Journal of Science and Technology 12 (02 ) December 2011
ISSN 1605 – 427X
© Sudan University of Science and Technology
www.sustech.edu
A Fuzzy Expert System for Rigging System Design
Adil Mohamed Elbasheer Ahmed,1 Mohamed Ibrahim Shukri,2 Waiel Fathi Abd El-Wahed3
1. Blue Nile University, Faculty of Engineering, , Sudan.
2. Valley University, Faculty of Engineering & Technology Nile, Sudan.
3. Menoufia, University , Faculty of Computers & Information, , Shebin El-Kom, Egypt
ABSTRACT:
This paper presents a fuzzy expert system that classifies cast component features and designs the
rigging system in sand casting. The heuristic rules encoded inside the system were extracted from
expertise and published literature. The rules consist of the conclusion for condition connected by IF-
THEN clauses to judge the relationship between the classification and rigging system design. Input
and output variables are defined by fuzzy sets which deals explicitly with uncertainty associated
with knowledge in the heuristic rules. Conclusions inferred from input parameters were combined
through inference and defuzzification process. A design condition chart was developed to evaluate
fuzzy system results. The rigging system design elements were analyzed based on: membership
values, design and cost status in the chart
KEYWORDS: fuzzy expert system, classification system, rigging system, membership, design control.
INTRODUCTION:
Fuzzy expert systems are being used The manufacturing of sand castings is vitally
successfully in an increasing number of depending upon the experience of foundry
application areas; they use linguistic rules and engineers, who perform task known as the
heuristic rules to describe systems. These methoding design. Methoding design
rules-based systems are more suitable for involves several decisions such as,
complex system problem where it is very configuration of casting and rigging system
difficult, if not impossible, to describe the design [5]. Methoding design still considered
system mathematically. One of the most as state-of-art [6] and not well documented in
important considerations in designing any technical literature, while few equations are
fuzzy expert system is the generation of the available. Anyhow, the heuristic rules
fuzzy rules as well as the membership inherited in the foundry contain linguistic
function for each fuzzy set. In most existing variables such as large, tall, small, etc, these
applications, the fuzzy rules are generated by vague areas were led the research to use a
experts in the area, especially for solving fuzzy expert system as a tool for
problem with vague area and expertise is classification of cast components and
becoming an indispensable [1, 2, 3, 4]. designing of rigging system.
124
Previous Approaches OBJECTIVES:
The overall objectives of this work are:
The review of exiting literature of computer
aided casting design reveals that computer 1. To design a fuzzy expert system for
simulation was widely used to optimize classification of cast component and rigging
rigging system design. Some researcher system design.
includes Kent [7], claimed that casting 2. To apply fuzzy logic in defining the
fuzzier areas in the designing process of
simulation does not provide the initial riser
cast component and rigging system in order
design for a casting, nor does it to facilitate the designing process.
automatically. Thus even if simulation is used 3. To develop an expert system built from
heuristic rule of foundry experts is becoming experience and knowledge compiled from
an indispensable. foundry experts and literature. The system
will classified cast component features into
The first popular task on casting simulation categories such as thin, tall, and medium,
procedure was the numerical optimization with confidence values.
method developed by Bradley and 4. To design the rigging system for
medium and small cast components with
Heinemann [8], on which they used hydraulic
reference to rules extracted from experts
models to simulate the optimization of gating and fuzzy system results.
system. Apparently, Mc David and Dantzig
claimed that this work has never been Overview: Rigging System Design
implemented [9]. Most of the model
simulation is done for one cast component The major considerations in the casting
design are the quality of the final product and
such as numerical optimization of casting the yield, which heavily depend upon gating
done by Hartmann [10]. Also computer-aided system and feeders (rigging system as
programmers are developed as commercial sometime refers to). Even after spending
software like Auto Cast developed by significant resources (man-hours, materials,
Advance Reasoning Technologies in machine overhead and energy) for casting
Collaboration [11]. development, one of the following situations
may arise during regular production [15].
There is very little literature available
1. Under design: resulting in high percentage
regarding expert system and fuzzy logic of of defective castings, due to the number or
casting design. There is active interest in the size or placement or selection of feeders and
casting industry in the development of expert gating and gating elements are inadequate.
system for the design of methoding systems. 2. Over design: leading to acceptable quality
Nitration et al [12] described a system which level, but poor yield and thereby high cost.
uses a simple geometrical solidification to In this case, the number and/or size of
feeders and gating elements are much
make a preliminary assessment of castability
higher than their respective optimal values.
for rectangular grid of cell, rather than on the 3. Borderline design: irregular defect levels
whole features. Recently Shukri and during regular production, although sample
Elbasheer have developed an expert system castings are defective-free.
for designing rigging system for small The terms gating system referred to all
casting by small set of rules written in visual passage or channels through which the metal
basic [13], while Bouhouch [14] used fuzzy enters a mould cavity [16]. There are three
types of gating system: Top gate, bottom gate
logic in developing a system for continuous and side gates. Nearly 40% of casting defects
casting process.
