=Paper=
{{Paper
|id=Vol-1821/W2_paper3
|storemode=property
|title=Developing a Prototype of Case-based System Utilizing Fuzzy Sets to Detect Faults of Injection Molding Process
|pdfUrl=https://ceur-ws.org/Vol-1821/W2_paper3.pdf
|volume=Vol-1821
|authors=Sara Nasiri,Ildar Allayarov,Madjid Fathi
|dblpUrl=https://dblp.org/rec/conf/wm/NasiriAF17
}}
==Developing a Prototype of Case-based System Utilizing Fuzzy Sets to Detect Faults of Injection Molding Process
==
https://ceur-ws.org/Vol-1821/W2_paper3.pdf
WM1017 - 9te Konferenz Professionelles Wissensmanagement
5.-7. April 2017 in Karlsruhe, Deutschland
Developing a Prototype of Case-based System Utilizing
Fuzzy Sets to Detect Faults of Injection Molding Process
Sara Nasiri, Ildar Allayarov, Madjid Fathi
Department of Electrical Engineering & Computer Science, University of Siegen
Institute of Knowledge Based Systems & Knowledge Management
Hölderlinstr. 3
57076 Siegen
sara.nasiri@uni-siegen.de
ildar.allayarov@student.uni-siegen.de
fathi@informatik.uni-siegen.de
Abstract: The purpose of this research is to develop a case-based system for
detecting faults of injection molding machine by utilizing the information retrieved
from troubleshooting guidelines. It is also reduced the system downtime. Case-
based fault detection system guides users to detect their failure through the retrieval
procedure. It utilizes fuzzy sets based on the relationship between parameters of the
process, part, and mold to define the weight of features which is important for
occurrence the faults. In this paper, a case-based system is also utilized the
occurrence weights of features to capture the problems in injection molding
processes and to recommend the way of fixing these faults.
Keywords: Fault detection, failure diagnosis, case-based reasoning, fuzzy logic,
injection molding machine
1. Introduction
Extracting and utilizing new knowledge from the given product information to improve a
new product is aimed at many types of research in the last decade i.e. using a rule-based
system for the reasoning about product information [1]. Fault detection systems (FDS)
can be developed with the help of several approaches, depending on a type of knowledge
needed in each particular case. Some of these techniques are artificial neural networks,
expert systems or case-based reasoning (CBR) compared in [2]. The FDS based on these
techniques are distinguished from each other by problem-solving, data acquisition and
their complexity level. Maintenance time should be kept as short as possible to meet the
high-performance output demands, especially for an automated production system.
needs the maintenance strategy. The maintenance downtime process is characterized by
maintenance delay, access, diagnosis, logistics, repair or replacement and finally
checkout [2]. The time of each activity within this process should be kept as short as
possible to loose minimal total downtime. In order to reduce the downtime, diagnosing
the occurred failures is the main step, on which we focus in this paper. CBR is one of the
appropriate techniques which help us to find the solution for a given problem and is
Seite 92 von 198
WM1017 - 9te Konferenz Professionelles Wissensmanagement
5.-7. April 2017 in Karlsruhe, Deutschland
applied in different domains [3]. A comparison of the properties of the surveyed artificial
intelligence methods shows that CBR can find the similar cases by using the previous
similar problems [4], [5] and [6].
In this paper, we developed the prototype of fault detection using fuzzy CBR by
focusing on the application scenario of injection molding machine. Injection molding is
a cyclic process that contains clamping, injection, cooling, and ejection to produce
plastic parts. For achieving a higher quality of products, three main points should be
considered: a) type of materials; b) process setting parameters (e.g. temperature); c)
basic parts of injection molding machine [7].
The structure of this paper is as follows: in the second section, we give an overview of a
fault detection system considering the types of fault and failures. In the third section, the
injection molding machine and its processes are explained. Case-based reasoning, fuzzy
sets and fault diagnosis is discussed in section four. Finally, section five explains
conclusion and a future work.
2. Fault Detection System in Injection Molding Machine
FDS is a monitoring system which can identify the occurrence of the faults in
machinery, recognize their types and locations. Fault detection and diagnosis play a
serious role in engineering systems, which is enhancment of production quality and
reduction of costs for testing. Modern fault detection systems distinguish several types of
problems. unpermitted deviation of at least one characteristic property
[8].
