Beruflich Dokumente
Kultur Dokumente
1, 2011
41
42
Introduction
43
precisely defined. Usually, the relations between OPs and QCs are highly non-linear
and complex in nature. A better way to define such relationships is the use of artificial
intelligence. Tahera et al. (2008) adopted fuzzy logic to deal with qualitative information
for a product with single QC. For the case of a product with multiple QCs there is a need
for a solution method to find the optimal input parameters.
In this paper, a process targeting model that uses fuzzy logic to define the relationship
between multi-output QCs and OPs is proposed. The cost function is defined using a
novel unsymmetrical interval-based Taguchi (UIT) loss function. An evolutionary-based
search algorithm is utilised to find the optimal process parameters.
The rest of this paper is organised as follows: Section 2 presents a comprehensive
literature survey related to targeting problem and Section 3 describes the fuzzy-based
process targeting model together with the asymmetric interval-based Taguchi typequadratic loss function. A local search algorithm based on an evolutionary approach to
determine the optimal parameters is presented in Section 4. An illustrative example from
the literature to demonstrate the utility of the model is provided in Section 5. The results
are presented and discussion in Section 6. Section 7 concludes this paper and suggests
future research directions.
Literature review
Targeting problem has been a serious problem in the literature of quality control for the
past 60 years (Springer, 1951; and Yang et al., 2008). Springer (1951) was the pioneer in
developing process targeting models. The can-filling problem with unsymmetrical costs
was considered with the objective to minimise the total expected cost. Hunter and Kartha
(1977) proposed a similar model where an undersized product is scrapped at a constant
price and over size product costs proportional to the can content. Bisgard et al. (1984)
extended the model of Hunter and Kartha (1977) by modifying the model so that
undersized products are sold at a price proportional to the can content. Golhar (1987)
proposed a model where undersized products are reprocessed at a fixed cost. Golhar and
Pollock (1988) proposed a model, in which the can content is constrained by an upper
limit. Any product which exceeds the upper limit will be reprocessed with the same
reprocessing cost as that for undersize product. Golhar and Pollock (1992) considered the
effects of variance on the profit. A variance reduction method for minimising the total
expected cost was proposed. Product uniformity via a Taguchi quadratic loss function is
proposed by Arcelus (1996). Al-Sultan and Pulak (2000) extended Golhars (1987) model
for the case of two-stage manufacturing process, also it can be viewed as a modified
version of Al-Sultans (1994) model with 100% inspection. Williams et al. (2000)
developed a process targeting model for a process that incurs random shifts. Roan et al.
(2000) incorporated production set-up and raw material procurement in the classical
process targeting problem. Rahim et al. (2001) proposed three new approaches for the
economic selection of a target variance integrated with a target mean. Lee et al. (2001)
determined the optimum process mean of a quality characteristic and the screening limits
of a correlated variable under single and two stage screening. Using a surrogate variable
associated with the product quality Lee and Elsayed (2002) developed a process targeting
model to determine the optimum process mean and screening limits for the surrogate
variable. Cho (2002) developed a process targeting model for two QCs. Teeravaraprug
and Cho (2002) studied the multi-variate quality loss function to incorporate the
44
customers overall perception of product quality into design. Duffuaa and Siddiqui
(2002, 2003) developed a process targeting model for three class screening problem by
incorporating product uniformity and measurement errors. Chen (2003) determined the
optimum process mean for the larger-the-better Weibull quality characteristic. Chen and
Chou (2004) proposed a modified Wen and Mergens model that includes the quadratic
loss function and a one sided specification limit. Rahim and Tuffaha (2004) used Taguchi
quadratic loss function to integrate the targeting and production run problems. However,
for the multi-output quality characteristic systems, conventional quadratic Taguchi loss
function cannot be used. Chou et al. (2002) proposed a multi-variate Taguchi loss
function. An extensive survey for multi-variate Taguchi function has been carried out in
Suhr and Batson (2001). Bowling et al. (2004) presented a Markovian approach and
developed a general form of a Markovian model for determining the optimum process
target levels within the framework of a multi-stage serial production system. Chung-Ho
(2005) modified Wen and Mergen model for determining the optimum process mean for
a process with a Log-normal distribution. Chung-Ho (2006) modified Wen and Mergen
model for determining the optimum process mean using a mixed quality loss. Kolus
(2005) extended process targeting models to processes in series with dependent QCs.
