Content uploaded by Eduard Shevtshenko
Author content
All content in this area was uploaded by Eduard Shevtshenko on Sep 30, 2015
Content may be subject to copyright.
Proceedings of TMCE 2012, May 7–11, 2012, Karlsruhe, Germany, Edited by I. Horváth, Z. Rusák, A. Albers and M. Behrendt
Organizing Committee of TMCE 2012, ISBN 978-90-5155-082-5
1013
RELIABILITY ANALYSIS MODULE DEVELOPMENT FOR PRODUCTION ROUTE
ELABORATION
Marina Pribytkova
Department of Mechanical EngineerinG
Tallinn University of Technology
Estonia
marina.pribytkova@gmail.com
Jevgeni Sahno
Department of Mechanical Engineering
Tallinn University of Technology
Estonia
jevgeni.sahno@gmail.com
Tatyana Karaulova
Eduard Shevtshenko
Department of Mechanical Engineering, Tallinn University of Technology
Tallinn, Estonia
{tatjana.karaulova eduard.shevtshenko}@ttu.ee
Meysam Maleki
V. Cruz-Machado
UNIDEMI, Department of Mechanical and Industrial Engineering
Faculdade de Ciencias e Tecnologia da Universidade Nova de Lisboa
Portugal
{maleki, vcm}@fct.unl.pt
ABSTRACT
This paper is intended for readers interested in
production routes reliability improvement and
related decision making in manufacturing
enterprises. Our current research is devoted to the
structure and functionality of reliability analysis and
results from collaboration between Tallinn
University of Technology and UNIDEMI research
team members. The proposed framework facilitate
decision making process through the reliability
analysis module, which combines the data stored in
ERP system with data received from failure mode
and effects analysis table. The reliability analysis
module structure shows how to extract the
knowledge, required for selection of appropriate
corrective action, which improves the reliability of
production route. The robust decision making
process is supported by Bayesian Belief Network,
that is built on faults classifier basis. The developed
framework was verified on experimental data,
obtained from machinery manufacturing enterprise.
The reliability analysis module enables: to calculate
the max/min boundaries of error probability for
selected production route; to define the most critical
faults, that influence production route reliability; to
select the most efficient corrective actions for
production route reliability improvement.
KEYWORDS
Reliability, FMEA, Bayesian belief network, ERP,
decision making, production route
1. INTRODUCTION
There are a large number of leading manufacturing
small and medium enterprises (SME) that sell
finished goods to end consumers. SME’s are
frequently unable to effectively perform enterprise
processes in-house so they use supply chain partners
to subcontract the routine processes or develop
extended enterprises as needed.
The important success characteristics of a partnership
are the quality, cost, and timeliness of products and
services provided [18]. Achieving success is
complicated by the differences among the
participating organizations. In this research we
develop a unified framework to decrease the issues in
managing production operations of the integrated
1014
Pribytkova, Sahno,, Karaulova, Shevtshenko, Maleki, Cruz-Machado
organization. The developed framework can be also
applied to improve the reliability of collaborative
processes in the entire Supply Chain Management
(SCM) network.
In most cases the quality and cost of finished
products depend on the selection of proper
technological processes which engineers apply in
production process. To ensure appropriately reliable
technology, modern enterprises are constantly
developing information systems (IS) including
functions such as data base structure development,
storing of production technology data, knowledge
management for tracking of production processes and
notification about possible faults. Broadly speaking,
enterprise resource planning (ERP) systems are used
to collect and process the enterprise transactional and
operational data, while data warehousing (DW)
concerns the loading, systemization and use of this
data for business analysis. IS organizes business
knowledge within an enterprise and allows its
sharing as required with external collaborative
partners.
The purpose of current research is to evaluate the
reliability of a production process and pinpoint
potential areas for reliability improvement in
industrial enterprises and collaborative networks.
The different methodological approaches are used in
current research: the DOE-NE-STD-1004-92
standard is used as a base for developed classifier of
faults; the Failure Modes and Effects Analysis
(FMEA) method enables to calculate the Risk
Priority Number (RPN) for every fault occurred; the
Bayesian Belief Network (BBN) analyse what effect
the improvement of different fault groups will cause.
The proposed framework is applied by data collected
from machinery enterprises.
