You are on page 1of 10

ELSEVIER Computers in Industry 30 (1996) 77-86

The selection of the casting process using an expert system


S.M. Darwish, A.M. El-Tamimi
Mechanical Engineering Department, King Saud Uniuersily, P.0. Box 800, Riyadh 11421, Saudi Arabia

Received 7 December 1994; revised 23 November 1995; accepted 23 November 1995

Abstract

This paper presents an expert system for the selection of casting processes. The selection criteria are based on the
production, design, manufacturing as well as trade-off attributes. The objective of the present work is to aid the designer in
adequately selecting casting processes for the production of components.

Ke_ywords: Expert system; Casting selection; Design attributes; Production attributes; Manufacturing attributes; Trade-off attributes

1. Introduction reference to experts in foundries or research in the


literature for the appropriate guidance.
Today’s manufacturing industry is finding that as To capture and safeguard know-how and provide
well as reducing costs and improving quality, they an aid, suitable for the design and manufacture of
are now having to reduce time scales in order to castings, expert systems are needed. This expert
compete effectively in an area of shrinking product system is intended to place specialist information at
life cycles and rapid advances in technology. This the designer’s fingertips, and links design, manufac-
calls for more specialist ‘up-front’ expertise and a turing, and production. There are databases already
concurrent engineering approach in order to elimi- available for the selection of materials, but these are
nate problems later. not exclusively for casting and therefore are not
Casting, like other manufacturing processes, ap- capable of selecting designs with compatible casting
pears to be an appropriate domain for knowledge processes. A knowledge base for alloy and process
systems [l-7] for many reasons including: no one selection has been developed [7]. However, the use
person can realistically be expected to know or of this knowledge base needs a lot of experience of
remember all aspects, and there is a large discrep- alloy composition and properties before any decision
ancy between the best and worst designs. can be made. There are also software packages
In fact, many factors are basic to the production (basically simulation programs) to aid the production
of castings with optimal mechanical and physical of castings, but these are for assistance in the re-
properties at the lowest cost; these are material re- design and production of components that have been
quirements, design attributes, production and manu- designed.
facturing attributes. HLowever, most general designers As one would expect from a mature technology,
have only rudimentary knowledge of casting technol- such as casting, the literature search identified huge
ogy and can not design for casting without continual sources of information covering all aspects of the

0166-3615/96/$15.00 Copyright 0 1996 Elsevier Science B.V. All rights reserved.


SSDI 0166..3615(95)00089-5
78 SM. Danish, A.M. El-Tamimi / Computers it] Industry 30 (19961 77-86

process. In this work, the information was initially


noted in text form and then organized under four
categories. These are:
Design attributes (material, weight, thickness).
Production attributes (production volume, produc-
ACQUISITION
tion rate).
Manufacturing attributes (surface quality and di-
mensional accuracy).
Trade off attributes (casting strength and cost).
Expert systems usually deal with the computeriza-
tion of problem solving that analyze large quantities
of material on a particular domain and provide an-
swers simulating human reasoning [I -71. The
knowledge of such systems is acquired from one or 4
more human experts of a certain field. Fig. 1, shows EXPLANATION
FACILITY
the structure of a typical expert system. Thus expert
systems act as artificial expert to arrive to the same
conclusion as the human expert. In the present work,
the expert system technique is adopted for the selec- I T
tion of casting processes. This is expected to aid the USER INTERFACE

designer in making the selection of casting pro-


cesses. .

Fig. 1, Expert system structure.