Chen (2006) extended the Pulak and Al-Sultans model by incorporating quadratic
quality loss function in order to determine the optimal process parameters. Chan and
Ibrahim (2006) assumed assignable cause for the shift in process means, and developed a
method to set the target values of a product with the multiple QCs. Chen and Khoo
(2008) extended the Chen (2006) model by incorporating inspection errors. Tahera et al.
(2008) proposed a fuzzy logic approach to determine the process parameters for a product
with single QC. The review of the literature indicated that there is no fuzzy process
targeting model for product with several quantitative and qualitative QCs.
Model formulation
The industrial (or) manufacturing process under consideration has the following
characteristics:
x
the required multiple QCs may not be measured easily, however OPs are easy to
measure
relationship between QCs and OPs may not be mathematical ones, due to
consideration of qualitative attributes
the process deteriorates randomly due to variations and after a certain period, the
OPs are used to find if there is any need to set the initial process parameters.
The problem is to determine the optimal process target when the parameters are not
directly measurable and can be obtained indirectly through OPs.
The assumptions of this model are: the process starts form in-control state, and
inspections are carried out as per known and pre-defined schedule. Since, the inspections
45
are carried to measure the OPs, the process will not be halted in the middle. It is also
assumed that the operators knowledge and experience are available at hand to formulate
fuzzy rules.
To formulate the above said problem, consider a product with multiple OPs and
multiple QCs. Let the OPs be represented by X [ x1 , x2 , , xn ] , and QCs be
represented by Y [ y1 , y2 , , ym ] , as in Figure 1. The OPs are those parameters which
can be measured easily, they are primarily input parameters, and can also consist of few
output parameters. Generally, not all the output parameters are QCs, that is, only few of
the output parameters are QCs. Practically speaking, due to technical or physical
constraints, it is not always possible to measure the QCs easily, or sometimes, they
cannot be directly measured by any means. To calculate the QCs, a suitable relationship
including OPs is to be defined. As an example, the volume in a can V in the can filling
problem is measured as the product of mass m and density U of the product inside the
can ( V U u m ). Therefore, the QC (volume in the can V ) is written as a function of an
input parameter (density U) and an output parameter (mass m of the can). It is to be
noted that, in this example both mass and volume are output parameters; however, mass
can be calculated easily, and is one of the OPs. In contrast to that, volume is the only
required QC (RQC), which cannot be measured directly. Depending upon the
complexities of the process, one RQC may have to be written as a function of another
RQC. However, an exact mathematical relation between OPs and RQCs, without much
assumptions, is very difficult to obtain. Moreover, a mathematical relationship can be
defined for quantitative aspects; for qualitative aspects, they cannot be defined. In reality,
the QCs are controlled by adjusting OPs which involves years of experience of the
operator. Although, a mathematical relationship between OPs and QCs is hard to
obtain; but, due to the advance technical and scientific knowledge, the experience of the
operator can be used to develop a relationship between OPs and QCs. These types of
relations are known as fuzzy relations. In the following sub-sections, the detail method of
formulation is described.
Figure 1
Products
Product
output
characterstics
46
Figure 3
Aggregation of various output values into one fuzzy output and defuzzification
of output value
47
A brief discussion is provided for the construction of fuzzy controller, interested readers
are advised to refer Tahera et al. (2008) for detail methodology, and to refer Fakhreddine
and Clarence (2004) for the in-depth theory of fuzzy logic. To make a fuzzy controller,
that is, to represent QCs in terms OPs, initially the crisp values of the OPs is converted
into fuzzy values. This can be achieved by defining membership functions for the values
of X . For each value, one or more membership function maybe defined based on the
available relations between various OPs and QCs. By defining the membership
functions, each crisp value of X is converted into corresponding fuzzy value, between
[0, 1]. Two or more fuzzified OPs are combined using a connective or. The resulting
fuzzified value (single value) is used to calculate the corresponding fuzzified values
(set of values) of the QCs by a connective and. Then, all fuzzified values of a particular
QC are aggregated using an aggregation method. Finally, the crisp or defuzzified value of
the QC is obtained using a defuzzification method. In particular, we have selected
Mamdani implication system, where the connective or is replaced by min operator, the
connective and is replaced by max operator, the aggregation method is overlapping
the sets or selecting the maxima, and the defuzzification rule is centre of gravity method.
An algorithm to make a mamdani type fuzzy controller is given in Table 1.
