Technology Cluster: TRANSPORTATION
Technology Platform: NOVEL ENGINE DEVELOPMENT
Modeling, design, modification and testing of
length is x . The piston on a left to right stroke traverses
EMF linear generator of a new free-piston gas
from –x to +x . The friction force required to move the
generator. (Project Leader: Prof. Khalid Md. Nor.)
slider is F . The quantity of heat added to one cylinder
during a cycle is Q . The force balance for the left-to-
Figure 1 shows the scope of each project, breakdown
right (expansion) stroke can be written as in Equation
of the activities along with the approximate duration
(1) with x positive in the left-to-right direction:
for each phase of the project. The Program is funded
by the Ministry of Science, Technology and
Environment (MOSTE) under the Intensified Research
In Priority Areas Programme (IRPA) Grant No. 03-99-
where P and P are the instantaneous pressure in the
02-0000-PR0025/4-00. Universiti Teknologi PETRONAS
combustion and kickback cylinder, A and A are the
acts as the Programme Leader for the project.
areas of the piston in the combustion and kickback
chamber, respectively, F is the frictional force, and m
RESULTS AND DISCUSSION
is the mass of the translator.
Results of numerical studies of piston velocity in a
A numerical experiment about the proposed
linear engine incorporating combustion and kickback
variational algorithm has been done using the
are presented. Initial study of equal-bore cylinders
following values of parameters: P
shows a larger piston velocity after combustion due
100,000 Pa, Q = 196 J, x = 0.025 m, b = 0.05 m, F =
to the lower air pressure at the kickback end of the
900 N, n = 1.33. The velocity and the position had
engine. By using variational principle, computations
been found by using a numerical integration applied
are performed to determine a velocity function as long
during expansion and compression strokes,
as the piston head does not exceed the maximum
respectively. Using the procedure, the initial bore scale
position. This function is an important factor to avoid
of 1 has been chosen by assuming that the kickback
knocking that reduces the compression ratio and
bore must be larger than the combustion bore. The
engine thermal efficiency.
iteration results are the maximum translator position
at each bore scale as depicted in the left side of Figure
Consider a two-stroke linear engine model as shown
3. From the figure, it can be seen that the maximum
in Figure 2. To derive the governing equation a priori,
the following independent variables are specified. The
combustion and kickback cylinder bore are b and b
respectively. The maximum theoretical half-stroke
Figure 3: Kickback Bore Scale and Maximum Translator Position
Figure 2: Schematic View of the Linear Engine [6]
During Iteration [6]
VOLUME THREE NUMBER ONE JANUARY - JUNE 2003 PLATFORM
Technology Platform: NOVEL ENGINE DEVELOPMENT
Technology Cluster: TRANSPORTATION
Figure 4: Function of Position for a Variation of Bore Scale (left) and Optimum Function of Velocity and Position [6]
position is directly at the x position. The bore scale
Johansen T. A., Egeland O., Johannessen E. A. and Kvamsdal R.,
"Free-Piston Diesel Engine Timing and Control – Towards
tends to 1.4877, as x
tends to x . The plot of
Electronic Cam- and Crankshaft", in IEEE Transactions on Control
translator position versus time for each bore scale and
Systems Technology, vol. 10, no.2, March 2002
the function of optimal velocity and position are
Peter van Blarigan, "Advanced Internal Combustion Electrical
shown in Figure 4.
Generator", Proceedings of the 2002 U.S. DOE Hydrogen ProgramReview, NREL/CP-610-32405
Peter Van Blarigan, "Rapid Combustion Electrical Generator",in Reciprocating Engines Peer Review, Illinois, 2002
The concept of a linear generator with a piston
M. I. Ahmad Kamal Ariffin, "LG Progress Report", UKM CAE
reciprocating freely between a combustion chamber
Group, March 26, 2003.
and an air-kickback chamber is presented. Areas of
research and a breakdown of activities for the ongoing
design and development project of a prototype linear
Abd Rashid Abd Aziz graduated with a B.S.
generator are provided. Results of a numerical
and M.S. in Mechanical Engineering fromthe University of Miami in 1988 and 1990
experiment reveal the optimal piston stroke length
respectively. He continued his PhD with a
and the requirement for unequal combustion and
full scholarship from the university and
kickback bore sizes. These two parameters – bore and
obtained his doctoral degree in 1995. Uponreturning to Malaysia, he lectured shortly at
stroke – are not only important factors for engine
Universiti Putra Malaysia prior to taking up
design, but also critical constraints in the design of the
lectureship position at Universiti Teknologi
rotor and stator for the electrical generator.
PETRONAS (UTP) in 1996. Currently a senior lecturer in theMechanical Engineering Program at UTP, Dr Abd Rashid is leadingthe Linear Generator Program, a multi-institutional R&D project
with funding from the Ministry of Science, Technology andEnvironment (MOSTE).
J. Allen, D. Law and D. Kemp, Patent Filing, 9930380.2, 1999.
Arshad W. M., Bäckström T., Thelin P. and Sadarangani C.,"Integrated Free-Piston Generators: An Overview", inProceedings of the Nordic Workshop on Power and IndustrialElectronics, Stockholm, 2002
PLATFORM VOLUME THREE NUMBER ONE JANUARY - JUNE 2003
Technology Cluster: TRANSPORTATION
Technology Platform: FUEL COMBUSTION
The Effects Of Horizontal Mounting Position Of Atomizer
To Spray Angles In A Confined Tube
Shaharin Anwar Sulaiman
Universiti Teknologi PETRONAS,
Bandar Seri Iskandar, 31750 Tronoh, Perak Darul Ridzuan, Malaysia.
Sprays released from a nozzle mounted in a horizontal tube have many engineering applications. An important
characteristic of a spray is the spray angle, which is formed by two straight lines drawn from the discharge orifice
of the spray nozzle. An experiment was carried out to study the effect of the horizontal mounting position of the
atomizer to the spray angle. To investigate the variations in spray angle, water was sprayed from a nozzle which
was mounted on the centerline of a horizontal clear Perspex tube. The experiment was carried out using four
different discharge pressures for the nozzle, while air flowed inside the tube at a constant velocity of 6 m/s. With
the use of a color video camera, the image of the spray was captured and analyzed prior to the measurement of
spray angle. The experiment was repeated by shifting the spray nozzle horizontal position along the tube
centerline. From the experiment, it was found that the farther downstream the nozzle was mounted with respect
to the tube inlet, the wider the spray angle became. In addition, the increase of spray angle was more significant
for the case of nozzles with higher nominal spray angles.
Keywords: spray nozzle, combustion processes, atomizer
be desirable [2]. Usually a wide-angle hollow cone
spray is preferred in gas turbine combustion. In a
Spray characteristics are determined by a few
different application, Widger [3] found that in the dust
parameters, namely the droplet size, concentration,
extraction in coal mines, the airflow rate would reach
velocity, and spray angle. The spray angle is defined as
a peak if the spray angle was misaligned by 15° from
the angle between the formation of two straight lines
the centerline of the tube axis.
drawn from the discharge orifice of the spray nozzle.
Being an important parameter, spray angle has specific
In this paper, an experiment was conducted on sprays
purposes in various engineering applications. For
confined in a horizontal tube. The spray nozzle was
example, in gas turbine combustion, the spray angle
mounted at the upstream of the tube, on its centerline.
has a strong influence on ignition, stability limits and
Air in the tube flowed at a constant velocity as a result
exhaust smoke. A reduction in spray-cone angle, for
of suction effect generated by the fan installed at the
instance, is beneficial to ignition, especially at high
downstream of the system. The main interest in this
altitude [1]. In another instance, if the gas turbine
experiment was to observe the difference in the spray
combustion chamber length were to be minimized,
angle when the position of the nozzle was varied along
spray angles of the range between 60° and 90° would
the centerline of the confinement tube. The tube used
This paper was presented at the 2nd World Engineering Congress, Kuching, Sarawak, Malaysia, July 2002.
VOLUME THREE NUMBER ONE JANUARY - JUNE 2003 PLATFORM
Technology Platform: FUEL COMBUSTION
Technology Cluster: TRANSPORTATION
was of a clear material therefore enabling the inside
mixed with the spray droplets flowed further
image of the spray to be captured for the measurement
downstream into the rear chamber, followed by the
of spray angle. This was done using a video camera
cyclone for the removal of moisture from the air prior
and a frame grabber interface to produce two-
to discharge into the atmosphere.
dimensional image that could be analyzed in a
computer. The measured spray angles were then
The spray was released by the waterside system
analyzed and compared.
through the atomizer nozzle, which was mounted in
the confinement tube as shown in Figure 2. Water from
the tank was pumped into the atomizer nozzle. The
pump, which was driven by a three-phase motor, had
The experiment was carried out using a constructed
a maximum pressure capacity of 22.9 MPa and a
test rig as shown in Figure 1. The test rig comprised of
maximum flow capacity of 13 litres per minute. The
a ‘waterside system' and an ‘air-side system' to allow
pressure during the experiment was set with the use
the production and flow of spray in a clear confined
of control valve in the piping system.
tube. Both systems were equipped with discharge
facilities to maintain proper experimental and working
The pump outlet and the atomizer were connected by
conditions. All apparatus were calibrated in the proper
a flexible hose of 13-mm diameter. In between the two,
manner to minimize errors.
a steel manifold and a surge tank were installed at a
distance of 40 times the pipe diameter from the pump
The apparatus for the air system comprised of a
outlet to prevent fluctuations of pressure in the water
confinement tube, two air chambers, an air inlet, and a
supply. The steel manifold had a water inlet and outlet,
cyclone separator. These components were air tight
as well as another two outlets each connected to the
so that the air inside the tube was isolated from the
surge tank and a pressure gauge. The surge tank was
ambient air except at the inlet. Air driven by the
only partly filled up with water during the experiment,
suction fan, connected by an air duct at a different
to leave a void to damp out any fluctuations.
room, flowed from the air inlet tube to the front
chamber. The average air velocity inside the tube was
In this experiment hollow cone spray atomizers were
6 m/s. In the confinement tube, which had a mean
selected for use. Table 1 shows the configuration of
internal diameter of 192-mm and a length of 1000-mm,
the spray selection. The selected nominal spray angles
air mixed with the spray droplets released by the
were 40°, 60° and 80°, rated at volume flow rates of 0.2,
atomizer nozzle. The confinement tube was made of
0.4 and 0.6 liters per minute respectively, and discharge
clear Perspex to enable video imaging. The air that had
pressure of 1 MPa, thus resulting in the use of nine
Figure 1: The experimental rig
Figure 2: The schematic layout of the waterside system.
PLATFORM VOLUME THREE NUMBER ONE JANUARY - JUNE 2003
Technology Cluster: TRANSPORTATION
Technology Platform: FUEL COMBUSTION
Table 1: Configuration of the selected spray atomizers.
Q (L/min)
Nominal Spray Angle (catalogue)
Figure 4. The orientation of camera installation.
atomizers. For identification, atomizers were labeled as,
The spray angle was obtained by direct measurement
for example 0.4-80, meaning rated at 0.4 L/min flow
on the spray images captured using the JVC model TK-
and 80° nominal spray angle, or the Lechler 214.245.30
C600E color video camera with RS Components Ltd.
atomizer, if referred to Table 1.
625-772 zoom lens of 18-108/2.5. The image of a steady
spray was recorded for ten seconds in videotape using
The atomizer holder was fabricated such that the
the 4-head double density Computar CTR-3024 video
nozzle could be shifted horizontally along the
recorder. To get an average value, three frames of spray
centerline of the confinement tube. In addition, proper
images were transferred into the computer in the form
positioning and alignment with the tube axis were
of JPEG (Joint Photographic Experts Group) files with
made by adjustment at the holder. Three mounting
the use of frame grabber interface and the Image Folio
positions were selected in this experiment, as indicated
v.3.0 and the Video Work v.1.5 software.
in Figure 3.
The spray pictures were then copied into AutoCAD,
The video camera was mounted on a tripod at a fixed
where lines were carefully drawn between a point
position thus allowing the camera to capture a vertical
adjacent to the atomizer outlet, and the farthest point
area coverage of 75-mm wide by 55-mm high at the
on the spray boundary where the edge could be clearly
center plane of the tube, which was satisfactory for the
defined, as shown in Figure 5. The spray boundaries
measurement of spray angles resulted from all the nine
were drawn carefully to ensure satisfactory
nozzles. Figure 4 shows the orientation for the camera
representations of the free edges that had a sharply
P1 = 65 mmP2 = 127 mmP3 = 176 mm
Figure 5. A sample measurement of spray angle by using
AutoCAD.
Figure 3. The nozzle horizontal position.
VOLUME THREE NUMBER ONE JANUARY - JUNE 2003 PLATFORM
Technology Platform: FUEL COMBUSTION
Technology Cluster: TRANSPORTATION
defined interface, which separated fluid with vorticity
occur in the case of 0.2-80, as shown in Figure 8, where
and fluid without vorticity. Having the two boundary
at the downstream most position and at a small range
lines drawn, the spray angle was measured by selecting
of discharge pressures, the actual spray angle is very
the AutoCAD command for angular dimensioning.
low. The reason could be that the sprays were not fully
developed at those pressures.
Another interesting observation is that the gradients
From this experiment a set of 108 data were obtained
of the curves in Figures 6, 7 and 8 become relatively
from the use of nine atomizers at three horizontal-
positions and at four pressure levels. The four discharge
pressures were 0.5, 1.0, 1.5 and 2.0 MPa. Each data set
was obtained by averaging three angle measurements
taken from three images captured by the video camera.
The data, which were grouped by nominal angles, were
then plotted on graphs of angle (measured) versus
pressure. Figures 6, 7 and 8 show the relationship
between the atomizer pressure and the resulting spray
angles, as well as the horizontal mounting position.
Since each chart is categorized by the same nominal
spray angle, nine curves are displayed together in every
Figure 6.
Spray angles at various positions and pressure.
Nominal angle of 40°.
From the results shown in Figures 6, 7 and 8, it is clear
that there are differences in the spray angles as the
horizontal mounting position is varied. The changes
in the spray angles are however small. It was also found
that for most cases the farther downstream the nozzle
is mounted, the wider the resulted actual spray
becomes. However, exceptions occur in the cases of
atomizers 0.4-80 and 0.6-80, where the spray angles
become narrower when operating at higher discharge
Figure 7.
Spray angles at various positions and pressure.
From the graphs it is also obvious that the changes of
Nominal angle of 60°.
the atomizer horizontal-positions cause a relatively
smaller variation in the measured spray angle for the
cases of low nominal spray angle. This is clearly shown
as a narrow band of curves in Figure 6. For example, in
the case of nominal spray angle of 40°, the variationsof spray angle are only around ± 2.5°. Nevertheless thevariations are relatively larger for atomizers with higher
nominal spray angle, where the measured spray angles
could have a variance of ±10°. This shows that the effectof the atomizer horizontal-position is less significant
for the case of narrow nominal spray angles, but is more
Figure 8. Spray angles at various positions and pressure.
obvious for wide-angle nominal sprays. Exceptions
Nominal angle of 80°.
PLATFORM VOLUME THREE NUMBER ONE JANUARY - JUNE 2003
Technology Cluster: TRANSPORTATION
Technology Platform: FUEL COMBUSTION
steeper as the nozzle is moved farther downstream,
particularly for atomizers 0.4-80 and 0.6-80. In each
curve, the gradient is an indication of the influence of
Variations in the horizontal mounting position of the
discharge pressure to the spray angle. This means that
atomizer along the tube centerline caused little
the farther downstream these atomizers are
changes in the spray angles. The changes were
positioned, the more sensitive their spray angles would
relatively smaller when atomizers of small nominal
be to the change in the liquid discharge pressure.
spray angle were used. On the other hand, the changes
were larger in the cases of atomizers of high nominal
Even though the spray angles at the downstream most
spray angle, or also known as the wide-angle spray. For
positions are found to be the highest, it was found that
most atomizers, the actual spray angles became wider
for atomizers 0.6-40 and 0.2-60, the lowest spray angles
when they were mounted farther downstream inside
did not appear at the upstream position, but rather at
the confinement tube. However a few exceptions
the middle position.
applied for some of the atomizers, where the possible
reasons for the occurrences have been discussed.
Another significant observation was that the spray
angles were more sensitive to the change in the liquid
From the experimental results, there have been some
discharge pressure when installed farther downstream.
inconsistencies where errors are predicted. A few
possible sources of errors are discussed. First, during
More work needs to be done under this area so that a
the experiment, the imaging was done only from one
more reliable and a wider range of information could
angle, which was from the side. Consideration should
be obtained. The work should involve an increase in
also be given for the angles measured from images
the number of variables for reliability purposes.
captured from the top and another angle, so that the
values could be averaged out.
Lefebvre, A.H., Gas Turbine Combustion, Hemisphere
It was also realized that during the experiment, the 100-
Publication, New York. (1983)
W incandescent light was installed in such a manner
that it was slightly offset from the centerline of the
Chung, I.P., DunnRankin, D., Ganji, A., "Characterization of aSpray from an Ultrasonically Modulated Nozzle," Atomization
camera lens. As a result, the top portion of the spray
and Sprays, Vol.7, No.3, pp.295-315. (1997)
image had a slightly different brightness and contrast.
Widger, I.R., "Improvement of High Pressure Water Sprays Used
This could lead to an error in distinguishing the spray
for Coal Dust Extraction in Mine safety," PhD Thesis, Mechanical
boundary. To overcome this problem, a brighter light
Engineering Department, UMIST, UK. (1993)
bulb, a 1000-W light for example, should be used and
installed farther from the lens, so that a more balanced
Shaharin Anwar Sulaiman graduated with
image contrast and brightness could be obtained.
a B.S. in Mechanical Engineering from IowaState University, USA in 1993. He earned hisM.Sc. in Thermal Power and Fluids
More works are recommended in the future to obtain
Engineering from the University of
comprehensive results in this study. For example, the
Manchester Institute of Science andTechnology (UMIST), UK in 2000. He worked
number of horizontal positions should be added so
as a Mechanical & Electrical (M&E) Project
that results will be more reliable. In addition, the
Engineer in YTL Construction for five years
distance between the atomizer and the confinement
until 1999. Currently he is a lecturer in the Mechanical Engineeringprogram at UTP. He is also a certified Professional Engineer and a
tube inlet should be extended so that the effects at
Corporate member of the Institution of Engineers Malaysia (IEM).
farther downstream could be observed. Furthermore
it is predicted that the airflow inside the tube is a
developing flow thus at farther downstream a more
consistent velocity profile is expected.
VOLUME THREE NUMBER ONE JANUARY - JUNE 2003 PLATFORM
Technology Platform: SYSTEM OPTIMIZATION
Technology Cluster: OIL AND GAS
Industrial Stream Composition Forecasting
Using Neural Network
A.H.S Maulud1, I. Ismail1, R.D. Menez2 and M.A. Shafawi2
1Universiti Teknologi PETRONAS,
Bandar Seri Iskandar, 31750 Tronoh, Perak Darul Ridzuan, Malaysia.
2MTBE (M) Sdn. Bhd., P.O. Box 1, 26080 Balok, Pahang Darul Makmur, Malaysia.
Artificial neural network has been reported as a successful tool to model continuous systems. Its application in
industrial scale has been reported but not widely. The success of its application is strongly influenced by the
understanding of the understudied process. Since the determination of network structure and neuron training
is considered well established with appropriate cross validation, the major task is to determine the appropriate
inputs without complicating the network model. In most modelling studies, the model has been shown to be a
successful tracker by evaluating the sum of squared error or mean of squared error. However, there is no standard
performance criterion for acceptance in industrial application. This paper will discuss some important aspects
on prediction of component concentration by using neural network method in one important stream in a
petrochemical complex. The result will be compared to a linear regression model.
Keywords: artificial neural network, process control, industrial modelling studies
produces side products which are either to be
consumed by the plant itself or sold for additional
The petrochemical industry is a very important sector
revenue. The operation of petrochemical plant is quite
in which a commodity petroleum product is converted
sensitive since it involves complex reactions and very
to higher value materials. The petrochemical complex
tight product specification. Thus, changes in certain
generally consists of a reactor section where a specific
stream composition may affect the overall plant
reaction is taking place. This is then followed by the
performance. Thus, close monitoring of certain stream
separation train to separate the product and unreacted
composition is very crucial. Normally samples have to
feed. The unreacted feed will be recycled to improve
be taken at certain period and sent for laboratory
the overall conversion.
analysis for this purpose. Since the complex volume is
very large, it is expected that it is adequate to monitor
In a typical petrochemical reactor, the reaction takes
these stream composition at certain interval period as
place under the presence of high technology catalyst
the parameter changes are not that frequent. However,
to improve the conversion and selectivity. Beside the
any upset or parameter change between the sample
main reaction, there will be a few side reactions.
cannot be properly monitored. Installation of on-line
Sometimes, the production of the main product also
analyser may incur a high investment. In addition, it
This paper was presented at the International Conference on Advances in Petrochemicals and Polymers in the New Millennium,
Bangkok, Thailand, 22-25 July 2003.
PLATFORM VOLUME THREE NUMBER ONE JANUARY - JUNE 2003
Technology Cluster: OIL AND GAS
Technology Platform: SYSTEM OPTIMIZATION
requires a good maintenance for a continuous reliable
of the product will be recycled and combined in the
result. Thus, alternatively an inferential analyser can be
feed. The ratio of total feed A and B to product E in the
used for this purpose.
feed must be maintained at specific value for optimum
operation requirement. This ratio is greatly influenced
For this case study, the reaction involves complex
by the purity of product E in the recycle stream. It is
reaction mechanisms, which may lead to some non-
necessary to predict the product E purity in the recycle
linearity characteristic. In view of this situation, the
stream. Therefore, a real time monitoring of plant
neural network approach could be utilised to provide
performance should be carried out.
a good prediction since neural network is able to
predict any continuous non-linear systems [2][3].
3.0 NEURAL NETWORK OVERVIEW
Application of neural network to predict certain
parameters in industrial scale have been reported for
A neural network concept has been widely discussed
example by Montague [4] and Barry [5].
in the literature [3, 4]. Since this paper is intended for
neural network application in industry, only a brief
2.0 PROCESS DESCRIPTION
description of neural network structure will be
In this paper, it is not possible to describe the process
in detail due to commercial confidentiality. Instead, the
A typical neural network model composes of
salient features will be illustrated for overall process
interconnected neuron layers (an input layer, one or
understanding. The process consists of reactor and
more hidden layers, and an output layer) between the
separation sections as shown in Figure 2.1.
input and output variables. A typical neural network
structure is shown in Figure 3.1.
There are two feeds (Feed A and Feed B) which
simultaneously react in the reactors to produce
The neuron or node is simply a processing unit. In most
Product C and Product D as shown in the reaction
cases, input neurons have no processing function. Most
information processing is done by hidden neurons.
