was assumed that all of the images in the database could be allocated as either stegogramme or cover image class without errors. The frequency of the images could be said to be a false negative, false positive, true positive and true negative area, and was used to predict the threshold value before the efficiency of the cut-off value choose was calculated [5].

There is a diagonal line, known as random guess line, that divides the ROC space to determine which plot is the best to be chosen. The plot above the line can

Figure 2. Cut-off value =0.02

-0.05 0 0.05 0.1 0.15 0.2

2

4

6

8

10

12

14

q

Den

sity

PDF of Stego and Non-Stego

StegoNon-Stego

FN

TN

FP

TP

0 1 2 3 4 5 6 7 84

5

6

7

8

9

10

1-Specificity

Sen

sitiv

ity

Receiver Operating Characteristic

worse

better

Figure 3. ROC plot

Table 3. Result for each cut-off value

Cut-off value = 0.02

Sensitivity (Se) 0.7454

Specificity (Sp) 0.5987

Cut-off value = 0.05



Cut-off value = 0



Cut-off value = -0.02





be considered as good result, while the plot under the line is the bad result. Below is the figure of the ROC curve plotted according to the four selected thresholds chosen.

From the ROC plot above, it could be seen that there were four points plotted which represented each chosen threshold. One point was plotted under the random guess line, which led to the bad result. The other three points were plotted above the guessing line. The point under the diagonal line which would give a bad result was ignored. The three points above the line were considered. The accuracy of each point was calculated and the result is shown in Table 4.

To calculate the accuracy, true-positive rate was divided with false-positive rate. From the table above, it was concluded that the best value to choose as a threshold was 0.02 as it gave the highest accuracy among others. To select the best threshold, some factors needed to be considered: the probability of correct classifications was to be maximised, i.e. the specificity and sensitivity, and the probability of incorrect classifications was to be minimised, i.e. the false-positive and false-negative fractions.

cONcLuSION

Steganography is one of the unique ways of communication. Transmitting a secret message or file is easier using this technique. Nowadays, there are lots of steganographic software available, some of which can be downloaded from websites and some of which can be purchased. Different software uses different types of image formats. Some of them use image as a carrier file and some of them hide the message in the audio or video format. In this research, focus was on hidden messages in image files.

Many detection algorithms have been developed to break the steganography software. Some of the message hiding techniques have been broken, and there are lots more which remain unbreakable.

Based on the results, analyses were carried out using the detection algorithm. A message was hidden in a carrier file, and the stegogramme was tested with the pairs algorithm. The number of the bitflips was increased as the hidden message file was increased. From the result, the class of the image was analysed. The class was either the image was a stegogramme or not. If only bitflips number were depended on, it would not be distinguishable whether the value is a cover image or vice versa.

The cut-off point was used to set the threshold that could distinquish between cover image and stegogramme. In this research, four cut-off points were tested to find the best point that could distinquish between both classes. From the four cut-off points, the best point was chosen as threshold.

To calculate the most accurate cut-off point that could distinguish between cover image and stegogramme, a Receiver Operating Characteristic (ROC) graph was plotted. From the graph, only the points above the diagonal or guessing line were considered. The accuracy was calculated to find the best threshold.

From observations, a good distribution graph could be obtained by increasing the number of the images used. In this research, only 50 images were first used to produce the distribution graph, but the spread of the graph was not good. An increase in the number of images to 100 gave a better plot of the graph. The output was good and the distribution graph could be used to find the fraction of false negative, false positive, true positive and true negative.

For future research, other types of image formats could be used and tested by the Pairs Analysis algorithm.

Table 4. Accuracy percentage for each threshold

Threshold (x) Accuracy

0.02 65%

0 60%

-0.02 56%



REFERENcES

[1] J. Fridrich, M. Goljan and D. Soukal, Higher-Order Statistical Steganalysis of Palette Images, University of Binghamton, 2003.

[2] A. Ker, Quantitative Evaluation of Pairs and RS Steganalysis, Oxford University, 2004

[3] A. Westfeld and A. Pfitzman, Attacks on Steganographic System, Dresden University of Technology, Germany.

[4] E. Bentley, “RLO: Sensitivity and Specificity”, Graduate Entry Medical School, University of Nottingham, 2007.

[5] G. Vanagas, “Receiver Operating Characteristics Curves and Comparison of Cardiac Surgery Risk Stratification Systems,” Interactive Cardio-Thoracic Surgery, 2004.

[6] L.K. Westin, Receiver Operating Characteristics Analysis: Evaluating Discriminance Effects among Decision Support System, Umea University, Sweden.

[7] David M. Lane, HyperStat Online Statistics Textbook, Rice University, 2007. Available http://davidmlane.com/hyperstat/

Emelia Akashah received her BSc in Computer Science with First Class Honours and obtained the MSc degree in Information System from University of Surrey, UK in 2007. She is currently a lecturer at the Department of Computer and Information Sciences, Universiti Teknologi PETRONAS, Malaysia.



TRANSIENT WELL PERFORMANcE MODELING FOR RESERVOIR PRESSuRE DETERMINATION

Hon Vai Yee#, Suzalina Zainal# and Ismail M. Saaid*#PETRONAS Research Sdn. Bhd.

*Universiti Teknologi PETRONAS, 31750 Tronoh, Perak Darul Ridzuan, Malaysia

ABSTRAcT

Continuous reservoir pressure data acquisition through intrusive well intervention has always been a challenge to the industry due to a combination of economic, operational and logistical constraints. This paper presents an alternative methodology wherein the reservoir pressure can be computed from the well’s steady-state flow properties and wellhead shut-in data using an advanced transient well algorithm.

The key research element was the incorporation of a mass transfer rate model into dynamic wellbore modeling for shut-in well. A mathematical workflow which incorporated the time dimension as inherited in the mass transfer theory was devised. This is a major advance on prior state-of-the-art models that assume instantaneous mass transfer upon equilibrium. This enhanced dynamic well model is able to simulate the fluid redistribution during well shut-in, by determining an accurate volume of each fluid phase at any cell in the well over time. Hence, the bottom-hole pressure over time can be calculated by adding the static head to the wellhead pressure. The bottom-hole near wellbore reservoir pressure is obtained when the well has reached equilibrium. The model also rigorously accounts for various transient behaviors of the reservoir fluids during shut-in including the reservoir fluid influx.

A number of engineering concepts were evaluated and all the elements that became the foundation of this novel methodology were presented. A sample well was used to demonstrate the robustness of the model in simulating the hydrocarbon fluid redistribution during well shut-in. A better description of wellbore dynamic behavior was obtained, ensuring an improved accuracy in transient well performance modeling.

This enhanced dynamic well model provides a vital tool for reservoir pressure determination, replacing subsurface data acquisition with surface data acquisition without well intervention. The model enables operating costs reduction associated with subsurface data acquisition, in addition to improving key data availability which will optimize production plans and recovery improvement opportunities.

Keywords:

This paper was presented at the SPE North Africa Technical Conference and Exhibition, Cairo, Egypt, 14–17 February 2010.

INTRODucTION

The behaviors of the oil and gas wells during steady-state, shut-in and start-up periods are

different. Many theories and principles of steady-state well performance model [2] are relatively well understood compared with the shut-in and start-up models. There are many important phenomena



occurring in the wellbore over the transient periods during shut-in and start-up. The optimisation of each flow process depended greatly on grasping the underlying physics to mathematically describe and model the involved flow processes.

When a well is shut-in, there is a countercurrent two-phase flow, in which liquid flows downward and the gas phase moves upward. The behavior of countercurrent flow is a combination of simultaneous flow of two phases in the upward and downward directions. Due to the effects of gravity and density differences, the preferential movement of the gas may cause severe segregation, resulting in short contact duration between the gas and the liquid. The amount of mass transfer between the gas and liquid is time dependent. Hence, it was critical to properly model the mass transfer rate during shut-in to obtain an accurate amount of gas and liquid in the well over the shut-in period.

This research aims to develop an enhanced dynamic well model by integrating mass transfer rate model with dynamic well model for an improved accuracy in transient well performance modeling.

BAcKGROuND

Mass transfer plays an important role in many industrial processes, such as the removal of pollutants from plant discharge stream by absorption, the stripping of gasses from wastewater, neutron diffusion within nuclear reactors, the diffusion of absorbed substances within the pores of activated carbon, the rate of catalyzed chemical and biological reactions [6]. However, the application of mass transfer in the oil and gas industry is not widely applied as evidenced from very few reported literature on this subject.

Mass transfer requires the presence of two regions at different concentrations. It refers to the movement of a chemical species from a high concentration region toward a lower concentration region. The primary driving force for fluid flow is the pressure difference,

whereas for mass transfer, it is the concentration difference [12].

MASS TRANSFER RATE MODEL

A mass transfer rate model to calculate the gas diffusion in a shut-in well was devised for integration with dynamic well model [10]. The mass transfer rate model comprised of a mathematical workflow which incorporated the time dimension as inherited in the mass transfer theory. The calculation of the mass transfer rate was based on Fick’s Law of diffusion [10]. The derivation of the mass transfer rate calculation was purely fundamental and coupling with the basic gas law chemistry. Figure 1 illustrates the overall workflow for mass transfer rate calculation and its relation with the gas and liquid properties.

The mass transfer rate calculation begins with the calculation of solution gas at bubble interface due to the effect of pressure build-up during well shut in, the Standing black oil correlation is used [9].

1. Solution gas at bubble interface, Rs, scf/stb

(1)

where SGPG is the gas specific gravity of produced gas, PBU is the build-up pressure, psig, API is the oil gravity, °, and T is the reservoir temperature, °F.

In physical chemistry, 1 mole of ideal gas occupies 22.4 dm3 at standard condition (1 atm and 273.15 Kelvins). This is also known as gas molar volume [5]. This was the concept used to deduce the amount of molar gas residing in the gas bubble interface/film from the solution gas, Rs. This molar gas is converted into gas concentration, Ci (mol/ltr) after the unit conversion (from scf/stb) and gas volume adjustment to the standard condition at 60°F and 1 atm (George, 2006).

205.1

00091.00123.01018

= − TAPIBU

s

PSGPGR



2. Gas concentration at bubble interface, Ci, mol/ltr

⋅⋅⋅=

288

273

4.22

1

159

3.28si RC (2)

where Rs is the solution gas at bubble interface, scf/stb, 1 scf is equivalent to 28.3 liter, 1 stb is equivalent to 159 liter, chemical standard condition is at 1 atm, 273K, and oil and gas standard condition is at 1 atm, 288K.

Similar calculation was performed to estimate the solution gas in liquid, Rliq [9] and the gas concentration in the liquid, Cliq.