2. STATE OF THE ART
ERP system is a comprehensive transaction
management system that integrates cross functional
business process information and places data in a
single database. Current ERP systems have some
known limitations in providing effective SCM
support: their insufficient extended enterprise
functionality in crossing organizational boundaries;
their inflexibility to ever-changing supply chain
needs; their lack of functionality beyond managing
transactions, and their closed and non-modular
system architecture [6]. In fact, one can argue that
very little academic research has done on ERP
beyond the research on the reasons for
implementation and on the challenges of the
implementation project itself [14, 29].
To address these limitations of ERP systems,
integration with DW systems is widely used to
provide a centralized source of information available
for the partners by request. The DW input data can
be provided from any source of information making
it suitable for the management systems of partner
enterprises. DW improves the productivity of
corporate decision makers through consolidation,
conversion, transformation, and integration of
operational data, and provides a consistent view of an
enterprise [4].
Thanks to the upsurge of export trade for the past few
years, companies have been trying to enhance the
reliability of their products to take advantage of such
a good opportunity for business development. An
obvious key to this strategy is prevention and failure
elimination to which Failure Mode and Effects
Analysis (FMEA) can be applied [18].
FMEA is a preventive technology for reliability
design and analysis which applies structured
systematic procedures and methods to locate the
potential failure modes of products at an early stage.
Causes of failures and impacts of such failures upon
the subsystem and the system are examined for
adoption of proper preventive measures and
improvement proposals. It is usually performed in the
beginning of a product life cycle to increase the
reliability of products or process and to reduce the
costs for follow-up corrective and improvement
actions [17].
In order to discover the knowledge from a FMEA
data we apply Bayesian networks (BN) in our current
research. BN is a graphical probabilistic model
through which models the acquisition of knowledge.
BN is a modern application of statistical approaches
to Artificial Intelligence and Data Mining. It is
particularly suited to taking uncertainty into
consideration when the uncertainties can easily be
described manually by experts in the field.
Compared with previously published approaches,
current approach suggests searching the production
route from Data Mart, where the data from different
sources is stored. Combining ERP and PDM systems
data with FMEA data enables to search for suitable
production routes based on reliability, time and cost
criteria’s. Bayesian network enables to increase the
robustness of decision making process for corrective
actions selection. Suggested approach also enables to
RELIABILITY ANALYSIS MODULE DEVELOPMENT FOR PRODUCTION ROUTE ELABOR.
1015
analyse the effectiveness of selected corrective
actions before implementation, and to measure their
influence on production route reliability. Failure
Mode and Effects Analysis
There are a large variety of methods that enable
evaluation of the reliability of a product or process
and FMEA is the one of them. FMEA is commonly
used in a variety of industries for Risk Management,
where simple quantification of risk is insufficient,
and where identification of root causes of risks and
means of mitigation are paramount. FMEA is one of
the most useful and effective tools for developing
designs, processes and services. The goal of FMEA
is to align the risks as closely as possible with its
source [26]. This enables the determination of the
root cause of the risk. FMEA is a reliability
procedure which documents all possible failures in a
system design within specified rules. It determines,
by failure mode analysis, the effect of each failure on
system operation and identifies single failure points,
which are critical to mission success [14].
Figure 1 Data on the causes of failures [24]
It can also rank each failure according to the
criticality category of failure effect and probability
occurrence. FMEA utilizes inductive logic in a
"bottom up" approach. Beginning at the lowest level
of the system hierarchy, (e.g., component part), and
from a knowledge of the failure modes of each part,
the analyst traces up through the system hierarchy to
determine the effect that each failure mode will have
on system performance. The outcome of the FMEA
is a list of recommendations to reduce overall risk to
an acceptable level, and can be used as a source for
designing a control strategy [7]. Accordingly to the
survey done by Drew Troyer based on U.S.
Department of Energy’s root cause standard DOE-
NE-1004-92 the human factor is responsible for 75 -
80% of what goes wrong in the factory, and the
equipment fault was the reason in less that 20% of
occurrences(Figure 1). FMEA enables the alignment
of corrective actions in the proper order to minimize
the number of faults in manufacturing process [23].