The outputs from the fuzzy controller are the QCs which are now defined in terms of
OPs, and symbolically they can be written as:
yi
\i X
where, yi is the ith QC, \ i is the ith fuzzy relation and X are the set of OPs.
So far, a method to incorporate both qualitative and quantitative characteristics in cost
function has been done. \ i is the required relation, the next step will be to develop a cost
function that will penalise the non-conforming OPs, thereby giving the most suitable
target values for the OPs.
Table 1
Algorithm:
Define fuzzy sets;
For all the parameters (both OPs and RQCs){
Relate each crisp value of input to a fuzzy set using
membership functions.
};
Aggregate different OP using min operator;
Obtain corresponding RQC using max operator;
For each RQC{Aggregate all the obtain values};
For each RQC{Defuzzify the aggregate values};
48
D T y
i i
i 1
where Di are constants that represent relative importance among the QCs and the ith QC
is represented by yi , further each Ti is defines as:
Ti yi
( Pi 1 , Pi 2 )
Ai
Uniform Penalty
Figure 4
0
Pi1 d yi d Pi 2
and
ki 2 yi Pi1 yi Pi 2 Pi 2 yi d USLi
Ai
otherwise
Target
Interval
ki1
LSLi
ki2
i1
i2
Quality Characteristics
USLi
49
where yi are the fuzzy outputs or QCs, [ Pi1 Pi 2 ] are interval limits for yi and are
assumed to be known, and ki1 , ki 2 are the penalties for the undersize and oversized
product with in the upper specified limits (USLi ) and lower specified limits (LSLi ).
Ai is the penalty for the ith QC of the product which is not conforming to the specified
limits. The final step in developing the cost function will be integrating fuzzy controller
and UIT loss function. In the following sub-section, an approach is proposed to finalize
the formulation of cost function.
D T X
i i
i 1
where
Ti X
( Pi1 , Pi 2 )
ki1 \ i X Pi1 \ i X Pi 2
ki 2 \ i X Pi1 \ i X Pi 2
Ai
LSLi d \ i X Pi1
Pi1 d \ i X d Pi 2
Pi 2 \ i X d USLi
otherwise
This model is one of the novel models that incorporate targeting of qualitative
characteristics, while minimising the cost. In addition the proposed cost function, that is,
UIT loss function is a novel formulation that has not been found in the literature. This
model is suitable for any type of targeting, especially targeting the process with products
of multiple QCs. Many common mathematical assumptions and relations are eliminated
and replaced by fuzzy relationships. Conventional optimisation methods cannot be
applied to solve for optimal OPs for the above formulation. Thus, an alternative solution
method is necessary to solve this model. A genetic algorithm is proposed to obtain the
optimal process targets. The reason for using genetic algorithm is due to the complex
nature of the UIT loss function, as seen in the above equation. In the following section, a
genetic algorithm-based approach is presented.
Method of targeting
An aptly suited local search heuristic for this model is an evolutionary algorithm. This is
due to the fact that, evolutionary algorithms can be easily coupled with fuzzy functions
50
(Francisco, 2008). The solution procedure along with the operators of evolutionary
algorithm is defined in the following steps:
Step 1 Generate a random initial population of OPs. It is assumed that the OPs values
will have a known upper and lower limit.
Step 2 For each individual (OPs vector) in the population fitness values are calculated
using the T function.
Step 3 Crossover and mutation (evolutionary operators) are applied for the present
population. For the sake of simplicity, standard two point crossover with roulette wheel
selection is utilised, and a standard random mutation is done.
Step 4 The crossovered and mutated populations fitness value is calculated again.
Step 5 The unfit population is winnowed using rank-based selection method.
Step 6 Steps 26 are looped until any of the termination criterions is met. This
algorithm terminates when either maximum number of generations is reached or when
there is no change in the fitness value of the best individual over the generations.
For a clear illustration, an overall flow chart describing whole solution process is shown
in Figure 5, and the pseudo code is given in Table 2.