3. Solution gas in liquid, Rliq , scf/stb

205.1

00091.00123.01018

= − TAPI

liq

PSGPGR (3)

where SGPG is the gas specific gravity of produced gas, P is the current pressure, psig, API is the oil gravity, °, and T is the reservoir temperature, °F.

4. Gas concentration in liquid, Cliq, mol/ltr

⋅⋅⋅=

288

273

4.22

1

159

3.28liqliq RC (4)

where Rliq is the solution gas in liquid, scf/stb, 1 scf is equivalent to 28.3 liter, 1 stb is equivalent to 159 liter, chemical standard condition is at 1 atm, 273K, and oil and gas standard condition is at 1 atm, 288K.

The difference in the gas concentration (Ci – Cliq) is the driving force for the mass transfer to take place from the bubble interface into the liquid.

Solution gas at bubble interface, Rs

Gas concentration at bubble interface, Ci

Gas concentration difference, ΔC

Molar flux , J

Dissolved gas moles, Ndiss

Solution gas in liquid, Rliq

Gas concentration in liquid, Cliq

Remaining gas moles after mass transferred, Molgas, i+1

Initial bubble volume, Vbubble,i

Initial bubble mass, Mbubble,i

Initial bubble mole, Molbubble,i

New gas mass, Mgas, i+1

New gas formation volume factor, Bg i+1

New gas density, ρgas, i+1

New bubble volume, Vbubble, i+1

New bubble diameter, Dbubble, i+1

New gas mole in liquid, Cliq,i+1

New solution gas in liquid, Rliq,i+1

Figure 1. Mass transfer rate calculation workflow



5. Gas concentration difference, ΔC, mol/ltr

liqi CCC −=∆ (5)

where Ci is the gas concentration at bubble interface, mol/ltr, and Cliq is the gas concentration in liquid, mol/ltr.

The molar flux was then calculated based on the film theory. The film theory is the simplest and oldest model to describe mass transport process applied to absorption with associated chemical reaction. Film theory assumes the mass transfer system is at steady-state condition, which is with no convection or turbulence at the interface where the mass transfer occurs [6]. It is expressed as:

6. Molar flux, J, kg mole/m2 sec

δC

DJ AB

∆⋅= (6)

where DAB is the gas-liquid diffusion coefficient, m2/sec, ΔC is the gas concentration difference, mol/ltr, and δ is the gas bubble film thickness, m.

The methodology to calculate the gas-liquid diffusion coefficient, DAB, is discussed in the next section.

The next step was to obtain the number of dissolved mole over a single bubble surface area at that particular time-step. It was calculated from:

7. Dissolved moles, Ndiss, mole

( ) 32 104 ⋅⋅∆⋅= rtJNdiss π (7)

where J is the molar flux, kg mole/m2 sec, Δt is the time-step, sec, and r is the gas bubble diameter, m.

The calculation was then continued with computing the initial gas bubble properties (the initial gas bubble volume, mass and mole). For

the mass calculation, the gas density was used to convert the gas volume to gas mass and vice versa.

Upon obtaining the initial gas bubble properties, the new gas bubble properties and the new gas bubble size were back calculated from the amount of gas which still remained in the gas bubble. The newly derived gas bubble size was used in the subsequent calculation in the next time step.

The new solution gas in liquid after mass transferred must be obtained as an input for the next time-step, it was back calculated from the new gas mole in liquid after mass transferred as in equation 4.

DIFFuSION cOEFFIcIENT

In deriving the molar flux from the film theory, the gas-liquid diffusion coefficient, DAB, was obtained as an input parameter. This gas-liquid diffusion coefficient could be obtained from three options:

1. Published self-diffusivity value2. Wilkie Chang Correlation3. Average diffusion coefficient

Option 1: Published Self-Diffusivity Value

The gas-liquid diffusion coefficient, DAB was calculated from the published self-diffusivity values [4]. These values are applicable for methane mixtures with hydrocarbons systems at high pressure. The self-diffusivity values that have been measured and published are the homogeneous mixtures of:

1. methane + hexane2. ethane + hexane3. methane + octane4. ethane + octane5. methane + decane6. ethane + decane, and 7. methane + hexane + benzene

over the whole concentration range, at 303.2K and 333.2K and 30 MPa, 40 MPa, and 50 MPa.



These values were obtained experimentally using nuclear magnetic resonance (NMR) technique. The estimated accuracy of the measurements was ± 5%.

The gas-liquid diffusion coefficient, DAB was calculated from this relation:

)( 212 ABx

AB DxDxBD += (8)

where DAB is the diffusion coefficient, m2/s, x1 is the mole fraction of component B, x2 is the mole fraction of component A, DB is the self diffusion coefficient of component B, m2/s, DA is the self-diffusion coefficient of component A, m2/s and B2

x represents the relation between the thermodynamic driving force (the chemical potential gradient) and the driving force used in the definition of the diffusion coefficient, in which

PT

x

x

fB

,2

22 ln

ln1

∂∂

+= (9)

where f2 is the activity coefficient of component 2 / component A on mole fraction scale, x2 is the mole fraction of component A, T is the temperature, Kelvin and P is the pressure, MPa.

Option 2: Wilke chang correlation

For dilute solutions of non-electrolytes, empirical correlation given by Wilke and Chang [11] is recommended to estimate the diffusion coefficient, DAB. It is normally used for liquid-liquid diffusivities,

however at high pressures; this relationship is also valid for diffusivities of gasses into liquids. It is expressed as [7]:

( )6.0

18103.117

A

BAB

V

TMD

⋅⋅⋅⋅×

=−

µϕ (10)

where DAB is the diffusion coefficient of solute A in solvent B, m2/s, MB is the molecular weight solvent B, kg/kmol, T is the temperature, K, μ2 is the solvent viscosity, kg/m·s, VA is the solute A molar volume at normal boiling point, m3/kmol and Φ is the association factor for solvent B.

The solute A molar volume was calculated based on the table of atomic volumes for basic elements as given in Table 1.

Meanwhile, the solvent B association factor, φ, was obtained from Table 2.

Option 3: Average diffusion coefficient

The third option was to obtain an average diffusion coefficient based on the mole fraction:

∑−

=n

iABAB iDiyAverageD

1

)(*)( (11)

where DAB was obtained with Wilke-Chang correlation and y(i) is the mole fraction of component i.

Table 1. Table of atomic volumes for basic elements

Atomic Volume

Element Volume (m3/1000 atoms)

Carbon 0.0148

Hydrogen 0.0037

Chlorine 0.0246

Bromine 0.0270

Iodine 0.0370

Sulphur 0.0256

Nitrogen 0.0156

Table 2. Table of association factor

φ Solvent Polarity

2.6 Water Polar

2.26 Water (Hayduk and Laudvic, 1974) Polar

1.9 Methanol Polar

1.5 Ethanol Polar

1.0 Benzene, ether, heptone Non-polar



The gas-liquid diffusion coefficient determined from one of these options above enables the calculation of molar flux in equation 6.

SHuT-IN MODELING

The developed mass transfer rate model was integrated into dynamic wellbore modeling for shut-in well. The modeling of the transient fluid behavior after shut-in consisted of four interrelated mechanisms: (1) calculation of the reservoir fluid influx as the bottom-hole pressure increased to near wellbore reservoir pressure, (2) calculation of the mass transfer rate between the fluids during wellbore phase redistribution period, (3) compression of the fluids as the pressure increased and the bottom-hole pressure approached reservoir pressure, and (4) modeling of the segregated static column.

Once the well is shut-in at the wellhead, the fluid flow from the formation into the bottom of the wellbore continues unabated until compression of the fluids in the wellbore causes the bottom-hole pressure to rise. If the wellbore fluid is highly compressible and the well rate is low, the reservoir fluid influx period can be long. Conversely, high-rate wells producing little gas have negligible reservoir fluid influx periods. The influx period may last from minutes to many hours depending on the nature of the fluid properties and the capacity of the flow string. Hence, the calculation of the reservoir fluid influx over the shut-in period is crucial in dynamic wellbore modeling.

The second phenomenon occurs after shut-in is the wellbore phase redistribution, which is the fluid segregation whereby liquid moves downward as gas bubbles rise. Wellbore phase redistribution is the cause of pressure changes that are purely due to the segregation of liquid and gas phases in the wellbore. Mass transfer effect is significant during this period, as the gas bubbles and liquid must be in contact for mass transfer to take place.

Pressure-Volume-Temperature (PVT) models and Equations of State (EoS) were applied to calculate the fluids properties changes over the shut-in period. The bottom-hole build-up pressure profile was obtained by adding the hydrostatic head to the wellhead pressure over the shut-in period. The hydrostatic head over time was obtained by modeling the gas-liquid interface movement and the associate fluids’ gradient changes along the well. The calculated wellhead pressure was then matched with the recorded closed-in pressure for the model’s calibration.

RESuLTS AND DIScuSSIONS

To demonstrate the enhanced dynamic well model capability in mass transfer rate calculation, results from a stand-alone gas bubbles diffusion model is presented. The gas bubble size distribution along the wellbore immediately after shut-in was determined from the gas Reynolds number [1]. Diffusion coefficient of 4.52 x 10-9 m2/sec (for a methane-oil system), gas specific gravity of 0.54, gas bubble film thickness of 0.001 m, and an average gas bubble rise velocity of 3 ft/sec are used. Figure 2 shows that a gas bubble with an initial diameter of 5 mm formed at the bottom of the wellbore during shut-in, will travel approximately 2100 ft to dissolve completely into the liquid within 700 seconds after shut-in. As the gas bubble diameter decreases, the required distance and duration for a complete diffusion will decrease accordingly and vice versa as illustrated in Figure 3.

This result clearly shows that the diffusion of gas into oil occurs over an elapsed time. A minimum flow length which must not exceed the length of the well is required for complete gas diffusion. Else, the gas will stay in gas form which will translate into a lower hydrostatic head and thus, a lower bottom-hole build-up pressure. Therefore, it is important to apply mass transfer rate model to capture this well behavior. If the assumption of instantaneous mass transfer is translated into dynamic wellbore modeling, the simulated gas and oil phase redistribution in the well would be inaccurate.



Figure 4 illustrates an example of the shut-in profiles generated from the enhanced dynamic well model. A classic bottom-hole pressure build-up profile is obtained, taking into account the effects of reservoir fluid influx, mass transfer, fluids redistribution and compression. A bottom-hole near wellbore reservoir pressure of 2370 psi is achieved when the well reaches equilibrium at approximately 8900 seconds after the well is shut-in. Meanwhile, a decreasing reservoir fluid influx over time is observed. The amount of reservoir fluid influx is significant during the initial shut-in period. It decreases continuously until the fluids in the wellbore are sufficiently compressed where the bottom-hole wellbore

pressure is large enough for the flow from the reservoir to be negligible.