125
are attributes to faulty design of gating and significant human expertise for their
poor pouring practice [170]. solution”.
Feeder acts as a liquid metal reservoir Fuzzy set theory is a concept that can bring
provides liquid feed metal until the end of the the reasoning used by computers closer to
solidification. Riser also serves as a heat that used by human beings. Fuzzy set was
reservoir, creating a temperature gradient that invented to deal with vague, imprecise and
induces directional solidification. Two uncertain problem [24]. Fuzzy logic is
criteria determine whether or not a riser is concerning with quantities and reasoning
adequate [18]. Common types of feeders are: using linguistic variable. In Driankov, Hellen
top feeder and side feeder. and Reinfrank they said that Zadeh states “by
linguistic variable we mean a variable whose
Fuzzy Expert System values are words or sentences in natural or
artificial language such as tall, medium,
A fuzzy expert system is an expert system short” [25],
that uses fuzzy logic instead of conventional
logic. It uses a collection of fuzzy Fuzzy logic provides a general concept for
membership functions and rules to facilitate description, classification and measurement.
reasoning. Since it uses rules, it falls into the Fuzzy logic comprises fuzzy sets, which are a
category of rule-based expert system, rules way of representing nonstatistical
can be easily demonstrated human thinking as uncertainty. In fuzzy logic, fuzzy set
they easily formulated. Fuzzy expert systems membership occurs by degree over the range
are used to provide non-expert with some interval “[0, 1]”, which is represented by a
expertise [19, 20, 21]. membership function. This is called a degree
of belief or membership value or confidence
Fuzzy expert system is an element of value [26]. The function can be linear or
artificial intelligence (AI). One of the earliest nonlinear; in this paper, the following
popular definitions of (AI) was “making functions Γ-function, L-function, A-function
computer think like people [22]. Feigenbaum and П-function are used
[23] defined an expert system as “an
intelligent computer program that uses In a fuzzy expert system, the process of
knowledge and inference procedures to solve reasoning consists of three steps as illustrated
problems that are enough difficult to require
in Fig. (1).
Fig.1: Fuzzy expert system architecture
126
Structure of the fuzzy expert system Step1: Fuzzification process:
This fuzzy expert system (hereafter called The verb “to fuzzify” has two meanings:
fuzzy system) is developed aiming to to find the fuzzy version of membership
classify casting component features and concept, and the second is to find the
design of rigging system. The system grades of membership of linguistic values
including number of heuristic rules or of a linguistic variables corresponding to
fuzzy rules inside rule sets. These rules input number scalar or fuzzy [27].
judge the relationship between Fuzzification is the process which
classification and rigging system design. determines the degree of membership of
The fuzzy system has been built by using the input values to define fuzzy set [28].
Expert Rule Knowledge Builder software.
Step 2: Inference process:
Fuzzy logic for classifications
Inference from a set of fuzzy rules
Thickness classification involves fuzzification of the conditions of
the rules, then propagating the confidence
The linguistic categories referring to the factors (membership values) of the
levels of thickness classification is conditions to the conclusions (outcomes)
defined: (1) thick, (2) thin (3) very thin. of the rules.
The domain of these set is an arbitrary
“shape factor” scale from 1 to 50 .The Step 3: Defuzzification process:
function is used for this classification is
illustrated in Fig. (2). Thus the feature “Defuzzification” refers to the reduction of
with shape factor equal 20 is “Thin” with a range of conclusion being reached with
full membership, while a feature with 32 different memberships to a single point.
would be both “Thin” and “Very thin” The conclusions based on inference engine
with partial membership to each. This type were defuzzified based on the output fuzzy
of classification will be used for selection set [26].
of required type of feeder.
Fig. 2: Fuzzy membership function of thickness classifications
127
Height classification
The linguistic classification referring to The function is used for this classification
the levels of height classification is is illustrated in Fig. (3). This classification
defined as follows (1) short, (2) medium aids the selection of gating system type as
(3) tall. The domain of these set is an shown later.
arbitrary “height” scale from 1 to 200 mm.
Fig.3: Fuzzy membership function of height classification
Size classification sets is an arbitrary “base dimension” scale
from 1 to 800 mm. The function used for
The linguistic categories referring to the this classification is illustrated in Fig.(4).
levels of size classification of features as This classification aids the selection of the
defined by: (1) very small (2) small (3) number ingates of gating system type as
large (4) very large. The domain of these will be shown later.