2. The selection of casting processes expert system

The selection procedure described in the present development, and maintenance, and an end-user in-
work is intended as a general purpose aid to the terface complete with pull-down menus and win-
designer in making the selection of casting processes dows enhance the ease of use and appearance.
for a given part. The selection procedure involves The processes stored in the knowledge base are
identifying the relevant possible alternatives (casting those used for the production of a cast engineering
processes) and ranking them according to their per- component. These are:
formance. In the proposed procedure, the casting 1. Manual sand casting.
material, weight, minimum thickness, surface rough- 2. Automatic sand casting.
ness, dimensional accuracy, production rate, and pro- 3. Die casting.
duction volume are all considered by the system. It is 4. Shell mould casting.
intended to provide a short list of alternatives which 5. Plaster mould casting.
will contain the best combination. 6. Investment casting.
The selection of casting processes expert system 7. Permanent mould casting.
was written using Rule Master, a software package 8. Centrifugal casting.
for building expert systems. Rule Master is charac- A summary of the processes capability informa-
terized in that knowledge can be entered either tion is given in Tables l-3 [g-12].
through examples, or directly as rules, backward and After establishing the information tables, the Rule
forward chaining as well as uncertainty, are sup- Master shell was used to build the preliminary cast-
ported and complete English language explanations ing processes selection expert system (PCPSES). It
assist the end-user in understanding the line of rea- was found convenient to structure the expert system
soning used to arrive at a solution. A modular, by developing four different modules (a design mod-
hierarchical structure assists in expert system design, ule, a production module, a manufacturing modules,
S.M. Danish, A.M. El-Tamimi/ Computers in Industry 30 (1996) 77-86 79

and a trade off module). Then these modules were of cast material (ferrous or non ferrous). In case of
nested by providing several call modules. Fig. 2, ferrous metals the state of the melting point (high or
shows the flow chart Iof the main module. low) is also questioned. The second sub-module is
the minimum thickness where the user has to pick
2.1. The design module one of five coded classes as shown in Table 4. It is
worth mentioning that these classes are identified
This module is concerned with the product design according to the capabilities of casting processes
characteristics which include the cast material, the [8-121. The third child in the design module is the
minimum thickness of castings and the casting casting weight sub-module, where four classes were
weight. Fig. 3 shows the flow chart of the design identified, coded, and offered to the user to choose
module. This module ‘contains three child (sub) mod- the appropriate class according to his application, as
ules. The first sub module asks the user for the type shown in Table 5 [8-121.

Table 1
Characteristics of casting processes
Casting method Range of metals Tolerances cm/cm Mass range Minimum Surface Economical lot size
thickness (mm) roughness Ra(um)

Sand casting No limit 0.03-0.05 Few gms to Mg 2.5-5 5-25 Up to mass prod.
Die casting Al. Mg, Zn, Cu 0.0015 (0.003 Cu) up to 50 kg 0.5-5 l-2 Min. 1000
Permanent mold C.I. Al, IMg, Cu 0.01-0.025 up to 50 kg 3 2-3 Minimum 1000
Shell mold No limit 0.02-0.05 up to 20 kg l-3 2-5 Minimum 100
Investment No limit 0.003-0.005 up to 10 kg 0.5-0.8 1.5-2 100-5000
Centrifugal No limit 0.03-O. 1 up to Mg 6 5-25 A few hundred

Table 2
Comparison of various casting processes

Property Sand casting Die casting Permanent mold Investment casting Centrifugal casting
Porosity 7 5-3 3-4 2-l l-2
Surface quality 6-7 2-3 4-5 1 5-4
Dimensional accuracy 6-7 l-2 5-4 3-2 5-4
Strength 7-5 l-2 4 7-5 3
Mold cost l-2 6-7 5 4 3
Production rate
Small lot size 2 6-7 4 3 5
Large lot size 3 l-2 3 5 4
Min. sec. Thickness 5-7 2-3 4 1 6

Table 3
Cost aspects of various casring processes
Process Tooling costs Direct labor cost Finishing cost Scrap cost
Sand casting Low High High Moderate
Shell molding Low to Moderate Low Low Low
Permanent mold Medium Moderate Low to Moderate Low
Die casting High Low Low Low
Investment casting Low to moderate High Low Low
80 SM. Dan&h, AM. El-Tamimi/ Computers in Industry 30 (19961 77-86

L-J Start

Selection
[designg,:
according to
1, -

+ Yes

Production attributes

1 Define weight range

Manufacturing attributes

Advice visible

Trade-off attributes

Fig. 2. Main module of casting process selection.