One of the fundamental principals for quantity and quality check in a model is volume balance. A volume balance is an application of conservation of volume to the analysis of physical systems. The conservation of volume is honored throughout the shut-in period as exemplified in Figure 5. The total fluids’ volume is equal to the tubing volume of 245 ft3 and it remained constant over the shut-in calculation.

As a result of the combination effects of reservoir fluid influx and pressure build-up, the total gas and

0

500

1000

1500

2000

2500

3000

0.00 1.00 2.00 3.00 4.00 5.00 6.00 7.00 8.00

Di�usion of Various Gas Bubble SizeSize = 1 mm

Size = 1 mm

Size = 1 mm

Size = 1 mm

Gas Bubble Diameter, mm

Dis

tanc

e Tr

avel

led

(ft)

0

100

200

300

400

500

600

700

800

0

500

1000

1500

2000

2500

0.00

1.00

2.00

3.00

4.00

5.00

6.00

Gas Bubble Di�usion

Gas Bubble Diameter, mm

Dis

tanc

e Tr

avel

led

(ft)

Tim

e (s

ec)

Figure 2. Gas bubble diffusion over distance and time

Figure 3. Distance of diffusion in association with gas bubble size



liquid volumes in the well changes over time. The gas volume is decreasing whereas the liquid volume is increasing. The gas is continuously compressed by the increasing wellbore pressure whereas the continuous reservoir fluid influx into the wellbore increases the total liquid volume in the well.

cONcLuSION

Dynamic wellbore models that simulate the wellbore fluid behaviors during shut-in are assuming instantaneous gas-liquid diffusion. These models are bound to have considerable drawback which results in inaccurate prediction of bottom-hole data.

Mass transfer concepts are suitable to depict the diffusion behavior between gas and liquid in shut-in well. The fundamental law of mass transfer by Fick is the basis for the enhanced dynamic well model. The binary diffusion coefficient, which has time dimension inherent, is one of principle properties in the calculation of mass transfer rate.

The derivation of the mass transfer rate model for shut-in well application is entirely fundamental with integration of basic gas law chemistry and established black oil correlation.

0.000

0.005

0.010

0.015

0.020

0.025

0.030

0.035

1800

1900

2000

2100

2200

2300

2400

2500

0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000

Shut-in pro�les

PBUQ in�ux

Shut-in Time (sec)

Bott

om-h

ole

Pres

sure

Bui

ld-u

p (p

si)

Rese

rvoi

r Flu

id In

�ux

(cu

ft)

Figure 4. Pressure build-up and fluid influx over time

70

90

110

130

150

170

190

210

230

250

0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000

Fluid Volume Pro�les

Total gas

Total liquid

Total �uid

Shut-in Time (sec)

Flui

d Vo

lum

e (c

u ft

)

Figure 5. Gas, liquid and total fluid volumes over time



The enhanced dynamic well model is able to estimate a reliable pressure build-up response during well shut-in and obtain the bottom-hole near wellbore reservoir pressure upon equilibrium. The calculation considered the gas-liquid mass transfer rate, fluids’ PVT characteristics, reservoir fluid influx, gas and liquid phase segregation over the well shut-in period.

AcKNOWLEDGEMENT

The authors would like to thank PETRONAS for the kind support and permission to publish this paper.

REFERENcES

[1] Coulson, J. M. and Richardsons, J. F.: “Chemical Engineering – Volume 2: Particle Technology & Separation Processes”, Butterworth-Heinemann (4th Edition), (1996) 98-109.

[2] Duns, H., Jr. and Ros, N. C. J.: “Vertical Flow of Gas and Liquid Mixtures in Wells,” Proc. Sixth World Pet. Congress, Frankfurt (Jun. 19-26, 1963) Section II, Paper 22-PD6.

[3] George Facer.: “A2 Adexcel Chemistry”, Philip Allan Updates (2006).

[4] Helbaek, M., Hafskjold, B., Dysthe, D. K. and Sorland, G. H.: “Self Diffusion Coefficients of Methane or Ethane Mixtures with Hydrocarbon at High Pressure by NMR”, J. Chem. Eng. Data (1996) 41, 598-603.

[5] James, E. B. and Fred, S.: “Chemistry: Matter and Its Changes”, John Wiley & Sons (4th Edition), (2004) 474.

[6] James, R. W., Charles, E. W. and Robert, E. W.: “Fundamentals of momentum, heat, and mass transfer”, John Wiley & Sons (1983) 471.

[7] Robert E. Treybal.: “Mass-Transfer Operations – Third Editions”, McGraw-Hill Book Company (1981), 21-123.

[8] Slider, H. C.: “Worldwide Practical Petroleum Reservoir Engineering Methods”, Pennwell Books, 2nd Edition (1983) 157, 159.

[9] Tarek A.: “Hydrocarbon Phase Behavior”, Gulf Publishing Company (1989) 177-221.

[10] Vai Yee H. and Suzalina Z.: “Enhanced Dynamic Well Model for Reservoir Pressure Determination”, Malaysia patent pending, filed November 17, 2009.

[11] Wilke C. R. and Chang P.: “Correlation of Diffusion Coefficients in Dilute Solutions”, A.I.Ch.E. Journal (1955).

[12] Yunus A. Cengel.: “Heat and Mass transfer: A Practical Approach, 3rd Edition (SI Units)”, McGraw-Hill Education (Asia) (2006) 773-827.



INTRODucTION

Natural gas has emerged as a major source of clean energy due to environmental factors and volatility of crude oil price. A gas processing plant (GPP) faces challenges on three fronts namely: 1) at plant inlet, multiple streams of feed gas from various producers are mixed causing fluctuations in feed gas flow rate and composition; 2) within GPP, unscheduled shutdowns due to equipment malfunction often occur; 3) at a GPP outlet, strict specifications of several products are regularly enforced by its customers where penalties will be imposed if these specifications are violated [1]. In the aspect of business, GPP enters into diverse agreements with gas producers. As a result, prices of feed gas vary depending, among others, on quality of gas and tenure of the contracts. In contrast, price of sales gas is tightly regulated by government. Prices of liquids

namely ethane, propane, butane and condensates are floated to market values.

These challenges forces GPP to improve its operational efficiency in order to maintain profitability. An identified area of improvement is during change of plant operating mode from natural gas liquids to sales gas, or vice-versa. The change of plant mode poses a short-term (weeks) and continuous scheduling problem in which pre-configured set points are directly implemented by regulatory controllers. While this practice has been accepted in the past, efforts are currently undertaken to improve it. This type of problem differs from batch scheduling, which receives considerable attention in operational research. Excellent reviews of batch scheduling have been published by Floudas et al. [2] and by Mendez et al. [3]. Continuous scheduling is often integrated with control giving rise to mixed integer dynamic

INTEGRATED ScHEDuLING AND RTO OF RGP WITH MPc AND PI cONTROLLERS

Nooryusmiza Yusoff* and M. Ramasamy Universiti Teknologi PETRONAS, 31750 Tronoh, Perak Darul Ridzuan, Malaysia

* [email protected]

ABSTRAcT

This paper proposes an integrated framework of scheduling and real-time optimization (RTO) of a refrigerated gas plant (RGP). At the top layer, a dynamic model of RGP is subjected to scheduling of plant operating mode from natural gas liquids to sales gas, and vice-versa. Setpoints from scheduling are passed to the steady-state RTO layer. Optimal trajectories of setpoints are obtained using sequential quadratic programming algorithm with constraints. These trajectories are disjointedly implemented by model predictive control (MPC) scheme and proportional-integral (PI) controllers for comparison. Four case studies for each mode scheduling are performed to illustrate efficacy of the proposed approach.

Keywords: Gas plant, scheduling, real-time optimization, model predictive control

This paper was presented at the International Conference on Chemical and Bioprocess Engineering in conjunction with Symposium of Malaysian Chemical Engineers, 12 - 14 August 2009, Kota Kinabalu



optimization (MIDO) problem. An example can be found in Chatzidoukas et al. [4] who formulated a MIDO problem on gas-phase copolymerization in fluidized bed reactor. The authors simultaneously optimized grade transition time of a copolymer and schemes of feedforward-feedback control. In another related work on polymerization, Kadam et al. [5]integrated real-time optimization (RTO) with model predictive control (MPC) within a dynamic framework of grade transition problem. A method of tracking the necessary conditions of optimality with a solution model was employed to preserve a feasible and optimal operation under uncertainty. The addition of RTO layer between scheduling and control layers was necessary in order to improve plant economics, production or other suitable objectives. Since scheduling is normally performed at a much larger time-scale (days to weeks) as compared to RTO (hours to days) and MPC (seconds to minutes), integration of the three automation layers to enhance economic benefits was difficult.

This paper proposes a potential means to address this issue through an integrated approach of scheduling and RTO. Set points were re-calculated based on current plant conditions. The optimal set points may be implemented using regulatory or advanced controllers such as the MPC scheme. A refrigerated gas plant (RGP), which is the low temperature separation unit and sales gas compression unit of the GPP, was employed as a test bed. Steady-state and dynamic models of RGP were simulated under the HYSYS environment. MPC actions were calculated using MATLAB. Communication between HYSYS and MATLAB was executed via the component object module technology. Efficacy of the proposed approach is illustrated by several cases of mode scheduling from natural gas liquids to sales gas and vice-versa. As its name implies, an operating mode refers to a state whereby RGP is producing larger amounts of the respective product.

INTEGRATION OF ScHEDuLING AND RTO

In a typical scheduling scenario, new plant set points are pre-determined from early design specifications

or heuristics. The set points are manually adjusted by experienced personnel to the desired levels. This practice has several drawbacks: 1) current state of the plant may change due to sustained large disturbance or major revamp activities and thus invalidate design set points, and 2) manual adjustment of set points may lead to excessive energy utilization especially when the target trajectory is not optimal. One way to overcome these drawbacks is to integrate scheduling tasks with real-time optimization (RTO) before passing set points to the control layer. The proposed approach is illustrated in Figure 1.

The new methodology leverages on availability of first-principle models in both steady-state and dynamic modes. This is necessary to maintain accuracy when data are transferred between the two models. In particular, scheduling is to be carried out in a dynamic model DM1 until a new steady-state is reached. Data from DM1 are to be passed to the steady-state model for the target optimization task. For practical reasons, only values of key variables will be exchanged to minimize mismatches between the dynamic and steady-state models. The number of variables sent from the dynamic to the steady-state model will be larger than the reverse. This is to ensure that: 1) there is rigorous calculation at the RTO layer, and that 2) feasible set points are passed to the

Figure 1: Structure of integrated scheduling and real-time optimization (RTO) approach (solid line). Set points may be implemented via model predictive control (MPC) scheme or, alternatively, regulatory controllers. Current approaches of scheduling implementation are illustrated as dashed line.