Fig.4: Fuzzy membership function of size classification
Fuzzy Rules for rigging system design Rules for feeder design
The fuzzy rules that judge the relationship The design of feeder includes: selection,
between the classification and rigging positioning and determining the number of
system are built as follows: feeder. For the production rules or the
fuzzy rules for selection the suitable type
of feeders is defined as shown in Fig (5).
128
Fig. 5: Fuzzy rules for selecting of feeders
The linguistic categories referring to the four (5) five feeders. The domain of these
levels of required number feeders to feed set is an arbitrary “base dimension” scale
the feature with molten metal as defined from 1 to 800 mm. The function used for
by: (1) one feeder (2) two (3) three (4) this classification is illustrated in Fig (6).
Fig .6: Fuzzy membership function of number of feeders
Rules for gating system design
The design of gating system includes: literature, the relation between the
selection of suitable type of gating system height classification and the type of gating
and determining the required number of system and between size classification and
ingates. From foundry experience and number of ingates were determined.
129
Fig. 7: Fuzzy rules for selecting of gating system
The fuzzy rule selecting the type of gating fuzzy rules for determining the numbers of
are defined as shown in Fig (7) while the gates are defined as shown in Fig (8).
Fig. 8: Fuzzy rules for determining the number of gates
How the fuzzy system works The system starts by fuzzification of
classification input features attributes into
The fuzzy expert system works as follows:
linguistic variable with their confidence
values. Then fuzzification output of the
1. Determine the fuzzy membership
classification is used as premises of fuzzy
values activated by the inputs.
2. Determine which rules are fired in the rules that judge the design of rigging
design.
rule set.
3. Trace rule activation membership RESULTS:
values back through the appropriate
output fuzzy membership function. The fuzzy system receives the features
4. Utilize defuzzification to determine input attributes from open knowledge base
the value for each output variable. connectivity. The system fuzzified the
5. Make decision according to output input of cast component features attributes
values. an example as shown in table (1). The
output of the system includes:
RESULTS and EVALUTION:
classification of features thickness, height
The fuzzy system is designed to give clear and size classification, the suitable gating
and feeder and number of ingate and
cut definitions of fuzzier area in a cast
feeder required with their confidence
component and rigging system elements. values as shown in tables( 2, 3, 4)
respectively.
130
Table (1): List and numerical of feature attributes
Table (2): Classification variables & membership values
Table (3): Feeder design variable &membership values
Table (4): Gating system design & membership values
Evaluations of rigging system design:
The great benefit of this work is that the To evaluate the rigging system design
design of gating system and feeder can be elements a plot of the membership values of
evaluated before production cycles, and the selected types of gate and feeder, number of
three design situation (under, over design gates and feeders is performed by the fuzzy
and borderline) can be determined. expert system. This design condition chart
131
shows the design level of confidence. Fig Hence the design of gating systems and
(9) shows an illustrated example which feeder’s elements are evaluated as follows:
gives the following results:
i. Side feeder: the design is over design,
1. The selected type of feeder is “side and then the design is good and high cost.
feeder” with confidence value “1.0”. ii. Two feeders: the design is over design,
2. The determined number of feeders is “2” partially good and high cost.
with confidence value “0.582”. iii. Bottom gating: over design partially
3. The selected type of gating system is good design and high cost.
“bottom gate” with confidence value iv. Two ingates: border design acceptable
“0.867”. design and cost
4. The determined number of ingates is “2”
with confidence value “0.533.
Fig. 9: Evaluation chart of rigging system design
CONCLUSION: of gates and feeder, number of ingates and
feeders and sizing of gating and feeder
This fuzzy expert system is developed to elements. Design condition chart is
aid the foundry man to perform the design developed to evaluate fuzzy system results.
process. The system identifies the cast
component features and classifies them The evaluation of rigging system based on:
and takes decision on the design of rigging membership values, design condition and
system. The system is developed as cost. The great benefits of this work are
artificial intelligence base system not only classify casting and rigging
incorporated expert system and fuzzy system design, but also predict of a good
logic. The fuzzy system classifies castings or poor design and high or low costs of
into: Thickness, height and size casting before production cycle.
classification. The fuzzy rules inside fuzzy
RECOMMENDATION:
expert system employs for transformation
of data and judgment of rigging system This work can be extended further to
design. The system designs gating system include other casting tooling features such
and feeder elements which includes: type as orientation of casting and parting line.
132
REFERENCES:
1. Hamid S. L. Mirhosseyni , Phil Webb. rigging system”, International Journal
A. (2009). Hybrid Fuzzy Knowledge- for Numerical methods in fluid, No. 28.
Based Expert System and Genetic 10. Hartmann, Kokot and Rudolf, (2003).