Table 1 Severity ranking table
Effect Severity of Effect
Rank
Hazardous
May endanger operator
(machine or assembly)
without warning
10
Hazardous
May endanger operator
(machine or assembly) with
warning
9
Very high
Downtime of more than 8
hours or 100% of product
may have to be scrapped
8
High
Downtime of 4 to 7 hours or
product (less than 100%)
may have to be sorted and a
portion scrapped
7
Moderate
Downtime of 1 to 3 hours or
a portion (less than 100%)
of the product may have to
be scrapped with no sorting
6
Low Downtime of 30 minutes to
1 hour or 100% of product
may have to be reworked
5
Very low
Downtime of 30 minutes
tend or the product may
have to be sorted with no
scrap 4
Minor
A portion (less than 100%)
of the product may have to
be reworked with no scrap
out-of-station
3
Very
Minor
A portion (less than 100%)
of the product may have to
be reworked with no scrap
2
No effect
Slight inconvenience to
operator or operation, or no
effect 1
FMEA also may be used for a statistically based
preventive maintenance schedule based on the
frequency and type of failure. FMEA is designed to
assist engineers in improving the quality and
reliability of design and provides several benefits
when properly used. Those benefits include:
• Improve product/process reliability and quality
• Increase customer satisfaction
• Early identification and elimination of potential
product/process failure modes
• Prioritize product/process deficiencies
1016
Pribytkova, Sahno,, Karaulova, Shevtshenko, Maleki, Cruz-Machado
• Capture engineering/organization knowledge
• Emphasizes problem prevention
• Documents risk and actions taken to reduce risk
• Provide focus for improved testing and
development
• Minimizes late changes and associated cost
• Catalyst for teamwork and idea exchange between
functions
FMEA processes lead to calculation of Risk Priority
Number (RPN) for every fault occurred. The
multiplied risk factor indices refer to Severity (S)
(Table 1), the outcome of a failure to Occurrence (O)
(Table 2), the chance of a failure to Detection (D)
(Table 3), the chance of a failure is not detected by
customers or the difficulty level of detection:
RPN = (S)× (O) ×(D) .(1)
Table 2 Occurrence ranking table
Occurrence
Criteria
Rank
Very high
≥100 per thousand pieces
10
50 per thousand pieces
9
High
20 per thousand pieces
8
10 per thousand pieces
7
Moderate
5 per thousand pieces
6
2 per thousand pieces
5
1 per thousand pieces
4
Low
0.5 per thousand pieces
3
0.1 per thousand pieces
2
Remote
≤0.01 per thousand pieces
1
Based on collected statistical data it will be possible
to assess the influence of every fault type and to
select the appropriate corrective actions.
2.1. Production route and process
design
A production route is defined as an equipment
sequence used for manufacturing a certain product
[28]. The optimal production route selection topic
lately received an attention of researchers [27]. One
approach is to select the machine centres based on:
lowest average cost, least average process time and
least aggregate cost and processing time [19]. In
order to minimise the total systems cost and to
maximize the machine reliabilities along the selected
production route the multi-objective mixed integer
programming model was developed. The similar
problem was also solved for transportation route
selection, where the unblocked reliability of route is
measured [8].
Table 3 Detection ranking table
Detection
Criteria
Rank
Almost
impossible
Cannot detect or is not
checked
10
Very
remote
Random checks only 9
Remote
Control is achieved with
visual inspection only 8
Very low
Control is achieved with
double visual inspections
only
7
Low
Control is achieved with
charting methods
6
Moderate
Control is based on variable
gauging out-of-station
5
Moderately
high
Gauging performed on set-
up and first-place check
4
High
Error detection in-station
3
Very high
Error detection in station by
automatic control
2
Very high
Discrepant part cannot be
made because item has been
error proofed by
process/product design
1
In order to identify design alternatives in batch
production P-graph model was used [12].The life
cycle analysis (LCA) principles within a formal
design process and optimisation framework enabled
to consider the assessment and minimisation of the
environmental impacts of the complete process
system [22].
2.2. Bayesian Network
A Bayesian network is a graphical model that
encodes probabilistic relationships among variables
of interest. When used in conjunction with statistical
techniques, the graphical model has several
advantages for data analysis [11].
• Because the model encodes dependencies among
all variables; it readily handles situations where
some data entries are missing.
• A Bayesian network can be used to learn causal
relationships, and hence can be used to gain
understanding about a problem domain and to
predict the consequences of intervention.
• Because the model has both a causal and
probabilistic semantics; it is an ideal
representation for combining prior knowledge
(which often comes in causal form) and data.
RELIABILITY ANALYSIS MODULE DEVELOPMENT FOR PRODUCTION ROUTE ELABOR.
1017
• Bayesian statistical methods in conjunction with
Bayesian networks offer an efficient and solid
approach for avoiding the over fitting of data.
Bayesian network is rooted in Bayes Theorem which
firstly reflected on scientific literature after death of
Thomas Bayes who actually developed it but did not
believe that mathematics scientists would accept his
idea as a scientific approach. Richard Price, a friend
of Bayes, took action to introduce his research to the
body of knowledge [3].
