Selecting the Optimum Inspection Method for Preventive Maintenance

Abstract

With technological advances, our lives have become increasingly dependent on various facilities. It is important to detect deterioration or failure symptoms using inspection or monitoring devices for enabling preventive maintenance to keep them working. For this purpose, it is necessary to select the proper inspection method to detect deterioration and failure symptoms of each equipment item. In this paper, we propose a systematic method to determine the optimum combination of inspection methods to minimize the sum of inspection costs and the losses due to equipment failure considering the detection capability of both humans and inspection or monitoring devices. The proposed method was applied to the equipment in a distribution center to demonstrate its effectiveness.

Full text

ScienceDirect
Available online at www.sciencedirect.com
Available online at www.sciencedirect.com
ScienceDirect
Procedia CIRP 00 (2017) 000–000
www.elsevier.com/locate/procedia
2212-8271 © 2017 The Authors. Published by Elsevier B.V.
Peer-review under responsibility of the scientific committee of the 28th C IRP Design Conference 2018.
28th CIRP Design Conference, May 2018, Nantes, France
A new methodology to analyze the functional and physical architecture of
existing products for an assembly oriented product family identification
Paul Stief *, Jean-Yves Dantan, Alain Etienne, Ali Siadat
École Nationale Supérieure d’Arts et Métiers, Arts et Métiers ParisTech, LCFC EA 4495, 4 Rue Augustin Fresnel, Metz 57078, France
* Corresponding author. Tel.: +33 3 87 37 54 30; E-mail address: paul.stief@ensam.eu
Abstract
In today’s business environment, the trend towards more product variety and customization is unbroken. Due to this development, the need of
agile and reconfigurable production systems emerged to cope with various products and product families. To design and optimize production
systems as well as to choose the optimal product matches, product analysis methods are needed. Indeed, most of the known methods aim to
analyze a product or one product family on the physical level. Different product families, however, may differ largely in terms of the number and
nature of components. This fact impedes an efficient comparison and choice of appropriate product family combinations for the production
system. A new methodology is proposed to analyze existing products in view of their functional and physical architecture. The aim is to cluster
these products in new assembly oriented product families for the optimization of existing assembly lines and the creation of future reconfigurable
assembly systems. Based on Datum Flow Chain, the physical structure of the products is analyzed. Functional subassemblies are identified, and
a functional analysis is performed. Moreover, a hybrid functional and physical architecture graph (HyFPAG) is the output which depicts the
similarity between product families by providing design support to both, production system planners and product designers. An illustrative
example of a nail-clipper is used to explain the proposed methodology. An industrial case study on two product families of steering columns of
thyssenkrupp Presta France is then carried out to give a first industrial evaluation of the proposed approach.
© 2017 The Authors. Published by Elsevier B.V.
Peer-review under responsibility of the scientific committee of the 28th CIRP Design Conference 2018.
Keywords: Assembly; Design method; Family identification
1. Introduction
Due to the fast development in the domain of
communication and an ongoing trend of digitization and
digitalization, manufacturing enterprises are facing important
challenges in today’s market environments: a continuing
tendency towards reduction of product development times and
shortened product lifecycles. In addition, there is an increasing
demand of customization, being at the same time in a global
competition with competitors all over the world. This trend,
which is inducing the development from macro to micro
markets, results in diminished lot sizes due to augmenting
product varieties (high-volume to low-volume production) [1].
To cope with this augmenting variety as well as to be able to
identify possible optimization potentials in the existing
production system, it is important to have a precise knowledge
of the product range and characteristics manufactured and/or
assembled in this system. In this context, the main challenge in
modelling and analysis is now not only to cope with single
products, a limited product range or existing product families,
but also to be able to analyze and to compare products to define
new product families. It can be observed that classical existing
product families are regrouped in function of clients or features.
However, assembly oriented product families are hardly to find.
On the product family level, products differ mainly in two
main characteristics: (i) the number of components and (ii) the
type of components (e.g. mechanical, electrical, electronical).
Classical methodologies considering mainly single products
or solitary, already existing product families analyze the
product structure on a physical level (components level) which
causes difficulties regarding an efficient definition and
comparison of different product families. Addressing this
Procedia CIRP 80 (2019) 512–517
2212-8271 © 2019 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/3.0/)
Peer-review under responsibility of the scientific committee of the 26th CIRP Life Cycle Engineering (LCE) Conference.
10.1016/j.procir.2019.01.090
© 2019 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/3.0/)
Peer-review under responsibility of the scientic committee of the 26th CIRP Life Cycle Engineering (LCE) Conference.
Available online at www.sciencedirect.com
ScienceDirect
Procedia CIRP 00 (2018) 000–000
www.elsevier.com/locate/procedia
2212-8271 © 2019 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/3.0/).
Peer-review under responsibility of the scientific committee of the 26th CIRP Life Cycle Engineering (LCE) Conference.
doi:10.1016/j.procir.2017.04.009
26th CIRP Life Cycle Engineering (LCE) Conference
Selecting the Optimum Inspection Method for Preventive Maintenance
N. Kuboki*, S.Takata
School of Creative Science and Engineering, Waseda University, Tokyo, Japan
* Corresponding author. Tel.: 03-5286-3299; fax: 03-3202-2543. E-mail address: nk-1208.w-63@fuji.waseda.jp
Abstract
With technological advances, our lives have become increasingly dependent on various facilities. It is important to detect deterioration or failure
symptoms using inspection or monitoring devices for enabling preventive maintenance to keep them working. For this purpose, it is necessary to