tion volumes were identified (according to the capa-
The design module results in a 60 possible cases bilities of casting processes) and offered to the user,
(3 for materials X 5 for thickness X 4 for weight as shown in Table 6 [g-12].
classes).
Table 4
Casting minimum thickness
2.2. The production module
Class No. Code Value (mm)

This module is concerned with the production 1 Tl < OR = 0.5


attributes namely: production volume and production 2 T2 0.5<OR=l.O
3 T3 1.0 < OR = 2.5
rate. Fig. 4 shows the structure of the production
4 T4 2.5 < OR = 6.0
module, which asks the user firstly about the volume 5 T5 > 6.0
of production, where four coded classes of produc-
S.M. Darwish, A.M. El-Tamimi/ Computers in Industry 30 (1996) 77-86 81

Table 5 Table 7
Casting weight classes The production rate classes

Class No. Code Value (Kg) Grade No. Code Designation

1 WI <OR=lOKg 1 H High
2 w2 lOKg<OR=20Kg 2 M Medium
3 w3 20Kg<OR=50Kg 3 L Low
4 w4 >50Kg

The second child in the production module is the


L-J S:art
production rate module, where three classes of coded
production rates were considered as shown in Table
7 [8-121. The production module produces 12 possi-
visible solution
ble alternatives which when multiplied by the combi-

r-, Define
Ql-+
quantity

code
Q4
range
visible solution

Non
I

Define surface finish

visible solution

.-

Goal
L--J+
Fig. 4. Production module.

Table 6
Casting production volumf:
Class No. Code Value (units)

1 Vl < =I00
2 v2 1OO<OROR=1OOO
3 v3 lOOO< =5000 Goal
4 v4 > 5000
Fig. 5. Manufacturing module.
82 SM. Darwish, A.M. El-Tamimi/Computers in Industry 30 (1996) 77-86

Table 8
Surface quality classes
Grade No. Code Designation

I HS High surface quality


2 MS Moderate surface quality
3 LS Low surface quality

nations of the design module (60) gives 720 possible


cases.

Define strength (S)


2.3. The manufacturing module level code (Sl-S3)

Fig. 5 shows the structure of the manufacturing


module. This module has two child modules namely Non
*
surface quality and dimensional accuracy modules.
The first of which, the surface quality sub-module,
considers three classes of surface quality (according

Table 9
Dimensional accuracy classes
Grade No. Code Designation

1 MD Moderate dimensional accuracy


2 LD Low dimensional accuracy
Fig. 6. Trade-off module

Fig. 7. General purpose straight centrifugal pump components. 1. Flange, 2. Sea1 plate, 3. Square ring, 4. Shaft seal, 5. Impeller, 6. Case, 7,
Base, 8. Motor pad, 9. Cap screw, 10. Drain fitting, 11. Water slinger.
S.M. Darwish, A.M. El-Tamimi / Computers in Industry 30 (1996) 77-86 83

to the capabilities of casting processes) which are 4. Conclusions


listed in Table 8 ([8.-121. The second child of the
manufacturing module is the dimensional accuracy An expert system, which is flexible in operation,
module, which considered two grades of dimensional has been developed to aid designers in selecting a
accuracy as shown in Table 9 [8-121. casting process for a given product.
Based on required casting characteristics, the pro-
2.4. The trade-off module posed casting process selection expert system pro-
duces a list of candidate process(s) to produce a
This module has two children the cost child and particular part. These choices help the designer in
the mechanical strength child. In the cost child the identifying alternatives early in the design process.
casting processes are arranged in descending order The system may be used either by those with
according to their cos#t [8-121, while in the mechani- much experience of designing castings to save their
cal strength module the casting processes were ar- valuable time, or those who need a lot of guidance.
ranged in descending order according to the mechan-
ical strength expected from each casting process (see
Fig. 6. Acknowledgements