Each hidden neuron will pass the weighted input
Feed A → Product C + Product E
through a summation function and another specified
Feed B → Product D + Product E
activation function. The most commonly activation
function used in hidden neurons is the sigmoidal
transfer function. The output of sigmoidal function
Product E is a side product. Unreacted Feed A and B
which lies in the range of 0 to 1 is given by:
will be recycled for overall conversion improvement.
Product E will be separated. Some of the products will
be consumed by certain unit in the plant while most
Figure 2.1: Process diagram schematic
VOLUME THREE NUMBER ONE JANUARY - JUNE 2003 PLATFORM
Technology Platform: SYSTEM OPTIMIZATION
Technology Cluster: OIL AND GAS
Figure 3.1: A typical neural network structure
by using Levenberg-Marquadt algorithm with mean
square error as optimisation performance function.
The ‘early stopping' generalization criteria will be
applied to improve the training generalisation.
There are a total of 653 daily average data dating from
Figure 3.2: Simplified neuron function
January 2000 to March 2002. The first 430 data will be
used as modelling data while the remaining 223 data
will used as testing data to measure the prediction
Figure 3.2 shows a simplified hidden neuron structure.
robustness on future data. Approximately 85% of
For output layer, the neuron has only summation
modelling data will be randomly selected as training
function in most cases. Normally the hidden and
data while the remaining will be used for cross
output neurons will have a bias input to improve the
validation for the ‘early stopping' training criteria.
An important aspect of industrial modelling is that it
4.0 MODELLING BASIS AND PERFORMANCE
should be able to fit the existing data and to predict
the future values within a specific limit. In this case,
the testing data will be treated as future data to
4.1 Modelling basis
measure the robustness of the model.
In this study, a multi-layer feed forward network
configuration programmed in Matlab 6.0™ software
4.2 Performance criteria
will be used. The network architecture is limited to a
In the process, a typical product E concentration in the
single hidden layer with non-linear function. The
recycle stream is 90%. For performance measurement,
neuron numbers will be adjusted to give the
the absolute error of product E concentration between
appropriate optimisation. Training will be conducted
the predicted and actual values (residue) will be used
PLATFORM VOLUME THREE NUMBER ONE JANUARY - JUNE 2003
Technology Cluster: OIL AND GAS
Technology Platform: SYSTEM OPTIMIZATION
since the problem statement is to predict the
parameters. Thus, we will not complicate the model
instantaneous value. Performance criterion such as
with unnecessary inputs and waste a lot of
sum squared error is more appropriate for cumulative
computational time. In order to avoid neuron
or overall performance. Since we are dealing with a
saturation, all input parameters are scaled to [-1, 1]
real industrial problem, it is expected that there are
many factors that may affect the process. Thus,
appropriate practical targets must be set. Based on
Three sets of different number of neurons in hidden
practical experience, 95% of prediction error value
layer will be evaluated based on the above input
should be less than 1% concentration unit.
parameters. The sets are 10, 15 and 20 neurons in
hidden layer.
In addition, the neural network prediction will be
compared with the prediction using the linear
6.0 RESULT AND DISCUSSION
6.1 Number of neurons
5.0 PARAMETERS SELECTION
From the initial result, it shows that a sufficiently
trained model has been achieved with less than 10
This is a multiple input and single output (MISO)
epochs . Figure 6.1 shows the mean squared error
problem. The output is product E concentration in a
(MSE) for training and cross-validation data with
recycle stream. For inputs, based on process
respect to the number of epochs.
knowledge and experience, there are 13 parameters
that may affect the product E concentration as follows:
Table 6.1 shows the training and validation error for
10, 15 and 20 neurons for hidden layer. Although 20-
iii. Reactor temperature 1
iv. Reactor temperature 2
Reactor temperature 3
vi. Separator pressure 1
vii. Separator pressure 2
viii. Separator pressure 3
xi. Combined feed flow
xii. Recycle product E flow
xiii. Recycle product E temperature
However, only four parameters (total feed flow, feed B
flow, recycle product E flow and recycle product E
Figure 6.1: Training and validation errors for 10 neurons in
hidden layer.
temperature) are selected initially, based on their
strong direct influence on the recycle product E
Table 6.1: MSE for different neurons (10-5)
Although we could include all available input
parameters in the beginning, it is pragmatic to use a
minimum input parameter and add them individually.
This will help us to identify the most significant input
VOLUME THREE NUMBER ONE JANUARY - JUNE 2003 PLATFORM
Technology Platform: SYSTEM OPTIMIZATION
Technology Cluster: OIL AND GAS
neurons has a least training error, the difference is not
concentration in recycle stream for the modelling data.
significant compared to 15-neurons error. In addition,
It is clearly shown that both neural and linear model
the 15-neurons has less error in validation data
predictions are able to track the product E
compared to 20-neurons. Thus, 15-neurons will be
concentration pattern adequately. However, the most
considered as the optimal number of neurons in
important aspect is its capability to meet the criteria
hidden layer for the time being.
as specified in Section 4.2.
From this point, all results and discussion are based
Figures 6.3 and 6.4 show the prediction errors (residue)
on a neural network with 15-neurons in hidden layer.
and absolute residue respectively. Our discussion will
mainly focus on absolute residue.
6.2 Modelling data
Figure 6.2 shows the prediction of neural and linear
From Figure 6.4, generally we can say that both neural
model compared to actual value of product E
and linear models are able to predict the product E
concentration accurately where majority absolute
residue is below than 1% concentration unit (0.01 on
the graph). Visual analysis also shows that neural
model gives a better prediction as compared to the
linear model. This analysis is verified in Table 6.2, which
shows values for 95% percentile of absolute residue
in all cases. The neural model has residues of less than
0.91% compared to 1.34% for the linear model on
modelling data.
Table 6.2: Absolute residue of the 95% percentile
Figure 6.2: Product E concentration
Figure 6.3: Prediction error (residue)
Figure 6.4: Absolute Prediction error
PLATFORM VOLUME THREE NUMBER ONE JANUARY - JUNE 2003
Technology Cluster: OIL AND GAS
Technology Platform: SYSTEM OPTIMIZATION
6.3 Testing data
model, degradation is been observed. This is verified
However, the true performance should be based on
by its 95% percentile of absolute residue, as shown in
test data. Figure 6.5 shows the absolute residue for
Table 6.3. The 95% percentile of absolute residue of
neural and linear models on a test data. From
neural model on the test data is 1.14%, which slightly
observations, the linear model is more robust than the
exceeds the performance criteria as stipulated in
neural model. Table 6.2 shows that the 95% of absolute
section 4.2.
residue of the linear and neural models is less than
1.15% and 2.15% respectively. This clearly shows that
Table 6.3: Absolute residue of the revised model at the 95%
percentile
the linear model is robust while the neural model is
unable to sustain its prediction capability on future
It is expected that some characteristics have not been
captured by the current input parameters. Thus, there
is a need to add more input parameters. Major residue
on testing error occurred on day 80 onward. From
Based on the test data to measure model performance,
input data analysis, it is found that there are significant
it is found that the original linear model and the
fluctuations on separator pressure 1 during this period.
revised neural model are almost equivalent in
From process knowledge, this situation will definitely
prediction performance. To evaluate the overall
affect the product E concentration in recycle gas.
performance, a comparison of the mean absolute
Therefore, separator pressure 1 will be included as
residue of the two models using the test data is
input parameter.
required, as shown in Table 6.4.
Table 6.4: Mean of Absolute residue
6.4 Revised model
Figure 6.6 shows the absolute residual of revised
Mean of Absolute Residue
model on test data.
In comparison with Figure 6.5, which has additional
input parameter, an improvement in prediction by the
neural model is observed. However, for the linear
Figure 6.5: Absolute residual on testing data
Figure 6.6: Absolute residue on testing data (revised NN model)
VOLUME THREE NUMBER ONE JANUARY - JUNE 2003 PLATFORM
Technology Platform: SYSTEM OPTIMIZATION
Technology Cluster: OIL AND GAS
From Table 6.4, both models have almost equivalent
Abdul Halim S. M. graduated from the
University of Tennessee with B.S. in
performance, where the revised neural model is
Chemical Engineering in 1993. He started
slightly better with 0.52% of mean of absolute residue.
his career with MTBE/PolypropyleneMalaysia Sdn. Bhd. as a process engineeruntil 1999. He obtained his M.Sc. in Applied
For further prediction improvement, it is possible to
Process Control from the University of
add additional input parameters. However, it must be
Newcastle-upon-Tyne in 2000. Since then,he has been lecturing at Universiti Teknologi
carefully evaluated since there is a trade-off between
PETRONAS (UTP). He is a member of Institution of Engineers
prediction improvement and model complexity.
Malaysia and a registered Professional Engineer with the Boardof Engineers Malaysia.
I. Ismail holds an M.Sc. in Control
Engineering (Sheffield, UK) and a B.S. in
Electrical Engineering (Wichita, Kansas) with
Modelling a system to industrial scale needs a good
11 years working experience in Project/
process understanding. This process knowledge is very
Maintenance / Technical Services
important in selecting the appropriate model input
Department as a Project Engineer /Maintenance Engineer / Instrument
parameters. Some input variables are related to each
Manager serving for PETRONAS Penapisan
other, which lead to redundancy and unnecessary
(Terengganu) and Ethylene/Polyethylene
model complexity. It has been shown that, with
(Malaysia). His current position is Electrical
Engineering Senior Lecturer at UTP. He is also a member of the
combination of process knowledge and proper
Institution of Engineers Malaysia and a registered Professional
modelling procedure, neural network is able to track
Engineer with the Board of Engineers Malaysia.
the product E concentration satisfactorily in recycle
M.A. Shafawi obtained a B. Chem. Eng. (Hons) from Universiti
gas in a petrochemical complex .
Teknologi Malaysia in 1995 with 8 years working experience as aprocess engineer in Technical Services Department of MTBEMalaysia Sdn Bhd (MMSB). He was also in the team which was
deeply involved in the commissioning of the PropaneDehydrogenation Plant (PDH) in the year 2001. His current
Cybenco, G. (1989). Approximations by superpositions of a
position is Senior Process Engineer (Advanced Process Control)
sigmoidal function, Math. Control Signal & System, Vol 2, pp303-
Hornit, K., Stinchombe, M., and White, H. (1989). Multilayerfeedforward networks are universal approximators, NeuralNetworks, Vol 2, pp359-366.
Montague, G., C. Gent, A.J. Morris & J. Buttress (1996) Industrialreactor modeling with artificial neural networks, Trans Inst MC,Vol 18 No 3.
Barry Lennox, Gary A. Montague, Andy M. Frith, Chris Gent andVic Bevan (2001) Industrial application of neural networks –an investigation, Journal of Process Control, Volume 11, Issue 5,October 2001, Pages 497-507
PLATFORM VOLUME THREE NUMBER ONE JANUARY - JUNE 2003
Technology Cluster: OIL AND GAS
Technology Platform: SYSTEM OPTIMIZATION
Debottlenecking of an Existing Fractionation Train
Using the Non-conventional Distillation Columns
Kamarul Ariffin Amminudin
Universiti Teknologi PETRONAS,
Bandar Seri Iskandar, 31750 Tronoh, Perak Darul Ridzuan, Malaysia.
Process debottlenecking is a common activity during the plant expansion project. To the operating companies,
the biggest issue is how to debottleneck the facilities at a minimum cost without sacrificing the operability and
safety aspects. The easiest approach is to develop a parallel process train, which is essentially a duplication of
the existing process train at a much lower capacity. Others may involve the revamp of the column internals such
as the change to a better type of internals like the advanced trays and structured packings. But too often these
options can run into a costly effort and may render the projects uneconomic after all. Do we have other options?
This paper will answer the above question by exploring other options to debottleneck the fractionation train.
The paper begins by addressing the issues while carrying out debottlenecking projects such as the need to
minimise the use of new equipment, to maximise the use of existing hardware in the existing infrastructure and
to reduce the plant downtime. All of these issues will influence our decision to propose the most cost-effective
solution to our debottlenecking problems. Hence, in considering these factors, we propose an option that uses
a non-conventional distillation column to debottleneck the fractionation system.
The concept of a non-conventional distillation column is not new. Today this type of column has found its
application in the petroleum refining, such as the use of side strippers, and the cryogenic air separation, such as
the use of side rectifiers. Other forms of these columns such as the prefractionator arrangement, the fully thermally
coupled column and the dividing wall column, however, have started to attract the interests of both process
designers and operating companies. The justifications for their use have been prompted by the less energy
consumption and greater efficiency when compared with the conventional fractionation system.
Thus this paper shows how the benefits of energy saving and efficiency improvements from the non-conventional
distillation column system can be translated into a huge debottlenecking opportunity for existing fractionation
train. Two case studies will be demonstrated in this paper. These case studies will highlight the issues and
opportunities by using such a new fractionation arrangement. The first case study involves the process
debottlenecking study of the natural gas liquids (NGL) separation train based on the gas plant in Europe and
the second one deals with the feasibility study of the recent project of the PETRONAS Gas Berhad (PGB) GPP-1
rejuvenation and revamp project, which was completed in December 2000.
Keywords: complex columns, process retrofits, prefractionator arrangement
This paper was presented at the First PETRONAS Petrochemical Plant Conference, Kuantan, Malaysia, 28-30 July 2002.
VOLUME THREE NUMBER ONE JANUARY - JUNE 2003 PLATFORM
Technology Platform: SYSTEM OPTIMIZATION
Technology Cluster: OIL AND GAS
sitting idle. Obviously, by doing so, we minimize the
use of new equipment.
Typically, a debottlenecking project is carried out when
• Minimize plant downtime whenever possible.
the plant has been in the operation for almost 2 to 3
• Give top priority on a straightforward and simple
years after commissioning. The main objective of the
debottlenecking project is to allow more throughputs
• Never compromise with safety issues.
to be processed into an existing facility. This is
common when the plant is capacity limited to produce
more products.
Let us consider an example. The plant is to increase its
Depending on the nature of the project, there are two
capacity by 20%. It is found that the existing separation
major debottlenecking projects. One is the process
train has experienced a bottleneck under this new
debottlenecking of the separation trains, as in the gas
scenario. What are the available options for us to
plants (LNG and NGL plants) and the other is the utility
debottleneck the separation trains cost effectively?
debottlenecking, as in the debottlenecking of the
furnace in the crude preheat train system of the crude
1. The quick and easy approach
oil refinery.
This is achieved by creating a duplication of
equipment or separation train at a much lower
For a given desired increase in capacity, the challenge
capacity than the existing ones. But in many cases
to the engineers is to debottleneck the existing plant
the cost of acquiring such new hardware can be
in the most cost effective manner. There are several
prohibitively high.
factors that need to be considered:
2. Focusing on the debottlenecking of the column
• Existing infrastructure and equipment, i.e. our
investments. The challenge is to exploit the existing
This is basically employing the traditional approach to
hardware to the maximum possible.
improve the hydraulic performance of the column
• Cost implications of doing the project itself.
internals, i.e. to get more hydraulic capacity through
• Plant shut down which results in the loss of
an existing distillation column. In this case, the
objective is to retrofit the column with the internals
• Duration of the shutdown itself. How do we plan
that allow more vapour or liquid traffic to pass through
ourselves to reduce the shutdown period?
without the loss of efficiency. In this kind of project,
• Attitudes towards change. How can we guarantee
one needs to perform a major "surgery" on the existing
that the retrofit design is simple and straightforward
column to replace the existing internals with the new
such that the plant is safe to operate after a
one that debottleneck the column. For example,
debottlenecking project?
advanced trays (Nye Trays and Multiple Downcomer
Tray) and structured packings, require the removal of
Given all the factors above, the following retrofit
the existing tray support rings and the existing
strategies for debottlenecking are the key success
factors to ensure that the implementation of the
project is cost effective:
The use of better internals obviously promotes
debottlenecking but this is normally achieved at the
• Maximize the use of existing hardware. The plant
expense of a huge cost. Of concern to us are the costs
inherently has a hidden capacity that can be tapped
of new types of internals, costs of removing the
more readily and if we carry out this exercise
existing internals and the opportunity cost of
systematically, we may avoid having equipment
extensive plant downtime.
PLATFORM VOLUME THREE NUMBER ONE JANUARY - JUNE 2003
Technology Cluster: OIL AND GAS
Technology Platform: SYSTEM OPTIMIZATION
3. Use of the thermal efficiency approach
separation work, resulting in higher energy
This approach involves the use of a side reboiler, side
condenser, and feed conditioning such as preheater
and precooling, to improve the thermal efficiency of
To overcome this problem, extra trays or packings are
the separation system. Any thermal efficiency
required between the side utility exchanger and the
improvement made by these units can be translated
main utility so that this helps offset the unnecessary
into a huge potential for debottlenecking as the liquid
increase in energy consumption. Nonetheless, our
and vapour loadings are reduced in certain sections
concern is the cost implications of acquiring these
of the column. For example, a side condenser
extra internals.
condenses part of the vapour before it arrives in the
main condenser and hence reduces both vapour and
INTRODUCTION TO NON-CONVENTIONAL
liquid traffic in the column section between the side
DISTILLATION COLUMN SYSTEM
condenser and the main condenser. This effectively
debottlenecks the column.
Normally, we deal with a conventional distillation
column where we have a single feed with two product
However, if the number of stages remains fixed, the
stream outlets, namely, the distillates and the bottoms.
use of side condensers or side reboilers always
However, anything that does not belong to the single-
contributes to a problem of not meeting the original
feed and two-product-column system is considered to
product specification. This is because the addition of
be the non-conventional distillation column, as shown
these side utilities reduces the driving force for the
in Figure 1.
Figure 1: Typical non-conventional distillation column system available in practice.
VOLUME THREE NUMBER ONE JANUARY - JUNE 2003 PLATFORM
Technology Platform: SYSTEM OPTIMIZATION
Technology Cluster: OIL AND GAS
Of interest in Figure 1, the side stripper has been widely
prefractionator allows greater freedom to match the
used in the crude oil distillation column unit, the
column feed composition with the composition in one
atmospheric tower and the side rectifier in the
of the prefractionator trays. These two major
cryogenic air separation units for many years. But
improvements can lead to a typical energy saving of
today, the least industrially applied system, the fully
about 30% compared to the performance of the
thermally coupled column, the prefractionator and the
conventional distillation column system.
dividing wall column are gaining acceptance in the
process industries (Hairston, 1999) even though the
Similarly, the fully thermally coupled column has an
concept was established some 50 years ago (Brugma,
equal energy saving benefit as in prefractionator
1937, 1942).
arrangement. But the former system has a direct mass
flow contact and the reboiler of the main column must
For example, BASF in Germany has 12 dividing wall
supply its vapour to both main and the prefractionator
column units currently in operation and they are
columns. Such an arrangement may reduce the
mainly used to separate mixture of the fine chemical
chances for debottlenecking especially in the section
products. In 1998, MW Kellogg revamped one of the
between the bottom thermal coupling points and the
existing distillation column units to become a dividing
wall column at the BP Amoco Coryton refinery in Essex,
England (MW Kellogg, 1998). Similarly the Sumitomo
Thus, to remove this counter productive effect, it is
Heavy Industries Co. of Japan also constructed the
desirable that some of the heat loads be shifted to the
dividing wall column (Parkinson, 1998). UOP, a process
prefractionator by a dedicated reboiler. Similarly, some
licensor, conducts several feasibility studies on the use
of the condenser load needs to be re-distributed to
of this type of system in their applications.
Apparently, the process industries are reluctant to
apply these types of columns commercially, especially
the dividing wall column. The reasons are mainly
rooted in the lack of design procedure, fear of
problems related to control and operations, and the
potential high risks involved in its application.
The major driving force behind the use of the non-
conventional column systems such as the
prefractionator, the fully thermally coupled column
and the dividing wall column is due to energy savings
and capital cost saving incentives.
Most of the energy savings in the non-conventional
column, as in Figure 2, is attributed to the
prefractionator. Here, a major source of separation
inefficiency in the conventional column is the remixing
effect. A significant reduction of this effect can be
achieved by providing a crude split in the
prefractionator. In addition, the prefractionator
minimizes the mixing losses at the feed tray. This is
because the non-sharp separation occurring at the
Figure 2: Prefractionator arrangement promotes greater energy
efficiency
PLATFORM VOLUME THREE NUMBER ONE JANUARY - JUNE 2003
Technology Cluster: OIL AND GAS
Technology Platform: SYSTEM OPTIMIZATION
the prefractionator as well. This makes the use of
Figure 5 shows the modifications required to the
prefractionator system the most acceptable choice.
existing column system to produce the prefractionator
arrangement. It highlights the simplicity involved in
CASE STUDY 1 – DEBOTTLENECKING OF NGL
achieving such a design. The original total condenser
from the debutanizer can be readily converted into a
partial condenser without incurring any substantial
This case study focuses on the application of the non-
cost. However, it is expected that more pipe work
conventional distillation column system in a retrofit
would be needed due to the re-arrangement of the
of an existing NGL separation train (Lestak and Collins,
columns. Pipe work is needed for new streams such
1997). The objective is to increase the capacity of the
as the connection between the new bottom stream
train by 20%. Figure 3 illustrates the existing
of the debutanizer and the deisobutanizer, and to re-
separation train and its current operating conditions.
route the feed and product streams to the
deisobutanizer. In addition, a new side draw product
However, at a much higher production rate, the train
stream to yield the nC4 product will be needed from
is experiences bottleneck as shown in the shaded
the deisobutanizer. The pipework costs depend
region of Figure 4.
heavily on the layout constraints that exist in the plant.
Two debottlenecking options are proposed. Option 1
Unlike the two columns, the depropanizer will be
involves the use of prefractionator system and Option
debottlenecked by using a conventional approach of
2 considers the use of fully thermally coupled column
revamping the internals and improving the thermal
system, as shown in Figures 5 and 6 respectively.
Figure 3: A simplified flowsheet of the existing NGL separation train
VOLUME THREE NUMBER ONE JANUARY - JUNE 2003 PLATFORM
Technology Platform: SYSTEM OPTIMIZATION
Technology Cluster: OIL AND GAS
Figure 4: Significant column bottlenecks occur here especially that involves deisobutanizer.
Figure 5: Prefractionator system that links both debutanizer and deisobutanizer
PLATFORM VOLUME THREE NUMBER ONE JANUARY - JUNE 2003
Technology Cluster: OIL AND GAS
Technology Platform: SYSTEM OPTIMIZATION
Figure 6: The reboiler from the deisobutanizer supplies the heat load for both columns
Figure 6 indicates more pipe work modifications are
needed to implement the fully thermally column
system. The pipe work modifications are primarily
needed to create the thermal coupling connections
between the debutanizer and the deisobutanizer. As
a result, the condenser and the reboiler from the
debutanizer are bypassed. Furthermore, as the load is
now shifted to the deisobutanizer, this
debottlenecking option may be more restricted as
compared to the prefractionator arrangement option.