A Bayesian network is a statistical model which
computes the posterior probability distribution of any
unobserved stochastic variables, given the
observation of complementary subset variables [9].
Bayesian networks have been proven to have strong
capabilities for expressing dependencies among
random variables. Considering this valuable
potential, this tool has attracted attentions in system
reliability area. Besides, experiences have shown its
superiority over traditional modelling and analysis
tools such as fault trees [5]. Bayesian network is also
known as belief networks or Bayes nets in short and
they belong to the family of probabilistic graphical
models which are employed to represent knowledge
about uncertain domain. Bayesian networks combine
principles from graph theory, probability theory,
computer science, and statistics [10]. Belief networks
are especially useful when the information about the
situation is vague, incomplete, conflicting, and
uncertain [25].
In this research the Bayesian Belief Network (BBN)
is used to analyse what effect the improvement of
different fault groups will contribute in the overall
system. A BBN is a graphical representation of a
probabilistic dependency model. It consists of a set
of interconnected nodes, where each node represents
a variable in the dependency model and the
connecting arcs represent the causal relationships
between these variables.
Particularly, Bayesian network B= (G, Ө) consists of
two parts. The first part is a Directed Acyclic Graph
(DAG) which includes nodes and arcs. DAG is
commonly used in statistics, machine learning, and
artificial intelligence that is the visual representation
of the network where variables of data set X1,…,Xn
are nodes and arcs indicates dependencies among
nodes [2].The second part of Bayesian network is the
conditional dependency distribution of Ө where
Өxi|πxi=PB(xi|πxi) and π
xi is the set of direct parent
variables of xi in G [8]. Finally the network B is the
following joint probability distribution:
(2)
BBN has following advantages: (a) it is a powerful
method for handling the missing value problem; (b)
due to the knowledge of casual relationship between
variables it is good in prediction; (c) it easily allows
the inclusion of prior knowledge; d) the probability
propagation may be used “backwards” also, when the
aim is to find the most probable scenario explaining
the evidence set [15].
3. RELIABILITY ANALYSIS MODULE
DEVELOPMENT
In life data analysis and reliability engineering, the
output of the analysis is always an estimate. The true
value of the probability of failure, the probability of
success (or reliability), the mean life, the parameters
of a distribution or any other applicable parameter is
never known, and almost certainly will remain
unknown to us for all practical purposes. Once a
product is no longer manufactured, and all units that
were ever produced have failed, and all of that data
has been collected and analysed, one could claim to
have learned the true value of the reliability of the
product. Obviously, this is not a common occurrence.
The goal of this work is to evaluate the reliability of
a production process and pinpoint potential areas for
reliability improvement. Realistically, all failures
cannot be eliminated, so another goal of reliability
engineering is to identify the most likely failures and
then to select appropriate actions to mitigate the
effects of those failures.
3.1. Faults classification for machinery
enterprises
Reliability engineering is dealing with the causes of
the faults in the factories. In our framework we
developed the classifier for the causes of the faults.
For this reason we used DOE-NE-STD-1004-92 as
a base standard. We have adapted the standard
classifier to the needs of the machinery enterprises
(Figure 2).
1018
Pribytkova, Sahno,, Karaulova, Shevtshenko, Maleki, Cruz-Machado
In the context of DOE-NE-STD-1004-92 there are
seven major cause (causal factor) categories, which
are divided into 33 subcategories for the machinery
industry. First three categories are necessary for
faults description for equipment, procedures
(technology), and personnel. Design and training
determine the quality and effectiveness of equipment
and personnel. These five elements must be
managed; therefore, management is also a necessary
element. And finally supplier/ subcontracting are also
important for successful functioning of an enterprise.
3.2. The specific details of the research
and development work
In this paper we suggest the framework for
improvement of profitability and sustainability of
manufacturing processes in industrial enterprises and
collaborative networks. This current research is
focused on the design of the reliability improvement
module that is a part of the proposed framework. The
module is verified through analysis of empirical data
received from the manufacturing enterprise. The
framework consists of the following levels: Input
data level, Operational data level, Reliability analysis
level, and Output data level as presented in Figure 3.
The framework specifies the data structure that will
be collected to DW from different sources, in order
to improve the reliability of manufacturing
operations during the design of new products [16].
Input data level
The first step is to send the required data to ERP
system. The CAD data is converged to Product Data
Management (PDM) system where the product Bill
of Materials (BOM) is prepared. The marketing,
sales orders, customer contact and contracts data is
stored to Customer Relationship Management
(CRM) system. The selected Input data level is
replicated to Operational data level.