Figure 5
X LSL, USL
X LSL, USL
Y
Y
51
Pseudo code:
Initialize X(0);
Evaluate X(0);
t=1;
while (termination criterion not satisfied){
Crossover X(t);
Mutate X(t);
Evaluate X(t);
Select X(t);
Archive the best X obtained;
T=t+1; }
Illustrative example
To demonstrate the utility of the proposed model, a problem related to roll compaction
process that fits the situation under consideration (multiple QCs non-linearly related to
OPs) has been selected. This process is one of the widely used processes in
pharmaceutical companies for size enlargement or granulation. In brief, this problem has
the following situation:
Roll compaction is similar to any rolling process, where the raw material is fed
through two counter rotating drums (rolls) for compacting or enlarging the size of raw
material. Specifically in roll compaction, the raw materials are usually dry powders, and
the objective is to achieve content uniformity with size enlargement. Although, this
process is similar to rolling process, it has many addition parameters, which make
this process complicated. Specifically, materials flow, elastic, permeability and
compatibility properties, friction against roll surfaces, material of roll surfaces, rolls
dimension, pressure, speed and type of feed method are some of the critical control
process parameters. Various approaches are used to study the targeting aspect of roll
compaction process. Peter (2004) presented a comprehensive literature survey on various
approaches used to solve the parameters setting of roll compaction process.
Recently, Mansa et al. (2005, 2008) proposed a fuzzy modelling of this problem. The
objective of there study was to set the roll speed and roll gap for a specific material in
order to achieve required porosity under a selected maximum pressure and nip angle. For
the sake of numerical illustration of our proposed formulation, this predefined model
from the literature is selected. A UIT loss function is applied to the selected model and
genetic algorithm is applied to solve for the optimal target values of the process control
parameters. The fuzzy rules for the relationship between OPs and QCs are defined in
Tables 37.
52
Table 3
Bulk density
Relation between bulk density, roll speed, roll gap and ribbon density
Roll speed
Roll gap
Ribbon density
Low
Low
Low
High
Low
Low
High
Low
Low
Mid
Low
Low
Low
Mid
High
Low
Low
High
Low
Low
Low
High
High
Low
High
Low
Low
High
High
Low
High
High
High
Mid
Low
High
High
Mid
High
High
High
High
Low
Low
High
High
High
Low
Table 4
Low
Low
High
High
Table 5
Roll speed
Relation between roll speed, roll gap and average maximum pressure
Roll gap
Low
High
Low
Mid
High
Low
High
Low
Mid
Low
Low
Mid
Mid
Low
Mid
High
Low
High
Low
Low
Low
High
Mid
Low
High
High
Low
Table 6
Bulk density
Relation between bulk density, roll gap and average nip angle
Roll gap
Low
Low
High
Low
High
High
High
Low
Low
High
High
Low
53
Low
High
High
Low
The behaviour of input and output variables is shown in Figures 68. In Figure 6, the
behaviour of ribbon density by varying bulk density and roll speed is shown. Similar
surfaces are plotted between roll speed and bulk density vs. average maximum pressure
and average nip angle as in Figures 7 and 8. These figures represented the same data
given in Tables 37, with incorporation of gauss membership function and Mamdani type
rule of inference.
Figure 6
Roll speed, bulk density vs. ribbon density (see online version for colours)
Figure 7
Roll speed, bulk density vs. average maximum pressure (see online version for colours)
54
Figure 8
Roll speed, bulk density vs. average nip angle (see online version for colours)
Once the controller is made with Mamdani type fuzzy system and Gauss type
membership functions, it is fused into the UIT loss function. The allowed ranges for input
and output variables are specified in Tables 89 are used as interval limits in UIT loss
function. Once the fuzzified cost function is obtained, it is plugged into genetic
algorithm, as shown in flowchart (Figure 5), to minimise the cost function and to obtain
the required optimal target values of the process parameters.
The advantage of this model over the conventional approaches is that it incorporates
optimisation of qualitative parameters. Also, our model has the capability of indirect
targeting, which is very practical from the managerial aspect. That is, sometimes a
production process may be halted for process inspection, but the indirect targeting
reduces the loss in production time by utilising the indirect relationship between OPs
and QCs. Moreover, fuzzy logic has the capability to infer for unknown situations that
arise apart from the situations given in the rule base. This inference property is not
obtainable in any conventional approach to targeting problem.