In this case, the reboiler and condenser may need
additional heat exchanger areas.
Figure 7 demonstrates the hydraulic performance
profile for the deisobutanizer in option 1
Figure 7: Deisobutanizer capacity is well below the flooding
(prefractionator system). It shows the system results
in a better capacity improvement as compared to the
fully thermally coupled column system, as seen in
Figure 8. This confirms that the prefractionator system,
separation train compared with the fully thermally
which provides load distribution between the two
coupled column system.
reboilers, offers greater additional capacity to the
VOLUME THREE NUMBER ONE JANUARY - JUNE 2003 PLATFORM
Technology Platform: SYSTEM OPTIMIZATION
Technology Cluster: OIL AND GAS
Its objective is to use the concept of non-conventional
column system into the existing GPP-1 LPGU
separation train.
Figure 9 illustrates the simplified GPP-1 LPGU/PRU
separation train with its operating conditions. The
debutanizer and the depropanizer were chosen for the
application of non-conventional distillation columns.
Since the depropanizer has a higher operating
pressure than the debutanizer, it is practical to
designate the depropanizer as a prefractionator. But
to achieve this in practice, significant pipe work
modifications will be required and the retrofit project
Figure 8: The deisobutanizer for the fully thermally coupled
may be costly . Figure 10 demonstrates some of the
column is capacity limited.
concerns raised in a proposed prefractionator option.
CASE STUDY 2 – DEBOTTLENECKING
This clearly demonstrates that the prefractionator
OPPORTUNITIES FOR GPP-1 LPGU/PRU
system is not always suitable for applications in the
debottlenecking projects as it may render the whole
project uneconomical. There will be additional cost
This feasibility study was carried out in August 2000
implications as more constraints arise. Moreover, the
as part of the GPP-1 rejuvenation and revamp project.
operational constraints of the existing separation train
Figure 9: Existing separation train for GPP-1 LPGU/PRU shows a much higher operating pressure for depropanizer than that in the
debutanizer
PLATFORM VOLUME THREE NUMBER ONE JANUARY - JUNE 2003
Technology Cluster: OIL AND GAS
Technology Platform: SYSTEM OPTIMIZATION
Figure 10: Constraints on the column pressure at A102 severely limits the air fan cooler reusability.
limit the maximum use of existing hardware. This is
• The potential benefits derived from the non-
indeed undesirable. Thus the prefractionator system
conventional distillation columns cannot be ignored.
is not recommended for GPP-1 separation train.
• In particular, prefractionator system promotes
The application of the prefractionator arrangement
• Not all applications are suitable for non-conventional
has started to attract many oil and gas companies. For
column system.
example, AspenTech in the UK carried out a feasibility
study project on the debottlenecking study using this
type of columns for a petrochemical company.
The author would like to express his appreciation to both OGPT
Similarly, a UK chemical company has also evaluated
and PGB for allowing some of the work here to be presented in
this kind of non-conventional applications. Future
this conference.
applications may be seen in LNG plants where the
intention is to reduce energy consumption.
Lestak, F.; Collins, C., "Advanced Distillation Saves Energy and
The following conclusions can be made:
Capital", Chem. Eng., July 1997, p72-76.
Amminudin, K.A.; Smith, R., "Design and Optimization of Fully
• Non-conventional distillation columns have more to
Thermally Coupled Distillation Columns – Part 2: Application
offer for process debottlenecking purposes.
of Dividing Wall Columns in Retrofit", Trans IChemE, vol 79, PartA, Oct 2001, p 716-724.
VOLUME THREE NUMBER ONE JANUARY - JUNE 2003 PLATFORM
Technology Platform: SYSTEM OPTIMIZATION
Technology Cluster: OIL AND GAS
Hairston, D.; "The Divide in Distillation", Chem Eng., April 1999,
Kamarul Ariffin Amminudin earned his
Ph.D. in the area of Process Integration fromthe University of Manchester Institute of
Brugma, A.J., Dutch patent no. 41850, Oct 1937 and US patent
Science and Technology, England. He is a
2295256, Sept 8, 1942.
senior lecturer for the chemical engineeringprogram at UTP. His expertise is in the area
Watkins, R.N.; "Petroleum Refinery Distillation", 2nd ed., Gulf
of process integration, pinch analysis,
Publishing, Houston, Texas, 1979.
process retrofit and wastewaterminimization. He has published his work in
Massimo and Pat, UOP, personal communication, UMIST,
journals in the area of complex columns and
Manchester, England, 1997.
process debottlenecking. Recently, he published an article aboutpinch analysis application in the process industry for energy
M.W. Kellogg Limited press release, 11 Sept 1998., M.W. Kellogg
saving measures in the bulletin of the Institution of Engineers
Ltd, Middlesex, England.
Malaysia. He was the consultant for several multinationalcompanies such as AspenTech, ICI, BP Amoco and MW Kellogg
Parkinson, G.; ed, "Chementator", Chem Eng., July 1998, p 21-
while work ing for his Ph.D. degree, especially providing
professional advice for separation system processdebottlenecking. In Malaysia, he was involved in a project withOGPT Technical Services (M) Sdn. Bhd. for the PETRONAS GasBerhad GPP-1 separation systems. Currently, he is helping toformulate water minimization strategies, together with PETRONASResearch Scientific Services Sdn Bhd, for PETRONAS Penapisan(Melaka) Sdn Bhd.
He is currently a Corporate Member of the Institution of EngineersMalaysia and a member of the American Institute of ChemicalEngineers.
PLATFORM VOLUME THREE NUMBER ONE JANUARY - JUNE 2003
Technology Cluster: INTELLIGENT SYSTEMS
Technology Platform: SMART CARD TECHNOLOGY
Face Recognition For Mykad Security
Ahmad Fadzil M. H. and Irfan A. R.
University Teknologi PETRONAS,
Bandar Seri Iskandar, 31750 Tronoh, Perak Darul Ridzuan, Malaysia.
The application of smart card has been significant during the past few years such as with the introduction of the
MyKad and the replacement of magnetic stripes ATM card to smart card. The objective of the project is to
design a face recognition system to be incorporated into MyKad for user identification purposes. The system
involves two security levels whereby the first level involves the Personal Identification Number (PIN) verification
that is stored in the MyKad/Smart Card. The secondary security level uses the face recognition system to confirm
the actual owner is seeking authorization. The user's face will be acquired using a digital camera that is interfaced
into the computer using a TV Capture Card. The face recognition algorithm was developed using Neural Networks
based on a variation in the backpropagation algorithm that is the Fletcher-Reeves method. The neural networks
consist of 600 input nodes, 10 hidden nodes and 1 output node. The number of hidden nodes selected varies
with test subject. The neural networks were trained to identify two classes of faces, which are the authorized
faces acquired from the digital camera, and the non-authorized faces gathered from the FERET Database. The
neural networks were tested and produced a performance between the range of 90-100% acceptance for the
authorized face and rejection for the non-authorized faces.
Keywords: face recognition, smart card, neural networks.
person's pose, perspective angle and expression.
Neural networks can be used to implement face
Biometrics can be defined as a biological
recognition [1, 2, 3, 4, 5].
measurement of the human features that is to
characterize an individual. Among the techniques of
The main goal of this work is to design a face
Biometrics employed in the society include the retinal
recognition system that will be incorporated into the
scans, fingerprints verification, face recognition and
smart card to emulate the application of MyKad. The
system is implemented by creating two-protection
layer for the Smart Card/MyKad. The first layer will
One of the most widely researched biometrics
involve the PIN verification mechanism that will be
applications is the face recognition. Face recognition
supported by the face recognition scheme in the
is the process of recognizing the owner of the
secondary level. By assigning these two protection
presented face. Recognition of faces is a complex
layers into the Smart Card, the level of security may be
problem due to the dependencies of the image on the
The paper was presented at the Malaysia-Japan Seminar on Artificial Intelligence Application in Industry (AIAI 2003), Kuala Lumpur, 24-25 June 2003.
VOLUME THREE NUMBER ONE JANUARY - JUNE 2003 PLATFORM
Technology Platform: SMART CARD TECHNOLOGY
Technology Cluster: INTELLIGENT SYSTEMS
2. SYSTEM DESIGN
matrix reshaping, input normalization, and training
Figure 1 shows the flowchart of the overall working
principle of the system.
In order to train the neural networks face samples are
required. The authorized face is taken in ten different
poses. This ten poses is then taken three times at
different time in order to increase the number of face
samples. The images are taken using a digital camera
with a resolution of 1.3 megapixels via a TV Capture
Card that is interfaced with a computer. Ensuring that
the distance between the camera and the test subject
is maintained at constant value also standardizes the
size of the faces. The ten poses are as follows
2) Normal with Smile
3) Upward by 20°4) Upward by 20° with Smile5) Downward by 20°6) Downward by 20° with Smile7) Left by 20°8) Left by 20° with Smile9) Right by 20°
Figure 1: Flow Chart of Security System
10) Right by 20° with Smile
The image captured using a digital camera is in the
The overall system starts whenever a person inserts
format of RGB colour images. An RGB format images
his/her MyKad into the card terminal. The person will
corresponds to an array of the size m x n x 3 whereby
then be prompted to insert his/her PIN number as the
m and n is the size of the image pixels. In order to
first layer of security for the system. The system will
reduce the data, images are converted to greyscale
then compare the PIN number inserted by the user
(Matlab function ‘RGB2GRAY'). An area of interest from
with the stored number in the Smart Card for
the faces is extracted by cropping the face images. The
verification and authentication. If the user enters the
two eyes from the images are used as the pointer for
wrong PIN number the system will prompt the user to
setting the region of interest for the faces. The area of
either re-enter the PIN number or to exit the system.
interest was standardized to the size of 240 x 160
After the PIN number is verified the system will load
the neural data stored in the smart card into the
system. The neural data will then be used to generate
In order to reduce of input nodes of the neural
a specific neural network that will perform the
networks a reduction in the image size is necessary. A
recognition process. The neural data that will be
technique called pixel averaging was used in the work
loaded into the smart card is the Input Weight, Layer
in order to reduce the size of the face images. In this
Weight, Layer Bias and Output Bias.
technique 4 pixels from the original image is taken and
averaged to form the new value of the pixel. This
The process of designing the face recognition
technique is repeated for consecutive level until the
algorithm involves the process of image acquisition,
next level of averaging will cause the reduce face
colour transformation, image rotation, image resizing,
PLATFORM VOLUME THREE NUMBER ONE JANUARY - JUNE 2003
Technology Cluster: INTELLIGENT SYSTEMS
Technology Platform: SMART CARD TECHNOLOGY
image to lose its salient features. After experimenting,
normalization of the input is performed by the
level 3 was found to be the maximum level of
following algorithm which was proposed by [3].
averaging to sufficiently recognize a person face. The
reduce size of the images is of the size 30 x 20 pixels.
Yi,j = Xi,j Ð 127
Figure 2 illustrates the specified technique.
Where Xi,,j is the pixel value on row i and column j ofthe image
Yi,j is the pixel value which is between -127 and 127
Figur Pixels
e 2: Pixel Averaging
L is the vector length of the image
Zi,j is the normalized value between -1 and 1
The next step of processing required to be
implemented on the reduced image is to change the
In order to use the FERET database for training the
2 dimensional array of the images into a vector (1
neural networks, the image provided by the database
dimensional array). (see Matlab function ‘RESHAPE')
must be firstly standardized in order to create a more
This transformation is necessary as the multilayer
uniform training specimen in addition to the images
peceptron is only capable of accepting vectors as its
captured using the digital camera. Two properties
were addressed in this process; the head size
standardization and the lighting adjustment. In order
The value of the pixels varies from 0 to 255. In order to
to standardize the face database to match the images
use a hypertangent function in the neural network the
captured using the digital camera. This is done by
vector input is required to be normalized between the
manually cropping the face images from the FERET
value of -1 and 1. This is because the hypertangent
database without any size standard imposed in the
function is only sensitive to changes when presented
cropping process. The image is then resized to the
with an input value between -1 and 1. Any value that
size of 240 x 160 pixels (Matlab function ‘IMRESIZE').
resides outside this region will be truncated and thus
The face size is then reduced to the size of the 30 x 20
causing information loss. The function of the
pixels using pixel averaging as been discussed. Due
hypertangent is given by
to the lighting variations in the face images a lighting
adjustment is required to be implemented on the
FERET database. It was found that the average value
of the image captured using the digital camera ranges
from 85 and 252 therefore the pixel value in the FERET
database was linearly adjusted to be between this
where x is the input value of the function
range (Matlab function ‘IMADJUST').
Therefore this value is required to be normalized
In order to create face recognition based on neural
before presenting it into the neural networks. The
network, a neural network was previously trained using
VOLUME THREE NUMBER ONE JANUARY - JUNE 2003 PLATFORM
Technology Platform: SMART CARD TECHNOLOGY
Technology Cluster: INTELLIGENT SYSTEMS
the authorized and nonauthorized faces. A three layer
The smart card program was written using Visual Basic
feed forward neural network has been implemented
6. The functions that were previously implemented in
consisting of 600 neurons 10 hidden layer and 1
MATLAB will be performed under the Visual Basic
neuron in the output layer. The activation function
platform. This is done by converting the m-file created
that was used in both layers is the hypertangent
into a DLL files which can be readily integrated into
function with outputs between -1 and 1. The training
any COM available application. The conversion is
algorithm that was used is a variation of the
implemented using the MATLAB COM Builder.
backpropagation algorithm that is the Fletcher-Reeves
method based on the conjugate gradient. In the
3. RESULTS & DISCUSSION
conjugate gradient algorithms a search is performed
along the conjugate directions of the gradient, which
Figure 3 shows the cropped and resized image of
produces generally faster convergence than steepest
the test subject.
descent directions. The weight and bias is initialized
using the Nguyen & Widrow method.
A problem which might arise is the overfitting the
neural networks. This occurs when the neural network
has been extensively trained using the training subject.
Unfortunately when the neural network is presented
with a new set of data, the neural network will not be
able to produce the targeted value for the test subject.
This is because the error in the training data has been
Figure 3: Reduced Images
driven to a very small value making the neural network
too exclusive for the training data without the
capability to generalize for new unseen data. In order
The decision of the neural network architecture was
to prevent the case of overfitting in the neural network,
based on a few factors. The number of the input nodes
a technique called early stopping has been
is determined by the size of face image that have been
implemented. In this method the available data is
reduced to 30 by 20 pixels. The number of the
divided into 3 subsets of data; training data, validation
optimum hidden nodes is determined by the lowest
data and test data. The training data is used to train
mean square error (mse) generated by the test data
the data as required in any normal neural network in
and training data. The mean square error of the test
order to enable the neural network to identify the
data is computed by the following formula.
pattern of the training data. The validation data is used
for monitoring the error made by the neural network
during the training process. When the neural network
starts to overfit the data this will cause the error in the
validation data to rise and after a specified number of
iterations, the training is stopped, and the weights and
As can be seen from Figure 4, the minimum number
biases at the minimum of the validation error are taken.
of mse occurs when the value of hidden nodes is equal
The test data will also be used to determine the
to 10, 15 and 17. In order to reduce the number of
optimum number of hidden nodes required for
weights generated in the neural network, the
particular neural networks. Finding the error
minimum number of the hidden nodes used is 10.
produced by the neural networks when simulated
using the test data does this.
The total input weight is obtained by multiplying the
number of hidden nodes with in the number of input
PLATFORM VOLUME THREE NUMBER ONE JANUARY - JUNE 2003
Technology Cluster: INTELLIGENT SYSTEMS
Technology Platform: SMART CARD TECHNOLOGY
The validation curve actually follows the training curve
that indicates a good correlation between the training
and validation data. This usually happens when there
are a good distribution data between the training and
the validation data.
The output from the neural network is thresholded so
that any output which is greater than 0.75 is
considered 1 and any value that is smaller than -0.75
is considered as -1 and any output that is located
between -0.75 and 0.75 is considered as undefined
Figure 4: Performance of Training & Test Data
It was also found that different authorized faces require
different number of hidden nodes. The number of
nodes which in this case is 600 x 10 = 6000 elements.
hidden nodes is determined based on the same
Each element requires a storage capacity of 8 bytes or
method as described earlier. This is due to the fact that
64 bits (double datatype). This will cause the system
different faces will eventually reside at certain region
to consume a total storage of 8 x 6000 = 48000 bytes
that might be different from the other authorized faces.
or 48 kbytes in order to store the input weight in the
The optimum number of hidden nodes for a particular
face will eventually differs at different iteration. This is
fully understood as the initialization function that is
The system was then again trained using 10 hidden
based on the Nguyen Widrow method randomises the
nodes and the performance curve is observed. Figure
initial weight, which will affect the convergence of the
5 shows the performance curve of the neural networks
neural networks to the targeted value.
training process.
The different success rate for different authorized faces
There are three distinct curves, which is which are the
is due to the fact that some of the authorized face
training curve, validation curve and the test curve. It
might have a high degree of similarity between the
can be seen that the training curve performance
non-authorized faces. If this condition occurs this will
reaches a value of mse which is lower than the goal of
cause some difficulties to the neural networks to
the training which is set at 1e-5.
establish the separation boundary between the
authorized and non authorized faces. Therefore in
order to improve the performance of this authorized
faces a higher number of sample faces has to be
gathered in order for the neural networks to better
classify the two classes. Figure 6 shows a chart
indicating the success rate for three different
authorized faces.
In order to integrate processing scheme, which was
implemented using MATLAB into the Visual Basic
environment a DLL extension file was created from the
m-file. This is done in order to liberate the system from
being processed using MATLAB and thus can operate
Figure 5: Neural networks performance curves
VOLUME THREE NUMBER ONE JANUARY - JUNE 2003 PLATFORM
Technology Platform: SMART CARD TECHNOLOGY
Technology Cluster: INTELLIGENT SYSTEMS
independently without the use of MATLAB analysis
tools. Figure 7 shows the implementation of rotating
an image in Visual Basic that was based from the
A face recognition system based on neural networks
function ‘IMROTATE' created in MATLAB.
has been successfully implemented. The process of
training the face samples involves the procedure of
transforming the coloured images, reducing the size
and normalizing the pixel value.
In order to prevent overfitting the neural networks, a
method of early stopping was employed in the system
whereby three subsets of data is used in this technique
namely the training data, validation data and test data.
This technique will also reduce the necessary training
data that is required to be presented into the neural
The optimum number of hidden nodes is determined
Figure 6: Overall Test Data for 3 Authorized faces
by finding the lowest mean square error given by the
neural networks on the test data that was presented
to the neural networks. The different number of
hidden nodes for different test subject is due to the
poor generalization of the available data towards
certain faces.
By employing the early stopping method and using
the optimum number of hidden nodes we were able
to achieve an efficient face recognition system with a
high recognition rate.
Further study should be conducted in the area of
reducing the storage capacity required by the neural
networks in the smart card in order to fully integrate
the face recognition system in the smart card.
Kingston Sew Eng Hock. 1994. Pemprosesan Imej: RangkaianNeural untuk Pengecaman Muka, B.Sc. Dissertation, UniversitiSains Malaysia, Malaysia.
Kong Peng Yong. 1995. Sistem Pengecaman Muka BerasaskanRangkaian Neural dan Kad Cerdik, B.Sc. Dissertation, UniversitiSains Malaysia, Malaysia
Figure 7: Rotating an Image
Abu Bakar Hassan. 1994. Human Face Recognition System basedon Neural Networks, M.Sc. Dissertation, Universiti SainsMalaysia, Malaysia
PLATFORM VOLUME THREE NUMBER ONE JANUARY - JUNE 2003
Technology Cluster: INTELLIGENT SYSTEMS
Technology Platform: SMART CARD TECHNOLOGY
Lim Cheah Choon. 1996. Pemproses Imej : Sistem PengecamMuka Berdasarkan Kecirian Muka, B.Sc. Dissertation, UniversitiSains Malaysia, Malaysia.
Henry A. Rowley, Shumeet Baluja & Takeo Kanade, 1998,"Neural Network-Based Face Detection," Transaction on PatternAnalysis and Machine Intelligence 20(1) : 23 - 38
Anil Jain, Ruud Bolle and Sharath Pankanti. 1999. BiometricsPersonal Identification in Networked Society, Massachusetts,USA, Kluwer Academic Publishers.
Shaogang Gong, Stephen J McKenna, Alexandra Psarrou. 2000.
Dynamic Vision : From Images to Face Recognition, London UK,Imperial College Press.
Laurene Fausett. 1994. Fundamentals of Neural NetworksArchitectures, Algorithms, and Applications, New Jersey US,Prentice-Hall, Inc.
Ahmad Fadzil M.H. graduated in 1983 from
the University of Essex, UK with an Honours
B.Sc. degree in Electronic Engineering. He
completed his M.Sc. in Telematics in 1984
and Ph.D. in Image Processing in 1991 at the
same university.
He has been a Lecturer in Signal Processingand Researcher in Image Processing at
Universiti Sains Malaysia (USM) since 1984. Between 1988-91, heserved at University of Essex, UK initially as a Senior ResearchOfficer in Image Processing and subsequently as a Lecturer. Onhis return to Malaysia, he was made Dean of the School of Electrical& Electronic Engineering at USM from 1992-96. In 1997, hebecame the founding Dean of the Engineering Faculty at thenewly established UTP. He became the Director of AcademicStudies at the university in 1999. Currently, he is the Director ofPostgraduate Studies, a position he assumed in 2003.
Dr Ahmad Fadzil is a Fellow and a Council Member of the
Figure 8: Sample faces at different poses
Institution of Engineers, Malaysia. He is a registered ProfessionalEngineer with the Board of Engineers, Malaysia. He is also aMember of the IEEE, USA. His research interests include imagecompression and image processing applications in telemedicineand artificial intelligence.
VOLUME THREE NUMBER ONE JANUARY - JUNE 2003 PLATFORM
Technology Platform: SMART CARD TECHNOLOGY
Technology Cluster: INTELLIGENT SYSTEMS
Enhancing Design Criteria For Novice Virtual Environment
Designers Through The Identification Of Usability Problems
Suziah Sulaiman and Rohiza Ahmad
Universiti Teknologi PETRONAS
Bandar Seri Iskandar, 31750 Tronoh, Perak Darul Ridzuan, Malaysia.
In this paper we investigate common problems made by novice designers in their virtual environment (VE)
designs. Our aim through this study is to alleviate some of the problems encountered by novice designers in
their attempts to design and implement their first VEs. The investigation involves the application of usability
evaluation techniques as used in Human-Computer Interaction (HCI). Some existing guidelines in VE design are
used as a basis for the preliminary work. In our work, we rank usability problems according to how serious they
are in the design process. Then, based on this ranking we suggest a set of candidate's criteria for VE design
meant specifically for novice designers. The findings from this study highlight the importance of these existing
guidelines and the need to modify some of them to suit the requirements of novice designers.