select the proper inspection method to detect deterioration and failure symptoms of each equipment item. In this paper, we propose a systematic
method to determine the optimum combination of inspection methods to minimize the sum of inspection costs and the losses due to equipment
failure considering the detection capability of both humans and inspection or monitoring devices. The proposed method was applied to the
equipment in a distribution center to demonstrate its effectiveness.
© 2019 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/3.0/).
Peer-review under responsibility of the scientific committee of the 26th CIRP Life Cycle Engineering (LCE) Conference.
Keywords: inspetion method; preventive maintenance; inspection and monitoring devices; causal relation; Genetic Altgorithm
1. Introduction
We developed a number of automated facilities to enhance
productivity in various industries. Nowadays, introduction of
automated equipment is indispensable due to manpower
shortage in the developed countries. The automation trend has
increased even in the developing countries because of soaring
labor costs.
With the rapid expansion of e-commerce in recent years, a
number of highly automated large-scale distribution facilities,
for example, have been built. Such facilities are required to
operate 24 hours a day, 7 days a week, to avoid suspension of
goods shipment. For this purpose, proper execution of
preventive maintenance is essential to reduce malfunctioning
and to avoid sudden breakdown of the facilities as much as
possible.
However, the inspection workload for preventive
maintenance of a large-scale distribution facility is enormous
because it encompasses a large amount of equipment such as
conveyors and sorters. As a result, facility operators face
manpower shortage and increasing inspection costs despite the
use of automated equipment to solve labor shortage and to cut
costs. To resolve this dilemma, businesses need to introduce
proper inspection devices or on-line monitoring devices to
reduce inspection workload.
However, the detection capability of inspection and
monitoring devices is not necessarily superior to that of human
inspection. Therefore, the advantages of introducing such
devices should be properly evaluated. In this research, we
propose a method to determine an optimum combination of
inspection and monitoring devices, as well inspection intervals,
to detect deterioration and failures occurring in the equipment.
In Section 2, we describe the key considerations for
determining the inspection method and how we deal with these
issues. In Section 3, we propose a procedure to determine the
optimum inspection method considering the factors discussed
in Section 2. In Section 4, we provide an application example,
where the proposed method is applied to a distribution facility
consisting of multiple equipment. Finally, we conclude the
paper in Section 5.
Available online at www.sciencedirect.com
ScienceDirect
Procedia CIRP 00 (2018) 000–000
www.elsevier.com/locate/procedia
2212-8271 © 2019 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/3.0/).
Peer-review under responsibility of the scientific committee of the 26th CIRP Life Cycle Engineering (LCE) Conference.
doi:10.1016/j.procir.2017.04.009
26th CIRP Life Cycle Engineering (LCE) Conference
Selecting the Optimum Inspection Method for Preventive Maintenance
N. Kuboki*, S.Takata
School of Creative Science and Engineering, Waseda University, Tokyo, Japan
* Corresponding author. Tel.: 03-5286-3299; fax: 03-3202-2543. E-mail address: nk-1208.w-63@fuji.waseda.jp
Abstract
With technological advances, our lives have become increasingly dependent on various facilities. It is important to detect deterioration or failure
symptoms using inspection or monitoring devices for enabling preventive maintenance to keep them working. For this purpose, it is necessary to
select the proper inspection method to detect deterioration and failure symptoms of each equipment item. In this paper, we propose a systematic
method to determine the optimum combination of inspection methods to minimize the sum of inspection costs and the losses due to equipment
failure considering the detection capability of both humans and inspection or monitoring devices. The proposed method was applied to the
equipment in a distribution center to demonstrate its effectiveness.
© 2019 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/3.0/).
Peer-review under responsibility of the scientific committee of the 26th CIRP Life Cycle Engineering (LCE) Conference.
Keywords: inspetion method; preventive maintenance; inspection and monitoring devices; causal relation; Genetic Altgorithm
1. Introduction
We developed a number of automated facilities to enhance
productivity in various industries. Nowadays, introduction of
automated equipment is indispensable due to manpower
shortage in the developed countries. The automation trend has
increased even in the developing countries because of soaring
labor costs.
With the rapid expansion of e-commerce in recent years, a
number of highly automated large-scale distribution facilities,
for example, have been built. Such facilities are required to
operate 24 hours a day, 7 days a week, to avoid suspension of
goods shipment. For this purpose, proper execution of
preventive maintenance is essential to reduce malfunctioning
and to avoid sudden breakdown of the facilities as much as
possible.
However, the inspection workload for preventive
maintenance of a large-scale distribution facility is enormous
because it encompasses a large amount of equipment such as
conveyors and sorters. As a result, facility operators face
manpower shortage and increasing inspection costs despite the
use of automated equipment to solve labor shortage and to cut
costs. To resolve this dilemma, businesses need to introduce
proper inspection devices or on-line monitoring devices to
reduce inspection workload.
However, the detection capability of inspection and
monitoring devices is not necessarily superior to that of human
inspection. Therefore, the advantages of introducing such
devices should be properly evaluated. In this research, we
propose a method to determine an optimum combination of
inspection and monitoring devices, as well inspection intervals,
to detect deterioration and failures occurring in the equipment.
In Section 2, we describe the key considerations for
determining the inspection method and how we deal with these
issues. In Section 3, we propose a procedure to determine the
optimum inspection method considering the factors discussed
in Section 2. In Section 4, we provide an application example,
where the proposed method is applied to a distribution facility
consisting of multiple equipment. Finally, we conclude the
paper in Section 5.