Scheduling

RTO

MPC

Regulatory Control

Plant



controllers for implementation at the plant. Another dynamic model DM2 is to be used to represent the plant. The only difference between the DM1 and DM2 models is that the latter is at the state prior to scheduling.

This nonlinear dynamic model of a plant can be implicitly described by the following set of differential-algebraic equations:

(1)

(2)

(3)

where x and z are differential and algebraic state variables, respectively. Process output is denoted by y whereas model and design parameters by p. Equations (1) and (2) are solved over fixed time horizon ],[ 0 fttt∈ given y, p and initial conditions x0.

The above dynamic model was used for scheduling (DM1) and control (DM2) implementation. RTO was performed based on a steady-state model represented by Equations (1) and (2) without the transient term. In general, an RTO problem can be written in the following form:

(4a)

subject to:

(4b)

(4c)

(4d)

(4e)

(4f)

where fE is an economic objective function; gE and hE are sets of inequality and equality constraints,

],[),( 0 fm ttt,t,,,fdt

d ∈=

pzyxx

],[),,,,,(0 0 fm ttttg ∈= pzyx

00 )( xx =t

Efssss yu ,

max

0),,,,( <pzyxu ssssssssEg

0),,,,( =pzyxu ssssssssEh

),,,( pzxuy ssssssm

ss f=

maxmin yyy ≤≤ ss

maxmin uuu ≤≤ ss

respectively. Steady-state process output yss and input uss are bounded within their corresponding minimum and maximum values. State variables xss and zss are concurrently updated via steady-state plant model. The optimization problem (Equation 4) can be solved using a sequential quadratic programming algorithm with constraints [6]. Optimization variables must be prudently selected so that they are able to surrogate the relevant control variables employed in the dynamic model.

APPLIcATION TO RGP

This section describes application of the proposed integrated approach of scheduling and RTO. Since the RGP was employed as a test bed, its process is briefly described and the results from several case studies are discussed.

RGP Model

In this work, steady-state and dynamic models of RGP developed previously [7, 8] were employed (Figure 2). The models are based on a first-principle modeling approach. Accuracy of the models reached about 95% when validated against the actual plant data. Size of the models were large with 762 differential-algebraic equations. However, modeling complexity was reduced by simulating the RGP process with a modular approach under the HYSYS environment.

In short, the RGP processes mixed feed gas at the normal plant throughput of 280 ton/h. Feed gas at 20°C and 60 bars was cooled by exchanging heat with sales gas in three coldboxes (E-101, E-103, E-105), a propane refrigeration cooler (E-102) and an air cooler (E-106). To enhance vapour-liquid separation, feed gas was flashed in two stages. Most vapour was expanded in a turboexpander (KT-101) whereas some was in a Joule-Thompson valve depending on the throughput level. Liquids were fed to various stages of a demethanizer (C-101). Top product of the demethanizer and that from the expansion process were sent to an absorber in a gas subcooled process (GSP) unit to improve recovery of natural gas liquids. Bottom product of the demethanizer was further



processed to separate the liquids into ethane, propane, butane and condensates. Top product of the absorber containing sales gas was recompressed twice to meet minimum specification of 30 bars.

Mode Scheduling

To illustrate the efficacy of the proposed approach, four studies were performed in the case of scheduling of the RGP operation mode from natural gas liquids to sales gas mode as follows (Case A):

1. Mode scheduling only with PI controllers (base case).

2. Mode scheduling only with MPC controllers.3. Integration of mode scheduling and RTO with PI

controllers.4. Integration of mode scheduling and RTO with

MPC controllers.

The above studies were repeated for mode scheduling of sales gas to natural gas liquids (Case B). In total, eight case studies were conducted. Each case was simulated for 510 min.

For scheduling only cases, set points were pre- determined from either design specifications or guidelines. In the case of scheduling from natural gas liquids to sales gas mode, RGP temperature as indicated by the top of absorber increased by about 20oC. A higher plant temperature was achieved partly by increasing the top of demethanizer pressure from 22 to 24 barg. Since the turboexpander KT-101 was discharged at the same pressure, more feed gas was diverted to the Joule-Thompson (J-T) valve for expansion without overloading the booster (K-101) and sales gas (K-102) compressors to meet specification of sales gas pressure. Coupling the effect of higher feed gas flow to the gas subcooled process (GSP) unit by 7.8 ton/h, value of sales gas increased substantially for RGP to yield a higher profit margin.

In the integrated approach, set points were optimized at the current plant state before implementation by controllers. Values of set points and constraints for all cases are shown in Tables 1 and 2, respectively. Pre-cooling of feed gas was shifted more heavily on coldbox E-101 with the exit stream temperature decreased by 3.1°C. Load on cooler E-102 reduced

Figure 2. RGP process and instrumentation diagram.



a little due to higher set point for its exit stream temperature by 0.4°C. Similarly, feed gas vapour entering coldbox E-103 was subjected to less cooling. The vapour exited the coldbox by 3.7°C hotter. This phenomenon caused the demethanizer tray 35 temperature to rise significantly to 15.2°C from it previous condition of 5°C. Consequently, more ethane and propane escaped the demethanizer as vapours at the top instead of liquids at the bottom.

Real-time Optimization

Seven variables were selected as optimization variables. Descriptions and bounds of these variables are shown in Table 1. For practical reasons, only the inequality constraints were specified. This way, a feasible solution could be quickly obtained within the RTO sampling interval of 200 min. The long sampling interval was necessary to match the open-loop settling time of the MPC controller. The total number of constraints was thirty-four including product specifications as well as limits in plant throughput, equipment and processes (Table 2). The solution to the optimization problem (Equation 4) is a set of targets that are passed to the control layer for implementation.

control

The control layer consists of two sub-layers namely MPC and regulatory control layers. For safety and reliability reasons, direct communication between the MPC and plant is currently prohibited [9]. MPC

action can only be implemented through regulatory controllers, which may take over from MPC if and when the need arises. Between the two controllers, the MPC was the preferred one due to its capability in handling multivariable controls with constraints [10]. Both PI and MPC controllers were tuned for set point changes in order to fairly compare their performances.


In the case of scheduling from sales gas to natural gas liquids mode, the RGP was cooled to -92°C. This was achieved partly by lowering the temperature of the streams exiting at all three coldboxes and cooler as specified in Table 1. Feed gas flow to the GSP unit was also increased significantly from 1.2 to 34.5 ton/h. In addition, the top of demethanizer pressure was decreased from 24 to 22 barg. This procedure induced higher recovery of liquids. Optimal set points for maximising values of natural gas liquids were obtained while maintaining operational stability at new conditions. Set points of streams exiting cooler E-102 and coldbox E-103 were increased by 0.9 and 1.7oC, respectively. At the same time, set point of demethanizer tray 35 temperature was reduced to 0.4oC from the previous state of 5oC. This action reduced both cooling and reboiling loads, and thus operational expenses. Feed gas flow to the GSP unit was decreased by almost one-half to reduce heat exchange between processed gas and sales gas at coldbox E-105. On the other hand, more feed gas flowed to turboexpander

Table 1. Values and bounds of optimization variables

Variable Unit

NGL Mode SG Mode Bounds

DescriptionNormal Optimal Normal Optimal Lower Upper

y1 oC -30.5 -30.4 -22.0 -25.1 -30.0 0.0 After coldbox E-101 stream temperature

y2 oC -40.0 -39.1 -30.6 -30.2 -42.0 0.0 After cooler E-102 stream temperature

y3 oC -53.9 -52.2 -42.6 -38.9 -60.0 0.0 After coldbox E-103 stream temperature

y4 oC 5.0 0.4 5.0 15.2 0.0 20.0 Demethanizer tray 35 temperature

y5 ton/h 34.5 18.8 1.2 9.0 0.0 40.0 Mass flow to gas subcooled process

y6 ton/h 195.3 209.8 246.4 161.0 0.0 310.0 Mass flow to turboexpander

y7 oC -80.9 -70.0 -70.0 -70.0 -100.0 -70.0 After coldbox E-105 stream temperature



Table 2. Values and bounds of constraint variables

Variable Unit

NGL Mode SG Mode Constraint

DescriptionNormal Optimal Normal Optimal Min Max

x1 MJ/m3 38.1 38.3 38.8 39.2 35.1 48.1 Gross heating value of sales gas

x2 - 0.57 0.57 0.58 0.59 - 0.75 Specific gravity of sales gas

x3 mol/mol 0.001 0.002 0.002 0.002 - 0.020 Carbon dioxide content in sales gas

x4 ton/h 224.7 228.6 235.4 241.1 206.0 - Mass flow of sales gas

x5 bar 33.5 33.5 33.5 33.5 30.0 - Pressure of sales gas

x6 oC 35.7 33.4 28.5 32.2 - 50.0 Temperature of sales gas

x7 ton/h 280.0 280.0 280.0 280.0 100.0 310.0 Mass flow of feed gas

x8 kW/oC 1442 1098 1024 1707 - 2000 Coldbox E-101 capacity

x9 oC 8.3 10.9 10.0 6.0 5.0 - Coldbox E-101 LMTD

x10 kW/oC 341 283 258 155 - 800 Coldbox E-103 capacity

x11 oC 13.6 15.2 15.5 17.0 5.0 - Coldbox E-103 LMTD

x12 kW/oC 132.0 34.4 0.7 113.6 - 400 Coldbox E-105 capacity

x13 oC 19.2 22.6 28.2 5.0 5.0 - Coldbox E-105 log LMTD

x14 kW 3283 2954 1715 1588 - 4000 Cooler E-102 duty

x15 kW 4548 3359 2525 2362 - 4700 Demethanizer C-101 reboiler duty

x16 kW 2248 2465 2850 1935 - 4000 Turboexpander KT-101duty

x17 kW 4653 4192 2859 3957 - 4700 Compressor K-102 duty

x17 kW 28.0 25.9 15.8 13.8 - 30.0 Pump P-101 duty

x19 kW 9.3 7.6 10.5 8.4 - 15.0 Pump P-102 duty

x20 % 62.0 47.7 31.2 26.4 - 85.0 Flooding at Section 1 of demethanizer




x24 % 31.4 28.9 30.8 28.6 - 50.0 DC backup at Section 1 of demethanizer




x28 % 68.5 71.7 72.3 70.7 - 85.0 Flooding in absorber C-102

x29 % 17.7 17.2 16.9 16.9 - 50.0 DC backup in absorber C-102

x30 oC 12.1 8.4 5.4 6.2 5.0 - Air cooler LMTD

x31 - 1.00 0.98 1.00 0.66 0.00 1.00 Ratio of gas to expander over that to JT valve

x32 - 0.75 0.61 0.51 0.50 0.00 1.00 Ratio of gas to coldbox E-103 to that bypasses it

x33 - 1.00 0.87 0.78 0.80 0.00 1.00 Ratio of gas to coldbox E-101 to that bypasses it

x34 - 0.150 0.081 0.005 0.035 0.005 0.150 Fraction of gas to gas subcooled process

LMTD=log mean temperature difference; JT=Joule-Thompson; DC=downcomer



KT-101 that was mechanically linked to booster compressor K-101. In turn, this action also reduced expenses since operating the sales gas compressor K-102 is much more expensive than maintaining the turboexpander-compressor.