Keywords: human-computer interaction (HCI), usability evaluation, virtual reality (VR), virtual environment (VE)
expertise [4]. To instil the usability elements as part of
the design poses another big challenge especially to
Designing effective and usable content of a virtual
novice VE designers. The question of how we inform
environment (VE) is becoming increasingly important
novice designers to incorporate usability in their
with the proliferation of many applications in real-
design is an issue to be discussed.
world visualisation. It is essential to look at the usability
of the VE application design as not to get users
This paper identifies common problems made by
frustrated when exploring the environment. Much
novice designers in their VE design. The intention is to
groundwork in the domain of evaluating VE
guide designers not to repeat the same mistakes again.
application starts from the traditional human-
The investigation is extended to ranking the usability
computer interface usability [3,5,6]. Applying Human-
problems based on how critical or serious they are in
Computer Interaction (HCI) concepts which
the VE design process.
encompass notions from inter-related fields enable
one to obtain the immersive feeling in the VE
2. BACKGROUND OF STUDY
In designing an effective VE application, there are many
Usability issues are concerned with obtaining products
factors that need to be considered. These factors
which are user friendly or easy to use. International
include the technology used, the time, the cost and
Standards Organisation (ISO) defines usability as ‘…the
This paper was presented at the International Conference on Engineering Education, UMIST, United Kingdom, August 18-21, 2002.
PLATFORM VOLUME THREE NUMBER ONE JANUARY - JUNE 2003
Technology Cluster: INTELLIGENT SYSTEMS
Technology Platform: SMART CARD TECHNOLOGY
effectiveness, efficiency and satisfaction with which
In conducting usability evaluation, there are no single
specified users can achieve specified goals in particular
techniques which could capture all usability problems.
environments' (ISO DIS 9241-11). Designing for
Often, combination of techniques proves to be
usability takes into consideration elements which
beneficial [12]. However, at times, a ‘quick and dirty'
include specifying user characteristics, requirements
technique is required in order to obtain a fast feedback
capture, usability specification and iterative design and
while maintaining the quality. Usability inspection
prototyping [8]. The integration of usability in design
techniques which include heuristic evaluation meet
will result in acceptance of the system by the users as
this requirement and are reported to be cost-effective
reported in many success cases (e.g. [13]).
[11]. Heuristics evaluation conducted by 5 experts
could capture 75% of the usability problems [10].
Evaluation techniques as used in the area of Human-
Many researchers (e.g. [9]) have applied this result to
Computer Interaction (HCI) could be applied to virtual
their work and reported the same findings in terms of
reality (VR) especially in designing effective virtual
the number of experts and usability problems found.
environment. Adopting matured techniques from
Reference [9] reported that experienced and novice
other disciplines to VR could accelerate the
usability specialists differ in terms of judgement in
development of virtual environment (VE). This is
evaluating a graphical user interface. The experienced
necessary as a VE is difficult to design and use [14]. VE
evaluators were able to use the knowledge of previous
systems differ from previously computer-centred
user testing results and the knowledge of similar
systems in the extent to which real-time interaction is
designs to identify problems which were missed by
facilitated. In terms of characteristics, VE systems
the novices.
require a 3-D visual space, a multi-modal interface and
an immersive environment [1]. These characteristics
are not required by other computer-centred systems
as they could be displayed on 2-D. Real-time
interactions, which may lead to users' immersive
In our work, we replicate Lee et al.'s [9] and Nielsen [10]
feelings, are not necessary too. One possible way to
heuristics evaluation approach. Using these
facilitate development of a VE is through an
approaches, we meant:
enhancement on the evaluation technique. Evaluation
• to identify common problems/ mistakes made by
activity as described in the Star Life Cycle [7] is central
novice designers, and
to the whole design process. Feedback received from
• to rank the problems accordingly using Nielsen's [10]
the evaluation process is used to inform the design
approach to decide how critical the problems are in
team about how well the proposed design fits the
needs of users.
3.1 Subjects
Reference [2] has formulated some guidelines for
Five usability experts who are members of staff at
evaluating usability of a VE based on HCI. The purpose
Universiti Teknologi PETRONAS were selected for this
of her guidelines is to advise developers when
study. An ‘expert' is defined to be someone who had
designing VE. These guidelines contain elements to
three or more years of experience evaluating students'
be considered when designing the objects, user
projects. All usability experts chosen have a
actions and system control of a VE. Reference [5] took
background in either computer science and/ or IT
a user-centred design and evaluation approach in
ensuring the usability of their virtual environments.
Their technique is based on user task analysis, expert
guidelines-based evaluation, formative and
A training document was created for the guidelines
described in Deol [2]. This document describes the
VOLUME THREE NUMBER ONE JANUARY - JUNE 2003 PLATFORM
Technology Platform: SMART CARD TECHNOLOGY
Technology Cluster: INTELLIGENT SYSTEMS
guidelines and provides examples for each. Three
major problem by the evaluator as it could hinder users
small VE applications developed by final year IT
from achieving immersive feeling in the virtual
students were selected. This selection was based on
environment. All usability problems were compiled
their scores obtained during an exhibition on campus
and checked for any similarities among them.
during the semester.
Problems which are of the same nature, i.e. sharing a
common theme/area are grouped together. From a
Guidelines developed by Deol [2] which are used
total of 48 usability problems in Table 1, we have
during the evaluation, specify ways to design objects,
identified only 9 groups of usability problems. These
how to control them and their interactions in the
groups of usability problems (in random order) are:
environment. These guidelines serve as useful
objects should look real; objects should behave
checklist to designers in ensuring the usability of the
naturally like in real world; objects should be easy to
design. They also provide a basis for encouraging
distinguish; interactions should follow real world
usability earlier in the development lifecycle.
convention; allow users to be in control of the
application; allow easy navigation as in normal real
world; allow users to move freely in the environment;
3.3 Design and Procedure
objects should have correct scale as in real world;
All five expert evaluators went through the following
purpose of objects and environment should be made
• Answer a questionnaire about their background
• Learn about the guidelines
Table 1: Total number of usability problems found by each
• Evaluate the VE applications
Evaluators were tested individually. They were asked
to think aloud as they carried out the task and were
prompted to continue thinking aloud when they fell
silent. Evaluators were given a maximum time limit of
1 hour to complete the evaluation. They were
encouraged to use all their time even if they had
finished early. During the evaluation, evaluators were
asked to describe any problems encountered and
recommend a possible solution. Their feedback was
recorded on hardcopy forms. They were asked to use
the guidelines to assist them in identifying those
4.2 Ranking the usability problems
We presented the 9 groups of usability problems noted
in the previous section to the same evaluators where
4. DATA ANALYSIS
they have a joint discussion. This activity is part of
heuristics evaluation developed by Nielsen [10] which
4.1 Compilation of usability problems
is used in Lee et al.'s [9]. The objective of this discussion
All usability problems reported by evaluators were first
is to gain consensus among all the five evaluators on
analysed by the authors. (A complete listing of
the findings that we have compiled. It should be noted
usability problem reported by individual evaluators is
that the authors only took the facilitators role and were
included in Appendix 1.) The process started with
not part of this discussion. During the discussion, the
counting the total number of usability problems
evaluators reviewed the compiled usability problems
identified by each evaluator. This is shown in Table 1.
(those performed by the authors) and suggested that
Each usability problem reported is considered as a
some of the groups could be combined as they are
PLATFORM VOLUME THREE NUMBER ONE JANUARY - JUNE 2003
Technology Cluster: INTELLIGENT SYSTEMS
Technology Platform: SMART CARD TECHNOLOGY
related. This has reduced the groups of usability
Navigations are made easy, as users are able to move
problems from 9 to 6 only. Each evaluator was asked
freely in the environment. This co-relates with our
to rank these 6 new groups on the basis of how critical
human characteristics which have been applied in
they are in terms of the effect they may have on
virtual reality such as viewpoint, navigation,
reducing the usability in VE design. A rating scale of 1
manipulation and immersion. The ability to access the
to 5 was used in which 1 means less critical and 5
environment without restraints could then assist users
means most critical. The average score given by all
in achieving the main objective of VE design which is
evaluators for each problem group was calculated and
the immersive feeling. Evaluators further suggested
is presented in Table 2.
that for smooth navigation, there should be indicators
to notify the users that certain areas should be
Table 2: Ranking of usability problems
restricted. As an example, mouse cursors changed
when user collides with objects such as walls or trees.
Usability problem groups
From the findings, evaluators have pointed out that
issues in designing VE objects are secondary to
Make the purpose of objects
interaction. They argued that virtual reality is a
and environment clear
component of three elements: imagination,
Allow users to be in control of
interaction and immersiveness. The design of the
objects which appear and behave like the
conventional situation constitute a step forward
Allow easy navigation in which
towards achieving the immersive feeling in the VE
users could move freely
environment. In designing the objects, emphasis
(as in real world)
should be on 3D modelling which includes lighting,
Make objects behave and
texture, collision detection, choice of colour and scale.
interact naturally as in
These are all important to create the realism aspect in
the objects. The behaviour of the objects is another
Make objects look real and
area that needs to be looked into so that they resemble
have correct scale as in
or function as closely as possible to those in the real
Make objects easy to
5. RESULTS AND DISCUSSION
In our last section, we obtained six usability problem
Table 2 shows the revised version of these six groups
groups, which were derived from the discussion of the
of usability problems and how critical they are in the
five evaluators in the study. During the study each
design. Evaluators pointed out that the most important
group was also named accordingly so that these new
element in designing VE is the sense of purpose of the
groups could match Deol's [2] set of guidelines. Table
environment and objects used in it. Users need to
3 shows both Deol's guidelines and our new list of
understand the goal and objectives of a particular
usability problems. It should be highlighted here again
environment. This will give them a mindset of what
that Deol's set of guidelines is used in this study as a
the environment or objects are trying to imitate or
basis to identify our new list of usability problems.
mimic. Once users have realised or understood the
environment, it would be easier for them to be in
control of the VE application.
VOLUME THREE NUMBER ONE JANUARY - JUNE 2003 PLATFORM
Technology Platform: SMART CARD TECHNOLOGY
Technology Cluster: INTELLIGENT SYSTEMS
Table 3: Deol's guidelines and our new list of usability problems
OUR NEW LIST
DESIGN OF OBJECTS
Make objects easy to distinguish
Make the purpose of objects and
environment clear
Make objects easy to identify
Allow users to be in control of the
Make the interactivity and significance of
Allow easy navigation in which users
could move freely as in real world
Make objects easy to access
Make objects behave and interact
naturally as in real world
Make objects look real and have correct
scale as in real world
DESIGN OF USER ACTIONS
Make objects easy to distinguish
Show what actions are available
Make the purpose of actions clear
Show how to carry out actions
Make actions easy to execute
Show the effect of completed actions
DESIGN OF SYSTEM CONTROL
Show that control has begun or ended
Show why control has taken place
Show what actions are available during
Deol [2] has categorised the guidelines into two: (i)
mechanism to advise novice designers in designing
Design of Objects and (ii) Design Interaction. Design
their first VE application.
Interaction is further segregated into: Design of User
Actions and Design of System Control. Each category
From Table 3, some observations are made which
consists of several criteria. On the other hand, our new
findings could strengthened the importance of Deol's
list does not have any categories. This list consists of
[2] guidelines in system design but at the same time
usability problem groups taken from Table 2. It is more
needing some adjustment to suit novice designers.
condensed in terms of the number of criteria (group)
as compared to that of Deol. Each group is assigned a
Overlapping of guidelines
value that indicates how critical the group is in the
There appears to be an overlap in some of our groups
design. A higher value implies that the usability
with Deol's. Our group overlaps with one or more of
problem group associated to it should be given more
Deol's guidelines as shown in Table 4.
priority when designing VE as compared to those
assigned with lower values. This approach could be a
PLATFORM VOLUME THREE NUMBER ONE JANUARY - JUNE 2003
Technology Cluster: INTELLIGENT SYSTEMS
Technology Platform: SMART CARD TECHNOLOGY
Table 4: Overlapping of guidelines
Table 6: New groups
Our Group
Make objects behave and interact naturally as in
Make objects easy to
Make objects easy to
Make objects look real and have correct scale as in
Make the purpose of
Make the purpose of
environment clear
Make the interactivity
and significance of
In this paper we have highlighted the importance of
usability in designing virtual environments (VEs).
Concepts in the matured inter-disciplinary fields of
Human-Computer Interaction (HCI) could be applied
in VE design. This paper raises the question of how we
• Addressing a more higher level issues
inculcate and assist young and creative novice
Some of our usability problem groups seem to be
designers so that they could integrate usability as early
addressing a more higher level issues when compared
as possible in their development cycle. The approach
to Deol's. A combination of more than one criterion in
taken in this investigation is by identifying common
Deol's could be a subset of our group. This is shown in
mistakes made by novice designers. The set of
problem groups identified and discussed in this paper
were converted to a set of "good design guidelines"
• Identification of new groups
for novice users. These guidelines will hopefully
There are two other groups identified in our list that
eliminate the occurrence of common mistakes the
are different from Deol. These groups may seem trivial
future designs of VE.
but could be useful especially for beginners in order
to guide them in designing appropriate object
These new guidelines include making the purpose of
modelling which resembles the real world. The new
objects and environment clear; allowing users to be
groups are shown in Table 6.
Table 5: Addressing higher level issues
Our Group
Allow users to be in control of the application
Show that control has begun or ended
Show why control has taken place
Allow easy navigation in which users could
Show what actions are available
move freely as in real world
Make the purpose of actions clear
Show how to carry out actions
Make actions easy to execute
Show the effect of completed actions
Show that control has begun or ended
Show why control has taken place
Show what actions are available during control
VOLUME THREE NUMBER ONE JANUARY - JUNE 2003 PLATFORM
Technology Platform: SMART CARD TECHNOLOGY
Technology Cluster: INTELLIGENT SYSTEMS
in control of the application; allowing easy navigation
Lee W.O., Dye K. and Airth D. (1995), Evaluating DesignSpecifications Using Heuristic Evaluation, In Human-Computer
in which users could move freely as in the real world;
Interaction, Interact '95, Norby K., Helmersen P.H., Gilmore D.J.
making objects behave and interact naturally as in the
and Arnesen S.A. (eds.), Chapman & Hall
real world; making objects look real and have correct
[10] Nielsen J. (1993), Usability Engineering, Academic Press
scale as in the real world; making objects easy to
distinguish. In this, we have emphasized the
[11] Nielsen J. (1994), Guerrilla HCI: Using Discount Usability
Engineering to Penetrate the Intimidation Barrier, In Cost-
importance of existing guidelines in VE design. We
Justifying Usability, Bias R.G. and Mayhew D.J. (eds.), Academic
believe that this additional new set of criteria will help
novice designers to focus even more on ensuring
[12] Preece J., Rogers Y., Sharp H., Benyon D., Holland S. and Carey T.
usability when building new VEs. This investigation has
(1994), Human-Computer Interaction, Addison-Wesley
also laid some groundwork for future research on
[13] Preece J., Rogers Y. and Sharp H. (2002), Interaction Design:
usability issues in the design of VEs.
Beyond Human-Computer Interaction, John Wiley & Sons
[14] Vince (1995), Virtual Reality Systems, Addison-Wesley
Banerjee P. and Zetu D. (2001), Virtual Manufacturing, John
Wiley & Sons, Inc
The authors would like to thank Dr. Mohamed Benbrahim, Dr. Abas
Deol K.K. (1998), "Designing Virtual Environments for Usability",
Md. Said and Mr. Halabi Hasbullah for their contributions to the
URL: http://www.soi.city.ac.uk/ dj524/kully.html [accessed
early drafts of this paper. Also thank you to all evaluators who
15th November 2001]
took part in this study and all students taking the VR course duringJune 2001 and January 2002 sessions at UTP.
Deol K.K., Sutcliffe A., Maiden N. (1999), A design advice toolpresenting usability guidance for virtual environments. Paperpresented at the Workshop on User Centred Design andImplementation of Virtual Environments, York, UK, September
Suziah Sulaiman received her M.Phil. degree in Human Factors
in Computing from South Bank University, London, in 1998. Shejoined UTP as an IT/IS lecturer in July 1998. Among her research
Earnshaw R.A., Vince J.A., Jones H. (1995), Virtual Reality
interests include human computer interaction, virtual
Applications, Academic Press Limited
environment and haptic interactions. Currently, she is pursuingher Ph.D. studies at University College, London.
Gabbard J.L., Hix D., Edward Swan II, J. (1999), User-CentredDesign and Evaluation of Virtual Environments, IEEE Computer
Rohiza Ahmad received her M.S. degree in
Graphics and Applications
Computer Science from the University ofNorth Texas, U.S.A. in 1989. She is currently
Hix D., Edward Swan II J., Gabbard J.L., McGee M., Durbin J.,
a lecturer in the IT/IS program at UTP. Earlier,
King T. (1999), User-Centred Design and Evaluation of a Real-
she held lecturing positions at various
Time Battlefield Visualisation Virtual Environment, IEEE
higher institutions in Malaysia includingUniversiti Teknologi MARA, Kolej Yayasan
Hix D. and Hartson H.R. (1993), Developing User Interfaces:
Pelajaran MARA and the Sepang Institute of
Ensuring Usability Through Product and Process. New York:
Technology. Her main interests are in the
areas of educational technology, databasedevelopment and applied object-oriented
Jordan P.W. (1998), An Introduction to Usability, Taylor & Francis
PLATFORM VOLUME THREE NUMBER ONE JANUARY - JUNE 2003
Technology Cluster: INTELLIGENT SYSTEMS
Technology Platform: SMART CARD TECHNOLOGY
APPENDIX 1:
USABILITY PROBLEMS FOUND BY EVALUATORS
Hit No. Usability problems
Hit No. Usability problems
Objects are easy to distinguish but they do
Some objects could not be distinguished/
not look so real. Adding shadow to object
identified e.g. windows
can make them real.
Effects of lighting should be considered.
Can walk through walls!
Some chairs are too affected by lights,
Colour of the objects are too bright
Consider the use of white colour as "wall"
Shelves and cabinet are not easy to
Navigation is not easy. User does not know
the restricted area to navigate. There
No interactivity e.g. water running from
should be some indicators whereby the
user should be notified with. For example,
No indication to show that control has
the mouse cursors should be changed
when the user collides with the restricted
Objects are not easy to distinguish
area such as walls, trees
Objects in the water are not easy to
Some objects could not be distinguished.
For example, windows and doors. It should
be designed distinctively with respect to
Texture of the grass is not real
the textures and shapes
The movement of objects – some objects
A small movement of the mouse make the
in the application move rapidly for
entire environment move
example the canoe and the butterfly. These
User has less control on the mouse
objects seem to move at specific location
and continuously without having any
The landscape texture and design should
be improved in the sense that:
the design should be made with less
slopes and curves
make it more horizontal
the structure of the building must be
straight with the structure of the land
Scale in comparison with the real world.
Objects should be scaled comparatively
with the real world
VOLUME THREE NUMBER ONE JANUARY - JUNE 2003 PLATFORM
Technology Platform: SMART CARD TECHNOLOGY
Technology Cluster: INTELLIGENT SYSTEMS
Hit No. Usability problems
Hit No. Usability problems
Objects are not easily accessed and
Objects are easy to identify but difficult to
navigate, there are no doors or exit to
distinguish when focus to a particular
enter into the building
Difficult to walk around the environment
Separation between wall and floor is not
smooth (We can see the separated line)
The objects are not real (without any
No interactivity – switch button to on/off
the light does not work
Click on the mouse, sometimes makes
Selection of shape used – sometime it is
movement forward and backward
not suitable e.g. tree
Some objects are not easily distinguished
Combination between 2 objects are not
well organised (we can see the space, it is
Does not have precise navigation/
movement by the mouse – users cannot
navigate through the environment as
No interactivity. Should have a sun to show
the shadow of the object
Movements are too fast and unpredictable
Objects are not real – esp trees, canoes,
butterfly (not easily identified)
Does not show the motives of the
Hit No. Usability problems
The interactivities assigned to the objects
are not intuitive (e.g. click on the door –
Walls are not easy to distinguish
result: viewpoint by the lake instead of
Interactivity & significant of object is not
being inside the building)
clear (Pintu tak terbuka masa masuk the
The interactivity and significance of the
objects are not clearly understood
Do not portray real environment since you
Actions are not easily executed, effect of
can walk through the walls to access the
the actions was not as users intended to
Lights are not easy to distinguish too
Does not indicate the purpose of the
Difficult to control the system
Object (butterfly) scale is not appropriate
There are no clear identification whether
with the actual project scale
the objects have interactivities
Viewpoint (box) is quite irritating to my
Shades of objects (grass) is not realistic to
show that certain area is higher/lower
PLATFORM VOLUME THREE NUMBER ONE JANUARY - JUNE 2003
Technology Cluster: INTELLIGENT SYSTEMS
Technology Platform: SMART CARD TECHNOLOGY
Identification of Handwritten Digits
Ahmad Fadzil M. H. and Intan Mastura A. M.
Universiti Teknologi PETRONAS,
Bandar Seri Iskandar, 31750 Tronoh, Perak Darul Ridzuan, Malaysia.
The paper describes the development of an imaging scheme that recognizes and identifies handwritten digits
using a combination of image processing and neural network techniques. The scheme is designed to recognise
postcodes so that mail-sorting task can be performed with minimum human intervention, reducing labour cost
while increasing speed and accuracy. The imaging scheme is a combination of four processes, namely, data
acquisition of postcodes, image pre-processing, image segmentation and neural network-based digit
identification. Image pre-processing process is essential to obtain a binary image for segmentation. Here, noise
reduction, thresholding and greyscale to binary conversion operations are performed. Postcode image samples
are then segmented into separated digit regions. The segmentation process involving normalization and
morphological operations produces skeleton image of a digit. This is followed by a histogram operation that
produces vertical, horizontal, right- and left-diagonal histograms as input to a neural network identification
process. A database of 500 handwritten samples of postcodes is used in trials in which samples are used for
image processing, training of the neural network and as test data for performance measure. During segmentation
process, 90% of all samples are successfully segmented into single digits. In tests, it is found that the recognition
rate is 100% for training data (100 samples) and 80% for test data (50 samples).