N. Kuboki et al. / Procedia CIRP 80 (2019) 512–517 513
2 Author name / Procedia CIRP 00 (2019) 000–000
2. Key considerations in inspection method selection
2.1. Factors to be considered in inspection method selection
In this research, we consider the selection of an inspection
method for a facility consisting of multiple equipment, such as
conveyors and sorters in the case of a distribution center.
The equipment consists of multiple components such as
belts and pulleys, and deterioration and failures could occur in
each component due to aging and the stress generated by
operations of the facility.
There are three kinds of inspection to detect deterioration
and failure: a walk-around visual inspection by human
operators, on-line monitoring by fixed sensors, and a walk-
around inspection by human operators using sensory devices.
In this paper, deterioration is defined as changes in physical
and chemical attributes of items, such as deformation, breakage,
wear, and corrosion (we use “item” as a collective term for
equipment and components in this paper). On the other hand,
failure is defined as a change in the item’s condition to a state
in which the required function cannot be performed.
To select an appropriate inspection method, we need to
consider the following factors. 1) Because of the causal
relations among deterioration and failures, we cannot select an
inspection method for each deterioration and failure
independently. We need to consider the fact that detection of
certain deterioration and failure occurrences could prevent
other such occurrences. 2) Even when a proper inspection
method is applied, the inspection results are not necessarily
correct. One could miss the deterioration and failure symptoms
or raise a false alarm. Furthermore, the detection capability of
the inspection depends on not only the capability of the devices
but also the relations between inspection interval and the
progress rate of deterioration. 3) In selecting the inspection
methods, we should evaluate the effects of inspection, such as
the cost of inspection, preventive maintenance cost, breakdown
maintenance cost when the inspection fails to detect the
deterioration and failure, and opportunity losses generated by
stoppage of the facility.
There is a chain of causal relations; for example, the bearing
wear of the drive pulley causes a meandering motion in the belt,
which damages the belt edge due to friction with the conveyor
frame. Detection of the bearing wear of the drive pulley could
therefore prevent the belt meandering. This means that it is not
necessary to inspect belt meandering. However, when the belt
meandering is caused not only by the bearing wear in the drive
pulley or when the detection capability for the bearing wear of
the drive pulley is insufficient, the belt meandering needs to be
inspected. We need to consider such relations in selecting the
inspection methods.
The inspection produces four types of results as shown in
Table 1 with certain probabilities depending on the equipment
conditions and inspection capability. The inspection could
either successfully detect or miss deterioration and failure
symptoms. When the equipment is in a normal condition, the
inspection could provide a correct result or raise a false alarm.
In each case, the inspection result leads to different effects. The
probability of these inspection results depends on not only the
capability of the inspection means but also the relations
between the inspection interval and the rate of deterioration. If
the inspection interval is long in relation to the rate of progress
of the deterioration, no inspection would be carried out during
the appearance of failure symptoms and before the failure
occurs.
These four types of inspection results produce different
effects in terms of costs or opportunity losses. In determining
an overall optimum inspection method for the facilities, we
need to evaluate the total effects taking these factors into
account in an appropriate manner.
A number of studies have been conducted regarding
selection of the inspection method. For example, a method for
selecting the most cost-effective arrangement of sensors based
on graph theory has been proposed considering the causal chain
and using the state transitions of components due to
deterioration and failures [1, 2]. Many studies on inspection
methods have considered the relation between the inspection
interval and the rate of progress of deterioration [3].
Considering the failure detection capability of sensors, Wang et
al. proposed a sensor selection method to select the optimal
sensor arrangement, in terms of cost, with the required failure
detection capability [4]. Tsutsui et al. proposed a maintenance
planning system based on effect evaluation, in which
maintenance cost and expected losses, such as production loss
due to stoppage of the facility, were considered [5].
However, few works have studied a largely optimum method
to select a combination of inspection methods for all equipment
in the facility, considering all factors described above.
2.2. Causal relations of deterioration and failures
To enumerate the deterioration and failure modes, which
should be considered in determining the inspection method, we
need to identify the lowest level of items subject to maintenance
(hereafter, ISM).
Potential deterioration and failures of ISMs as well as their
causal relations are identified by deterioration and failure
analysis. The result is represented in the form of a causal
relation chart, shown in Figure 1. Deterioration occurred in the
j-th equipment item, which could be the origin of the causal
chain, depicted as ܦܨ
௝ଵ, ܦܨ
௝ଶ on the left side of Figure 1. The
failures induced by the deterioration are represented in the
middle of the figure as ܦܨ
௝ଷ, ܦܨ
௝ସ. After completing the causal
chain analysis, we identify the state quantities, which represent
the degree of deterioration and failures, and enumerate the
inspection method applicable to them, as depicted on the right
side of the figure. In general, multiple state quantities and
multiple inspection methods could be considered for a certain
deteriorationܦܨ
௝ଵ; they are expressed as ܵܳଵ
௝ଵand ܵܳଶ
௝ଵ…, and
ݔଵ
௝ଵand ݔଶ
௝ଶ….
Basically, causal relation charts are created for each ISM.
However, the deterioration and failures of one ISM could be
Table 1. Four types of inspection results
inspection results
success fail
condition
normal no alarm false alarm
symptomatic detection missing