Figure 3 Reliability analysis framework
Figure 2 The diagram of faults classification and relationships for probability success definition
Faults Classification
5E. Inadequate presentation or materials
5D. Insufficient refresher training
5C. Inadequate content
5B. Insufficient practice or hands-on experience
5A. No training provided
5. Training deficiency
4D. Technological parameters problems
4C. Dimensions related problems
4B. Drawing, specification, or data errors
4A. Inadequate design
4. Design problem
3. Personnel error
3D. Verbal communication problem
3C. Violation of requirement or procedure
3B. Inattention to detail
3A. Inadequate work environment
6F. Other management problem
6E. Policy not adequately defined, disseminated, or enforced
6D. Improper resource allocation
6C. Inadequate supervision
6B. Work organization/planning deficiency
6. Management problem
6A. Inadequate administrative control
7. Supplier/ subcontractor problem
7C. Defective product or material
7B. Time delivery error
7A. Communication problems
2D. Error in tool or cutting data selection
2C. Error in equipment or material selection
2B. Lack of procedure
2A. Defective or inadequate procedure
1C. Software failure
1B. Defective or failed material
1A. Defective or failed part
1D. Equipment failure
1E. Bad equipment work
1F. Contamination
1J. Critical human error
1. Equipment problem
2. Procedure problem (technology)
RELIABILITY ANALYSIS MODULE DEVELOPMENT FOR PRODUCTION ROUTE ELABOR.
1019
Operational data level
The operational data of enterprise is managed by an
integrated cross-functional ERP system. The
integration is made through a data base shared by all
functions and data processing applications in the
company. The operational data required for analysis
and reporting is replicated to DW [13]. Also this
level is applied for relevant production route search
for the item (product) created in ERP system. For this
purpose the product definition data: part/assembly
type, name or profile, dimensions, material standard;
is combined with technological data: production
route, Work Station (WS), production orders that are
received from DW.
Reliability analysis level
When the appropriate production route is discovered
the process of route modification for particular order
is started in Reliability analysis module [21]. This
level enables to perform it by combining the FMEA
method with the BBN approach. FMEA provides
data about all possible failures at WS and BBN
allows to prioritize work with these failures and to
estimate improvement of reliability of the production
route. At this level analysis starts from receiving the
percentage of WS faults from DW where this data is
collected. For this purpose the number of products
with defects, produced in specific WS, are divided by
total number of product produced in this WS. If the
suggested percentage of faults is within the level
required by customer, work with reliability analysis
module is finished and we go back to Operational
data level. If the percentage of faults is too high the
causes are analysed. For this purpose the posterior
probability boundary is calculated based on
assumption that the error took place. The calculation
of the max/min boundaries of error probability for
the selected operation of production route shows the
most critical fault types that influence production
route reliability.
It enables decision maker to select the most efficient
corrective actions for the causes with maximum
influence of production route operation reliability
improvement.
Output data level
After the required level of reliability is achieved the
decision maker chooses the most suitable production
route that further imported to ERP system and then
into production.
4. RELIABILITY IMPROVEMENT
PROCESS
Reliability improvement process in our research is
implemented by using the Reliability Analysis
Module (see figure 3)
This process consists from the following steps:
Step 1.Definition of failure types.
The preparation process is started by definition of the
possible failure categories and adaptation of
classifiers under the requirements of the enterprise.
The possible failure categories were selected
accordingly to the standard DOE-NE-STD-1004-92,
and the particular causes of failures were selected
empirically based on machinery enterprise data.
Figure 4 The Header of FMEA table. (Fields marked by
* is new in FMEA)
Step 2.Establishment of FMEA.
This process was started from analysis of particular
machinery enterprise requirements. During this
process the existing FMEA form was modified
accordingly to the requirements of machinery
enterprise. The columns marked with asterisk were
added to the header of FMEA table (see Figure 4).
Step 3.Analysis of collected FMEA data and
faults probability calculation.
Based on received data The RPN value is calculated
by multiplying failure Severity, Occurrence and
Detections parameters. The probability of error for
every fault group is calculated by following equation:
%100×= ∑
∑
Total
PC
PR
RPN
RPN
P
(3)
Where PRP – Probability of production route errors;
∑
RPNPC – RPN value for a particular cause of
errors;
∑
RPNTotal – Total RPN value of a production
route.