Keywords: handwritten digit identification, mail sorting, morphological operation, histogram operation, neural
mail and facsimile, the use of mail letters is still
pervasive in society; this is especially true during
In an automated mail-sorting process, it is essential to
festive seasons. Without automation, the number of
recognize only the postcodes, which contain a series
sorters during peak times may need to be increased
of digits. Normally, the recipient's address is
to ensure no backlog in delivery.
handwritten. Therefore, pattern recognition and image
In recent years, Optical Character Recognition (OCR)
processing techniques are used to recognize the
scheme is used to automate sorting [1, 4, 5, 6]. However,
handwritten digits in postcodes.
OCR implementation is only applicable for characters
that are either computer generated or typewritten.
Traditionally, mail letters are sorted manually. Even
Therefore, handwritten digit recognition is still
though there are other ways to send mails such as e-
necessary for ordinary mails, cheques etc.
This paper was presented at the International Association of Science and Technology for Development (IASTED)
International Conference on Applied Informatics (AI 2003), Innsbruck, Austria, 10-13 Feb 2003.
VOLUME THREE NUMBER ONE JANUARY - JUNE 2003 PLATFORM
Technology Platform: SMART CARD TECHNOLOGY
Technology Cluster: INTELLIGENT SYSTEMS
The main goal of this work is to develop a writer-
independent recogniser that is essentially an imaging
scheme that recognizes and identifies handwritten
digits using a combination of image processing and
neural network techniques. As an application example,
the imaging scheme was developed on MATLAB to
identify handwritten postcodes in order to sort mail
2. SYSTEM DESIGN
Figure 1 shows the block diagram of handwritten digit
identification scheme. The scheme consists of four
processes, which are data acquisition, image pre-
processing, image segmentation and digit
The data acquisition is a scanning process in which a
sample is converted to a format that can be processed
by computer [2]. During the development of the
system, 500 handwritten postcode samples were
Figure 1: Block diagram of handwritten digit identification
collected, scanned at 200 dpi in grey scale and stored
imaging scheme.
in JPEG format. 300 samples are used as training and
Isodata algorithm is an iterative technique given by
the following equation:-
Image pre-processing mainly consists of two parts,
which are noise reduction and thresholding
operations. Noise reduction is necessary in order to
remove noise in a sample. There are three methods
The histogram is initially segmented into the
available in removing noise namely, linear filtering,
foreground and background portions. The sample
median filtering and adaptive filtering [3]. In linear
mean (mf, 0) of the grey values associated with the
filtering the value of output pixel is a linear
foreground pixels and the sample mean (mb, 0) of the
combination of the pixel values in the neighbourhood.
grey values associated with the background pixels are
Median filtering is similar with averaging filtering
computed. A new threshold value q1 is now computed
except that it takes the median of neighbourhood
as the average of these two sample means. The process
pixels rather than the mean to perform filtering
is repeated, based upon the new threshold, until the
operation. An adaptive filter depends on image
threshold value does not change any more.
variance. Masking operation has been added for
further noise reduction process. In this operation,
Triangle algorithm is as illustrated in Figure 2. A line is
sliding operation is performed to remove isolated pixel
constructed between the maximum of the histogram
(noise spikes) in image array.
at brightness bmax and the lowest value bmin of theimage. The distance d between the line and the
In the thresholding operation, threshold level is
histogram h(b) is computed for all values of b from b =
determined automatically [11]. Two histogram-
bmin to b = bmax. The brightness value bo where the
derived algorithms have been investigated namely,
distance between h[bo] and the line is maximal is the
isodata and triangle algorithms [4].
PLATFORM VOLUME THREE NUMBER ONE JANUARY - JUNE 2003
Technology Cluster: INTELLIGENT SYSTEMS
Technology Platform: SMART CARD TECHNOLOGY
of the weighted distances from the centroid points of
rows to the straight line. The estimated slant is the
slope of the straight line, which is given by
G(j) is the weight associated with the jth row and (xÕ,
Figure 2: The triangle algorithm
yÕ) is the centroid of the image.
threshold value, that is, q = bo. This technique is
The morphological operations namely ‘close', ‘thin' and
particularly effective when the object pixels produce
‘fill' functions are used to produce skeleton images of
a weak peak in the histogram. Both techniques were
the digits. The ‘close' function performs binary closure;
a dilation followed by an erosion. The ‘fill' function fills
isolated interior pixels (individual 0's that are
Image segmentation process consists of four parts,
surrounded by 1's). Function ‘thin' transforms objects
which are segmentation, normalization, slant
to lines. It removes pixels so that an object without
correction and morphological operations. The process
holes shrink to a minimally connected stroke, and
separates the postcode sample (number) into its
object with holes shrinks to connected ring halfway
individual digits. In order to perform this operation
between each hole and the outer boundary.
automatically, two methods can be used which are
either implicit or explicit methods [5]. However, the
In order to identify the segmented digits contained in
implicit method is used since it is accurate for digits
a postcode, a 3-layer feed-forward neural network has
that are well separated from each other. Generally, the
been implemented; 88 inputs, 14 neurons in hidden
component in image array is either 1 (white) or 0
layer and 10 outputs. Backpropagation algorithm is
(black) after thresholding is performed. By using these
used as a learning algorithm. The inputs for the neural
properties, a sliding operation is performed in order
network are based on histograms of zeros for a given
to separate the digits. Next, the normalization
digit sample. Specifically, the histograms for number
operation resizes the image to a uniform size that is,
of zeros for each row, column, left-diagonal and right-
15x15 pixels. The MATLAB ‘IMRESIZE' function uses
diagonal for a digit sample are determined. These four
bilinear interpolation to perform this process [3].
integer parameters are concatenated and used as an
Slant correction is necessary to eliminate any
input to the neural network for digits identification
distortion in the digit due to its slant characteristic. A
common example is the digit ‘1'. Methods using
invariant features and removing variations were
Before training, it is important to normalise the inputs
investigated [6]. By simply rotating the digit a slant
and targets so that they always fall within a specified
can be corrected. The angle of rotation is determined
range [7]. The MATLAB functions ‘premnmx' and
by the centroid of the digit. The centroid (xoj, j) for
‘postmnmx' can be used to scale inputs, targets and
every row of a given image is calculated. Hence, H row
outputs so that they fall in the range [-1, 1]. The
centroid points are obtained. H is the height of the
network output ‘an' will correspond to the normalized
image. Then linear regression is used to best fit a
targets ‘tn'. The un-normalized network output ‘a' is in
straight line of the form x = a + by minimizing the sum
the same units as the original targets ‘t'.
VOLUME THREE NUMBER ONE JANUARY - JUNE 2003 PLATFORM
Technology Platform: SMART CARD TECHNOLOGY
Technology Cluster: INTELLIGENT SYSTEMS
Before training session is performed, the
network weight and biases are initialized.
Once the network weights and biases have
been initialized, the network is ready for
training. The training process requires a
set of examples of proper network
behaviour - network inputs ‘p' and target
outputs ‘t'. During training the weights and
biases of the network are iteratively
adjusted to minimize the network
performance function ‘net.performFcn'.
The performance function for this project
is sum square error (SSE). SSE is the sum-
squared error between the networks
outputs ‘a' and the target outputs ‘t'. After
Figure 5: Sample undergoing segmentation and morphological operation
training, simulation with test data is
Some of samples are clearly written while others are
not. This is due to different types of media used for
3. RESULTS & DISCUSSION
writing such as the pen or pencil. Blurry images might
be produced during the scanning process.
Figure 3 shows some of the samples being collected
for this project. All samples are in greyscale format.
Figure 4 shows a postcode sample undergoing image
Each sample has a different handwriting pattern,
background level and signal-to-noise level.
After performing segmentation and normalization,
90% of the sample set is effectively segmented and
resized. The information contained in segmented
digits is preserved except for number ‘1'. It is observed
that this type of digit experiences distortion especially
to its slant characteristic. Therefore, slant correction is
necessary to overcome this problem. It was found that
the remaining 10% cannot be segmented properly
because the digits are connected to each other. Figure
5 (last page) shows a sample undergoing the
Figure 3: Samples collected in data acquisition.
segmentation and normalization operations.
Figure 4: Image preprocessing on a postcode sample.
PLATFORM VOLUME THREE NUMBER ONE JANUARY - JUNE 2003
Technology Cluster: INTELLIGENT SYSTEMS
Technology Platform: SMART CARD TECHNOLOGY
Figure 6 shows the image after slant correction is
performed on segmented digits. It can be seen that
there is an improvement in the slant characteristic of
digit ‘1'.
Figure 7: Concatenated histograms for a sample digit ‘3'.
The ‘traingdx' function has been used as a training
algorithm. It is a network training function that
updates weight and bias values according to gradient
descent momentum and an adaptive learning rate.
This network uses hyperbolic tangent sigmoid transfer
function as its neuron model. The weight value and
bias value have been initialized to 0.01. Sum square
error has been used as performance function. The
Figure 6: Digit sample undergoing slant correction.
network will be continuously trained until sum-
squared error goal of 0.01 is reached. The momentum
Figure 7 shows the example of integer sequence being
constant has been set to 0.95 and the maximum epoch
fed to neural network. For horizontal and vertical
to 5000. After the network has been initialized, the
histogram, each contains 15-integer sequences. The
training process is performed.
right diagonal and left diagonal, each contains 29-
integer sequences. Hence, the neural network will have
Figure 8 shows the error being reduced during training
88 inputs. This integer sequence is the digit
of 100 samples. In this particular case, the process is
representation that will be fed into the neural network.
stopped at 93 epochs when error is less than 0.01.
Figure 8: Training performance for digit ‘0'
VOLUME THREE NUMBER ONE JANUARY - JUNE 2003 PLATFORM
Technology Platform: SMART CARD TECHNOLOGY
Technology Cluster: INTELLIGENT SYSTEMS
Network simulation is conducted and it is found that
100% of training sample data is recognized. For the
El-Yacoubi A, Sabourin R & Suen C.Y., An HMM-Based Approach
test data, a recognition rate of 80% is achieved.
For Offline Unconstraint Handwritten Word Modelling AndRecognition, IEEE Transactions On Pattern Analysis And Machine
Awcock G.W & Thomas R, Applied Image Processing (London:
In the handwritten digit identification imaging
scheme, 4 processes have been used namely, data
MATLAB Image Processing Toolbox, (The MathWorks, Inc., 1997).
acquisition, image pre-processing, image
segmentation and digits identification.
Wakahara T & Odaka K, Adaptive Normalization of HandwrittenCharacters Using Global/Local Affine Transformation, IEEETransaction On Pattern Analysis And Machine Intelligence, 1999.
Using segmentation, normalization, slant correction
Lee J.S & Ching Y.S, Analysis of Class Separation & Combination
and morphological operations, 90 % of samples are
of Class-Dependent Features for Handwriting Recognition.
successfully segmented. This corresponds to digits
IEEE transaction on Pattern Analysis and Machine Intelligence.
that are clearly separated from each other. Further
Jinhai C. and Zhi Q.L., Integration of Structural and Statistical
investigation on segmentation process is necessary for
Information for Unconstrained Handwritten Numeral
digits that are not well separated from each other.
Recognition IEEE transaction on Pattern Analysis and MachineIntelligence, 1999.
Normalization is performed after segmentation
Kartalopoulos S.V, Understanding Neural Networks and Fuzzy
process in order to obtain a uniform size of segmented
Logic: Basic Concepts And Applications, (IEEE Press, 1996).
images. Slant correction is performed to eliminate any
MATLAB Neural Network Toolbox (The MathWorks, Inc., 1998).
distortion in segmented images in terms of its slant
characteristic. In morphological operation, the
Solihin Y and Leedham C.G. 1999, Integral Ratio: A New Class ofGlobal Thresholding Techniques for Handwriting Images, IEEE
segmented image is being thinned in order to produce
Transaction On Pattern Analysis And Machine Intelligence.
skeleton image.
For digit identification, histogramming and neural
Ahmad Fadzil M.H. graduated in 1983 from
the University of Essex, UK with an Honours
network techniques are performed. Histogramming
B.Sc. degree in Electronic Engineering. He
is used in order to determine inputs to an 88-14-10
completed his M.Sc. in Telematics in 1984and Ph.D. in Image Processing in 1991 at the
neural network. The inputs are normalised to be within
same university.
the range [-1,1]. It is found that the recognition rate
for test data is 80%. Improvement in recognition rates
He has been a Lecturer in Signal Processingand Researcher in Image Processing at
can be achieved by using more samples during
Universiti Sains Malaysia (USM) since 1984. Between 1988-91, he
served at University of Essex, UK initially as a Senior ResearchOfficer in Image Processing and subsequently as a Lecturer. Onhis return to Malaysia, he was made Dean of the School of Electrical
The imaging scheme can be used in various
& Electronic Engineering at USM from 1992-96. In 1997, he
applications such as mail sorting via postcodes,
became the founding Dean of the Engineering Faculty at thenewly established UTP. He became the Director of Academic
handwritten cheques etc.
Studies at the university in 1999. Currently, he is the Director ofPostgraduate Studies, a position he assumed in 2003.
Dr Ahmad Fadzil is a Fellow and a Council Member of theInstitution of Engineers, Malaysia. He is a registered ProfessionalEngineer with the Board of Engineers, Malaysia. He is also aMember of the IEEE, USA. His research interests include imagecompression and image processing applications in telemedicineand artificial intelligence.
PLATFORM VOLUME THREE NUMBER ONE JANUARY - JUNE 2003
Technology Cluster: INTELLIGENT SYSTEMS
Technology Platform: INTELLIGENT BUILDING
Application Of Dry Walls In Residential Houses:
A Buildable System
Noor Amila W.A.Z., Nasir Shafiq, Ibrahim Kamaruddin
Universiti Teknologi PETRONAS,
Bandar Seri Iskandar, 31750 Tronoh, Perak Darul Ridzuan, Malaysia.
Flexibility, durability, cost-effectiveness, and buildability are the major criteria in the value based construction
and design of residential houses, which is the current approach emerging in the industry. Flexibility in the
construction of wall system is an important concern to homeowners in relocating the walls to suit their changing
needs. Dry walls system, which are finished with paint or tiles, are easy to install, durable, well insulated and
would be the preferred option in residential development to meet the requirements of value based design. A
case study based on seven types of dry walls system is presented. These walls can be removed and replaced
with relative ease and minimal mess compared to the conventional brick wall, and hence most buildable system,
which may add value to the construction of mass scale housing development.
Keyword: dry walls, buildable design, partition wall, fire rating, sound insulation
home environment. Details of installation,
concealment of services and joint treatment are given
Under the current trends in the global economy the
major emphasis in the design of the system.
construction industry is exploring new ways and
methods to meet the challenges in the industry. Non-
This paper presents a case study of a buildable
structural walling in a building structure has a relatively
showhome constructed at the Construction Industry
larger quantum as compared to the other building
Training Institute (CITI), Building and Construction
elements. Traditionally, brick masonry walls have been
Authority (BCA), Singapore, as part of its ongoing
extensively used in the construction industry. Despite
efforts to promote wider use of buildable designs for
the advantages of the brick masonry wall system, there
are some constraints in its application such as speed
of construction, buildability, cost and appearance.
2. CITI/BCA DRY WALLS SYSTEMS: AN OVERVIEW
Currently, the dry walls system is becoming popular
In July 2000, BCA opened its buildable showhome at
as an alternative to the brick masonry wall system in
Construction Industry Training Institute (CITI). The
many parts of the world such as USA, UK and
buildable showhome was put together using seven
Singapore. The design of the dry walls system is aimed
different dry wall systems (BCA, 2000), to demonstrate
at demonstrating the actual condition within a typical
the actual condition within a typical home
The paper was presented at the Inaugural Management in Construction Researchers Association (MICRA) Seminar, Penang, October 2002.
VOLUME THREE NUMBER ONE JANUARY - JUNE 2003 PLATFORM
Technology Platform: INTELLIGENT BUILDING
Technology Cluster: INTELLIGENT SYSTEMS
environment. These walls are easier and faster to build
dB. This is comparable to single layer brick wall. These
than the traditional brick wall. They have better quality
factory produced prefabricated walls also ensure good
and finishes as they are being manufactured under
quality and consistency.
controlled factory conditions, .
To demonstrate the performance of these dry walls,
These walls were finished with paint or tiles according
each wall was put through a series of 11 tests at the
to the design of a typical residential unit. Unlike
Singapore Productivity and Standard Board (PSB)
conventional brick wall, cement and sand plaster were
Testing Centre in accordance to BS5234: Partition Wall
not needed for these walls. The installation process
Part 2: 1992. These tests included stiffness test, impact
was simple, easy and required less labour. These dry
test, crowd pressure test, door slam test and anchorage
walls systems are the preferred option in residential
test. In addition, the walls also passed the non-
development in advanced countries such as USA,
combustibility test in accordance to BS476: Part 4: 1970,
Japan, Australia, and many parts of the European
and fire rating test in accordance to BS476: Part 20:
countries (BCA, 2000).
An important concern to the homeowner is the
To set the minimum quality standard for greater
flexibility in relocating the walls. Dry walls have the
confidence and acceptance of dry walls, BCA and PSB
advantage of giving homeowners flexibility to change
jointly published the new Singapore Standard: Code
the layout of their homes to meet their changing
of Practice for Partition Wall (2000), which is based on
needs. The dry walls used in the showhome can be
BS5234. The same may be considered as the reference
removed and placed with relative ease and minimal
code for the construction industry practitioners in
mess compared to conventional brick wall.
Strength, sound insulation and quality are other major
2.1 Dry Wall Systems: Materials And Dimensions
considerations to homebuyers. The walls were also
subjected to heavy loading by cabinets. The sound
Table 1 shows the material type and physical
insulation of these dry walls is in the range of 30 - 40
properties of seven different types of dry walls as
presented in the BCA showhome.
Table 1: Material used and physical properties of dry wall systems
(kg/m2) for Productivity
3m height m2/man-day
Cement, cellulose fibre and limestone and
chemical admixture
Cement, sand and lime
Vermiculite sandwiched between
Gypsum with recycled cellulose fibre
Gypsum with rock dust
Sand, cement and foaming agent
PLATFORM VOLUME THREE NUMBER ONE JANUARY - JUNE 2003
Technology Cluster: INTELLIGENT SYSTEMS
Technology Platform: INTELLIGENT BUILDING
2.2 Tests Procedure And Resuts
iv. Door Slam (DS): a 35 or 60 kg door is slammed
with a force of 15 kg. No damage should be
Table 2 shows the results of different tests on dry wall
observed. The test simulates a door slam caused
systems conducted at PSB Mechanical Test Centres.
by wind, or people closing the door energetically,
The procedures and objectives of different tests as
which is characterized by a perfect function of the
conducted are as follows:
Stiffness: the wall is subjected to a horizontal force
Crowd Pressure (CP): the wall is subjected to a
of 500 N applied at a height of 1.5 m. Maximum
sustained load transmitted through 2.5 m
deflection and residual deformation should be
horizontal beam, at a height of 1.5 m. No damage
within specified limits as specified by the relevant
or collapse should be observed. The test simulates
codes. The test simulates bending caused by
loads caused by a crowd leaning against a wall.
people leaning on the wall.
vi. Lightweight Anchorage (LA): the wall is subjected
Small Hard Body Impact (SHBI): the wall is
to a force transmitted through a bracket, with a
subjected to a 3 kg impactor with a 50 mm
shim plate positioned between the bracket and
diameter steel sphere head, swung
the surface of the wall. No damage should be
perpendicularly against the wall. The extent of
observed and shim plate should not be dislodged.
damage is assessed. The test simulates impact
The test simulates loadings from lightweight
caused by sharp or pointed objects, such as trolleys
fixtures such as wall pictures, clothing hooks and
and wheelchairs.
basic wall shelving.
iii. Large Soft Body Impact (LSBI): the wall is subjected
vii. Heavyweight Anchorage (HA): The test simulates
to a 50 kg load in the form of spheroconical bag
heavyweight fittings such as wall cupboards,
swung against the wall. Permanent deformation
washbasins and water closets.
should not exceed 2 mm, with no structural
damage. The test simulates the impact caused by
Table 3 shows the data obtained from different wall
people falling against the wall.
manufacturers in order to describe the functional
requirement of different wall system.
Table 2: Wall Comparison by Strength and Robustness Test in Accordance with BS5234: Part 2
USG Fiberock
Severe Duty (SD) is a measure of its suitability for use in areas prone to vandalism and rough use.
Minimum passing value for Heavyweight Anchorage is 1500 N.
VOLUME THREE NUMBER ONE JANUARY - JUNE 2003 PLATFORM
Technology Platform: INTELLIGENT BUILDING
Technology Cluster: INTELLIGENT SYSTEMS
Table 3: Wall Comparison by Functional Requirements
USG Fiberock
Fire Rating (minutes)
BS476: Part 20: 1987
Sound Insulation (dB)
BS2750: Part 3: 1980
Surface Appearance
BS5234: Part 2: 1992
BS476: Part 4: 1970
3. DISCUSSION OF RESULTS
Table 3 shows that the fire rating properties of brick
wall is superior to that of the dry walls system except
As shown in Table 1, the unit weights of dry walls are
in the case of CSR ALC Panel. However, all dry walls
approximately one half to one twelfth that of the brick
systems meet the fire rating requirement of the British
wall, which will result in a lighter design of the
Standard for residential building, which is specified as
structural system.
The productivity of dry walls system ranges from four
The sound insulation property of brick wall system is
to six times of that of the conventional brick wall
superior to the dry walls system except for Styroplast.
system. This results in a faster speed of construction,
However, the permissible sound insulation
which makes it more cost effective.
requirement for residential building is below 60 dB,
which qualifies all the dry walls systems.
Based on a number of actual structural design and
construction practices by the authors, it is obtained
As mentioned earlier, the expected quality of the dry
that the net reinforcement content of the overall
walls can be more superior to the brick wall system
structural system has been reduced between 20 – 25%
since it is manufactured within factory-controlled
by maintaining the same sizes of structural
components for all the above dry walls systems.
Similarly, the construction time has been reduced by
up to 40% because of high productivity rate and faster
construction and erection time.
The major conclusions made from of this study are as
Table 2 shows that the performance of the dry walls
system is comparable to the brick wall system with
Dry walls systems are easier and faster to build and
respect to all aspects as defined by the British Code
more cost effective than the traditional brick wall
BS5234: Part 2. Only the CSR ALC shows a fair
performance for Heavyweight Anchorage with the
value of 1287 N.
Dry walls systems offer more flexibility in
relocating the walls to suit the requirements as
desired by the homeowners.
PLATFORM VOLUME THREE NUMBER ONE JANUARY - JUNE 2003
Technology Cluster: INTELLIGENT SYSTEMS
Technology Platform: INTELLIGENT BUILDING
iii. Dry walls systems ensure higher quality, better
Noor Amila W.A.Z. earned her first degree
in Housing, Building and Planning (Building
appearance and lighter weight than the traditional
Engineering) from Universiti Sains Malaysia,
brick wall system.