During simulation, set points were only introduced to the plant after 30 min to show the plant’s previous steady-state level. At the end of experiments in all cases, new steady-states were reached. As an example, dynamic trajectories and final states of RGP profit are shown in Figure 3. The top two figures refer to mode scheduling from natural gas liquids to sales gas and the bottom two figures refer to the other mode scheduling. Results of scheduling only cases are presented on the left of Figure 3 whereas those from an integrated approach are on the right. In general, change of plant mode from natural gas liquids to sales gas resulted in a higher profit margin. The reverse is true for the other mode of scheduling. Economic results are tabulated in Table 3. For fair comparison of different online procedures, instantaneous values were averaged out over the entire simulation time as [11].

(5)

where FE and EF , respectively, denote instantaneous and average economic parameters namely profit, revenues and expenses over time horizon [t0,tf].

Profit is taken as a function of revenues and expenses [7]. Revenues are derived from the values of sales gas

∫−=

ft

tE

f

E dttFtt

tF0

)(1

)(0

and natural gas liquids. Expenses are due to costs of feed gas and operation. The operational costs include those which emanate from refrigeration and reboiler duties, compressor fuel gas, turboexpander maintenance and pumping actions.

case A (Natural Gas Liquids to Sales Gas Mode)

Four studies were performed in this case of mode scheduling from natural gas liquids to sales gas mode. Case A1 is the base case where scheduling was implemented by PI controllers. For the same set points implemented by the MPC controller as in Case A2, RGP profit increased by 0.1%. This was achieved by an increase in revenue from sales gas by the same quantum but decrease in revenue from liquids by 0.5%. At the same time, operational expenses decreased by 1.3% due to the efficiency of the MPC controller in bringing the plant to a new state optimally.

Cases A3 and A4 show that additional benefits can be achieved if scheduling set points are optimized at the RTO layer before they are implemented by controllers. Profit increased by 0.4 and 0.5%, respectively, for Cases A3 and A4. Revenue from sales gas increased by 0.9% in both cases. However, sharp declines in revenue from liquids were noticed at 4.6 and 4.7% for Cases A3 and A4, respectively. In terms of operating expenses, PI controllers managed to obtain a reduction by 1.3% whereas the MPC scheme by 4.4%.

case B (Sales Gas to Natural Gas Liquid Mode)

Similar to Case A, four case studies were carried out in Case B. This time, scheduling was performed from sales gas to natural gas liquids mode. Case B1 was used as a basis to be consistent with studies in Case A. In Case B2, negligible benefit was achieved even though set points were implemented by the MPC controller. This happened because economic parameters almost cancelled out each other with 0.6% reduction in operating expenses was matched with 0.1% reduction in revenue.

Table 3. Average values (RM/min) of economic parameters over 510 min simulation time.

Case Profit

Revenues Expenses

Sales Gas Liquids Feed Gas Operation

A1 1921.75 2954.99 417.24 1435.55 14.93

A2 1923.35 2958.38 415.27 1435.57 14.73

A3 1929.70 2982.02 398.03 1435.61 14.74

A4 1930.58 2982.76 397.60 1435.50 14.28

B1 1895.38 2843.57 508.91 1435.55 21.54

B2 1895.70 2844.22 508.45 1435.56 21.41

B3 1900.81 2856.18 500.19 1435.62 19.95

B4 1901.27 2857.68 499.08 1435.58 19.92



0 100 200 3001700

1800

1900

2000

2100

1700

1800

1900

2000

2100

Pro

fit (

RM

/min

)

Pro

fit (

RM

/min

)

Pro

fit (

RM

/min

)

Pro

fit (

RM

/min

)

Time (min)

Case A1

Case A2

0 100 200 300Time (min)

Case A3

Case A4

0 100 200 3001700

1800

1900

2000

2100

Time (min)

Case B1

Case B2

0 100 200 3001700

1800

1900

2000

2100

Time (min)

Case B3

Case B4

An integrated approach of scheduling and RTO were represented in Cases B3 and B4. In terms of profit, the benefit was 0.3% for both cases. This was achieved using slightly different means by the PI and MPC controllers. The former increased revenue from sales gas by 0.4% whereas the latter by 0.5%. The PI and MPC controller actions reduced revenue from liquids by 1.7 and 1.9%, respectively. However, this gap between both controllers shrank to 0.1% in terms of benefit derived from operating expenses. For this particular case, the reduction indicates that the optimal set points sent by the integrated approach may be executed by either PI or MPC controllers since the latter may have exhausted all efforts in obtaining an optimal trajectory for its manipulated variables.

cONcLuSION

A new approach of integrating scheduling and real-time optimization is illustrated using a refrigerated gas plant (RGP) as a test bed. Efficacy of the proposed approach was demonstrated in several cases of scheduling of RGP from natural gas liquids to sales gas, and vice-versa. Both PI and MPC controllers were employed to implement newly calculated set points. The MPC controller was found to bring additional benefits because the RGP was taken to another state optimally.

AcKNOWLEDGEMENT

Financial support from Universiti Teknologi PETRONAS is greatly appreciated.

Figure 3. Dynamic trajectories of RGP profit for Cases A and B.



REFERENcES

[1] K. A. Bullin, “Economic Optimization of Natural Gas Processing Plants Including Business Aspects,” Ph.D. Dissertation. Texas A&M University, USA, 1999.

[2] C.A. Floudas and X. Lin, “Continuous-Time Versus Discrete-Time Approaches for Scheduling of Chemical Processes: A Review,” Computers and Chemical Engineering, vol. 28, 2004, pp. 2109-2129.

[3] C. A. Mendez, J. Cerda, I.E. Grossmann, I. Harjunkoski and M. Fahl, “State-of-the-art Review of Optimization Methods for Short-term Scheduling of Batch Processes,” Computers and Chemical Engineering, vol. 30, 2006, pp. 913–946.

[4] C. Chatzidoukas, J.D. Perkins, E.N. Pistikopoulos and C. Kiparissides, “Optimal Grade Transition and Selection of Closed-Loop Controllers in A Gas-Phase Olefin Polymerization Fluidized Bed Reactor,” Chemical Engineering Science, vol. 58, 2003, pp. 3643-3658.

[5] J. Kadam, B. Srinivasan, D. Bonvin and W. Marquardt, “Optimal Grade Transition in Industrial Polymerization Process via NCO Tracking,” AIChE Journal, vol. 53(3), 2007, pp. 627-639.

[6] R.M. Chamberlain and M.J.D. Powell, “The Watchdog Technique for Forcing Convergence in Algorithms for Constrained Optimization,” Mathematical Programming Study, vol. 16, 1982, pp. 1-17.

[7] N. Yusoff, M. Ramasamy and S. Yusup, “Profit optimization of a refrigerated gas plant,” ENCON, Kuching, Sarawak, Malaysia, 2007.

[8] N. Yusoff, M. Ramasamy and S. Yusup, “A simulation study on dynamics and control of a refrigerated gas plant,” FOCAPO, Boston, MA, 2008.

[9] P. Tatjewski, “Advanced Control and On-line Process Optimization in Multilayer Structures,” Annual Reviews in Control, vol. 32, 2008, pp. 71-85.

[10] H. Huang and J.B. Riggs, “Comparison of PI and MPC for Control of a Gas Recovery Unit,” Journal of Process Control, vol. 12, pp. 163-173.

[11] S. Ferrer-Nadal, I. Yelamos-Ruiz, M. Graells and L. Puigjaner L. “An Integrated Framework for On-line Supervised Optimization,” Computers and Chemical Engineering, vol. 31, 2007, pp. 401-409.

Dr. Nooryusmiza Yusoff is a Senior Lecturer at the Universiti Teknologi PETRONAS (UTP), where he is currently serving as Associate Head of Chemical Engineering Department. He graduated with BSc Degree in Chemical Engineering (1997) from Northwestern University, USA; MSc Degree (2001) from University of Calgary, Canada; and PhD (2010) from UTP,

Malaysia. He has authored two books: 1) Statistical Analyses of Ozone Temporal Trends (2006), and 2) Profit Optimization (2010). His areas of research interest center on process modeling and simulation, advanced process control as well as process systems engineering.

Dr. M. Ramasamy is presently an Associate Professor in the Department of Chemical Engineering at Universiti Teknologi PETRONAS (UTP). He graduated from Madras University in the year 1984. He obtained his masters degree with specialization in Process Simulation, Optimization and Control from Indian Institute of Technology, Kharagpur in 1986.

In 1996, he received his Ph.D. from Indian Institute of Technology, Madras. His areas of research interests include modeling, optimization and advanced process control. Dr. Ramasamy has guided several undergraduate and postgraduate students and published/presented several technical papers in international refereed journals, international and national conferences. He has delivered a number of special lectures and also successfully organized seminars/symposiums at national level. Presently, he is the project leader for the PRF funded project on “Optimization of Energy Recovery in Crude Preheat Train” and eScience funded project on soft sensor development.



A REVIEW OF RISKS AND MITIGATION MEASuRES IN BuILD-OPERATE-TRANSFER PROJEcTS

Kalaikumar Vallyutham*, Syed Kamarul Bakri Syed Ahmad Bokharey , Narayanan Sambu Potty and Nabilah Abu Bakar

Universiti Teknologi PETRONAS, 317500 Tronoh, Perak Darul Ridzuan, Malaysia *[email protected]

ABSTRAcT

Infrastructure investments are important in developing countries, it will not only help to foster the economic growth of a nation but it will also act as a platform in which new forms of partnership and collaboration can be developed mainly in East Asian countries. In the past two decades, many infrastructure projects have been completed through the build-operate-transfer (BOT) type of procurement. The development of BOT have attracted participation from local and foreign private sector investors, to secure funding and to deliver projects on time, within the budget and to the required specifications. The private sector was preferred by the governments in East Asia to participate in BOT projects due to lack of public funding which had caused the private sector or promoter of the BOT projects to be exposed to multiple risks. These risks are discussed in this paper. Effective risk management methods and good managerial skills were required in ensuring the success of the project. The review indicated that mitigation measures should be employed by the promoter throughout the concession period and support from the host government is also required in ensuring the success of the BOT project.