Table 8
Problem A
[441, 453]
[0.5, 9]
[0, 5]
Problem B
[446, 454]
[0.5, 9]
[0, 5]
Problem C
[447, 463]
[0.5, 9]
[0, 5]
50
Table 9
Problem A
[0.72, 0.74]
[320, 330]
[4, 6]
Problem B
[0.70, 0.72]
[340, 350]
[4, 6]
Problem C
[0.72, 0.74]
[320, 350]
[3, 6]
55
Three set of problems are defined for the there different type of materials as given in
Tables 89. Based on the information of acceptable interval ranges of OPs and QCs, the
proposed genetic algorithm produced feasible solutions. The results are summarised in
Tables 1011. Since these results are not from exact solution method, it cannot be
considered as global optimal solutions. However, meta-heuristics like evolutionary
algorithms may produce a local optimal solution. Moreover, problem specific heuristics
may be designed to guarantee the global optimal solution.
The results show that the required output values are in between the targeted intervals
as desired. Porosity is one of the required quality characteristic which is dependent on
both input as well as output characteristic. The porosity is calculated using the formula:
BD
Porosity 1
PD
where BD is bulk density and PD is the particle density. From the results, it can be
seen that, this method can be used for the targeting of those values that are mutually
dependent on output as well as input values. Also, the proposed method can be used for
targeting attributes like colour, aroma and taste.
Although, the results obtained are within acceptable limits, that is, the values in
Table 10 are within the feasible interval. However, some values may not be frequently set
for the process means due to ease of management (e.g. to set some of the process means
may require lot of time and effort). Even some values may not be changed to the one
obtained by the genetic algorithm due to integrality constraints or module constraints.
The management, integrality or module constraint can also be incorporated in genetic
algorithm very easily. While, a roll compaction process is selected for illustration, the
proposed method can be designed and applied for almost any process that has a product
(output) with RQCs.
Table 10
Problem A
444.9697
3.4414
2.6283
Problem B
446.0037
8.8550
0.0642
Problem C
463.0000
1.5635
0.0319
Table 11
Problem A
0.7250
324.9
5.4
Problem B
0.7111
345.9
4.4
Problem C
0.7297
334.7
5.6
56
Conclusion
In this paper, a fuzzy logic-based process target model for products with multi QCs is
presented. A novel UIT loss function is used to penalise the deviation from the target.
A genetic algorithm is designed to obtain the optimal process targets. The proposed
approach is suitable when the relationship between process input parameters and output
QCs is fuzzy and can not be measured directly. The fuzzy logic modelling approach can
be used for both qualitative as well as quantitative QCs. The utility of the proposed
model is illustrated by an example from roller compaction process. This model can be
applied to a wide spectrum of industrial applications like food and beverages,
petrochemicals and pharmaceuticals, cements and paints. Apart from that, this approach
can be made as an online process control technique due to the fast convergence time of
genetic algorithms. Future research in this area may include application of this model
in different areas of manufacturing and a study to assess the performance of various
meta-heuristics in solving this model. Nonetheless, one of the critical points that were
assumed in the problem formulation was the shape of membership function. Future
research may also include the optimisation of shape and number of rules, using advance
techniques like neuro adaptive fuzzy learning. Type-2 fuzzy systems may also be used
when the process data are available as an enhancement to the rule base.
Acknowledgements
The authors acknowledge King Fahd University of Petroleum and Minerals for the
support in conducting this research. The valuable comments and suggestions made by the
editor and reviewers on the previous version of the manuscript are deeply appreciated.
References
Al-Sultan, K.S. (1994) An algorithm for the determination of the optimum target value for two
machines in series with quality sampling plans, Int. J. Production Research, Vol. 32, No. 1,
pp.3745.
Al-Sultan, K.S. and Pulak, M.F.S. (2000) Optimum target values for two machines in series with
100% inspection, European Journal of Operational Research, Vol. 120, No. 1, pp.181189.
Arcelus, F.J. (1996) Uniformity of the production vs. conformance to specifications in the canning
problem, K.S. Al-Sultan and M.A. Rahim (Eds.), a book chapter published in Optimization in
Quality Control, pp.243257.
Bandemer, H., Gottwald and Siegfried (1996) Fuzzy Sets, Fuzzy Logic, Fuzzy Methods with
Application. New York: John Wiley and Sons.
Bisgard, S., Hunter, W. and Pallensen, L. (1984) Economic selection of quality of manufacturing
product, Technometrics, Vol. 26, pp.918.
Bowling, R., Khassawneh, M.T., Kaewkuekool, S. and Cho, B.R. (2004) Markovian approach to
determining optimum process target levels for a multi stage serial production system,
European Journal of Operational Research, Vol. 159, No. 3, pp.636650.