Penang, in 1998. Later she pursued her M.Sc.
in Building Technology from the sameuniversity in 1999. She began her career as
iv. Certain dry walls systems considered in the study
a lecturer in L&G Twintech Institute of
are functionally better than the brick wall system.
Technology in 1999. Currently, she is alecturer in the Civil Engineering program at
Nasir Shafiq earned his first degree in Civil
Engineering from NED University, Karachi,
BCA Buildability Series, 2000, The Buildable Showhome.
Pakistan, in 1982. He did his M.Eng. inStructural Engineering and Construction at
British Code BS5234: Partition, Part 2: Code of Practice for
the Asian Institute of Technology (AIT ),
Testing of Partition Walls.
Bangkok, Thailand, in 1987 and Ph.D. inConcrete Durability at the University of
British Code BS476: Part 4: Code of Practice for Non-
Leeds, UK, in 1999. He worked as structural
Combustibility Test
engineer in Civil Aviation Authority, Pakistan
for 3 years. He served as Assistant Professor in the Civil Engineering
British Code BS476, Part 20: Code of Practice for Fire rating
Department, NED University from 1987 to 1994. In 1994, he joined
Meinhard (Singapore) Pte. Ltd. as Senior Structural Engineer. In2000, he worked as Principal Structural Engineer and HOD
Singapore Standard, 2001: Code of Practice for Partition Wall
(Engineering) for 2 years at Hainal-Konyi (Singapore) Pte. Ltd.
Currently, he is an Associate Professor in the Civil Engineeringprogram at UTP.
Ibrahim Kamaruddin is an Associate
Professor in the Civil Engineering program
of UTP. He obtained his Ph.D. from the
University of Leeds, UK, on work pertaining
to the use of synthetic fibres in road
construction materials. His research interest
is in construction materials focussing on
highway building materials and bituminous
composites. He was a lecturer at Universiti
Teknologi MARA prior to joining UTP. He is currently the Directorof Undergraduate Studies at UTP.
VOLUME THREE NUMBER ONE JANUARY - JUNE 2003 PLATFORM
OTHER RESEARCH AREAS
Study On The Capillary Performance In Wire Bonding
Operation For Integrated Circuit (IC)
Norani M. Mohamed1, Sivabalan Sasthiryar2
1Universiti Teknologi Petronas,
Bandar Seri Iskandar, 31750 Tronoh, Perak Darul Ridzuan, Malaysia.
2 Agilent Technologies (Malaysia),
Bayan Lepas Free Industrial Zone, 11900 Penang, Malaysia.
Wire bonding of gold wire in microelectronic devices has remained the oldest and most commonly used method
in interconnecting technology. These interconnections are made possible by using the ceramic capillary. Here,
investigations were conducted on the performance of the capillaries and the quality of wire bonding produced.
The study focused on two different types of capillaries namely, PIKO, which is made of zirconia composites with
special additives, and NORM, which is made of fine grain alumina ceramic.
Using a spool of 1.2 mils gold wire (purity 99.9968%), the wire bonding process was carried out with a wirebonder
machine, with thermosonic power. The bonded integrated circuits (IC) then underwent a series of stress tests.
Subsequently, the capillaries used were inspected visually by electron microscopy and the materials characteristics
were analysed.
Results have revealed that PIKO showed better performance and reliability (up to 2500 x temperature cycles)
than NORM. In fact, PIKO capillary produced a much better and longer lasting wire bonding (<300,00 bonds)
without compromising the quality of the microelectronic devices. In addition, PIKO capillary only required lower
operating power and force at wire bonding machine than NORM which itself is utility cost saving for
semiconductor industries. Moreover PIKO purchase cost is much lower than NORM.
Keywords: solid state physics, semiconductor, wire bonding, integrated circuit
surface and gold wire. With the ultrasonic bonding
technique, a bond is made directly between the wire
Wire bonding has been used extensively in
and the bond pad. Thermosonic tailless ball and stitch
semiconductor products to connect the internal
bonding [1] is a widely used assembly technique. A
semiconductor die to the external leads. Wire bonding
fine gold wire ranging from 18 mm to 33 mm in
techniques include thermosonic or gold ball bonding
diameter is fed down through the capillary. Normally
and ultrasonic or aluminum wedge bonding. In
wire bonding capillaries are formed out of ultrapure
thermosonic bonding, gold ball is formed at the end
99.99% Al2O3 powder for standard non-pitch
of the wire and a bond is made between the bond
application. But with the advancement of ceramic
This paper was presented at the XVIII Regional Conference on Solid State Science & Technology, Pulau Langkawi, 7-9 September 2001.
It is also published in the Journal of Solid State Science & Technology 2001.
PLATFORM VOLUME THREE NUMBER ONE JANUARY - JUNE 2003
OTHER RESEARCH AREAS
material technology, a new generation of material has
were used to describe the reliability and quality of wire
emerged. This choice of capillary material for ultra-
bonding on IC packages. These tests are in accordance
fine pitch application comes in the form of zirconia
with the standard of the Joint Electronic Development
composites with special additives. The sintered
Engineering Council (JEDEC). The first test, termed
capillary has a well-defined surface finish with all its
temperature cycle (TMCL) or thermal shock, is where
specified dimensions moulded by the state-of-the-art
IC packages are subjected to a number of cycles at
ceramic injection moulding technology [2]. The
temperatures of –25 °C to 85 °C within 10 minutes
continuous growth in the development of new
interchange. The second test, known as wet high
packaging technology has posed a greater challenge
temperature operating live (WHTOL), is a test that is
for wire bonding tools.
conducted at 25 °C in 85% relative humidity whileoperating at 5 V current. The third test, referred to by
the term high temperature operating live (HTOL) is
similar to the second test with the exception of being
in a dry environment. For each test, 33 IC units were
To determine the bonding quality and performance,
pull test and ball shear were conducted with the use
Figure 1: Tailless ball and stitch bonding
of the 2400A Dage Multitester. All the completed wire
bonded IC packages were observed under SEM
The new packaging technology specifically for ultra-
Cambridge Model 360 Stereoscan in order to
fine pitch application has made Ball Grid Array (BGA)
determine the bonding quality and density
package the choice for high pin count devices besides
characteristics. Examination of the capillaries by SEM
the commonly used Quad Flat Pack package. This has
was carried out at PECO Industries Pte. Limited, Korea.
resulted in further shrinkage of the die and
consequently, a much closer pad pitch of less than 100
mm. Smaller ball size of less than 55 mm and
optimisation of the stitch length are major
Results of the stress tests can be illustrated in three
consideration in the design of the capillary.
tables encompassing TMCL test, WHTOL test and HTOL
test. The functional reliability and capability of the
The work here focused on the performance and quality
device bonded by PIKO and NORM capillary can be
of two types of capillaries, PIKO and NORM. This also
determined by thermal cycle (TMCL) test. Table 1
includes the study on the quality of the wire bonded
shows that PIKO and NORM have the same wire
on integrated circuit (IC) packages.
bonding performance without failure until 1000 x
(cycles). Beyond 1000 x, the performance of NORM
started to fail with an increase from 2 to 7 IC units at
2500 x. Unlike NORM, failures with PIKO are few and
Two types of capillaries, PIKO and NORM were used in
consistent, exhibited by one or two units only. At the
the wire bonding process. A spool of 1.2 mils gold wire
final stage of 2500 x, a failure of 7 IC units of NORM
(purity of 99.9968%) was used for bonding the
gives 21.2% whereas PIKO shows only 3.0% failure.
integrated circuit package, in this case a photosensor
Failure analysis shows that all PIKO failures are related
operating at 5.0 V. Wire bonding was carried out with
to disintegration of dies rather than the bonding
the wire bonder machine, Kaijo FB131 with
performance, which is due to the capillary used. On
thermosonic power. After the ICs are bonded, they will
the other hand, NORM failure is a typical wire bonding
undergo a series of stress test. Three major stress tests
failure where the wire over stress and snaps resulting
VOLUME THREE NUMBER ONE JANUARY - JUNE 2003 PLATFORM
OTHER RESEARCH AREAS
disintegration of die, giving percentage failure of only
3.0%, far below NORM capillary. As for WHTOL test, no
significant difference can be observed between PIKO
and NORM (Table 2). Both PIKO and NORM give 3%
failure after 1000 hrs and the cause was not due to the
bonding but to disintegration of the die. With HTOL,
where the test was done without the humidity, IC units
Figure 2: Wire bonding exhibiting open short
bonded by both capillaries have not shown any sign
of failure at any stage of the test.
Section B of Table 1 describes the result of TMCL test
at 1,000 cycles with varying number of bonds starting
The strength of wire bonding will be reflected by the
from 0 to 900,000. The first failure with NORM is
bondability measurement. Such measurement were
observed at 300,000 bonds. As the number of bonds
obtained by conducting pull test on 12 wires taken
increased to 900,000, the number of units that failed
from each photosensor package that have undergone
increased tremendously resulting in 78.7% failure.
every stress test. Figure 3 shows a graph of the pull
Almost all failures of NORM is in the form of open short
test for 12 wires bonded by PIKO and NORM that had
which is related to the bonding. IC units of PIKO did
undergone HTOL test. Minimum requirement of pull
not show any failure until the number of bonds
test is 8.0 g. Both showed similar strength distribution
reached 700,000 where one unit failed and then the
at 0 to 168 hrs. Apparently after 1000 hrs., PIKO showed
continued to 900,000 bonds without further failure. It
higher strength with an average of 12.4 g as compared
is observed that the failure of PIKO is associated with
to NORM with 8.54 g.
(a) PIKO bonded wires
(a) PIKO bonded wires
(b) NORM bonded wires
(b) NORM bonded wires
Figure 3: Graph of the pull test conducted on (a) NORM and (b)
PIKO bonded wires after being subjected to HTOL stress test at
Figure 4: Graph of the pull test conducted on (a) PIKO (b) NORM
different time duration.
bonded wires after being subjected to WHTOL stress test.
PLATFORM VOLUME THREE NUMBER ONE JANUARY - JUNE 2003
OTHER RESEARCH AREAS
As depicted in Figure 4, both PIKO and NORM exhibited
category would be for the highest cycles of 1000 and
same strength after undergoing WHTOL test for 0 and
2500 giving the lowest mean value of 6.0 g.
168 hrs. The NORM wires, which had been subjected
to the test for 500 hrs, had shown a decrease in
The result of the pull test conducted on 12 wires
strength giving a value of 9.68 g as compared to 14.1
bonded by the capillaries that have been used for a
g for PIKO. The strength of the NORM 500 hrs.-wires
varying number of bonds is illustrated in Figure 6. It
continuously degraded to 7.51 g whereas PIKO
was found that the performance of PIKO wires is
remained stronger at 12.0 g.
excellent as at 900 K (900,000 bonds), the pull strength
is still above 8.0 g. Generally, PIKO wires shows
In the pull test for the wires subjected to TMCL test
consistent strength after bonding of 50 K to 900 K,
(Figure 5), PIKO wires showed satisfactory and
giving average value of 14.0 g. As shown in Figure 6b,
consistent strength values even at 1500 x. In contrast,
the values for NORM wires can be clearly classified into
NORM wires seemed to be divided into three
3 groups. The first would consist of the lower range of
categories. The first class would be for 0 x (no cycles)
bond, 50 K to 150 K, with average value of 15.0 g. The
and 200 cycles with the mean value of 14.0 g. The
second is the middle range, 300 K to 500 K, where the
second class involved the 500 cycles and 1000 cycles
average value remains above 8.0 g. A very much lower
with the mean value not exceeding 8.0 g. The last
value, between 7.0 g and 6.0 g can be obtained for the
highest bond range, 600 K to 900 K.
(a) PIKO bonded wires
(a) PIKO bonded wires
(b) NORM bonded wires
(b) NORM bonded wires
Figure 5: Graph of the pull test conducted on (a) PIKO (b) NORM
Figure 6: Graph of the pull test conducted on (a) PIKO (b) NORM
bonded wires after being subjected to TMCL stress test.
bonded wires after being used for varying number of bonding.
VOLUME THREE NUMBER ONE JANUARY - JUNE 2003 PLATFORM
OTHER RESEARCH AREAS
Shear test was conducted with the purpose of testing
giving average values above 40.0 g. However, the shear
the formation of the gold ball at the die pad, forming
strength tends to decay for the ones subjected to 500
intermetallic connection during the device operational
hrs. and 1000 hrs of WHTOL. PIKO balls have also shown
period Minimum requirement of 40.0 g should be
some decrease but the rate is very much smaller.
observed in order to qualify the performance bonding.
On the whole, gold balls bonded by PIKO showed
Shear test for TMCL (Figure 9) shows a significant
better bonding strength compared to NORMs in all 3
difference between the strength of PIKO and NORM
stress tests. Average shear strength for PIKO bonded
balls. PIKO balls can still maintain shear strength above
wires after undergoing HTOL test was found to exceed
45.0 g for 1000 cycles TMCL, and this is considered to
55 g (Figure 7(a)). In the case of NORM bonded wires,
be a very good result. At the early stage with no cycle
the strength value decreases linearly until for a
in TMCL test, NORM balls seem to exceed 65.0 g, similar
duration of 1000 hrs, the value dipped below 35 g
to PIKO. However, this value reduced by 50% at the
(Figure 7(b)).
later stage of 1000 cycles with the average above
Figure 8 shows the graphs of shear test for PIKO and
NORM balls, which had undergone WHTOL tests. Balls
Surface analysis conducted by SEM showed that the
bonded by PIKO and subjected to <100hrs. WHTOL
stitch and ball bonding by PIKO are satisfactory even
tests seem able to maintain the bonding strength
though they have undergone TMCL stress test at 2500
above 35.0 g. NORM balls showed good shear strength
cycles. At this stage NORM bonding has failed resulting
only after undergoing 0 hr. and 168 hrs. of WHTOL,
in an "open short", as shown in Figure 2. PIKO and
(a) PIKO bonded wires
(a) PIKO bonded wire
(b) NORM bonded wires
(b) NORM bonded wire
Figure 7: Graph of the shear test conducted on the (a) PIKO (b)
Figure 8: Graph of the shear test conducted on the (a) PIKO (b)
NORM bonded wires after being subjected to HTOL stress test.
NORM bonded wires after being subjected to WHTOL stress test.
PLATFORM VOLUME THREE NUMBER ONE JANUARY - JUNE 2003
OTHER RESEARCH AREAS
(a) NORM capillary
(a) PIKO bonded wires
(b) PIKO capillary
Figure 10: Outer and inner tip of (a) NORM and (b) PIKO
capillary after 900,000 bonds.
(b) NORM bonded wires
Figure 9: Graph of the shear test conducted on the (a) PIKO (b)
Figure 11: A line marked by the diamond tip (140g) on the (a)
NORM bonded wires after being subjected to HTOL stress test.
PIKO and (b) NORM capillary.
Table 1: Summary of experiments and observations
ZrO2 with additives [3]
Translucent White [3]
Density: 5.5 g/cm3 [4]
Density: 3.99g/cm3 [4]
Average Grain size: <0.25 mm
Average Grain size: <2 mm
Finishing: Fine Matte & Polished
Finishing: Fine Matte & Polished
Bending strength: 2400 Mpa
Bending strength: 850 Mpa
No damage exhibited by the die
Certain level of die damage occurs
No damage of capillary
Damage to capillary through gold build-up
Better reliability up to 2500 cycles TMCL
Reliability degrades after 1000 cycles TMCL
Consistence bonding quality until 600,000 bonds
Consistence bonding quality until 30,000 bonds
Low power and force used during operation
Higher power and force needed
Lifetime is much longer (>300,000 bonds)
Lifetime is shorter (<300,000 bonds)
VOLUME THREE NUMBER ONE JANUARY - JUNE 2003 PLATFORM
OTHER RESEARCH AREAS
NORM capillary have also been analyzed in order to
detect any changes on the surface. As evidence in
Figure 10, there is little change in view of cracks and
The authors would like to express their gratitude for
wear at the tip of PIKO capillary after 900,000 bonds
the technical support given by Agilent Technologies
as compared to NORM capillary. Thus this proves that
Malaysia, PECO Industries Pte. Ltd., Korea and Universiti
the original condition of PIKO capillary is longer lasting,
Sains Malaysia. Financial support from Universiti
maintaining good bondability performance. The
Teknologi PETRONAS is gratefully appreciated.
comparatively even and straight groove on the surface
(shown in Figure 11), displayed by PIKO infers higher
density and finer grain material.
Babur, N., (1996), "Total Energy Management Strategies forWafer Fabs.", Acom Engineering and Consulting, p152.
Table 1 shows a summary of the experiments and
observations, which indicates the superiority of PIKO
Smith, W.M., (1998), EPRI: Presentation on EPRI's ElectronicsIndustry Target Semiconductors /Tools, p19.
and areas needed to be improved by NORM capillary.
Report: "Capillary Technology" PECO KOREA Pte. Ltd.
Busch, J., (1998), "Energy Efficiency Potential in SemiconductorCleanrooms" Lawrence Berkeley National Laboratory, p163.
The results have shown that PIKO capability and
performance with respect to material strength and
Norani M. Mohamed graduated from the
wire-bonding quality is far better compared to NORM
University of Essex with an Honours B.Sc.
degree in Physics in 1983. She then
capillary. It was found that PIKO capillary is much more
continued her M.Sc. in Lasers & Their
reliable and is suitable for high-speed wire bonding
Application in 1984 at the same university.
process, which require fine pitch bonding quality.
Later in 1992, she completed her Ph.D. inPhysics, specifically in the area of UltrathinLangmuir-Blodgett Films.
She began her career as a lecturer at the School of Physics, USMin 1984, teaching courses such as Modern Physics and Lasers. In1986, she moved to the School of Materials & Mineral ResourcesEngineering at the new engineering branch campus of USM inIpoh. Here, she taught courses such as engineering materials,semiconductor physics, semiconductor technology,semiconductor & opto-electronic devices. During her tenure, shewas appointed the Deputy Dean for the distance learningprogram, and then as the Program Chairman for PostgraduateStudies and finally as the Program Chairman of AdvancedMaterials. She joined UTP in May 2001 and is currently theassociate professor at the Electrical & Electronic Engineeringprogram. Her research interests include thin film technology andsemiconductor processing (both hardware and software). She isnow actively involved in developing nanotechnology research atUTP.
PLATFORM VOLUME THREE NUMBER ONE JANUARY - JUNE 2003
OTHER RESEARCH AREAS
The Training of Scientists and Engineers
in the 21st Century: a Social Science Perspective
Victor T. King
Department of Politics and International Studies
University of Hull, United Kingdom.
There are complex relationships between scientific enquiry and practice and the societies and cultures in which
they are embedded. Social relations and cultural values shape and are shaped by scientific and technological
innovations. For this reason university programs for scientists and engineers in the 21st century must also devote
some attention to the social sciences and humanities. The well rounded graduate is someone who is equipped
not only with knowledge of his or her subject but also with a social and cultural imagination and a sensitivity to
the ways in which scientific endeavor and technological innovations are perceived, evaluated and received into
society. The internationalization of scientific knowledge and the global processes affecting work, business and
social life also suggest that scientists and engineers require an increasingly international outlook and the ability
to operate effectively in cross-cultural situations.
Keywords: science, technology, society, culture, training
technology on the one hand and the social sciences,
and for that matter the humanities, on the other are, I
What might an anthropologist usefully say to scientists
think, becoming closer as we recognize the
and engineers? My remarks are based on the premise
importance of interdisciplinary and cross-disciplinary
that social scientists do have theoretical interests in
approaches to understanding the ever-increasing
the processes of scientific and technological practice
complexity of the environment within which we live.
and innovation, on the ways in which science is
What has become of vital concern for us all in a world
organized as a profession, and on the culture of
in which there has been dramatic advances in genetics,
scientific endeavor. We must recognize that scientific
innovative health technologies, informatics and
activity is embedded in particular socio-cultural and
robotics is how these innovations are received on the
economic contexts, that all of us have perspectives on
ground and the moral and social implications of
science which are developed through our experience
adopting them.
of living and working in societies and cultures, and that
this in turn shapes the ways in which technology is
My own experience has been in the study of processes
acquired and used. Of course, this does not require
of development in Southeast Asia with specific
social scientists to understand the details of what
reference to rural and community development and
scientist and engineers think and do. Furthermore, the
environmental change (King, 1998,1999), and this has
relationships between the natural sciences and
involved me in working with scientists (irrigation
The author is currently the external examiner for General Studies program at Universiti Teknologi PETRONAS.
VOLUME THREE NUMBER ONE JANUARY - JUNE 2003 PLATFORM
OTHER RESEARCH AREAS
engineers, agronomists, hydrologists, geologists,
scientists in addition to studying scientific institutions
among others). We do tend to see the world rather
and the organization of scientific life produce
differently; we use a different technical vocabulary and
technologies as a result of the methods they use and
different methods; and often our priorities are rather
promote, and they do so in the course of their
different. Nevertheless we need each other, and in the
collaboration with, for example virtual reality
world of rural development, the scientist or
technology to enable appropriate simulation of
technologist ignores the social and cultural dimension
of change at his or her peril, and the social scientist is
often unable to do very much at all without the
Therefore, we need constantly to remind scientists and
expertise and knowledge of the engineer or the
engineers that both within and beyond the laboratory
and the classroom we all interact socially with others;
together we form, sustain and change social
THE SOCIAL AND CULTURAL CONTEXT OF SCIENCE
relationships; we live and work in social groups of
various kinds; we hold to certain shared cultural values,
So in training scientists and engineers for the 21st
ideas and beliefs, including religious ones, and we
century we need to remind ourselves that in facing
behave in certain ways so that, despite the enormous
and responding to a rapidly changing world, indeed a
variety of individual behavior, there are discernible
computer- and technological-based world in which
patterns in our values and actions. We also seek
the pace of change will continue to accelerate, we shall
meaning in what we do and what we are, and these
have to develop approaches to teaching and learning
meanings are discovered in the realm of cultural
(not just for students but for those at different stages
symbols and interpretations. So science and
of their working lives) to be adaptable, flexible and
technology operate within societies and cultures; they
responsive, to be international in outlook, and to be
are interwoven into our everyday lives, and, in my view,
able to address and understand the problems with
to train good scientists and engineers we should at
which they are presented from different perspectives
least be aware of the powerful influence which social
and with a broader based knowledge than traditional
relations and cultural values have on technology, and
educational systems have generally provided. I accept
also the powerful influence scientific discoveries and
that scientists and engineers at such institutions as
technological innovations in turn have on our
Universiti Teknologi PETRONAS will be preoccupied
everyday lives. Technology shapes and is shaped by
with developing and applying the specific subject or
social and cultural action; there is mutual adaptation
disciplinary knowledge acquired in the course of
one with another. With that observation I have been
scientific and professional training. However, this
especially pleased to see, in my capacity as external
process of knowledge acquisition is undertaken by
examiner of the General Studies program at Universiti
individuals as members of organizations, and these
Teknologi PETRONAS, that it imparts to its students
organizations, whether they be universities, public
skills and knowledge which go beyond the boundaries
sector institutions, private companies or professional
of science and engineering, and that those who study
associations, operate according to particular social
at the University have the opportunity to learn
principles, cultural values and moral considerations.
something about the social sciences and about their
Social and cultural anthropologists for example have
own society, culture and history.
undertaken important and revealing studies of science
and technology laboratories and institutes not just as
MEDICINE AND THE ARTS
places where scientific research is carried out, but
where scientists are also devoted to maintaining their
A recent article in the Times Higher Education
symbolic stature, legitimacy, identity, career structure,
Supplement (THES) in the United Kingdom entitled
and material resources. We should also note that social
‘Try an aspirin and a bit of Shakespeare?' posed the
PLATFORM VOLUME THREE NUMBER ONE JANUARY - JUNE 2003
OTHER RESEARCH AREAS
question ‘Does an appreciation of the arts make for a
and Technology (1973), that not only will "science and
better doctor, or should prospective medical students
technologyÖnot leave us alone" but they " appear to
stick to science?' The medical profession in the UK is
promote contradictory states of affairs". I shall return
divided in its views, although Sir George Alberti,
to this issue of contradiction in a moment. Science
President of the Royal College of Physicians, has said
and technology are now central to our lives. They have
recently that by studying the humanities ‘It will help
become the dominant cultural form; their values are
(doctors) to empathize and deal sensitively with
pre-eminent in all modern states and there has
patients and their families when they are often at their
emerged an increasingly close relationship between
most vulnerable and in need of understanding'. The
the state or government and science and technology.