1020
Pribytkova, Sahno,, Karaulova, Shevtshenko, Maleki, Cruz-Machado
Step 4.Forming of Bayesian Belief Network.
The BBN is build on received fault probability data
(see Figure 5). Accordingly to the FMEA the most
important faults categories are considered: Personal
Error, Equipment Problem, Procedure Problem,
Traning deficiency. BBN is drawn based on failure
probabilities with drawn from FMEA. This network
represents possible states of the given failures and
their corresponding errors. The probability of any
node being in one state or another without current
evidence is described in figure below. Probabilities
on some nodes are affected by the state of another
nodes depending on casualities.
(a) Bayesian Belief Network with low boundaries
(a) Bayesian Belief Network with high boundaries
Figure 5 The Bayesian Belief Network for the production route probability of error
This BBN can answer questions like: if personnel
error exists, was it more likely caused by inadequate
work environment, inattention to detail, or violation
of requirements. According to this network (see
Figure 5), there is boundary 42-62% probability of
having error which is respectly resulted by personnel
error (13-36%), equipment problem (25-30%),
prodecure problem (5-7%), and training deficiency
(less than 6-8%).
Based on FMEA data of machinery enterprise (see
Appendix A) the probability of possible faults is
calculated. Table 4 is extraction from appendix A.
According this table the total ∑RPN = 2044.
After the preparation is completed we are ready to
start using the reliability improvement module.
According to Figure 5, personnel error is the most
probable failure type. Particularly, inattention to
details which is one of personnel errors has the
highest probability. Therefore, corrective actions are
RELIABILITY ANALYSIS MODULE DEVELOPMENT FOR PRODUCTION ROUTE ELABOR.
1021
focused on this failure causes aiming to decrease it as
much as possible. Four corrective actions can be
applied in the current case: (1) Poka-Yoke (2) visual
instruction (3) improvement route card and (4)
additional training see Appendix B.
Table 4 Probability of possible faults (example)
Node Name
(number)
RPN
% of
error
Severity
range
Equipment Problem
(1)
∑765
37,42
Software Failure
(1C)
287 14,04 8 to 10
Equipment Failure
(1D and 1D1)
442 21,6 7 to 8
Critical Human
Error(1J
68 3,32 3 to 5
Procedure Problem
(2)
∑188
9,19
Defective or
inadequate procedure
(2A) 92 4,5 7 to 8
Error in tool or
cutting data selection
(2D)
96 4,69 4 to 8
Personnel Error (3)
∑1066
52,15
Inadequate work
environment (3A)
193 9,44 2 to 10
Inattention to detail
(3B)
468 22,89 3 to 8
Violation of
requirement or
procedure (3C)
337 16,48 3 to 8
Design problem (4)
∑16
0,78
Dimensions related
problem(4C)
16 078 3 to 5
Training
Deficiency(5)
∑240
11,74
Insufficient practice
or hands-on
experience(5B)
240 11,74 5 to 7
TOTAL RPN
VALUE
2044
Table 5 shows the impact of each of these corrective
actions on personnel error failure type. In order to
make this table, RPN of corrective action are taken
from FMEA and imported to the BN model. This
table presents available corrective actions and their
influence on the corresponding failure cause (in this
case inattention to detail) which will be presented to
the decision maker. Apparently, the final decision
will be made by the decision makers considering
information in Table 5, as well as the costs of each
action and the policy of enterprise.
Table 5 Influence of corrective actions on personnel
errors
Failure
cause
Corrective
action
Influence
on failure
cause
Influence
on
severity
Inattention
to detail
Poka-Yoke 15 % 7
Inattention
to detail
Visual
instruction
5 % 0
Inattention
to detail
Improve
route card
10 % 0
Inattention
to detail
Additional
training
11 % 0
In current case study the whole FMEA file was
taken as a base for the Bayesian network
building, however in real case all WS should be
analysed individually. As a rule FMEA is quite
volumetric and after selection of corrective
actions using separate BNs it is easier to find a
failure to implement these corrective actions for
in order to improve reliability of the WS.
5. CONCLUSION
In this paper a reliability analysis module was
developed in order to increase the reliability of a
selected production route. The reliability analysis
framework was developed for machinery
manufacturing enterprises. Bayesian Belief Network
was built based on imported data from FMEA table.