514 N. Kuboki et al. / Procedia CIRP 80 (2019) 512–517
Author name / Procedia CIRP 00 (2019) 000–000 3
related to those of other ISMs. In this case, one causal relation
chart could contain the deterioration and failures of multiple
ISMs. Therefore, we need a certain means to understand the
relations between the deterioration and failures of all ISMs. For
this purpose, we use a relation matrix in which the identified
deterioration and failures of all ISMs are listed in the first row
and column and the causal relation between the deterioration
and failures is indicated by a “1” in the corresponding cell of
the matrix as shown in Table 2.
2.3. Results of Inspections
Inspection results are divided into the following four cases:
1) degradation or failure symptoms exist, and are detected
(detection),
2) degradation or failure symptoms exist, but are not detected;
thus, the failure occurs (missing),
3) the items are in a normal condition, and the inspection
indicates this result (no alarm),
4) the items are in a normal condition, but the inspection
indicates deterioration or failure symptoms (false alarm).
The inspection results depend on the means of inspection
and the inspection interval. There are many means of inspection,
such as visual inspection by human, inspection using portable
devices, and on-line monitoring with fixed sensors. Whether the
inspection reveals the deterioration or failure symptoms
depends on the skill of the inspectors or capability of the
inspection or monitoring devices.
The detection capability of the inspection also depends on
the inspection interval. Figure 2 shows the relation between the
progress of deterioration and the inspection interval. In the
figure, the change in the state quantity with respect to time is
represented. Two horizontal broken lines are shown in the
figure. The lower line indicates the minimum state quantity
level where failure symptoms can be detected. The upper line
indicates the limit above which the required function cannot be
performed. When inspection is executed between t1 and t2, the
failure symptom can be recognized by the inspection, and
successful preventive action can be initiated. However, if an
inspection is not executed during this period because of a long
inspection interval, the deterioration or failure symptoms are
missed.
In analyzing the inspection results, it should be noted that
the “missing” of a certain deterioration or failure symptom
could induce deterioration and failures downstream in the
causal chain.
2.4. Effect evaluation of inspections
The effectiveness of the inspection method should be
properly evaluated for determining the inspection method. For
this purpose, we need to consider the following effects as the
results of inspections.
E1: Inspection costs, including the cost of the monitoring
device and labor cost.
E2: Cost of treatment, such as repair and replacements, for
preventive maintenance and breakdown maintenance,
including the cost of items and labor cost for the treatment.
E3: Losses associated with interruption of facility operations
due to inspection, preventive maintenance, and
breakdown maintenance.
E1 is the total cost for the inspection of each deterioration
or failure symptom. It includes the cost of inspection, the costs
of monitoring devices, and labor cost. Regarding the device
cost, whether it can detect multiple deterioration or failure
symptoms should be considered.
E2 is the cost of taking action depending on the inspection
results. The cost of preventive maintenance action is included
in the case of detection, while the cost of breakdown
maintenance action, which is usually more expensive than
preventive maintenance, is included in the missing case. The
false alarm case includes the cost of checking whether
deterioration and failures actually occurred.
E3 is also calculated depending on the inspection results.
Since preventive maintenance could be carried out at a
Deterioration Failure Inspection
method
State
quantity
Figure 1. Causal relation chart
Figure 2. Relation between the progress of deterioration
and the inspection interval
Detected
amount
Symptoms
occurred
Time
Equipment stop
or induce othe
r
deterioration or failures
t1 t2
Table 2. Matrix of causal relations
…
1
1
1
…