Keywords: BOT project, risks management, concessionaire, consortium.

INTRODucTION

Since the 1980s in East Asian countries, build-operate-transfer (BOT) projects have been popular and have been evolving vigorously. This is in alignment with the need for basic infrastructure to develop the countries. However, the lack of huge funds had urged host governments to utilize the BOT type of procurement. Table 1 shows the involvement of the private sector in infrastructure projects in East Asia and Pacific. The energy sector recorded the highest investment amounting to USD101,187 million followed by telecommunication, transport, and water and sewerage.

The BOT is a type of infrastructural project which is based on the granting of a concession by a principal, usually a government, to a promoter sometimes known as the concessionaire who is responsible for the construction, financing, operation and maintenance of a facility over the period of the concession before finally transferring the facility, at no cost to the principal, in a fully operational condition [2]. The facility will be operated by the concessionaire during the concession period to generate revenue, to settle debt payments and profit for the investment. The concession contract will bind the host government and promoter throughout the concession period which is not just limited to initiation, implementation, operation and

This paper was presented at the World Academy Of Science, Engineering And Technology, Rio de Janerio, Brazil, 29 – 31 March 2010.



maintenance, and handing over. Although private sector companies were given the opportunity to build and operate the facility, the role of the host government in supporting BOT privatized projects determined its success or failure.

NATuRE OF RISKS IN BOT PROJEcTS

The existence of risk in every single project is unquestionable. The type of risks are varied depending on the nature and size of the project. Risks in BOT project are solitary and different but somehow always related to the phases in a project namely, initiation, implementation and operational phases. In order to accomplish the objectives, the risks need to be managed wisely.

Proper measurements have to be taken into account to allow for uncertainness in judgement which might take place during the implementation of the project. Any BOT project is subjected to influences from within and outside the organization itself. The objectives of the project are governed by

the stakeholder who possesses expectations and interests in it and which could jeopardize the project if the risks are not properly handled and managed.

For instance, a host government would place public interest first for any project. The host government would be actively involved during the pre and post implementation stage by revising the design to ensure the safety of the finished product for the public. In addition to that, changes will be made to ensure that the project complies with all local regulations such as environmental safety requirements, building codes and other stipulated regulations. Thus, it becomes the responsibility of the promoter of the project to explore for the best solution to tackle those risks and manage it well within the context of the concession contract.

DEFINITION OF RISK

According to Merna and Thani [3], risk is characterized as the potential for unwanted negative consequences of an event or a measure of the probability and the

Table 1. Total Projects by Primary Sector and Subsector for East Asia and Pacific Region (USD million) [1]

Primary Sector Sub Sector Project count Total Investment

Energy Electricity 381 91,766

Natural Gas 191 9,421

Total Energy 572 101,187

Telecom Telecom 71 78,145

Total Telecom 71 78,145

Transport Airports 27 4,599

Railroads 20 15,198

Roads 194 36,162

Seaports 101 18,824

Total Transport 342 74,782

Water and sewerage Treatment plant 297 8,195

Utility 57 20,494

Total Water and sewerage 354 28,689

Grand Total 1,339 282,804



severity of adverse effects. Generally, numerous decisions would be made in a BOT project based on assumptions and expectations that might be encountered during the project. Whilst, definition of risk and uncertainty by Raftery [4] is:

“Risk and uncertainty characterize situations where actual outcome for a particular event or activity is likely to deviate from the estimate or forecast value.”

The given definition does not consider other important key elements that could influence the project’s activities. Definitions of risk and uncertainty in a BOT project could be extended to include uncertainties in financial markets, construction problems, demand forecasts, instability in a country’s economic situation, uncertainties in host government organizations, stakeholders’ expectations and other external aspects of the projects. Chapman [5] defined risk as exposure to the possibility of economic and financial loss or gain, physical damage or injury, or delay as a consequence of the uncertainty associated with pursuing a

particular course of action. Further to that, Smith [6] provided several absolute definitions of risks for construction projects from several references:

1) Association of Project Management (2002): ‘a combination or frequency of occurrence of a defined threat or opportunity and the magnitude of that occurance’.

2) HM Treasury (2001): ‘the uncertainty of outcome, within a range of potential exposures, arising from a combination of the impact and probability of events’.

3) BS 6079 (British Standard Institutions, 1996): ‘is the uncertainty inherent in plans and the possibility of something happening that can affect the prospects of achieving business or project goals’.

4) Smith (2002): ‘risk is adverse but and unknown by its nature can have both positive and negative effects’.

Uncertainty surrounding a factor or

event

Effect of factor or event on the

project outcome

Probabilityof occurrence of

the factor or event

Probability Distribu�on for the outcome values

Figure 1. The Concept of Risk [3]



Primarily, the outcome of implementation could be affected directly by the risks in a BOT project. Figure 1 shows a concept model of risk developed by Merna and Al-Thani [3] which included uncertainty, probability, effect and outcome.

SOuRcES OF RISKS

The progress of a project corresponds with the occurrence of risks. Risks could categorized into three major captions: financing, political and technical

risks [7]. The success of a project could be measured by the overall project cost, duration and quality of the final product or services delivered. Usually the risks correspond with these three parameters. The risks could also be clustered as global and elemental risks [3].

Global risks are defined as being exerted externally to the project environment. Adversely, elemental risks originate from the sources within the project structure which are manageable within the elements

Table 2. Summary of Types and Sources of Risks in BOT project

No Type of Risks Sources Mitigation Measures

1 Country Risk • Unstable government• Inadequate foreign reserve

• Carry out a thorough country risk profile and budgetary practices by reliable third party such as reputable management consultant or refer to the World Bank, ADB or the United Nations

2 Financial Risk • Wrong financial assumption and packaging

• Inadequate cash flow• Poor feasibility study

• Consult a top-notch financial consultant to conduct the projection or verifying the financial report of the project

3 Construction Risk

• Poor design report• Prolong construction schedule• Changes in factors of

production

• Ensure good design report and vetted by owner and consultant before the project commences

4 Inadequacy in Concession Contract

• Major terms are not included• Exit clause not done• Variation according to time

or economic condition not provided

• Compare the contract with proven and similar good concession contract

• Employ a good legal advisor who are familiar with the industry in drafting the contract

5 Shareholders’ Risk

• Unsupportive shareholders• Loggerhead between

shareholders

• Work closely with the major shareholders and know their aspirations

6 Market Risk • Change in market trend • Carry out a thorough market research before embarking on the project

7 Changes in Key Management Personnel Risk

• Poor working conditions and benefit

• Provide good benefits and working environment to selected key management personnel

8 Operation and Maintenance Risk

• Unreliable operations and maintenance team

• Poor machinery and equipment installation

• Poor technical feasibility reports and design

• Operations and maintenance agreement must have benefits or benefit reduction to operation and maintenance company

• Provide a maintenance manual and update it regularly



of the project [3]. The nucleus element of a BOT project is the concession agreement which could compose a number of variables that influence the whole project cycle. Diversity in the variables could result in deviations from the objective and direction of the project. Essentially, this might jeopardize the project’s output and could lead to a negative impact on the investment if the promoter were to fail to address it effectively.

The promoter should thoroughly investigate the various sources of risks before making any decision in a BOT project. Based on the literature review, major types and sources of risks are summarized in Table 2. It also shows appropriate mitigation solutions to overcome the identified risks which might occur at different times during the concession period. It has become the responsibility of the promoter to assess and manage the risks with diligence to minimize or prevent any obstacle to the overall progress of the project.

Risks have a vital role in a BOT project. Delmon [8] stated other sources of risks that should be considered: capital budget, construction time, construction cost, operation cost, politics and policies, market conditions, stakeholders’ cooperation and credibility as well as global economic environment. From a study on the impact of risks to the performance of a BOT project, Zayed and Chang [9] concluded that an organized measure for risk management is sufficient for the concessionaire. The concessionaire should possess the capability to minimize (limit and confine) the impact to the project outcome by understanding, analyzing and responding to the risks.

A BOT project would be classified as a success by measuring the revenue to the concessionaire over the period. The promoter should be able to get rid of a reasonably high degree of risks and maximizing profit, without any tolerance to the objective and requirements. A well experienced promoter would be able to identify and understand the associated risks of a BOT project, and inherently address it effectively.

TYPES OF RISKS IN BOT PROJEcTS

Apart from the risks shown in Table 2, there are also other innumerable risks associated with a BOT project. These risks could be categorized as follow:

A. Financial risks – currency risks, internet rate risk, equity risk, foreign exchange risk, commercial risk, liquidity risk, counter party risk and economic risk

B. Political risks – sovereign risks and country risk

C. Technical risks – construction risk and operation and maintenance risk

D. Other risks – market risk, inadequacy of concession contract, shareholders’ risk and risks associated with changes among key management personnel

An inexperienced promoter might overlook the risks in the BOT project due to lack of information and uncertainty about future conditions, which could itself also be a risk. Consequently, alleviative measures must be established and supervised with care with means to reduce and manage the risks effectively to an acceptable level and minimizing the chances of a project failure.

A) Financial Risk

Merna and Njiru [2] defined financial risk as the impact on the financial performance of any entity exposed to risk. Investors and lenders are aware of certain risks and willing to face it, in order to gain the profit from their investment. The higher the risk, the higher the profit shall be gained. Developing countries in East Asia are facing obvious financial risk and the source of financial risk could be summarized as below: [10].

1) Currency risks2) Interest risks3) Equity risks4) Foreign exchange risk



5) Commercial risk6) Liquidity risk7) Counterparty risk8) Economic risk

1) Currency Risk

An investor or lender is aware of the existence of currency risks in any BOT projects. It occurs due to funding from international banks or foreign companies and thus exposed to the volatility of the exchange rates. Bing et al. [11] stated that fluctuations in currency are considered as an austere problem in international transactions.

The influence of currency risks can be minimized using several measures. Wang et al. [13] argued that, even though foreign firms are capable of hedging currency fluctuations in international money markets, it is somehow ineffective. It was suggested that they should make an agreement with the host government to determine the currency to be employed for payment. It was also proposed that to mitigate currency risks, [12] the promoter is advised to sign a dual-currency contract using local and foreign currencies. Normally in East Asia, soft currency is generated from the operation, whilst the funding was given in hard currency.

One of the preferred methods in reducing currency devaluation is by financing in the local currency. An example is the Shandong power project where an enormous amount of Renminbi tranche (equivalent to USD822 million) was funded by China Construction and Shandong International Trust and Investment Corporation, the first ever twinning with a large US dollar tranche [13]. During the negotiation stage between a host government and a concessionaire/promoter, a proper agreement should be achieved in order to mitigate currency risks – the host government should provide guarantee for currency mismatches to avoid any defect in project implementations and operations.