It makes possible to calculate posterior probabilities
of each fault group on the error probability of the
manufacturing process. It was resulted in discovering
the most critical fault group where corrective actions
were analysed and the most effective one was
selected to decrease error probability in that specific
fault group. Results were verified by data received
from machinery enterprise. According to this
research BBN can be effectively employed by FMEA
practitioners. It has the potential to calculate the
posterior probability of error after implementation of
corrective actions. Decision makers may benefit from
its output to make the most relevant decision in their
manufacturing processes.
The current research was concentrated on Reliability
analysis level of production routes. The future
research will be focused on the Operational data
level, where design of data warehousing will be
1022
Pribytkova, Sahno,, Karaulova, Shevtshenko, Maleki, Cruz-Machado
presented for production route definition, search, and
analysis.
6. FUTURE WORK
In future work we will continue working on detailed
aspects of proposed framework. We will develop the
structure for the integration of data from PLM, ERP,
Data Mart and GeNie systems and create the
prototype of current solution. It is planned to verify it
by simple product data. When the design process for
the new product is started it will be possible to use
the comprehensive search engine in order to discover
all possible production routes from historical data,
based on the product parameters. After the total list
of possible production routes is discovered the
Kohonen (SOM) will be constructed based on such
criteria’s as: production route time, cost and
percentage of faults for every Work Station used in
operations will be presented to decision maker for
final selection of the best alternatives.
ACKNOWLEDGMENTS
Authors of this paper would like to acknowledge
financial funding from Estonian Ministry of
Education and Research for targeted financing
scheme SF0140113Bs08, DoRa 5, Grant ETF9460
and MIT-Pt/EDAM-IASC/0033/2008 project. The
authors also appreciate Yan Wang from Georgia
Tech for his concern on our discussions.
REFERENCES
[1] Abad-Grau, M. M., Arias-Aranda, D.,
(2006),“Operations strategy and flexibility:
modeling with Bayesian classifiers”, Industrial
Management & Data Systems, Vol. 106, No. 4, pp.
460-484.
[2] Baesens, B., Verstraeten, G., Poel, D. V., Egmont-
Petersen, M., Kenhove, P. V., Vanthienen, J.,
(2004), “Bayesian network classifiers for
identifying the slope of the customer lifecycle of
long-life customers”, European Journal of
Operational Research, Vol. 156, No. 2, pp. 508-523.
[3] Bayes, T., Price, R., (1763), “An Essay towards
Solving a Problem in the Doctrine of Chances. By
the Late Rev. Mr. Bayes, F. R. S. Communicated by
Mr. Price, in a Letter to John Canton, A. M. F. R.
S.”, Philosophical Transactions of the Royal Society
of London, Vol. 51, pp. 370-418.
[4] Berson, A., Smith, S.J.,(1997), “Data Warehousing,
Data Mining, & OLAP”, Tata McGraw Hill.
[5] Boudali, H., Dugan, J., (2005), “A discrete-time
Bayesian network reliability modeling and analysis
framework”, Reliability Engineering & System
Safety, Vol. 87, No. 3, pp. 337-349.
[6] Caiazzo, F., Pasquino R., Sergi V., (1997), “Two-
dimensional objects recognition and fuzzy
classification approach”, Proceedings of the third
A.I.TE.M. Conference, Fisciano (SA), pp. 277-286.
[7] Chin, K.S., Wang, Y.M, Poon, G.K.K., Yang, J.B.,
(2009), “Failure mode and effects analysis using a
group-based evidential reasoning approach”,
Computers & Operations Research, Vol. 36, No. 6,
pp. 1768-1779.
[8] Das, K. Lashkari, R.S., Sengupta, S. (2007).
“Reliability consideration in the design and analysis
of cellular manufacturing systems”. International
Journal of Production Economics.
[9] Gambelli, D., Bruschi, V., (2010),”A Bayesian
network to predict the probability of organic farms’
exit from the sector: A case study from Marche,
Italy”, Computers and Electronics in Agriculture,
Vol. 71, No. 1, pp. 22-31.
[10] Gopnik, A., Tenenbaum, J. B., (2007), “Bayesian
networks, Bayesian learning and cognitive
development”, Developmental science, Vol. 10, No.
3, pp. 281-7.
[11] Heckerman, D., (1999) “A Tutorial on learning with
Bayesian Networks”, Learning in Graphical Models,
MIT Press, Cambridge, MA.
[12] Iskandar, H., Rajagopalan, S. (2008). “Designing
sustainable alternatives for batch operations using
an intelligent simulations-optimisation framework”.
Chemical engineering research and design
Vol.86.,809-822.