N. Kuboki et al. / Procedia CIRP 80 (2019) 512–517 515
4 Author name / Procedia CIRP 00 (2019) 000–000
convenient time, from the operational viewpoint, the loss due
to operation interruption is usually lower for detection than for
the missing case, which requires urgent breakdown
maintenance. In the case of false alarm, the time for precise
checking of component deterioration through disassembling,
for example, should be considered.
For evaluating E2 and E3, we need to estimate the
occurrence probabilities of the three cases of inspection
results—detection, missing, and false alarm—because “no
alarm” does not lead to costs of E2 and E3. They depend on the
occurrence probabilities of deterioration and failures, the
detection capability of inspection, and the causal relations
among deterioration and failures.
In this research, we assume that the occurrence probabilities
of deterioration and failures can be estimated from various data
such as maintenance history and reliability tests. We also
assume that the deterioration and failures downstream of the
causal chain are provoked by missing in upstream deterioration
or failure symptoms in the inspection.
Based on the occurrence probability of each deterioration
and failure and the detection capability of each inspection
method, we can calculate the number of times the inspection
results (detection, missing, and false alarm) would occur．
Figure 3 shows how to calculate these numbers.  denotes
the number of occurrences of the q-th deterioration and failure
of the j-th item of equipment, 
, within a certain period of
time. When the probabilities of the inspection results (detection,
missing, and false alarm) for 
, obtained by the inspection
method , are expressed by 
, 
, 
, the number of
times of the inspection results, 