THES article also draws attention to a conference
States increasingly concern themselves with science:
sponsored by the Nuffield Trust some three years ago
policies are designed to encourage scientific research
which called for medical students to be recruited from
and the application of science to productive, military
more liberal educational backgrounds and for the
and welfare policies; politicians and government
medical curriculum to incorporate the humanities.
officials seek scientific advice and information for
Indeed, the first Professorship in Medicine and the Arts
decision-making and policy formulation. The most
is about to be established at King's College in the
direct way in which the state and science and
University of London.
technology interact is through the national education
and training system, and through state policies for
Of course, there is always a balance to be struck in a
science in schools, colleges, universities and other
curriculum and it is not the intention of those who
argue for the arts and humanities to replace instruction
in medical science with the study of drama, dance,
Since the Industrial Revolution we have become much
music, painting, literature and poetry. Rather it is to
more conscious of the role scientific and technological
broaden the curriculum in relevant ways, as well as to
change plays in society and the effects it has, and in
open scientific training to those students with a non-
consequence much more concerned about the need
science background, or those with mixed science and
to plan for change and to try to control it, however
humanities qualifications. A recent Australian study
imperfectly. We have had to become more aware of
also showed that those students with some arts
the power of technological innovation for very obvious
background made good doctors and did slightly better
reasons. Prior to the Industrial Revolution inventions
in their training than other graduate entrants. What is
tended to be sporadic, exceptional and more localized
also being proposed is that by broadening the field of
in their effects, and therefore societies could usually
learning and by introducing the experience of
adjust to and accommodate them without major and
students from diverse backgrounds into an academic
rapid disruptions to social and cultural arrangements.
program one has the opportunity to present problems
However, increasingly from the nineteenth century,
and issues in interesting and novel ways so that they
innovations have become more general, regular, and
stick in students' minds and that this broader training
widespread and more dramatic in their impacts.
gives students different ways of conceptualizing,
Technology has also acquired its own legitimacy based
understanding and approaching particular issues. I
primarily on its contribution to the efficiency of
would suggest that the same arguments apply to the
business and other operations, its role in promoting
relationship between science, technology and the
economic growth and increasing national wealth, and
social sciences.
its central position in the process of modernization,
enabling developing countries to catch up with others
SCIENCE, TECHNOLOGY AND THE STATE
or at least compete more effectively with them.
Commercial and manufacturing developments based
Leslie Sklair has stated pertinently in his book
on the new technologies have become increasingly
Organized Knowledge. A Sociological View of Science
VOLUME THREE NUMBER ONE JANUARY - JUNE 2003 PLATFORM
OTHER RESEARCH AREAS
vital to a nation's international competitiveness and
and popular music and film are everywhere. Cultural
its capacities for creating wealth.
differences persist and ethnic and national identities
are constantly reaffirmed, although they might well
In Malaysia, the government has repeatedly
change in form and content.
emphasized the importance of reorienting the
education system towards a greater emphasis on
SCIENCE AND A CONTRADICTORY STATE OF
science and technology and technical education. It is
a frank recognition of the need to be able to compete
economically, to provide industries and the service
Nevertheless, in the writings of nineteenth and early
sector with skilled labor, to ensure that there is an
twentieth century social thinkers on processes of
effective response to changing industrial structures,
industrialization and modernization there was a deep
to realize major benefits from technology transfer and
ambivalence about their consequences for human
to encourage local innovation. From the 1990s and
societies, values, and well-being. Indeed, there has
beyond, the universities, including Universiti Teknologi
always been this tension between science and the
PETRONAS, have been charged with a direct and
social sciences. These writers recognized that there
positive role in Malaysia's industrial development in
was an increasingly close relationship between science
order to respond to rapid global scientific and
and technology which produced the machine-based
technological changes and to help meet the
manufacturing society and its associated social and
government's objective of increasing the share of
cultural arrangements. They acknowledged the
manufacturing in Malaysia's total output and to move
positive contributions of modernity, also expressed
more and more into high-tech, high value-added,
among earlier eighteenth century Enlightenment
thinkers in terms of progress, economic growth, the
development of rational thought and action, and the
It is unnecessary to dwell on the phenomenon of
encouragement of political democracy.
globalization, other than to note the impact that the
information revolution is having on us all, the powerful
Yet the founding fathers of sociology and
influence of the international media, and the dramatic
anthropology, among them Max Weber, Emile
increase in communications and physical mobility.
Durkheim, and Karl Marx all expressed misgivings
These developments are obvious to us all. There are
about what the technology might do to people: the
very positive gains to be had from these innovations,
negative effects of widespread urbanization,
and they certainly act to keep governments on their
industrialization, and modernization expressed in the
toes in ensuring that their nation can rise to the
concepts alienation, anomie, class conflict, social
challenges presented by an increasingly
inequality, exploitation, and imperialism. Then again
interconnected world dominated by large
after the optimism of American social science in the
international organizations. It also requires us to be
1950s, sociologists in particular, began to point to the
more outward-looking, to be able to work with people
less desirable consequences of the globalization of
from other cultural backgrounds and to know
capitalism; the underdevelopment and
something about the world beyond our own national
impoverishment of certain parts of the world and
borders. What is clear, and despite the predictions to
certain groups in society and the enrichment of others;
the contrary, is that processes of globalization and the
the destruction of the environment, industrial
powerful cultural influences conveyed by international
pollution, state-sponsored militarism and
communications and transnational companies are not
authoritarianism, the breakdown of established social
making us all the same, even though American fast-
groups and cultural values, and the disaffection of
food outlets, international retailers and designer-ware,
young people.
PLATFORM VOLUME THREE NUMBER ONE JANUARY - JUNE 2003
OTHER RESEARCH AREAS
My position too is that modern society, and its recent
and national harmony. So the questions are: How do
manifestation in post-modern, consumerist society, is
we humanize and indeed socialize science and
not an unqualified blessing; the development of a
technology? How do we make it more relevant and
modern technology-based society is neither a solution
understandable in human terms to ordinary people?
to all our problems nor is it an unmitigated disaster as
This demands first of all a consideration of what
some of the more militant environmentalists and eco-
motivates people and what is important to them. In
warriors would have it. As Leslie Sklair proposed it has
short, it requires us to realize that different people and
a contradictory character. It both creates new
different social groups may well put different cultural
opportunities and generates new problems and
meanings and different cultural understandings on a
challenges. Industrial technology strengthens the
piece of technology. This social conscience, this
economy and increases wealth on the one hand, and
awareness of the broader context within which science
dislocates some businesses and jobs on the other;
and technology operate, this ability to communicate
improvements in transport increase our physical
effectively what science and technology are and what
mobility, yet they have led to pollution, increased noise
they do to those affected by it, must come, at least in
level and road-deaths and overcrowding in our cities;
part, from the ways in which scientists and engineers
new electronics and communication devices can be
are trained and the kind of curriculum which they
used to combat crime but create opportunities for new
crimes or furnish new methods for committing old
crimes; mass communications technology benefits
TECHNOLOGY, HEALTH AND SOCIETY
education, journalism and the media, but it aggravates
unrest and social tensions by widening the gap
There has been some interesting social research
between expectations and reality, and creates
undertaken during the past 15 to 20 years on the
problems for sustaining moral standards.
complex relationships between technology, society
Developments in genetic engineering in humans,
and culture, and this has intensified recently. The field
animals and plants give the promise for major
is usually referred to as socio-technology. For example,
advances in the treatment of diseases and for the
in May 2001, the UK's Economic and Social Research
increase in food production, but, given the recent
Council launched its 5-year Innovative Health
widespread public rejection of GM foods in the
Technologies (IHT) Program with a budget of 5 million
European Union and controversies about cloning and
pounds sterling. IHTs comprise genetics, imaging
the development of genetic databases, they generate
technologies, new drugs, cloning, stem cell research
major ethical and moral dilemmas and concerns about
and e-health (new information sourcing on health
the unknown impacts on human health and the
issues via the Internet). These innovations raise many
apparent need in some of these areas for more
questions which can be addressed by social science.
experiments on animals.
The research is directed to exploring what role these
new technologies currently play and will play in
Bearing these contradictions in mind, it should be one
redefining the way we manage and experience health
of the concerns, I think, of scientists and engineers to
and medicine, and our concepts of what it is to be
cultivate a social conscience, a sensitivity to social and
‘healthy' and to ‘receive healthcare'. They also arouse
cultural issues and at least to consider and keep in
much public interest and concern and raise questions
mind the possible effects of what scientists invent,
about the ways in which new technology is explained
develop, produce, apply, or implement on their fellow
(in terms of its risks, for example, as well as its benefits),
human-beings; on their work patterns, prospects for
the ways in which consumers then make sense of a
employment and self-realization, their family life and
new technology, how the media shapes the public
the socialization of their children, their life outside
response, and the social processes involved in its
work, their cultural values, and the prospects for social
adoption. It is noticeable that there has been
VOLUME THREE NUMBER ONE JANUARY - JUNE 2003 PLATFORM
OTHER RESEARCH AREAS
considerable volatility in the way in which consumers
culture'. I am prepared to grant that there have been
have reacted to certain health-related research, in
some success stories in managing cultural change, but
genetic research and cloning for example. This in turn
there have been some spectacular failures as well,
has led to governments becoming increasingly
often because there was insufficient realization that
involved in establishing special advisory committees
cultures are created through the process of interaction
and commissions with ethical and moral
between individuals and groups as they try to make
responsibilities in order to address public concern.
sense of their own position and role in the organization
in which they work, and their cultural conception of
their own organization.
Let me turn now to another area of research again
It is not simply a question of managers dictating a
broadly within the socio-technology field but which
particular corporate culture, although obviously they
includes what is referred to as organizational or
have a greater opportunity to guide and influence
corporate culture; in other words, the importance of
others. Cultures are dynamic, variable, flexible and
the relationships between technology and the ways
negotiated so that individuals and groups depending
in which it is received, used and managed in
on their position and function in an organization, and
companies and public sector organizations. An
on other variables such as gender and ethnicity will
important impetus for the increasing interest among
often put different meanings and interpretations on
social scientists in the social and cultural context of
the symbols, language, behavior and action which
technology and industry will be one familiar to
comprise culture.
Malaysians, and it fed into the field of management
science and business studies. This impetus was
THE INTRODUCTION AND ACQUISITION OF
provided by the startling success of Japanese
companies in the 1980s onwards. The distinctive
culture of Japanese firms was seen as vital to their
It is into this complexity that new technologies are
ability to compete and win market share. Obviously
introduced and individuals and groups will interpret
the remedy proposed at that time was that if you could
‘technology' using their own cultural assumptions and
identify successful organizational cultures then
will usually develop new interpretations as they
managers could copy these and implant them in their
interact with technology and with other colleagues
own companies. It also alerted us to the increasing
who are using it (Preece, 1995). This is where the
importance of knowing something about other
important concept of ‘needs' comes in and the research
cultures, other ways of organizing and doing things;
which has been undertaken by social scientists on
in other words the importance of a cross-cultural and
consumerism. We all assume that as individuals or
members of organization when we buy something
(such as a piece of new technology like a computer),
Nevertheless, this early research was soon criticized for
we do so to satisfy a clearly defined need. It is a rational
its lack of attention to the different types of culture
decision. Yet in research done in the UK in the late
required for different types of company of different
1980s and the early 1990s on the purchase and use of
size, structure, product type or range, organization,
home computers, it became clear that the definition
history and environment, or indeed to the fact that
of needs in a cultural context is complex. The
different groups within the same company develop
enormous demand for this new technology from the
their own sub-cultures. This field of corporate culture
mid-1980s could not be explained simply in terms of
and management organization led to the publication
the satisfaction of pre-existing needs. Many customers
of a large number of books and manuals for managers
had little idea of the capabilities of the computers
which claimed that ‘you can learn how to manage
when they bought them. Users began to develop
PLATFORM VOLUME THREE NUMBER ONE JANUARY - JUNE 2003
OTHER RESEARCH AREAS
needs and uses after purchase and they related to their
university administrators, centralized strategic
machines in different ways giving them meanings and
planning, increasing efficiency in delivering teaching
values or simply devaluing them and leaving them
and research, and generating funds from sources
unused, responses which had not been anticipated
outside government. These processes contributed to
either by the designers or the retailers.
the ever-increasing demands on higher education and
the expectation that universities as institutions
With regard to the adoption of new technology by an
primarily funded by the tax-payer should be
organization made up of many groups and individuals
transparent in their procedures and publicly
we can appreciate the increased complexity of the
processes of technology acquisition. If we are sensitive
to social and cultural matters, to individuals as seekers
The system was designed to provide standardized
after meaning, it is misguided to assume that the
kinds of management information across the then
meaning of a piece of technology and the needs it is
university sector involving Student Records, Finance,
supposed to meet are objectively determined, set
Physical Resources, Payroll and Personnel, Research
down in advance and are constant. This is seldom the
and Consultancy, and Management Information. It
case and there is usually a tension between managers
engendered considerable problems, and I remember
who wish to deploy technology in standardized ways
well my departmental secretaries being critical of the
and company workers who wish to localize it and
extra workload which the system generated and the
evaluate it in their own terms. We probably all know
fact that the schedule in which information had to be
of cases where the introduction of technological
inputted completely disrupted the arrangements that
innovations has been resisted or sabotaged, and this
they had in place to spread their work more evenly
is invariably because of issues to do with social
across the academic year. The MAC initiative largely
organization and cultural evaluation. In other words
failed to realize its objectives fully for two major
acquisition is a much more uncertain, drawn-out and
open-ended process than we might like to think.
Firstly, at the time that it was introduced it contained
To illustrate these points I have chosen an interesting
within its design features certain assumptions about
sociological study undertaken by Janice McLaughlin,
the way in which universities were organized which
Paul Rosen, David Skinner and Andrew Webster in the
did not match the perceptions of the people who
mid-1990s and described in their book "Valuing
worked in the institution, and the way they organized
Technology" (1999). Their research compared
their working lives. It assumed that university decision-
technology acquisition in three different UK
making was centralized when invariably academic
organizations: a staff management system in a retailing
departments enjoyed considerable independence in
company, a clinical information system in a hospital
teaching and research and the ways in which they
and a management and administrative system in a
organized the delivery of their academic programs and
university. For obvious reasons I am especially
the conduct of their scholarly activities. The MAC
interested in the case of the university because as a
system demanded that universities deliver certain
senior academic manager in my own university at the
standard kinds of data in a defined way which began
time I had to live and work through the introduction
to undermine the autonomy of academic
of this system. It was called the Management and
departments. It was also difficult to cope with the
Administrative Computing (MAC) system, which was
range of different kinds of university in size, subject
initiated in the late 1980s, and generally it failed to do
mix, structure and history, and this was compounded
what was expected of it. As the researchers point out
by the great variety of working practices, needs and
its introduction was part of the general trend in
missions of academic departments. Interestingly there
universities towards the professionalization of
were particular groups and categories of employee in
VOLUME THREE NUMBER ONE JANUARY - JUNE 2003 PLATFORM
OTHER RESEARCH AREAS
the university selected for the case-study which were
argued that new times require new approaches and
more successful than others in resisting and modifying
new ways of thinking, and that to equip ourselves to
the MAC system.
meet novel situations we have to combine the
strengths of a variety of disciplines embracing the
Secondly, by the time it was implemented the MAC
humanities, social sciences and natural sciences. This
system was already being overtaken by changes in the
was not an altogether unexpected statement, but it is
sector; the raising of the polytechnics to university
true nonetheless, and it is much easier said than done,
status, the shifts to semester-based, modular courses,
particularly when academics still tend to have strong
and the move away from the use of terminals linked
commitments to their own academic discipline. Yet
to mainframe computers towards Windows-based
these commitments have to be overcome if we wish
desktop PCs supported by central servers. The MAC
to provide a broad-based, innovative curriculum and
system had simply not anticipated these changes or
produce well rounded and well equipped graduates
the rapid pace with which they would be introduced.
to meet the challenges of the 21st century.
For these reasons, parts of the system were abandoned,
some other elements were modified to such an extent
King, Victor T. (ed), Environmental Challenges in South-East
in the local setting that they had moved considerably
Asia, Richmond, Curzon Press, and Copenhagen, Nordic
away from the overall objective of providing a national
Institute of Asian Studies, 1998.
system, and today, despite the enormous resource in
King, Victor T., Anthropology and Development in South-East
finance and staff time invested in the system, it is being
Asia. Theory and Practice, Kuala Lumpur, Oxford University
either further modified or set aside and replaced with
Press, 1999.
other more effective alternatives.
McLaughlin, J., Rosen, P., Skinner, D., Webster, A., ValuingTechnology, London and New York, Routledge, 1999.
Preece, David, Organizations and Technical Change. Strategy,Objectives and Involvement, London and New York, Routledge,
This case study provides a lesson for us all. It is obvious
that we shall have to continue to embrace new
Sklair, Leslie, Organized Knowledge. A Sociological View of
technological innovations; it is inevitable, although in
Science and Technology, London, Allen and Unwin, 1973.
specific cases technology is not necessarily the best
method of solving particular managerial or
Victor T. King is a professor and a center
organizational problems. However, we should not lose
director of South-East Asian Studies at the
sight of the fact that what we do is part of a cultural
University of Hull, UK. He held severalacademic positions at the university prior
and moral universe and that technological
to his appointment as the center director of
development however wondrous it is does not have a
South-East Asian Studies. His researchinterests include socio-economic
life and momentum of its own. It is given meaning
development and applied anthropology,
and value by people. Finally, in my view, scientists and
especially in Malaysia, Brunei and Indonesia;
engineers will be better equipped if they recognize
religion, symbolism and material culture in South-East Asia. Hisrecent work focuses on resettlement and land development
fully and openly that they work within and for society.
schemes in East Malaysia, environmental change in Sarawak,
It also requires them to acknowledge the international
tourism and development, and business culture in South-East Asia.
His ongoing projects, among others, include the study of socio-
dimensions of their work and to keep their minds open
economic development in the state of Perak, Malaysia.
to other academic disciplines and to other ways of
seeing and understanding the world.
In a recent debate in my own university about
academic reorganization, one of my senior colleagues
PLATFORM VOLUME THREE NUMBER ONE JANUARY - JUNE 2003
OTHER RESEARCH AREAS
Integration of Multimedia in
Visualising Geometric Transformations
for Mathematics Education: A Preliminary Review
1 Wan Fatimah Bt Wan Ahmad
2 Halimah Badioze Zaman
1 Universiti Teknologi PETRONAS,
Bandar Seri Iskandar, 31750 Tronoh, Perak Darul Ridzuan, Malaysia.
2 Faculty of Information Science and Technology,
Universiti Kebangsaan Malaysia, 43600 Bangi, Selangor.
The growth of personal computers usage and the introduction of the Multimedia Super Corridor (MSC), coupled
with the availability of a wide range of educational software have changed forever the face of education in the
country. Since then, computers have revolutionised the educational delivery system. Multimedia in education
has taken centre stage in the educational technology initiatives under taken in recently. This is because personal
computers allow us to create and use documents with multimedia objects such as pictures, photographs,
animations, texts, voices etc. The use of computer system together with the interactive multimedia techniques
will create the learning environment and resources for our education system. This paper intends to highlight
the definitions and benefits of multimedia in mathematics education. Technology generally, can help all students
understand mathematics and interactive multimedia courseware particularly, may give students with special
instructional needs to access mathematics in a way that they might not otherwise experience. Students of
mathematics tend to find the subject difficult due to the inability to visualize the abstract of mathematical
concept. With multimedia, it is hoped that this problem can be overcome.
Students can naturally use their own experiences with shapes to learn about transformations such as slides
(translations), turns (rotations), and flips (reflections). The purpose of this paper is to discuss the issues that has
been raised in teaching and learning of geometric transformations and the research that has been done. The
authors will also highlight a conceptual model of a multimedia based tutoring system on geometrical
transformations; pedagogical approaches suitable for teaching and learning mathematics and the modules
that will be developed through a prototype. The modules will be based on the student learning approach that is
the reinforcement, revision and enrichment modules. We hope that this tutoring module for recognising the
geometric transformations will be beneficial in teaching and learning of mathematics in the smart schools.