Chan, W.M. and Ibrahim, R.N. (2006) Designing the optimal process means and the optimal
production run for a deteriorating process, Int. J. Advanced Manufacturing Technology,
Vol. 31, Nos. 34, pp.367373.
57
Chen, C.H. (2003) Determining the optimum process mean for the larger-the-better Weibull
quality characteristic, Int. J. Applied Science and Engineering, Vol. 1, No. 2, pp.172176.
Chen, C-H. (2006) The modified Pulak and Al-Sultans model for determining the optimum
process parameters, Communications in Statistics Theory and Methods, Vol. 35, No. 10,
pp.17671778.
Chen, C.H. and Chou, C.Y. (2004) Determining the optimum process mean of a one-sided
specification limit, The International Journal of Advanced Manufacturing Technology,
Vol. 20, No. 6, pp.439441.
Chen, C-H. and Khoo, M.B.C. (2008) Joint determination of optimum process mean and economic
specification limits for rectifying inspection plan with inspection error, Journal of the
Chinese Institute of Industrial Engineers, Vol. 25, No. 5, pp.389398.
Chen, Y.T. and Kumara, S.R.T. (1998) Fuzzy logic and neural networks for design of process
parameters: a grinding process application, Int. J. Production Research, Vol. 36, No. 2,
pp.395415.
Cho, B.R. (2002) Optimum process targeting for two quality characteristics using regression
analysis, Quality Engineering, Vol. 15, No. 1, pp.3747.
Chou, C.Y., Liu, H.R., Huang, X.R. and Chen, C.H. (2002) Economic-statistical design of
multivariate control charts using quality loss function, The International Journal of Advanced
Manufacturing Technology, Vol. 20, No. 12, pp.916924.
Chung-Ho, C. (2005) Determining the optimum process mean under a log-normal distribution,
Quality and Quantity, Vol. 39, No. 1, pp.119124.
Chung-Ho, C. (2006) Determining the optimum process mean for a mixed quality loss function,
The International Journal of Advanced Manufacturing Technology, Vol. 28, Nos. 56,
pp.571576.
Darwish, M.A. (2009) Economic selection of process mean for single-vendor single-buyer supply
chain, European Journal of Operational Research, Vol. 199, No. 1, pp.162169.
Duffuaa, S.O. and Siddiqui, A.W. (2002) Integrated process targeting and product uniformity
model for three class screening, Special Issue of Journal of Quality Reliability and Safety
Engineering, Vol. 9, No. 3, pp.261274.
Duffuaa, S.O. and Siddiqui, A.W. (2003) Process targeting with multi-class screening and
measurement error, Int. J. Production Research, Vol. 41, No. 7, pp.13731391.
Elshafei, M., Siddiqui, A.W. and Duffuaa, S.O. (2006) Optimal process targeting of multicomponents multi-characteristics products, Proceedings of the 4th International Conference
on Manufacturing Research (ICMR 2006), September.
Fakhreddine O.K. and Clarence W.D-S. (2004) Soft Computing and Intelligent Systems Design:
Theory, Tools and Applications. Toronto, Canada: Pearson Education Limited.
Fazel, Z.M.H., Trken, I.B. and Torabi, K.O. (2007) Type-2 fuzzy modeling for desulphurization
of steel process, Expert Systems with Applications, Vol. 32, No. 1, pp.157171.
Francisco, H. (2008) Genetic fuzzy systems: taxonomy, current research trends and prospects,
Evolutionary Intelligence, Vol. 1, No. 1, pp.2746.
Fung, R.Y.K., Chen, Y. and Tang, J. (2006) Estimating the functional relationships for quality
function deployment under uncertainties, Fuzzy Sets and Systems, Vol. 157, No. 1,
pp.98120.
Golhar, D. (1987) Determining of the best mean contents for a canning problem, Journal of
Quality Technology, Vol. 19, No. 2, pp.8284.
Golhar, D. and Pollock, S. (1992) Cost saving due to variance reduction in a canning problem,
IIE Transactions, Vol. 24, No. 1, pp.8992.
Golhar, D. and Pollock, S. (1988) Determining of the optimal process mean and the upper limit for
a canning problem, Journal of Quality Technology, Vol. 20, No. 3, pp.188199.
58
Hariga, M.A. and Al-Fawzan, M.A. (2005) Joint determination of target value and production
run for a process with multiple markets, Int. J. Production Economics, Vol. 96, No. 2,
pp.201212.