, 

, 

, are given by
the following equations.




 (1)




 (2)




 (3)
When 
 is not the origin of the causal chain,  is
estimated as the sum of the numbers of the missing of

of 

 , which leads to 
 , considering the causal
relations among the deterioration and failures. For example,
 of 
 is calculated as the sum of the number of
missing, 

 and 

 of 
 and 
 as shown in
Figure 3.
When the values of E2 and E3 effects generated by each
detection, missing, and false alarm occurrence of deterioration
and failure, 
, are denoted as 
, 
 , and 
 (i=2 or 3),
the cumulative effects of E2 and E3 types, denoted as Eci (i= 2
or 3), with one of the possible combinations of inspection
methods are calculated by Equation (4).
 











 (4)
3. Procedure for inspection method selection
3.1. Outline of the procedure
The optimum inspection method, which is a combination of
inspection methods for all deterioration and failure occurrences,
is selected by the following procedure:
1) Analysis of the deterioration and failures and their causal
relations in the form of a causal relation chart,
2) Estimation of the number of inspection results, namely,
detection, missing, and false alarm,
3) Evaluation of effects,
4) Selection of the optimal combination of inspection methods
for deterioration and failures enumerated in Step 1 by using
Genetic Algorithm (GA).
3.2. Approximate optimization of the combinations of
inspection methods
To select the optimal inspection method for the facility, we
need to evaluate the effects of all combinations of inspection
methods for all deterioration and failures identified in the
deterioration and failure analysis (the inspection methods
include the case where no inspection is executed). However, all
possible combinations of inspection methods are difficult to
evaluate because of the huge number of candidates. We
therefore adopt GA for an approximate optimization of
inspection methods.
Table 3 shows the chromosome structure. The locus of the
gene corresponds to the deterioration or failure occurring in the
equipment, and each gene represents the inspection method
number, which identifies the inspection method applied to the
deterioration or failure. We use the cumulative effects defined
in equation (4) as the fitness value.
To generate the initial population, we first independently
select the best combinations of inspection methods for all
equipment by conducting a full search. They are then combined
to generate the chromosome, which is included in the initial
population to accelerate convergence. One chromosome is
generated by the above method, and the rest of the
chromosomes of the initial population are generated by
assigning inspection methods randomly to each deterioration
and failure.
Deterioration Failure Inspection
method
Occurrence :known
Detection :
Missing :
False Alarm :
Occurrence :known
Detection :
Missing :
False Alarm :
Occurrence :
Detection :
Missing :
False Alarm :
Figure 3. Calculation of deterioration and failure
Table 3. Chromosome expression
Gene locus
(Deteriorat ion or
failure) … …
Gene
(Inspection method
number)
… …
Equipment number Deterioration or failure number