Al-Samhan Factory for Saudi Pump Co. is highly


acknowledged for his collaboration.
3. The selection of a casting process with the
expert system/ example of application of the
system Appendix A. A typical run session of the design
module
Al-Samhan Company at Riyadh, Saudi Arabia
produces a general purpose straight centrifugal pump
(see Fig. 7). Th e 1’mn ‘t’mg conditions of the case of REMARKS ON THE SELECTION PROCESS
this pump are: ferrous material, low melting point, PROCEDURES
minimum thickness of class 3 (Table 4) weight 1) THE FIRST COMPULSORY STAGE SE-
corresponds to class 3 (Table 5), moderate produc- LECTION IS BASED ON:
tion rate (Table 7) production volume of class 2 * THE DESIGN ATTRIBUTES.
(Table 6) low surface finish (Table 8), and low [MATERIAL AND CAST MINIMUM
dimensional accuracy (Table 9). THICKNESS AND WEIGHT]
Appendices (A-C), show the run session con- 2) FURTHER SELECTION CAN BE CAR-
cerned with part 6 (see Fig. 7) of the pump. Ap- RIED OUT FOR THE FOLLOWING:
*
pendix A shows the run session of the design module PRODUCTION ATTRIBUTES:
of the casting process selection expert system, where [QUANTITY AND PRODUCTION
stipulating a ferrous casting, having low melting RATE]
*
point, minimum thickness of class 3 and weight of MANUFACTURING ATTRIBUTES:
class 3 leads to four possible casting processes; sand [SURFACE FINISH AND DIMENSIONS
casting (automatic and manual), shell mould, cen- ACCURACY]
*
trifugal casting and permanent mould casting. In TRADE-OFF ATTRIBUTES:
order to refine the selection the production module is [COST AND STRENGTH]
stipulated with a production volume of class 2 and (1)
moderate production rate which leads to sand casting
manual and automatic (see Appendix B). The selec-
tion was more refined in the manufacturing module PROCEEDING WITH SELECTION ACCORD-
by entering low surface finish which explicitly means ING TO
low dimensional accuracy. This confirms the sand * DESIGN ATTRIBUTE *
casting choice. (2)
84 S.M. Dam&h, A.M. El-Tamimi/Computers in Industry 30 (1996) 77-86

THIS IS THE DESIGN ATTRIBUTES MODULE R5 = THICKNESS RANGE > 6.0 AND
FOR MATERIAL TYPE;
FOR CAST WEIGHT; and Rl
FOR CAST THICKNESS R2
R3
continue menu access R4
R5
(3)
menu access

(7)
WHAT IS THE TYPE OF CAST MATERIAL?

F = FERROUS CASTING
N = NON-FERROUS CASTING
F WHAT IS THE CASTING WEIGHT [KG]?
N
RWl = WEIGHT RANGE < OR = 10
menu access RW2 = WEIGHT RANGE > 10 AND <OR =
20
(4) RW3 = WEIGHT RANGE > 20 AND < OR =
50
RW4 = WEIGHT RANGE > 50
WHAT IS THE MELTING POINT?
RWl
L = LOW MELTING POINT RW2
H = HIGH MELTING POINT RW3
L RW4
H
menu access
menu access
(8)
(5)

RECOMMENDED CASTING PROCESSES


FERROUS CAST WITH LOW MELTING POINT
SAND CASTING;
continue menu access
SHELL MOULD CASTING;
(6) CENTRIFUGAL CASTING;
PERMANENT CASTING;

continue menu access


WHAT IS THE MINIMUM CASTING THICK-
NESS [MM]? (9)

RI = THICKNESS RANGE < OR = 0.5


R2 = THICKNESS RANGE > 0.5 AND < OR =
1.0 Appendix B. A typical run session of the produc-
R3 = THICKNESS RANGE > 1.0 AND < OR = tion module
2.5
R4 = THICKNESS RANGE > 2.5 AND < OR = FURTHER SELECTION CAN BE CARRIED
6.0 OUT
S.M. Danvish, A.M. El-Tamimi/Computers in Industry 30 (1996) 77-86 85