Algorithm for efficient selection and “Numerical optimization of casting
assignment of Material Handling processes-Leveraging coupled process
Equipment, Expert Systems with simulation and multi-object
Applications, 36 11875–11887. optimization to manufacturing”
2. Edward S. Sazonov, Powsiri, Hota and level,21st CAD-FEM User's Meeting
Halabe, (2002). “Fuzzy Logic Expert 2003,International Congress on FEM
System for Automated Damage Technology, Berlin, Germany, Nov. 12-
Detection from Changes in Strain 14-2003.
Energy Mode Shapes,” Nondestructive. 11. B. Ravi, “Computer-Aided Design, Past
Testing and Evaluation, Vol.18 present and Future”, paper [On-Line].
(1).pp.1-20. 12. Natarajian R., Chu C. N.Kasheyap R.
3. Ibrahiam M. M. Elmany, Rabia, (2005). L., (1989). “An integrated environment
“Fault Detection of Computer for intelligent design of castings, in
communication Network Using an Expert System Application in Materials
Expert System” American Journal of Processing and Manufacturing”, Ed.
Applied Sciences 2 (10), pp. 1407-1411. Demeri M Y, The Metals and Materials
4. Ismail Saritas, Allahverdi and Sert, Society .
(2003). “A Fuzzy Expert System 13. Shukri and Elbasheer, “An expert
Design for Diagnosis of prostate system for designing gates and risers for
cancer” International Conference on small and medium size castings”, Sudan
Computer System Technology Tech. Engineering Society Journal, 52:39-47.
5. B. Ravi, (2003). Casting Design 14. Salah Bouhouche, Lahreche, Moussaoui
Knowledge Management, 51st Indian and Bast (2007). “Quality Monitoring
Foundry Congress. Using Principal Component Analysis
6. Brain, Knight, Cowell and Preddy “an and Fuzzy Logic Application in
Object-Oriented Support Tool for the Continuous Casting Process” American
Design of Casting Procedures”. Journal of Applied Sciences.4 (9):
Technical paper [On-line]. pp.637-644.
7. Kent, Ou, Hardin and Beckermann, 15. B. Ravi, “Design for Casting, Presented
( 2002) . Development of New Feeding- at SOURECON`99,”(1999). Hyderabad,
Distance Using Casting Simulation, Organized by the Institute of Indian
Metallurgical and Material Transaction Foundry.
B, Vol. 33B. 16. Ernest du. Marine, “New Opportunity
8. Bradley, F., Heinemann,S., (1993). in the New Millennium-Creative Ways
“Hydraulics-based optimization for Added Value Foundries as
modeling for gating design, Applied Developers-Experience of a Medium
Mathematical Modeling”, No.17.pp Size Supplier to the Machinery
406-414. Manufacturing Industry”, Paper
9. Mc David, R. M, Dantzig, J. A. (1998). Represented at The German Foundry
“Design sensitivity and finite element Congress 25-26 May2000,in Leipzig,
analysis of free surface flow with Cast Plan t + Technology Corporation
application to optimal design of casting International 1-,1.
133
17. B. Ravi, Srinivasan, (1992. ).“Computer htt://www.cit.gu.au/~noran.
Aided Gating and Metal Rising 24. A.,K.,L. and Chan, (2001) “Application
Simulation”, Technical Paper of Fuzzy Set Theory in Flexible
Proceedings of the 40th Annual Manufacturing System Design”,
convention of IIT, Madrass,14-16 Feb. Computational Intelligence in
18. Steel Founders' Society of America, Manufacturing Handbook, CRC.
“Feeding & Risering Guidelines for 25. Driankov, Hellen and Reinfrank,
Steel Castings”, Steel Founders' Society ( 1993).“An Introduction to Fuzzy
of America, 2001. Control”, Springer-Verlag, Berlin
19. http://www.jimbrule.com/fuzzytutorial. Heidelberg.
html. Last accessed 4 November 2006. 26. Chaeung, J. and Pauly , (2005).
20. Siler. W. and J. Buckley,( 2005). ”A Fuzzy Logic Expert System for
“Fuzzy Expert Systems and Fuzzy Estimating the Intrininsic Extinction
Reasoning”, John Wiley & Sons, Inc. Vulnerabilities of Seamount Fishes To
New Jersey. fishing”, Morato,T. and Pauly, D. (eds.)
21. ALY, S, and VRANA,(2006). “Toward Seamounts, Biodiversity and Fisheries
efficient modeling of expert systems: a Center Research, pp 33-55.
survey” Agric. Econom.-Czech, pp. 27. Zadeh, L. (1965). “Fuzzy sets”.
456-460. Information and Control, 8: 338-353.
22. Giarratano and Riley, (2005). “Expert 28. Peter, Szabdcs, Andor and Nador,
System: Principles and Programming”, (2005). “Fuzzy logic Motor Control
4th Ed., Course Technology Publisher. with MSP430 x 14x,Application”
23. Griffith University, “The Evolution of Report, SLAA235.
Expert Systems, available electronically” at:
134