516 N. Kuboki et al. / Procedia CIRP 80 (2019) 512–517
Author name / Procedia CIRP 00 (2019) 000–000 5
Regarding the genetic operations, 2-points crossover is
applied (crossover points are random) and the same number of
chromosomes as the current population are generated. With
regard to mutation, if a randomly selected locus of mutation
corresponds to a deterioration or failure, with portable
inspection devices being the applicable inspection method, they
are preferentially selected as the alternative inspection method.
Such preferential selection is executed with a probability of
50% of the total mutation operations. The rest of the mutation
is executed randomly.
For the next generation, individuals with the top 5% fitness
values among the population generated by the crossover and
mutation operations as well as the current population are
selected by the elite preservation strategy, and the remaining
95% of the individuals are selected by roulette selection.
4. Application example
4.1. Target facility to be maintained
We applied the proposed method to a distribution center
facility. It has 97 equipment, consisting of one sorter and 15
types of conveyors. We first extracted the ISMs by structural
development, and analyzed their deterioration and failures. We
identified 10 to 30 types of deterioration or failure for each
equipment item. The total number of deterioration and failure
states identified was 1075. For each deterioration or failure, 2
to 5 kinds of possible inspection methods were listed. Figure 4
shows an example of the results of deterioration and failure
analysis.
4.2. Optimization operation
We selected the optimum combination of inspection
methods for each deterioration and failure from the candidates
of inspection methods identified by the deterioration and
failure analysis, using GA as described in Chapter 3. The values
required for effects evaluation, such as cost of the item,
personnel expenses, and cost of monitoring device, were set
based on interviews with experts of the distribution equipment
manufacturer.
Regarding detection capability of the inspection methods,
we also rely on experts of the distribution equipment
manufacturer, because objective data were not available. We
asked them to evaluate the occurrence probabilities of detection,
missing, and false alarm on a three-point scale (high: 3,
medium: 2, and low: 1). Since detection and missing are
mutually exclusive events, we set their occurrence probabilities
such that they add up to 1. The probability of false alarm was
determined from the ratio of the evaluated levels of detection
and missing. Consequently, the occurrence probabilities of
detection, missing, and false alarm, denoted as 
, 
, and

 respectively, are calculated according to the following
equations, where ,  and  indicate the evaluated
occurrence probabilities when the inspection method  is
applied to the q-th deterioration of the j-th equipment item.


 (5)


 (6)


 (7)
Regarding inspection timing, we estimate the probability of
inspection execution between  and  in Figure 2. We assume
that  and  are variable and follow the normal
distributions 
 and 
respectively.  is
set to the replacement time recommended by the manufacturer,
and  is set to the value multiplied by  by fixed
magnification. This time, the magnification was set to 1.5. 
and  are assumed to be 5% of  and  Assuming
biannual inspections, we estimated the probability of
inspection execution between  and  by means of Monte
Carlo simulation.
Approximate optimization was performed with the
parameters for GA operations shown in Table 4. If we
generated a chromosome corresponding to each deterioration
or failure on a one-to-one basis, the size of the chromosome
would be 1075. To reduce this number in order to increase the
convergence speed in GA operations, we applied the same
inspection methods to similar type of equipment operating with
the same load and speed. Consequently, the size of the
chromosome decreased to 327.
We considered the following two conditions in performing
approximate optimization:
1) limiting the cost of inspection and monitoring devices;
2) limiting the total manual inspection time considering the
shortage in manpower, even if the cost of monitoring
devices increases.
In case 1), we set the cost of the monitoring device to 20
million yen or less. In case 2), we set the manual inspection
time to 500 hours or less.
These constraints were considered when chromosomes were
generated in GA operations. Those that did not satisfy these
constraints were treated as lethal genes and the generation was
Deterioration Failure
Belt
meandering
Contact of belt
with flame
Belt end
damage
Bearing wear
of pulley
Elongation
of belt
Inspection
method
Vibration Vibration
sensor
Noise meterSound
State
quantity
Position of
take-up pulley
Limit switch
Photoelectric
sensor
Amount of
belt elongation
Amount of
misalignment
Figure 4. An example of the results of deterioration and failure analysis
Table 4. Parameters of GA
Population number 100
Generation number 3,000
Crossover 2 points crossover (crossover points are random)
Mutation 5% probability for each generation
Selection method Elite method (5%)
+ Roulette method (95%)