Fault detection is the accepted term for recognition of any degraded settings or inactive
status of a system, plant or its operational parts. Fault-diagnosis approaches use the
analytic and heuristic signs. Hence they must be provided in a unified way like self
assurance-numbers, membership functions of fuzzy sets or probability frequency
function after a statistical evaluation over a while. Then either classification techniques
may be employed, in case a learned pattern-based procedure is recommended to
recognize appearing problems from symptom patterns or clusters. However, fault
detection using CBR is applied in maintenance [2] with the goal to find the process
parameters of injection molding machines [9], [10] but there is still more needs to do
researchers in this field. In this paper, CBR considering fuzzy rules and sets is utilized
for fault detecting of injection molding machine (IMM). It's a cyclic method of a high-
speed mold filling followed by means of cooling and ejection. Step one to the injection
molding is the clamping of the mold. The clamp is a component that holds a mold while
the melted plastic is being injected. The pressing mold is kept under a pressure during
the injected molten plastic is cooling. On the next step, the injection of the melted plastic
is performed. The plastic stays within the mold, where it is being pressed until it will be
cool and solid. The next step encompasses the holding period, which is ensuring that all
cavities of the mold are full of melted plastic. After a holding period, the cooling phase
starts and continues unless the plastic becomes strong within the mold. Subsequently, the
mold is opened and a newly produced plastic part is ejected from it. Then the section is
cleared from any residuals in the mold. IMM is likely one of the largest and rational
forming approaches present for processing plastic substances [11].
Seite 93 von 198
WM1017 - 9te Konferenz Professionelles Wissensmanagement
5.-7. April 2017 in Karlsruhe, Deutschland
3. Fault Detection utilizing Case-based Reasoning
The benefits of the FDS founded on a CBR approach are that comparable cases could be
obtained rapidly just by inputting a set of unusual levels (of symptom features) to make
sure that the solution could be produced accordingly.
In this paper, based on the troubleshooting guides [12], [13], [14] and [15], twenty
features in six categories are considered for injection molding processes which are
temperature(f5,f6,f9,f16 and f17), pressure(f1, f2, f3, f4 and f14), time, speed(f7 and f8),
gate size(f20), runner size(f19).
Fault diagnosis based on CBR could assist in dealing with ill-defined issues related to
the diagnostic activities which have two main steps:
1) Fuzzy Sets and Feature Weight
Case indexing and defining the weights of features is the first step of case retrieval. In
this research four fuzzy sets are identified for classifying the relationship between
process parameters of injection molding and a bundle of a part and mold parameters
which can be estimated as strong (S:0,7-0,9), medium (M:0,5-0,7), weak (W:0,3-0,5)
and very weak (V:0,1-0,3). Table 1 shows the relationship between the parameters of
pressure and a set of various part and mold parameters which are defined based on [16],
[9] weight. The FO (Feature's Occurrence) column is defined based on the probability of
occurrence of these features in all faults. The relationship of injection
pressure(f1) and holding pressure(f3) is illustrated in Table 2. The fuzzy classification
weights are calculated as follows:
(1)
where wi is the fuzzy weight for the ith feature, FO is the occurrence weight of feature,
PR is the relationship between features and parameters and n is the number of
parameters.
Table 1: Relationship between pressure, part and mold parameters of injection molding
Process Part and mold parameters
Parameters Molding Part Wall Part Part Projected Runner Runner Gate Gate
Wi
FO
Material size thickness complexity volume area type size type size
F1 0,4 S V S S W V W M S S 2,4
F3 0,2 S S S S M S W M S S 1,5
2) Case Base and Retrieval
In this section, the prototype of FDS is explained, which is implemented based on the
open source tool myCBR [17], [3]. At the first step, all attributes and instances are
defined. Then, their weights are calculated by equation (1). Defining the case base is the
third step which includes 15 most often appearing faults arising in different types of
IMMs. Hence to provide better fault detection results and to identify more exact
solutions for the IMMs of various vendors, the case base consists of the common
injection molding faults and solutions. This allows to apply them not only to the
machines of observed brands but also to all possible IMMs out of this research
considering different default settings of IMM features. Each and every brand of IMMs is
Seite 94 von 198
WM1017 - 9te Konferenz Professionelles Wissensmanagement
5.-7. April 2017 in Karlsruhe, Deutschland
producing its machines established on its own technical design requirements. As a result,
the machines of various manufacturers differ to each other in the default settings of IMM
features as injection and clamping pressure, injection and screw rotation speed, clamping
time etc. as they were set by the manufacturer while assembling. For instance, if a
standard injection pressure for injection molding of a polypropylene for Machine A is
1200 bar and for Machine, B is 1000 bar, then it is impossible to classify them both to
the same measurement group of injection pressure feature because of difference between
these two values. For this reason, all features of diverse IMMs will be unified in a way
of mapping their normal and abnormal feature values to the fuzzy alternatives, that
would confirm the IMMs of various manufacturers and could be directly utilized in the
FDS. The faults which are being researched for a development of this prototype were
chosen from several injection molding scientific guidelines and troubleshooting guides
of various manufacturers of IMMs [12], [13], [14] and [15]. These faults are caused due
to abnormal states of one or several process parameters or features of the machines. In
Table 2, three faults considering their causes are illustrated. This table is filled with L
(Low), N (Normal-Standard) and H (High) which is the feature range of occurred fault.