ON THE BASES OF Appendix C. A typical run session of the manu-


PRODUCTION ATTRIBUTE: [QUANTITY and facturing module
RATE]
FURTHER SELECTION CAN BE CARRIED
continue menu acc’ess
OUT ON THE BASES OF
(1)
MANUFACTURING ATTRIBUTE:
(SURFACE & DIMENSIONS}
continue menu access
WHAT IS THE PRODUCTION VOLUME]?
(1)
Vl = SIZE RANGE < OR = 100
V2 = SIZE RANGE > 100 AND < OR = 1000
V3 = SIZE RANGE > 1000 AND < OR = 5000
WHAT IS THE SURFACE QUALITY CLASS?
V4 = SIZE RANGE > 5000
HS = HIGH SURFACE QUALITY
Vl
MS = MODERATE SURFACE QUALITY
-v2 LS = LOW SURFACE QUALITY
v3
v4 HS
MS
menu access
-LS
(2)
menu access

(2)
WHAT IS THE PRODUCTION RATE
[CASTS/HOUR]?
LOW SURFACE QUALITY
H = HIGH PRODIJCTION RATE
M = MEDIUM PR.ODUCTION RATE continue menu access
L = LOW PRODUCTION RATE
(3)
H
M
L THE RECOMMENDED CASTING PROCESS IS
menu access SAND CASTING
(3) continue menu access

(4)
THE RECOMMENDED CASTING PROCESSES
IS
References
SAND AUTOMATIC CASTING FOR MOD-
ERATE TO HIGH PRODUCTION RATE [l] R. Davis, “Knowledge based expert systems planning and
SAND MANUA.L CASTING FOR LOW TO implementation”, Addison Wesley Training System, Addi-
MODERATE PR:ODUCTION RATE son Wesley, Reading, MA, 1987.
[2] T. Accidiacono, “Computerized reasoning”, PC Tech. Jour-
continue menu access nal, May (1988) 44-56.
[3] F. Hayes Roth, D.A. Waterman, and D.B. Lenat, Building
(4) Expert Systems, Addison Wesley, Reading, MA, 1983.
86 SM. Darwish, A.M. El-Tamimi/ Computers in Industry 30 (1996) 77-86

S.M. Darwish earned his Ph.D. in me-


[4] SD. Hu, Expert Systems for Software Engineers, Chapman
chanical Engineering at Birmingham
and Hall, London, 1987.
University, United Kingdom, in 1986.
[5] A. Gupta and B. Prasad, Micro-Computer Based Expert
He is currently an associate professor in
Systems, IEEE Press, 1988.
the Mechanical Engineering Department
[6] R.K. Miller and T.C. Walker, “Artificial intelligence appli-
at King Saud University, Saudi Arabia.
cation in manufacturing”, 2nd SEA I Tech. Publication,
His research interests include:
1988.
CAD/CAM, Expert system, Metal cut-
[7] N. Sirilertworakul, P.D., Webster, T.A. Dean, “A knowledge
ting and welding of metals. He worked
base for alloy and process selection for casting”, Int. J.
as a consultant for the R&D unit of the
Mach Tools Manufact. 33(3) (1993) 401-416.
Aluminium Company of Egypt (198%
[s] J.G. Bralla, Handbook of Product Design for Manufacturing, ,,._*\
IYYI).
McGraw-Hill, New York, 1986.
A.M. El-Tamimi earned his Ph.D. in mechanical engineering
[91 D.B. Dallas, Tool and Manufacturing Engineers Handbook,
at The University of Manchester (Institute of Science and Tech-
McGraw-Hill, New York, 1976.
nology UMIST), United Kingdom, in 1978. He is currently an
[lOI Metals Handbook, 8th ed, Volume 5: “Forging and casting”,
associate professor in the Mechanical Engineering Department at
ASM, 1980.
King Saud University, Saudi Arabia. His research interest include:
[ill R.W. Heine et al., Principles of Metal Casting, McGraw-Hill,
CAD/CAM and Manufacturing systems. He is a member of
New York, 1967.
ASME, IIE, IEE and IME.
[121H.W. Yankee, Manufacturing Processes, Prentice-Hall, En-
glewood Cliffs, NJ, 1979.

You might also like