2) Interest Rate Risk

In contrast, interest rates will affect the project in terms of borrowing and debt payments. Any fluctuations in interest rates will affect the lenders. An appropriate interest rate should be agreed upon for the project. The lenders would have to pay extra costs if the interest rates are higher or would have benefited if the interest rates are lower.

More foreign investors or those in the private sector could be attracted by providing an interest rate guarantee by the host government in a BOT project. This approach was adopted in an Indonesian BOT toll road concession where the government guaranteed on a maximum interest rate, minimum revenue guarantee, debt guarantee, tariff guarantee and minimum tariff guarantee [14].

3) Equity Risk

Performance of the concessionaire is crucial in seeking funds to implement a BOT project. Usually, equity risk is related to the performance of the company which is measured by the share price of the company. A higher price of the share will benefit the shareholder but a lower price will affect the prestige of the concessionaire. The capability of a company in raising capital for a BOT project is reflected on the share price.

It was believed that, equity and other long term investors will agree to provide the amount of funding for BOT project if the promoter has proven their financial capability of the project over its entire lifespan [15]. It was said that to attract domestic capital of debt and equity was difficult especially in East Asia when huge amounts of investments were involved in infrastructural projects. Nevertheless, the competence in carrying out detailed and comprehensive feasibility studies, and economic and risk assessment studies, would emplace the promoter in a better position to obtaining domestic equity finance for the funding a BOT project.



4) Foreign Exchange Risk

Fluctuations in foreign exchange are considered another major risk which might affect a BOT project during construction and operations. Foreign companies who are interested in investing in another country should be aware of the opportunities and threats associated with international currency transactions before proceeding.

The Malaysian government has managed to reduce foreign exchange risks by providing guarantees. The guarantees were to absorb the shortfall when (1) the adverse exchange rate movements exceed 15 percent on its offshore debt (2) adverse interest rate movements exceed 20 percent on its floating rate offshore debt [2,7]. This approach was adopted in the North-South Highway project which benefited the promoter.

5) Commercial Risk

Commercial risk is described as a risk that can jeopardize the financial performance to the project. In spite of that, commercial risk in a BOT project is characterized differently by Merna and Njiru [2] who classified it into three categories: risks related to the completion, risks during operations and risks related to input or output of a project.

A supply and off-take agreement between the supplier and the government is crucial in mitigating the risk. In an agreement, the related parties will agree upon the required amount of input, for instance, coal in a power plant project, and the output generation, which is the electricity. This will allow the supplier to stock the materials upfront at a lower price and the promoter would be able to generate the required output within the stipulated cost, without burden to the public by increasing tariffs. Apart from that, it will secure long term revenues for the promoter by selling the output to the client.

6) Liquidity Risk

In most BOT projects, revenues are generated from operations. To ensure the success of a BOT project, it should be able to generate sufficient amounts of revenue to settle the debt within the stipulated time frame. An amount of profit that can be generated from the operating facility is determined by conducting analyses on the projected revenue during the operational phase. Failure to generate the required revenue will result in liquidity risks.

7) Counterparty Risk

Inadequate support in terms of finance from the lender at specific time is defined as counterparty risk. It can be interpreted as a credit risk. According to Lam and Chow [16], credit risk is the risk when the counterparty (partner of the joint venture) to any financial transaction is not able to fulfill its commitment on the due date. The debt capacity of a promoter reduces when credit risk arises. In a concession contract, transactions between two or more parties contain a risk that one party will default on an obligation of the commitment. Failure in financing the required cash flow for a BOT project is the most common issue that arises.

8) Economic Risk

This risk is primarily related to the facility’s operations which consists of materials supply, labour supply, equipment availability, inflation, tariffs, fiscal policies and exchange rates [17]. Project cash flow is affected by any financial aspects that relate to the economic parameters. Increments in the cost of supply and maintenance, eventually will increase operational costs, which thus reduces revenue. This could be seen as a threat to the promoter.



B. Political Risk

1) Sovereign Risk

Sovereign risk is a risk related to the provision of loans to a foreign government, commonly used in the banking world [3]. The risk is governed by the political environment of a country where the investment will take place, specifically, the location of a BOT project. Sovereign risk occurs when the political environment is unstable and will affect the investor or promoter of the project.

In East Asia, some of the BOT projects faced difficulties due to political instability such as in Thailand where there were frequent changes in political leaders. Apart from that, countries which are governed by different ideologies such as Libya and Saudi Arabia are also facing sovereign risks. A BOT project might face serious risks when there are changes in government policies and regulations as a result of changes of a ruling government as can be learned from past experiences.

Changes in bureaucracy due to reshuffling could also impact on the decision making process in a concession contract. Some kind of guarantee by a host government could avert the risk. In addition to that, the concession contract agreement should be based on the international order systems to safeguard the promoter.

2) Country Risk

Country risk is totally different from sovereign risk. It is related to the overall investment climate in a specific country. The aspects that contribute to country risk are socio-economic conditions, internal or external conflicts inflicting the country, corruption, ethnic tension, policies and legislatures. Before any BOT project implementation, the promoter should conduct a thorough country risk profile and budgetary practices, by appointing a reliable third party (reputable management consultant or a good political analyst or both) to minimize the risk. Decisions could be made based on the study and

they could also seek assistance by referring to the World Bank or the Asian Development Bank.

Nowadays in East Asia, consortiums are established to undertake most of the BOT projects. These consortiums consist of Engineering Procurement and Construction (EPC) contractors, and Operation and Maintenance (O&M) contractors who will be responsible from the time of the feasibility study till implementation of the whole project and its operations for a stipulated period of time. Every foreign investment is subjected to country risk due to unstable government and its component, and inadequate foreign reserves.

c. Technical Risk

Technical risk could be classified into construction risk and O&M risk. Essentially, technical risk is the most common and well understood form of risk. Technical risk is the subject of close surveillance. To minimize the technical risk, the concessionaire is responsible to evaluate the risks in detail: to ensure that the project will be constructed in accordance to the design specifications of a host government’s requirements, and that the project will function well. Thus, reputable and established consultants together with an experienced contractor should be hired to implement a BOT project without any tolerance to the standard codes and practices.

1) Construction Risk

Unknown ground conditions, delay in procuring of construction materials, and price escalation of raw materials for construction such as an increase in the price of steel, copper or aluminum are the problems related to construction risks which occur during the construction phases. In addition to that, poor design reports, extending construction schedules and changing factors of production also contribute to construction risk.

The North-South Highway project in Malaysia is an example of a promoter who had to bear the increments in project costs. The initial estimated cost



was USD1.2 billion but due to hassles encountered during the construction phase such as land acquisition problems and poor road designs, the cost escalated to USD1.8 million [2,7,18]. The increase was almost fifty percent higher than the estimated cost. This scenario was not unusual in East Asia since the local contractors were still learning and relying on the expertise of foreign based contractors which was costly in terms of consulting fees. It is essential that the design reports be made available and be vetted by the owner and consultant before a BOT project commences. Preferably, the appointment of an independent third party to audit and comment on the design and construction methodology would help minimize the construction risk.

2) Operational and Maintenance Risk

During this phase there are several associated risks. One of them is when the performance of the facility is not up to the required level due to technical problems. Selection of inefficient machineries and equipment during the implementation phase and poor workmanship during the installation phase could result in poor performance. Spare parts for machinery and equipment for the facility are to be are easily available at affordable costs. Throughout the concession period, machinery and equipment should undergo some routine service due to wear and tear, so as to optimize performance. Newly available technology should be incorporated to ease the operations phase. Sometimes, initial costs could be high but in the long run it might benefit the consortium.

The operations and maintenance team requires specialized technical skills and abilities in operating the facility. An inefficient team would lead to unnecessarily high costs of operations which might bring less revenue to the consortium. It is very important that a proper agreement be established to ensure that the interest of the operator is secured. The efficiency of the facility’s operation could be increased by providing a maintenance manual, updated on a regular basis together with standard operating procedures.

D. Other Risks

Market risk is based on the demand on the facility upon the completion the project. Thus, the promoter has to carry out extensive market research before embarking on a project. The revenues are generated upon completion of the project and very much dependent on the end user. A poor feasibility study could cause the BOT project to fail if there is no demand or if it failed to attract consumers.

Significant risk occurs to the promoter when there is inadequacy in the concession’s contract. This problem occurs during the tendering stage whereby some promoters simply tender in without properly understanding the major terms and conditions of the project as stipulated by the host government. The standard terms should be adopted during the formulation of the construction contract, operations and maintenance contract and other ancillary documents, and agreed upon by both parties. Therefore it is always recommended to invite renowned tenderers to bid or negotiate for a project.

An exit clause is very crucial in a BOT project. Disputes might occur in a later stage due to unforeseen clauses of the project if the exit clause was not done properly and variations according to time or economic condition is not provided in the contract. Therefore, conflicts could be reduced by formulating a concession contract with reference to a similar and proven contract. Good legal advisors who are familiar with the industry should be employed by concessionor and concessionee to check the contracts thoroughtly before accepting it. The contract should benefit both parties without burden to the consumers.

Personnel in key management of the concession are crucial and consist of a range of expertise. An ideal concessionaire organization must hire highly skilled and competent management personnel. They will be the backbone of the consortium as their involvement will not be limited to project initiation, negotiation, implementation and project management or during transfer to the principal.



They would have to be familiar with the BOT process and with all contractual terms of the concession and actual project management as well. These personnel must be sustained to ensure that the management team remains efficient and productive. Poor working conditions and benefits could lead to their leaving the organization. In East Asia, few promoters provided benefits to their key management personnel, for example, employee share ownership scheme and profit sharing which would enhance the productivity and level of performance of the team and sustain their employment.

PROJEcT RISK MANAGEMENT IN BOT

This study found that risk management is very important in mitigating risks in a BOT project. It always has been the main objective of lenders and investors to ensure a good return on their investments. Thus, the success of a project’s implementation and operations is measured by the revenue generated or returns on investment. Continuing risks evaluation through the life span of the whole project is essential to manage and operate the stakeholders’ assets. Chances of a project’s success could be maximized by conducting risk management. Basically risk management was never intended to eliminate the risks completely but to identify and foresee the risks to control it at a manageable level and proper mitigation measurements adopted to maximize revenue of the project.

Risk Management Technique

The process towards the success of a BOT project starts with identifying and appropriating allocation of risks to the parties that have the greatest control over those risks. A number of studies revealed many suggestions on classifying all risks that could be put into a concession contract [11,19,20]. From the studies, it was found that different types of risks exist in a concession contract. The most meaningful aspect is to tackle and allocate the risks effectively to assure smooth project implementation.

Previous studies indicated that lenders or investors were exposed to higher risks for a BOT project due to the high front end development costs, lengthy negotiation processes and multiparty involvements [9]. Similarly, the promoter would also face equal significant weight risks over the concession period. This has not being shown in the study and in spite of that how those risks are being mitigated.