Table 2. Fault examples
Fault f1 f2 f3 f4 f5 f6 f7 f8 f9 f10 f11 f12 f13 f14 f15 f16 f17 f18 f19 f20
Sink
1 Marks N N L N H H L N N L L N L L L N N N N N
Flow
2 L L N N L L L N N N N N N N N L N N N N
Lines
Excessive
3 H H N L H H H N N N H N N N N N N N N N
Flash
This mapping principle is applied to all of the features of IMMs and will be utilized by
the user during filling of the query to detect the occurred fault which is the most similar
case which is calculated by utilizing myCBR. The retrieved result of an example query is
77%, 72%, 62% and 60% for fault 10, fault 1, fault7 and fault 2 respectively. The most
similar case is selected for reuse of its solution and it is an input of adaptation phase.
Therefore the solution of the most similar fault should be reviewed by IMM worker. In
this prototype null adaptation is used and it can be extended to develop adaptation
mechanism in the future work.
5. Conclusion
This paper presented the results of a research student project that applies fuzzy sets in
CBR as a methodology and utilizes the myCBR tool to develop a fault detection
prototype. It makes two contributions. First, it describes the injection molding machine
and fault detection system. The faults and main important features of these fault's
occurrence are defined and analyzed based on the troubleshooting guidelines.
Additionally, fuzzy case-based reasoning considers fuzzy rules and sets. The retrieval
testing and evaluation of the prototype of FDS shows that it can be utilized in injection
molding machine for fault detection. Although the current prototype as a CBR system
has a useful function, namely it could be extended in a way of collecting more specific
Seite 95 von 198
WM1017 - 9te Konferenz Professionelles Wissensmanagement
5.-7. April 2017 in Karlsruhe, Deutschland
cases containing faults of a certain product of particular IMM and hence enriching a case
base. Another aspect for extension is the implementation of the adaptation rules to
specify the recommended solutions. Finally, for evaluating of the retrieval phase, the
similarity measurement could be enhanced based on the specific cases and real test
results.
References
[1] M. Dornhöfer, M. Fathi and A. Holand, "Applying Rules for Representing and Reasoning of Objective
Product Use Information Exemplarily for Injection Molding Machines," in ICIEIS 2011, Part II, CCIS 252,
A. Abd Manaf et al. (Eds.), 2011.
[2] Y.-T. Tsai, "Applying a case-based reasoning method for fault diagnosis during maintenance," Proc.
IMechE Vol. 223 Part C: J. Mechanical Engineering Science, pp. 2431-2441, 2009.
[3] K. Bach and K.-D. Althoff, "Developing Case-Based Reasoning Applications Using myCBR 3," in Case-
Based Reasoning Research and Development, pp. 17-31, 2013.
[4] A. Aamodt and E. Plaza, "Case-based reasoning: Foundational issues, methodological variations, and
system approaches," AI Communications, pp. 7(1), 39 59, 1994.
[5] R. López and e. al., "Retrieval, reuse, revision, and retention in case-based reasoning," The Knowledge
Engineering Review, vol. 20, no. 3, pp. 215-240, 2005.
[6] M. M. Richter and R. Weber, Case-Based Reasoning: A Textbook, Springer, 2013.
[7] V. Goodship, Practical Guide to Injection Moulding, UK: Repra Technology and ARBURG Limited, 2004.
[8] R. Isermann, Fault-Diagnosis Systems, Verlag Berlin Heidelberg: Springer, 2006.
[9] S. L. Mok and C. K. Kwong, "Application of artificial neural network and fuzzy logic in a cased-based
system for initial process parameter setting of injection molding," Inteligent Manufacturing, 13, pp. 165-
176, 2002.
[10] K. Shelesh-Nejad and E. Siores, "An Intelligent System for Plastic Injection Molding Process Design,"
Materials Processing Technology, pp. 458-462, 1997.
[11] D. V. Rosato and M. Rosato, Injection Molding Handbook, Springer, 2000.
[12] MMM, "Mold Masters Milacron LLC Troubleshooting Guide," 2011.
http://www.moldmasters.com/media/documents/HRUM-EN-XX-LTR-V14.pdf.
[13] PPL, "Pentagon Plastics LTD Troubleshooting Guide," 2015. http://www.pentagonplastics.co.uk/wp-
content/uploads/2015/03/Trouble-Shooting-for Injection-Moulding-2014.pdf.
[14] RT, "Routsis Training - Scientific Molding Pocket Guide / IM Reference Guide," 2016.
http://www.traininteractive.com/download/pdf/routsis_injection_molding_reference.pdf.
[15] ECC, "EASTMAN CHEMICAL COMPANY Troubleshooting Guide," 2012.
http://www.eastman.com/Literature_Center/P/PPD407.pdf.
[16] Y. K. Lau, Basic Injection Molding Trouble Shooting Guides, Hong Kong: Hong Kong Plastics Technology
Centre, 1996.
[17] A. Stahl and T. R. Roth-Berghofer, "Rapid prototyping of CBR Applications with the Open Source Tool
myCBR," in ECCBR 08, Berlin, Heidelberg, 2008.
Seite 96 von 198