[13] Loun, K., Riives, J., Otto, T. (2011) “Evaluation of
the operation expedience of technological resources
in a manufacturing network”, Estonian Journal of
Engineering, Vol. 17, No. 1, pp.51–65
[14] MIL-HDBK-338B (1998),“Military Handbook:
Electronic Reliability Design”.
[15] Neapolitan, R.E. (2003) “Learning Bayesian
Networks, Prentice Hall.
[16] Sahno, J., Savimaa, R., Kangilaski, T., Opik, R.,
Maleki, M & Machado, V.C., (2011),“Data Mart
Framework for Production Route Selection”,Annals
of DAAAM for 2011 & Proceedings of the 22nd
International DAAAM Symposium, Vienna, Vol. 22
No. 1, pp. 1543-1544.
[17] Saravanan, S., Yadava, G.S.,Rao, P.V.,
(2003),“Machine Tool Failure Data Analysis For
Condition Monitoring Application”, Proceedings of
RELIABILITY ANALYSIS MODULE DEVELOPMENT FOR PRODUCTION ROUTE ELABOR.
1023
11th. National Conference on Machines and
Mechanisms, Delhi, pp. 552-558.
[18] Sharon, K.J., (1998), "Combing QFD and FMEA to
optimize performance", ASQC Quality Congress,
Vol. 52, No. May, pp.564-75.
[19] Subramanian, V., Lee, G.K., Ramesh, T., Hong,
G.S., Wong, Y.S. (2000) “Machine Selection Rules
in a Dynamic Job Shop”. Advanced Manufacturing
Technology
[20] Kangilaski, T. (2008). “ICT and Business
Alignment in Virtual Organization”, Proc. 6th IEEE
International Conference on Industrial Informatics,
INDIN, IEEE, pp. 1251–1256
[21] Karaulova, T., Pribytkova, M., Sahno, &
Shevtshenko, E., (2011),“Design of Reliable
Production Route System“, Annals of DAAAM for
2011 & Proceedings of the 22nd International
DAAAM Symposium, Vienna, Vol. 22, No. 1, pp.
539-540.
[22] Khan,F.I., Natrajan,B.R., Revathi,P. (2001).”Green
Pro: a new methodology for cleaner and greener
process”. Journal of Loss Prevention in the Process
Industries, Vol. 14., pp. 307–328.
[23] Troyer, D., “Human factors Engineering: the Next
Frontier in Reliability”,
http://www.machinerylubrication.com/Read/23904/
human-factors-engineering-reliability accessed
26.07.11
[24] Upton, D.M., McAfee, A.P., (2000),“A path-based
approach to information technology in
manufacturing”, International Journal of
Technology Management, Vol. 20, No. 3/4, pp.
354–372.
[25] Wang, Y., (2010), “Imprecise probabilities based on
generalized intervals for system reliability
assessment”, Vol. 4, No. 4, pp. 319-342.
[26] Wang, Y.M, Chin, K.S., Poon, G.K.K., Yang, J.B.,
(2009), “Risk evaluation in failure mode and effects
analysis using fuzzy weighted geometric mean”,
Expert Systems with Applications, Vol. 36, No. 2,
pp. 1195-1207.
[27] Yanyan, C., Ying, L., Huabing, Du. (2003).
“Intelligent Transportation Systems”. IEEE
Proceeding, Vol. 2., pp. 975-978.
[28] Zhao-Ling, Y., Xi-Gang, Y. (2000) “An approach to
optimal design of batch processes with waste
minimization”. Computers & Chemical
Engineering, No 24, pp.1437-1444.
[29] Zhu, Y., Li, Y., Wang, W., Chen, J., (2010), “What
leads to post-implementation success of ERP? An
empirical study of the Chinese retail industry”,
International Journal of Information Management,
Vol. 30, No. 3, pp. 265-276.
1024
Pribytkova, Sahno,, Karaulova, Shevtshenko, Maleki, Cruz-Machado
APPENDIX A: A FRAGMENT OF FMEA TABLE
RELIABILITY ANALYSIS MODULE DEVELOPMENT FOR PRODUCTION ROUTE ELABOR.
1025
APPENDIX B: POSTERIOR PROBABILITIES AFTER APPLYING EACH CORRECTIVE
ACTION
(1) Original state of BBN when max severity is applied Prior probability of each
corrective action
Posterior probabilities
when “Poka-Yoke” is
implemented
Posterior probabilities
when “visual instruction”
is implemented
Posterior probabilities
when “improve route
card” is implemented
Posterior probabilities
when “additional
training” is implemented
This page is deliberately left empty.
1026