Holden, T. and Serearuno, M. (2005) A hybrid artificial intelligence approach for improving yield
in precious stone manufacturing, Journal of Intelligent Manufacturing, Vol. 16, No. 1,
pp.2138.
Hunter, W. and Kartha, C. (1977) Determining the most profitable target value for a production
process, Journal of Quality Technology, Vol. 9, No. 4, pp.176181.
Iqbal, A., He, N., Li, L. and Dar, N.U. (2007) A fuzzy expert system for optimizing parameters
and predicting performance measures in hard-milling process, Expert Systems with
Applications, Vol. 32, No. 4, pp.10201027.
Kolus, A. (2005) Determining Optimal Target Values for Two Processes in Series, MS Thesis,
King Fahd University of Petroleum and Minerals, Dhahran, Saudi Arabia.
Lee, M.K. and Elsayed, E.A. (2002) Process mean and screening limits for filling processes under
two-stage screening procedure, European Journal of Operational Research, Vol. 138, No. 1,
pp.118126.
Lee, M.K., Hong, S.H. and Elsayed, E.A. (2001) The optimum target value under single and
two-stage screenings, Journal of Quality Technology, Vol. 33, No. 4, pp.506514.
Mansa, R.F., Bridson, R.H., Greenwood, R.W., Barker, H. and Seville, P.K.J. (2008) Using
intelligent software to predict the effects of formulation and processing parameters on roller
compaction, Powder Technology, Vol. 181, No. 2, pp.217225.
Mansa, R.F., Bridson, R., Greenwood, R.W., Seville, P.K.J. and Barker, H. (2005) Using
intelligent software to predict the effects of the formulation and parameter on roller
compaction, 8th International Symposium on Agglomeration, European federation of
chemical engineering society of powder technology of Japan.
Montgomery, D. (2005) Introduction to Statistical Quality Control (5th ed.), New York: John
Wiley and Sons.
Patyra, M.J., Mlynek, D.M. (1996) Fuzzy Logic: Implementation and Applications, New York:
John Wiley and Sons.
Peter, K. (2004) Roll compaction/dry granulation: pharmaceutical applications, European Journal
of Pharmaceutics and Biopharmaceutics, Vol. 58, No. 2, pp.317326.
Rahim, M.A., Bhadury, J. and Al-Sultan, K.S. (2001) Joint economic selection of target mean and
variance, Engineering Optimization, Vol. 34, No. 1, pp.120.
Rahim, M.A. and Tuffaha, F. (2004) Integrated model for determining the optimal initial settings
of the process mean and the optimal production run assuming quadratic loss functions, Int. J.
Production Research, Vol. 42, No. 16, pp.32813300.
Roan, J., Gong, L. and Tang, K. (2000) Joint determination of process mean, production run size
and material order quantity for a container-filling process, Int. J. Production Economics,
Vol. 63, No. 3, pp.303317.
Springer, C. (1951) A method for determining the most economic position of a process mean,
Industrial Quality Control, Vol. 8, No. 1, pp.3639.
Suhr, R. and Batson, R.G. (2001) Constrained multivariate loss function minimization, Quality
Engineering, Vol. 13, No. 3, pp.475483.
Tahera, K., Ibrahim, R.N. and Lochert, P.B. (2008) A fuzzy logic approach for dealing with
qualitative quality characteristics of a process, Expert Systems with Applications, Vol. 34,
No. 4, pp.26302638.
Teeravaraprug, J. and Cho, B.R. (2002) Designing the optimal process target levels for multiple
quality characteristics, Int. J. Production Research, Vol. 40, No. 1, pp.3754.
Williams, W.W., Tang, K. and Gong, L. (2000) Process improvement for a container-filling
process with random shifts, Int. J. Production Economics, Vol. 66, No. 1, pp.2331.
59
Xiong, G., Nyberg, T.R., Zhang, Z-Y. and Xiong, G-Y. (2002) A kind of genetic fuzzy control
algorithm and its application in distilling tower pressure controller, IEEE Region 10 Annual
International Conference, Proceedings/TENCON, Vol. 3, pp.14551458.
Yang, Y., Shaibu, A-B., Cho, B.R. and Mayorga, M. (2008) The simultaneous optimization of
process target levels and specification region for dual quality characteristics, IIE Annual
Conference and Expo, pp.194199.
Zadeh, L.A. (1965) Fuzzy Sets, Information and Control, Vol. 8, No. 3, pp.338353.