N. Kuboki et al. / Procedia CIRP 80 (2019) 512–517 517
6 Author name / Procedia CIRP 00 (2019) 000–000
repeated until the generated chromosome satisfied the
constraints.
4.3. Results and discussion
Table 5 shows the total manual inspection time, the labor
cost for inspection, the monitoring device costs, the cost of
maintenance treatments, and the losses associated with the
interruption of the operation in each case. As seen in the table,
labor costs and total manual inspection time are large but the
cost of monitoring devices is low when the device cost is
limited. On the other hand, total manual inspection time and
labor costs for inspection are small but the cost of monitoring
devices is high where the total manual inspection time is
limited.
Regarding inspection methods, wear of pulley bearing, belt
elongation, and belt meandering are all visually inspected in the
case of a constraint on the device cost. On the other hand, with
a constraint on inspection time, acoustic sensors and
photoelectric sensors are selected for wear of the pulley bearing
and belt elongation, respectively, and total manual inspection
time and labor cost are reduced. However, if the detectability
of the inspection or monitoring devices is not sufficient, the
number of missing events increases, resulting in higher
breakdown maintenance cost and losses due to interruptions of
operations. Since losses due to interruptions of operations are
very large as shown in Table 5, monitoring devices with high
probability of missing are difficult to select. Therefore, it is
important to secure the detection capability of the inspection
and monitoring devices for lightening the inspector’s workload.
5. Conclusion
In this research, we propose a method to select the optimal
combinations of inspection methods for a facility in terms of
the cumulative effects induced by inspection and maintenance
treatment costs as well as losses caused by interruption of the
operations. In applying the method, we consider the causal
relations of deterioration and failures and the results of
inspection, namely, detection, missing, and false alarm.
The proposed method was applied to a distribution center
facility. We executed approximate optimization with two kinds
of constraints: a constraint on the device cost and a constraint
on manual inspection time. Results show that inspection and
monitoring devices require a certain level of detection
capability to replace manual inspection.
References
[1] Zhang G, Vachtsevanos G. A methodology for optimum sensor
localization/selection in Fault Diagnosis. 2007 IEEE Aerospace
Conference Proceedings; 2007.
[2] Bhushan M, Rengaswamy R. Design of sensor network based on the signed
directed graph of the process for efficient fault diagnosis. Industrial &
Engineering Chemistry Research; 2000; 39. 4: 999-1019.
[3] Wang WB. An overview of the recent advances in delay-time-based
maintenance modelling. Reliability Engineering & System Safety; 2012;
106: 165-178.
[4] Wang J, Sun WJ, Li T, Cao YY, Wang XM. An Innovative sensor selection
and optimization method considering reliability and detectability. Proc. of
IEEE 11th Conference on Industrial Electronics and Applications (ICIEA);
2016: 946-951.
[5] Tsutsui M, Takata S. Life cycle maintenance planning method in
consideration of operation and maintenance integration. Production
Planning & Control; 2012; 23, 2-3: 183-193.
Table 5. Comparison of results under two conditions
Case 1 Case 2
Total manual inspection time (hour) 1,667 499
Labor cost for inspection (Thousand yen) 57,012 26,310
Monitoring device cost
(Thousand yen) 2,255 28,175
Cost of the maintenance treatment
(Thousand yen) 355,089 381,489
Loss associated with shutdown of equipment
(Thousand yen) 3,594,951 3,709,261