An allocation provided in the concession contract would require the promoter to allocate risks to the party who would be best able to manage it. Four basic steps were proposed to systematically manage the risks [6]:

1) Identify the risk sources2) Quantify their effect (risk analysis)3) Develop management responses to risks4) Provide for residual risk in the project estimates

Overall risk reduction would depend on the capability of the concessionaire to identify and formulate mitigation measures in risk management. This capability will generate confidence among the lenders and government to the concessionaire. Tiong [21] argued that a promoter would gain high profits when a host government is reluctant to provide any financial guarantee. The promoter would have realized that they needed to manage the risk by taking all mitigating measures to prevent any impact to the project. In the cases where the profits of a BOT project were controlled by the host government, there were BOT projects which made good returns to the stakeholders. This was not limited by the entrepreneurial capability of the concessionaire but included their ability to install early warning systems through mitigating measures in the system.

The success of a BOT project depends on the concessionaire’s strategies. The adopted strategies could be the use of an advanced well proven technology and contracting out the responsibilities, which enabled them to understand the nature of the risks and how to response to it in an effective manner. In addition to that, the success of an employed risk management system which could reduce the risks



to an acceptable level to all stakeholders is deniable. Key principles of a successful risk management are [22]:

1) Clearly identified and visible senior management support for the project

2) Explicit policies which are clearly communicated to all

3) The adoption of a transparent and repeatable framework of activities

4) The existence of a culture that supports and understands the concepts of controlling risk

5) Fully embedded management process which are consistently and rigorously applied and are clearly linked to the achievement of the objectives

6) Implementation of effective plans and regular reviews to ensure the benefits of the processes are realized and lessons are learned for future projects

A well structured risk management framework within the promoter’s organization is fundamentally important to support the entire process of risk management during the project initiation and implementation. An employed comprehensive risk management system in the concessionaire will be the guideline for any decision making to all initiated action plans during the life cycle development of the project.

cONcLuSION

Every BOT project is subjected to multiple risks. Thus it is the responsibility and liability of the promoter to mitigate the risks to ensure the success of a BOT project by recruiting a strong management team. The promoter requires support and cooperation from the host government and investors to assist them in accomplishing the project. The host government should play a more active role by providing guarantees for a BOT project that will benefit the public. It is always recommended that the lenders and investors conduct feasibility studies before financing a BOT project, similarly to the promoter before embarking on a project. It can be

concluded that a comprehensive risk management study for any BOT projects should be conducted and the mitigation plans must be followed strictly in ensuring the success of the project.

AcKNOWLEDGMENT

The authors would like to express their appreciation to Professor Dr. Andrew Gale who advised, assisted, guided and provided expert feedback in preparing the materials, and to Universiti Teknologi PETRONAS for the facilities in preparing the paper.

REFERENcES

[1] World Bank, (2008) PPI Database. Available from: http://ppi.worldbank.org/explore/ppi_exploreRegion.aspx?regionID=2 (accessed on 17 January 2010)

[2] T. Merna and C. Njiru, Financing Infrastructure Projects. Thomas Telford, London, 2002.

[3] T. Merna and F. Al-Thani, Corporate risk management: An organizational perspective. Wiley, London, 2005.

[4] J. Raftery, Risk analysis in project management. E&FN Spon, London, 1994, p. 4.

[5] C.B. Chapmen, Risk in investment, procurement and performance in construction. E&FN Spon, Chapman & Hall, London, 1991.

[6] N.J. Smith, Appraisal, risk and uncertainty. Thomas Telford, London, 2003.

[7] C. Walker and A.J. Smith, Privatized infrastructures: the BOT approach. Thomas Telford, London, 1996.

[8] J. Delmon, BOO/BOT projects: A commercial and contractual guide. Sweet & Maxwell Limited, London, 2000.

[9] T.M. Zayed, and L.M. Chang, “Prototype Model for build-operate-transfer risk assessment.” Journal of Management Engineering, vol. 18 (1), p. 7-16, 2002.

[10] T. Merna, “Financial risk in the procurement of capital and infrastructure projects.” International Journal of Project and Business Risk Management, vol. 2(3), p. 256-270, 1998.

[11] L. Bing, R.L.K. Tiong, W.W. Fan and D.A.S. Chew, “Risk management in international construction joint ventures.” Journal of Construction Engineering and Management, vol. 125(2), p. 277-284, 1999.

[12] S.Q. Wang, R.L.K. Tiong, S.K. Ting and D. Ashley, (2000), “Foreign exchange and revenue risk: analysis of key contract clauses in China’s BOT project.” Construction Management and Economics, vol. 18(3), p. 311-320, 2000.



[13] S.Q. Wang, R.L.K. Tiong, S.K. Ting and D. Ashley, “Evaluation and management of foreign exchange and revenue risks in China’s BOT project.” Construction Management and Economics, vol. 18(2), p. 197-207, 2000.

[14] A. Wibowo, A., “Valuing guarantees in a BOT project.” Engineering, Construction and Architectural Management, vol. 11(6), p. 395-403, 2004.

[15] S.M. Levy, Build operate transfer: paving the way for tomorrow’s infrastructure, John Wiley & Sins Inc. Canada, 1996.

[16] K.C. Lam and W.S. Chow, “The significance of Financial risks in BOT procurement Building.” Building Research and Information, vol. 27(2), p. 84-94, 1999.

[17] Y.Y.L. Florence and H. Linda, “Risks faced by Singapore firms when undertaking construction projects in India.” International Journal of Project Management, vol. 24(3), p. 261-270, 2006.

[18] J.E. Schaufelberger and I. Wipadapisut, “Alternate financing strategies for build-operate-transfer projects.” Journal of Construction Engineering and Mangement, vol. 129(2), p. 205-213, 2003.

[19] S. Ye and R.L.K. Tiong, “Effects of concession period design on completion risk management of BOT projects.” Journal of Construction Management and Economics, vol. 21(5), p. 471-482, 2003.

[20] K.T., Yeo and R.L.K Tiong, “Positive management of differences for risk reduction in BOT projects’.” International Journal of Project Management, vol. 18(4), p. 257-265, 2000.

[21] R.L.K. Tiong, “BOT projects: risks and securities.” Construction Management and Economics, vol. 8(3), p. 315-328, 1990.

[22] M.F. Dallas, Value and risk management: A guide to best practice. Blackwell Publishing, Singapore, 2006.

Kalaikumar Vallyutham graduated in 2002 with B.Eng (Hons). in Civil Engineering from Universiti Teknologi Malaysia. He earned his Msc in Structural Engineering from The University of Manchester in 2007. During his early years as graduate, he worked with a contractor and consultant in various projects. He joined UTP in 2005 as tutor. Currently he is a lecturer at Civil

Engineering Department at UTP.

Narayanan Sambu Potty recieved 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 Inda and Universiti Malaysia Sabah. His main research areas are steel and concrete structures, offshore structures and construction management.

PLATFORM i s a biannual , peer-

reviewed journal of Universiti teknologi

PEtRONaS. it serves as a medium for

faculty members, students and industry

professionals to share their knowledge,

views, experiences and discoveries in

their areas of interest and expertise. it

comprises collections of, but not limited

to, papers presented by the academic

staff of the University at various local and

international conferences, conventions

and seminars.

the entries range from opinions and

views on engineering, technology and

social issues to deliberations on the

progress and outcomes of academic

research.

Opinions expressed in this journal need

not necessarily reflect the official views

of the University.

all materials are copyright of Universiti

teknologi PEtRONaS. Reproduction

in whole or in part is not permitted

without the written permission from

the University.

NOTES FOR CONTRIBUTORS

Instructions to Authors

authors of articles that fit the aims, scopes and policies of

this journal are invited to submit soft and hard copies to

the editor. Paper should be written in English. authors are

encouraged to obtain assistance in the writing and editing

of their papers prior to submission. For papers presented or

published elsewhere, authors should include the details of

the conference or seminar.

Manuscript should be prepared in accordance with the

following:

1. the text should be preceded by a short abstract of 50-100

words and four or so keywords.

2. the manuscript must be typed on one side of the paper,

double-spaced throughout with wide margins not

exceeding 3,500 words although exceptions will be made.

3. Figures and tables have to be labelled and should be

included in the text. authors are advised to refer to recent

issues of the journals to obtain the format for references.

4. Footnotes should be kept to a minimum and be as brief

as possible; they must be numbered consecutively.

5. Special care should be given to the preparation of the

drawings for the figures and diagrams. Except for a

reduction in size, they will appear in the final printing in

exactly the same form as submitted by the author.

6. Reference should be indicated by the authors’ last names

and year of publications.

PLATFORM Editor-in-ChiefUniversiti Teknologi PETRONASBandar Seri Iskandar31750 TronohPerak Darul RidzuanMALAYSIA

P L A T F O R M

Volume 8 Number 1 Jan- Jun 2010


Separation Of Carbon Dioxide From CO2/CH4 Binary Mixture Using Silicone Rubber Membrane Farooq Ahmad, Hilmi Mukhtar, Zakaria Man and Binay K Dutta

2


Development Of Polymeric Concrete For Sustainable FuturesM.F. Nuruddin, A. Kusbiantoro and S. Qazi

10

Natural Convection Flow In Vertical Channel Due To Ramped Wall Temperature At One BoundaryM. Narahari and Vijay R Raghavan

17


Mix Design of Foamed Concrete With MIRHA Using Taguchi MethodMuhd Fadhil Nuruddin and Ridho Bayuanji

26


V2S: Voice to Sign Language Translation System for Malaysian Deaf PeopleOi Mean Foong, Tan Jung Low and Wai Wan La

35

Evaluating Pairs Analysis Threshold using Receiver Operating Characteristic (ROC) GraphEmelia Akashah P. A., Anthony T. S. Ho and Savita K. Sugathan

42


Transient Well Performance Modeling for Reservoir Pressure DeterminationHon Vai Yee, Suzalina Zainal and Ismail M. Saaid

48


Integrated Scheduling and RTO of RGP with MPC and PI ControllersNooryusmiza Yusoff and M. Ramasamy

57

A Review of Risks And Mitigation Measures in Build-Operate-Transfer ProjectsKalaikumar Vallyutham, Syed Kamarul Bakri Syed Ahmad Bokharey,

Narayanan Sambu Potty and Nabilah Abu Bakar

66

VOLUME EIGHT NUMBER ONE JANUARY- JUNE 2010 ISSN 1511 …press.utp.edu.my/wp-content/uploads/2018/12/platformv8n1-new.pdf · Dr. Shahrina Mohd. Nordin Subarna Sivapalan Sub-Editor:

Documents