robustness to demands. Existing approaches to MLT
reduction rely on flow control techniques that
required mathematical models to predict MLT.
Varied inspection can be viewed as a flow control
technique however, it does not require mathematical
modelling because of the fuzzy implementation–
which is useful in complex manufacturing situations
where the models are difficult to acquire.
2 LITERATURE REVIEW
2.1 Varied Inspection for Quality
Control in Mass Customization
Varied inspection is an aperiodic inspection method
compared to traditional methods. The inspection
system may choose to inspect or not inspect parts as
they pass through production based on factors such as
part quality, supply/demand, WIP, MLT,
bottlenecking, starving or other needs of the
manufacturer (Naidoo et al., 2016). The research was
focused on part quality and MLT reduction, whereas
previous research done by Naidoo et al., (2016)
focused on WIP reduction. MLT reduction was
desired as parts provide no profit while they remain
unfinished and in production. Through reducing the
amount of time on inspection, parts spend less time in
production thus reducing lead time. Shorter lead time
ensures better robustness to manufacturers in
supplying demands. Table 1 shows common
characteristics of varied inspection in terms of
advantages and disadvantages (Naidoo et al., 2016).
Table 1: Characteristics of Varied Inspection.
Advantages Disadvantages
Appraisal costs are reduced
through reduced inspection.
Could allow defective parts to
move throughout the system.
Can be used to prevent
bottlenecking by
increasing/decreasing the
number of inspected
products.
May result in external failure
costs when products fail at the
site of the customer.
Over-inspection is reduced.
High average consequence
costs.
Reduced average MLT as
reduced inspection reduces
overall production time.
Reduced WIP as some parts
are sent through the
production without
inspection.
The significant advantage of varied inspection (as
compared to 100% inspection and acceptance
sampling) was that the inspection frequency was not
fixed – it could be adjusted to suit the production
requirements. Varied inspection could be
implemented as a solution to slow inspection that
affects production rates (Davrajh and Bright, 2010).
However, Groover (2014) stated that this type of
inspection yielded high average consequence costs.
2.2 Fuzzy Logic Control for Production
Systems and Varied Inspection
FLCs had been used in production systems to
improve control since the 1990s (Homayouni et al.,
2009). However, Azadegan et al., (2011) stated that
there was minimal FL applications in the field of QC.
Complex manufacturing environments are difficult to
analytically model and probability theory cannot be
used to solve all manufacturing issues, which was
why fuzzy set theory was supported for control over
production (Tsourveloudis, 2000), (Gien, 1999). A
FLC was used in this research as it could handle
imprecise inputs and does not require a model of the
system to control it (Naidoo et al., 2016). Classical
control methods require accurate mathematical
models for effective control- fuzzy control is a
heuristic control approach thus the complex task of
obtaining mathematical models are not required. A
great advantage of FLCs is that it represents an
extension of human logic and can be based on human
evaluations, therefore it can replicate how a human
expert would control a system (Tsourveloudis, 2000).
FLCs have learning capabilities and can be improved
with other computational tools such as neural
networks and Evolutionary Algorithms (EA)
(Homayouni et al., 2009). Research done by Naidoo
et al. (2016) showed that a FLC could be used to
perform varied inspection for the purpose of WIP
reduction. This research was to investigate the effects
of varied inspection on MLT, where fuzzy controllers
are “Mamdani-type” with rules in the form of (1).
IF X is A AND Y is B THEN Z is C
(RuleWeight)
(1)
X and Y are the inputs with A and B linguistic values
respectively, and Z is the output with C linguistic
values. Linguistic values are the fuzzy sets that
consist of membership functions (Ioannidis et al.,
2004). The “RuleWeight” determines the strength of
the rule with ‘1’ having the strongest weight. The
fuzzy controllers designed used minimum for “AND”
and the centroid method for defuzzification. The
controllers were designed with the Fuzzy Logic
Toolbox
®
in Simulink
®
.