Keywords: multimedia, education, mathematics
This paper was presented at the Second Conference On Conference on Information Technology in Asia 2001 (CITA'01),
Universiti Malaysia Sarawak (UNIMAS), October 17-19, 2001
VOLUME THREE NUMBER ONE JANUARY - JUNE 2003 PLATFORM
OTHER RESEARCH AREAS
1.0 INTRODUCTION
3.0 DEFINITIONS OF MULTIMEDIA
Technological advances have changed and are still
During the last decade, multimedia technology has
changing many aspects of society and education.
been a dominant focus of the computer industry. In
These changes are seen in both the delivery method
its many forms, multimedia offers exciting possibilities
as well as content of many subjects offered in the
for education and entertainment.
educational institutions. The National Council of
Teachers Mathematics (NCTM) has taken a strong
Multimedia is the integration of more than one
interest of integrating technology in the teaching and
medium into some form of communication or
learning of mathematics [1]. Technology is particularly
experience delivered via computer. Most often,
useful for facilitating the teaching and learning of
multimedia refers to the integration of media such as
mathematics concepts for students [2].
text, sound, graphics, animation, video, imaging, spatial
modelling into a computer system [4]. Employing a
The use of computer for enhancement of learning is
relatively inexpensive desktop computer, users are
not a new concept. From the early 1970's there has
now able to capture sounds and video, manipulate
been a push for the computer to gradually take a
audio and images to achieve special effects, synthesize
leading role in education. In Malaysia, the use of
audio and video, create sophisticated graphics
computer in formal education was first introduced into
including animation, and integrate them all into a
schools in 1981 when 20 schools were selected to be
single multimedia presentation. Multimedia is not a
in the pilot ‘computer in education project' [3]. By late
form of instruction to learn from, but rather a tool for
1996, Smart School had become one of the seven
constructing and learning with [5]. Learners may create
flagship applications of the Multimedia Super Corridor
their own multimedia knowledge presentations that
(MSC). The Malaysian Smart School is a learning
reflect their own perspectives on understanding of
institution that has been systemically reinvented in
terms of teaching-learning and management practices
to help children cope with the Information Age.
Multimedia is considered as a breakthrough in
Through this program, computer and multimedia will
teaching and learning aids because it allows the user
become the main medium in the process of teaching
to be an active participant rather than a spectator [6].
and learning.
The author also discusses the difficulties commonly
found in middle school environment, which include
2.0 PURPOSE OF PAPER
the use of the technology and how the teachers can
integrate this new teaching aid into their curriculum.
The purpose of this paper is to highlight the definitions
and benefits of multimedia in mathematics education
Saga [7] discusses the integration of different media
specifically in the area of geometric transformations.
developed in Japan. The study tested the effectiveness
This paper also provides an overview of the issues that
of the system including the results related to level of
have been raised in teaching and learning of
interest, learning of factual information and degree of
geometric transformations and the researches that
have been done. The authors will also highlight a
conceptual model of a multimedia based tutoring
3.1 Benefits of Multimedia
system on geometrical transformations; pedagogical
Multimedia technology has been said to be able to
approaches suitable for teaching and learning
enhance teaching and learning and teachers and
mathematics and the prototype modules that will be
students can benefit from it tremendously, if the
technology is used well during the design and
development stage of the multimedia courseware.
PLATFORM VOLUME THREE NUMBER ONE JANUARY - JUNE 2003
OTHER RESEARCH AREAS
Among the benefits reported on multimedia
traditionally introduced to students through
technology used in education are:
constructing the figures with compass and ruler or a
Enhance students' achievement;
model. The introduction of geometric shapes begins
Help students master skills required for the
in infancy with blocks, puzzles, sorting toys, mobiles
and books. Students first learn to recognise a shape
Serve as a motivational tool, improving attitude
by its appearance as a whole. Students can create
towards learning, confidence, and self- esteem;
shapes on a piece of paper and represent them in
Enhance ability to remember and understand
drawings, learn about geometric properties by
combining apart shapes to form new shapes. Students
Enhance organisational and problem-solving
can naturally use their own experiences with shapes
to learn about transformations.
Help students become independent learners and
Transformations are required in many geometrical
Improve skills and knowledge of teachers [8].
constructions as part of mathematics curriculum.
Geometrical transformation involves mental or
The following benefits are identified by students:
physical manipulation of shapes to new positions or
Students can work at their own pace.
orientations. It is a study of movements such as
Students get constant feedback on progress.
translations, reflections and rotations or, as they are
More flexible access to education,
commonly referred to as slides, flips and turns.
Reduction of lecturer activity coupled with more
Explorations and activities with transformations
active student centred learning.
encourage development of spatial reasoning as
More time for the teacher to talk with individual
students practice visualizing what shapes look like as
they're slide, turned and flipped.
Fewer students waiting for assistance, there is
increased student motivation.
Visualization has been an area of interest concerning
Students cooperatively help one another, moving
mathematics education, particularly with the advent
around the room.
of computer technology that allows the possibility of
A more relaxed classroom atmosphere
a wide range of visual displays. Many researchers
Increased mastery, especially by weaker students
emphasize the importance of visualization and visual
Consistency in course content.
reasoning for learning mathematics [11, 12]. They have
Reduces ‘catch up' problems if a student is sick [9].
made use of visual forms of representation and
suggest that visual thinking can be alternative and
One of the areas of mathematics that can benefit from
powerful resource for students doing mathematics.
visual aspects of multimedia is geometry.
These visual representations also seem to have a
powerful influence on students' development of
GEOMETRY AND VISUALIZATION
geometric concepts.
Geometry is the study of spatial relationships that can
Apart from these, learning geometry is important
be found in the three-dimensional space we live in and
because other courses such as Physics, art, mechanical
on any two-dimensional surface in this three
drawing, Chemistry (for atomic and molecular
dimensional space [10]. Geometry deals with the
structure), Biology (for cell structure) and Geology (for
investigation of shapes and objects in two and three
crystalline structure) also use geometry as a
dimensions, their locations, relationships and
properties. Shapes are one of the most common ways
to represent geometric objects. Geometric objects are
VOLUME THREE NUMBER ONE JANUARY - JUNE 2003 PLATFORM
OTHER RESEARCH AREAS
1.1 Issues in the learning and teaching
ways to help their students to think and to analyze
problems, and in developing insight in their students.
Gorgorio [13] listed topics where students with poorly
The "van Hiele Model of Thinking in Geometry" [15], as
developed spatial abilities have difficulties in
in Figure 1, provides strategies and suggested activities
geometrical transformations such as translation,
for helping students focus on levels of thinking in
reflections, rotations, dilations and expansions. All
these topics are part of our curricular content.
At Level 0, students recognize the geometric shapes
Geometrical transformation is difficult because it
on the basis of their visual appearance for example,
requires high abstractive thinking and logical
triangles and squares. However, they do not recognize
deduction. Students usually encounter with these
properties of these figures yet. At Level 1, students will
difficulties when they study it. Students' performance
analyse the geometric shapes but at this stage they
will be even poorer when it comes to items that involve
are not able to explain the interrelationships between
the understanding of features and properties of
figures and properties. At Level 2, students can
shapes. National and international surveys revealed
establish interrelationships of the properties within the
that secondary school students could not identify and
figures and among the figures. Informal proofs can be
name shapes like kite, rhombus, trapezium,
followed but students do not see how the logical
parallelogram and triangle [10]. Students have shown
order/proof could be altered. At Level 3, students will
lack of ability to connect a diagram with its symbolic
understand the significance of deduction as a way of
representation. Learning the characteristics and
establishing geometric theory with axiom. At Level 4,
properties of a geometrical shape is essential because
students can compare different axioms. It is highly
they can form the basis of higher levels of thinking and
abstract and this level of study is suitable for study of
help in gaining a practical and intuitive grasp in
geometry in college.
mathematical space. Representation of geometrical
object involves the use of diagrams, along with words,
Van Hiele proposed a sequence of five "phases" of
apparatus and activities. The ability to interpret the
learning as in Figure 2 and Table 1.
figural language of geometry, manipulate figural
stimuli, associate theorems with previous experiences
In another research, Ha and Duc [16] have listed five
and translate non-figural stimuli into figural terms are
levels of activities in the learning process of geometric
very important in the learning of geometry [14].
transformations in order to develop skills in the solving
of transformations (refer to Table 2). According to Ha,
Learning geometry can also be difficult and studies
in Level 1 and Level 2, students can use software to
have shown that students need experience in
deal with problems effectively. However, in the other
"thinking" at different levels. The most influential
levels, students can use software and their deduction
research into learning of geometry was the work of
in order to solve the problems. The procedures
two Dutch, Pierre van Hiele and Dina van Hiele-Geldof,
involved are shown in Table 3.
during the late 1950s. They were interested in finding
Figure 1: Van Hiele's thinking model
PLATFORM VOLUME THREE NUMBER ONE JANUARY - JUNE 2003
OTHER RESEARCH AREAS
Figure 2: Learning phases
Table 1: Learning phases
Students will discuss and develop questions on a topic to be explored
Students explore the topic through materials that has been carefully sequences by the
Students express and exchange views about the structures that have been observed
Students encounter more complex tasks and gain experience in finding their own way of
resolving the problems
Students are able to internalise and unify relations into a new body of thought
Table 2: Activities levels
Recognize transformation through activity of discovering invariance.
Construct image of a figure through a given figure transformation. With the variant nature,
transformation is considered as a deductive tool but it is not systematic.
Use the invariance in the simple situation which is based on deduction.
Use figure transformation like useful tool to solve mathematical problem.
Use figure transformation among algebraic viewpoint. (The product of figure
Table 3: Solving procedure
Construct figure and distinguish the nature of problems and define supposition and
conclusion on problems.
Realize operations on the figure to find out the necessary correspondences, define the
relation between the elements in the problem and discover the most suitable figure to deal
with the problems.
Use the figure transformation to solve problem.
Check and make a conclusion.
VOLUME THREE NUMBER ONE JANUARY - JUNE 2003 PLATFORM
OTHER RESEARCH AREAS
Hoffer [17] has identified five geometric skills that is
understand mathematics and interactive multimedia
visual, verbal, drawing, logical and application while
courseware may help some students to access
De Villiers [18] referred to the processes in the study
mathematics in a way that they might not otherwise
of geometry as axiomatising, proving, defining,
experimenting, refuting, pattern finding, generalising,
specialising, classifying and theorem finding.
DEVELOPMENT OF A CONCEPTUAL MODEL
Four types of mathematical questions regarding
According to Piaget [21], "To know an object is to act on
navigation and maps can help students develop a
it. To know it is to modify, to transform the object and to
variety of spatial understandings: direction (which
understand the process of this transformation and, as a
way?), distance (how far?), location (where?) and
consequence, to understand the way the object is
representation (what objects are abstract?). In
constructed" (pg 176). Therefore, learning experiences
answering these questions, students need to develop
will require a ‘hands-on' involvement of the students.
a variety of skills that relate to direction, distance, and
Based on the issues raised in the teaching and learning
position in space. Computer can help students to
of geometric transformations, the authors have
abstract, generalise, and symbolise their experiences
proposed a prototype. The proposed prototype will
through the navigation process.
include the following topics: rotation, translation,
reflection and combinations of transformations.
Students using interactive multimedia courseware
pertaining to shapes very often have to choose a
A conceptual Instructional Design (ID) model of the
motion to solve a puzzle. These actions are
multimedia courseware to be developed is as shown
explorations with geometric transformations and are
in Figure 3, which includes medium, development
important part of spatial learning. They help students
approach and learning theories. Multimedia will be
become conscious of the motions and encourage
used in the presentation of the prototype. This means
them to predict the results of changing a shape's
that the presentation is not only in text but also will
position or orientation. Students should be given
be in text, graphics, audio and animation.
opportunities to practice the skills until they achieve
their own understanding. The interactive environment
The proposed multimedia prototype will be based on
provided by multimedia courseware or programs has
a tutoring system that will relate mathematics with real
the potential to foster students' movement from
life. The objectives of the courseware are that the
concrete experience with mathematics to more formal
students are able 1) to define, use and visualize the
levels of abstraction and to improve the students'
effect of translations, rotations and reflections and
mathematical thinking. Spatial visualization can be
understand the effects of repeated transformations
developed by building and manipulating a concrete
and 2) to draw the image of the object under the
and mental representation of shapes, relationships,
combinations of the transformations given. These
require the learners' mental and physical manipulation
of shapes to construct a new image under the given
Studies conducted by Khalili & Shashaani [19] and
Butler & Mautz [20] have shown that the use of
multimedia technology has a positive effect on
The teaching and learning theories will include the
learning and achievement from elementary schools to
cognitive theory, constructivism theory and Van Hiele's
college. Interactive multimedia courseware can
theory of thought. There will be three modules that
provide a rich environment for activities in which
the students can choose from, depending on the levels
students can compose and decompose shapes.
of the learners. The modules are reinforcement module,
Multimedia technology can help all students
revision module and enrichment module. The
PLATFORM VOLUME THREE NUMBER ONE JANUARY - JUNE 2003
OTHER RESEARCH AREAS
reinforcement module consists of the introduction of
system. The performance will be recorded and
geometrical transformations, steps and examples on
how to identify the various transformations, exercises
set by the systems. In the revision module, slow
learners and learners who are weak in certain topics
can relearn the topic until they fully understand. These
In the world of computers or the digital information
learners can work for as long as they want before they
age, the education systems should also provide
move to a new topic in the reinforcement module.
students with integrated and diverse ways of learning.
Enrichment module consists of exploration and
Students should be given the opportunities to
application problems. These questions are more
experience and learn in ways that have never been
challenging, require learners to think, to analyze and
imaginable before. Due to the fact that students learn
apply concepts learnt earlier. There will be activities
in various ways that are unique to their own interests
that learners have to explore and apply the
and experiences, the learning aids provided should
geometrical transformations. Learners will be assessed
also cater for this diversity in learning styles. Learning
in the assessment module that will be built into the
aids that are designed and developed should be both
Figure 3: Conceptual model of courseware
VOLUME THREE NUMBER ONE JANUARY - JUNE 2003 PLATFORM
OTHER RESEARCH AREAS
in conventional as well as electronic form (e.g.
[11] Triadafillidis, T.A. 1995, "Circumventing Visual Limitations in
Teaching the Geomerty of Shapes, Educational Studies in
multimedia). Students can interact with the various
Mathematics, 29: 225-235.
forms of multimedia in order to properly structure their
knowledge and allow them to become better learners
[12] Battisha, M.T. 1994, "On Greeno's Environmental/Model View
of Conceptual Domains: A Spatial /Geometric Perspective",
as well as problem-solvers.
Journal for Research in Mathematics Education,25(1). 86-99.
[13] Gorgorio, N. 1998, "Exploring The Functionality of Visual and
The paper reviewed a few issues that had been raised
Non-visual Strategies in Solving Rotations Problems".
in the teaching and learning geometric of
Educational Studies in Mathematics. 35: 207- 231.
transformations and the research that has been done.
[14] Del Drande, J.J. 1987, "Learning and Teaching Geometry, K-
The authors have also highlighted a conceptual model
12,USA", Spatial Perception and Geometry in NTCM 1987
of a multimedia based tutoring system on geometrical
Yearbook. 126-135.
transformations, pedagogical approaches suitable for
[15] Fuys, D., Geddes, D. and Tischler, R. 1988, "The van hiele Model
teaching and learning mathematics and the modules
of Thinking in Geometry among Adolescents",Journal for
that will be developed through a prototype. It is hoped
Research in Mathematics Education Monograph No.3 Reston,VA.:National Council of Teachers of Mathematics. 210-221.
that with the help of the developed prototype, the
students are able to access the information that can
interest them on the learning abilities and techniques
[16] Ha, T.T.T. and Duc, N.H., 1999, "Developing Student's Activities
in using Transformations to Solve problems on Plane
in geometric transformations.
Geometry with the Assistance of Informatics". ICE on ESH'99(30Nov -2 Dec 1999), SEMEO RECSAM, Penang, Malaysia. 164-170.
[17] Hoffer, A. 1981, "Geometry is More Than Proof", Mathematics
National Council of Teachers of Mathematics. 1989, "Curriculum
and Evaluation Standards for School Mathematics", Reston, VA.
[18] De Villiers, M.D. 1997, "Possibilities for Doing More Mathematics
Falba, C.J. and Williams, D.L. 1998, "Exploring Mathematics with
Via Congruence Geometry", Paper presented at MALATI/
interactive Computer Multimedia: An Investigation of Mirror
EMSLEP Geometry Thinkshop, University of Stellenbosch, South
Symmetry", Intervention in School & Clinic, 33(3). 187-190.
Yussuf Hashim. 1998, Teknologi Pengajaran. Shah Alam:
[19] Khalili, A.& Shashaani, L. 1994, "The effectiveness of Computer
Penerbit Fajar Bakti Sdn. Bhd.
Applications: A Meta-analysis", Journal of Research on
Computing in Education, 27(1), 48 - 61.
Von Wodtke, M. 1993, Mind over Media: Creative thinking Skillsfor Electronic Media, New York: McGraw Hill.
[20] Butler, J.B. & Mautz Jr., R.D. 1996, "Multimedia Presentations and
Learning: A Laboratory Experiment", Issues in Accounting
Lajoie, S.P. and Derry, S.J. 1993, Computers as Cognitive Tools,
Education, 11(2), 259 - 281.
Hillsdale, N.J.:Lawrence Erlbaum.
[21] Piaget, J. 1964. "Development and Learning", Journal of Research
Kanning, R.G. 1994, "What Multimedia Can Do in Our
in Science Teaching, 2, 176-186.
Classrooms", Educational Leadership, 51(7), 40 - 44.
Saga, H. 1993, "Are We Ready Enough to Learn from Interactive
Wan Fatimah Wan Ahmad received her
Multimedia?" Educational Media International, 29(3), 181- 188.
M.S. degree in Mathematics from CaliforniaState University, Long Beach, California,
U.S. Department of education. 1996b, (Available on online)
U.S.A. in 1987. She is currently a lecturer in
the IT/IS program at UTP. She was a lecturer
at Universiti Sains Malaysia, Tronoh beforejoining UTP. Her main interests are in the
Waddick, J. 1994, "Case Study: The Creation of a computer
areas of educational technology,
Learning environment as an alternative to traditional
multimedia and mathematics education.
Lecturing Methods in Chemistry", Journal of Education And
She was involved in a project under the Intensified Research in
Teaching Technology International (ETTI), 31(2)
Priority Areas (IRPA) in 1996-1999. The name of the project was"Multimedia in the Learning of Engineering Mathematics".
[10] Bennie, K. 1998. "Shape and Space": An Approach to the study
Currently, she is pursuing her Ph.D. studies at Universiti
of Geometry in the Intermediate Phase, Paper presented at
the 4th Annual Congress of the association for MathematicsEducation of South Africa, Pietersburg, July 1998.
PLATFORM VOLUME THREE NUMBER ONE JANUARY - JUNE 2003
PLATFORM is a biannual, peer-reviewed
journal of Universiti Teknologi
NOTES FOR CONTRIBUTORS
PETRONAS. It serves as a medium for
Instructions to Authors
faculty members, students and industry
professionals to share their knowledge,
Authors of articles that fit the aims, scopes and policies of
views, experiences and discoveries in
this journal are invited to submit soft and hard copies to the
their areas of interest and expertise. It
editor. Paper should be written in English. Authors are
comprises collections of, but not limited
encouraged to obtain assistance in the writing and editing
to, papers presented by the academic
of their papers prior to submission. For papers presented or
staff of the University at various local and
published elsewhere, also include the details of the
international conferences, conventions
conference or seminar.
and seminars.
Manuscript should be prepared in accordance with the
The entries range from opinions and
views on engineering, technology and
social issues to deliberations on the
The text should be preceded by a short abstract of 50-
progress and outcomes of academic
100 words and four or so keywords.
The manuscript must be typed on one side of the paper,
Opinions expressed in this journal need
double-spaced throughout with wide margins not
not necessarily reflect the official views
exceeding 3,500 words although exceptions will be
of the University.
All materials are copyright of Universiti
Figures and tables have to be labelled and should be
Teknologi PETRONAS. Reproduction in
included in the text. Authors are advised to refer to recent
whole or in part is not permitted without
issues of the journals to obtain the format for references.
the written permission from the
Footnotes should be kept to a minimum and be as brief
as possible; they must be numbered consecutively.
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.
Reference should be indicated by the authors' last names
and year of publications.
Chief Editor, PLATFORM
Universiti Teknologi PETRONAS
Bandar Seri Iskandar
31750 Tronoh
Perak Darul Ridzuan
P L A T F O R M
Volume 3 Number 1
Jan - Jun 2003
Technology Cluster: TRANSPORTATION
Technology Platform: Novel Engine Development
Design and Development of a Prototype Free Piston Linear Generator Engine
Abd. Rashid Abd. Aziz
Technology Platform: Fuel Combustion
The Effects of Horizontal Mounting Position of Atomizer to Spray Angles
in a Confined Tube
Shaharin Anwar Sulaiman
Technology Cluster: OIL AND GAS
Technology Platform: System Optimization
Industrial Stream Composition Forecasting Using Neural Network
Abdul Halim S. M, I. Ismail, R.D. Menez, and M.A. Shafawi
Debottlenecking of an Existing Fractionation Train Using the Non-conventional
Distillation Columns
Kamarul Ariffin Amminudin
Technology Cluster: INTELLIGENT SYSTEMS
Technology Platform: Smart Card Technology
THREE NUMBER ONE JANU
Face Recognition for MyKad Security
Ahmad Fadzil M.H. , Irfan A.R.
Enhancing Design Criteria for Novice Virtual Environment Designers
through the Identification of Usability Problems
Suziah Sulaiman, Rohiza Ahmad
Identification of Handwritten Digits
Ahmad Fadzil M. H., Intan Mastura A.M.
Technology Platform: Intelligent Building
Application of Dry Walls in Residential Houses: A Buildable System
Noor Amila W.A.Z., Nasir Shafiq, Ibrahim Kamaruddin
OTHER RESEARCH AREAS
Study on the Capillary Performance in Wire Bonding Operation
for Integrated Circuit (IC)
Norani M. Mohamed, Sivabalan Sasthiryar
The Training of Scientists and Engineers in the 21st Century:
a Social Science Perspective
Victor T King
Integration of Multimedia in Visualizing Geometric Transformation
for Mathematics Education: A Preliminary Review
Wan Fatimah Wan Ahmad, Halimah Badioze Zaman
Source: https://www.utp.edu.my/The%20University/Publications/Shared%20Documents/UTP%20Platform/utp%20platforrm%202003/Platform%20v3n1.pdf
A simple (ish) guide to the Psychoactive Substances Bill Date: 15/12/2015 What is it?: A Bill that, if it becomes law, will make it an offence to produce, supply or offer to supply any psychoactive substance with the exemption of nicotine, alcohol, caffeine and medicinal products.1,2 The main intention of the Bill is to shut down shops and websites that
PRODUCT TECHNICAL DOSSIER TRIBULUS TERRESTRIS EXTRACT 95% SAPONINS Product Code: P20046 Raw Material Full Name: Tribulus Terrestris Extract 95% Saponins Raw Material Full Botanical/Chemical/Latin/Trade Name/Synonyms: Tribulus Terrestris L This material is Food Grade Assay (by UV): Min 95.0% Saponins Ratio of Material: Approx. 10:1 Solubility in Water: 40-50% Soluble in Water Solubility in Alcohol: The Solubility of this product in 75-80% Alcohol 80% Particle Size: 100% pass through 80 Mesh Bulk Density: 40g/ml – 60g/ml Loss on Drying: Max 5% Ash: Max 5.0%
