COMPUTER-BASED TECHNIQUES FOR DECISION SUPPORT OF MOLD
Today, there is a clear trend to use more plastics and composites in engineering industry. The
time to market for plastic injection products is becoming shorter and more crucial, thus the
lead time available to making the injection moulds is decreasing. This will need an integrated
optimizing of product, process and material properties. To achieve product, tool and process
optimization, early decisions in design process must be supported.
Injection molding is an essential element of near net shape manufacturing process, where it
may account for over 25% of the total product cost and development time in engineering
industry, especially when order quantity is small.
This paper introduces a decision analyze approach to analyze the best praxis’s of companies
and use the results of analyzes for improvement the computer supported mold design process.
The pattern recognition technique is used together with the questionnaires developed by
authors to evaluate technically and economically acceptable variations of different plastic
product and injection mold parameters.
To collect the information about best praxis’s from different Estonian companies, the
developed questionnaires were used.
The implementation of the decision analyze results to customize Unigraphics MoldWizard
system is described.
Keywords: injection mold design, decision analysis and support system, questionnaires to
evaluate the best engineering praxis’s, pattern recognition methodology
Nowadays, mould design faces increasing deadline pressure and the design itself is
predominantly based upon the experience of the mould designer. Mold designers are required
to possess thorough and broad experience, because detailed decisions require knowledge of
the interaction between different parameters.
Today, many CAD systems provide the geometric modeling functions which facilitate the
drafting operations of mould design, but do not provide to mold designers the inexpensive
solution of necessary knowledge to develop good mould designs.
Computer-aided engineering packages are usually good at data processing for information-
intensive problems. However, in mould design a substantial practical knowledge is needed.
Therefore the methods, how to describe the data of best praxis and to put it into CAD system
to develop the good mould design, must be taken under consideration.
Previous works in applying computer techniques in the field of injection molding covered
many aspects of injection molding and mold design activities.
As an example Hui and Tan  presented a heuristic search approach based on sweep
operations to develop automated mold design systems for determining parting direction,
parting line, side core, etc. Lye and Yeong  established a computer-aided tool for the
selection of injections machines, the size and position of injection guns and the ejectors.
The knowledge based systems have demonstrated great potential to assist designers in
interacting with a CAD system for conceptualized design as well as for the final engineering
design of a mold by using engineering rules of thumb with extensive analytical procedures.
The major advantage of knowledge based system for mold design is the explicit
representation and manipulation of knowledge, representing human expertise and best
praxis’s of companies.
In this paper an approach to develop a supportive technique for CAD-based mould design is
described. The Unigraphics NX MOLDWIZARD has been chosen for base as example.
2 Approach and Methodology
Engineering Decision Analysis is a process that allows examining decisions in a structured
and efficient manner. Decision Analysis uses techniques such as Pattern Recognition ,
Artificial Intelligence , Machine Learning  and others in order to discover decision
patterns and relationships in data, to integrate existing analytical design methods and
instruments. It can be viewed as computer automated analysis and exploration of data and
knowledge (best praxis) used in companies to design of plastic parts, their manufacturing
technologies, including the design of injection molds.
Knowledge discovery, machine learning or data mining often involve large databases.
However, the databases are usually built for purposes other than knowledge discovery, and
there are often many redundant attributes in the databases. This increases the complexity of
knowledge discovery and decreases the quality of the discovered knowledge.
The knowledge discovery requires the following steps:
• Preprocessing of continuous attributes, optimizing the values of attributes, and
eliminating useless attributes.
• Finding a good reduct (sometimes called feature selection) from attributes of a
• Extracting knowledge from a decision system.
• Testing the rules and applying them to make decisions.
Our approach was to develop a simple system to support effective and rational decision
making in mold design process that uses existing CAD elements to represent collected
knowledge. To describe the decisions and evaluate the corresponding best praxis’s the authors
have collected data from different Estonian companies using elaborated questionnaires.
Standard Excel technique was used for modeling decisions.
Our proposed decision analysis technique helps:
• To model plastic part, technology planning and mold design problems as decision trees.
• Analyze decision based models & make better design decisions based on best praxis of
companies and using pattern recognition technique.
• Customize CAD system (UG NX MOLDWIZARD) for plastic part and mold design
• Communicate results of effective design options, underlying analysis, best design
decisions and supporting assumptions.
2.1 Knowledge acquisition
Overall mold design process includes product design, material selection, manufacturing
process selection, mold design and mold manufacturing process selection, shown in Figure 1.
Mold design process includes mold base design and core/cavity design. In mold base design
process engineers have to select mold type, mold dimensions, machining methods and mold
material according to plastic product parameters. In core and cavity design process engineers
have to select melt feed system, cooling and venting system and ejection system depending on
plastic product parameters.
Mold Design and Costing and
Figure 1. Overall process flowchart of decision support
In every selection stage, there have to be made several decisions to make these selections. The
decisions are follows:
Melt feed system related decisions:
1. Cavity number and layout, shrinkage, injection molding machine selection
2. Location, size, type and cross section of gate
3. Layout, length, diameter, type and cross section of runners
4. Location, length, diameter, type and cross section of sprue
5. Number and location of parting line, ejection direction
Cooling and venting system related decisions:
1. Length, diameter and layout of cooling channels, distance between channels itself and
between channels and mold cavity
2. Venting gaps or vents location and size
Mold base assembly related decisions:
1. Type of mold
2. Mold dimensions
3. Locating ring size
4. Auxiliary equipment
Ejection system related decisions:
1. Ejection system type, layout, components (plates, pins, bushes) and their type,
position, dimensions and number
2. Ejector return system type and location
Mold materials related decisions:
1. Cavity and core insert material
2. Cavity and core plate material
3. Mold plates materials
4. Other components materials
Machining methods related decisions:
1. Cavity and core machining and finishing methods
2. Gate, runners and cooling channels machining methods
3. Mold plates machining methods
In our approach we developed questionnaires to collect information and data from Estonian
companies. The questionnaires were divided into following four parts:
1. Product information – information about product dimensions, shape, material,
production and different environmental requirements etc.
2. Process information – information about tool, injection machine and injection process
parameters such as preheating, cooling time, ejection time etc.
3. Tool information – information about mold type, mold dimension, cavity, runners,
gate, ejection system, cooling system, mold base components, materials and
4. Cost information – information about part cost, plastic material cost, mold cost and
For every decision making stage there has to be developed most reasonable and suitable
technique to evaluate collected knowledge. In this paper we choose, as an example, mold base
assembly related decisions making (mold type and dimension selection) to show how
proposed methodology could be used in Unigraphics NX MOLDWIZARD environment.
3 Evaluation of collected knowledge
Mold base assembly design during various stages is different. Mold base design includes
mold type selection, mold dimension selection, machining methods selection and mold
material selection. Therefore, these different design stages cannot be represented just by one
kind of knowledge representation. Depending on what kind of decisions we have to make,
reasonable method or mechanism must be chosen. Each manufacturer has been developed its
own mold type standards. Our example based on DME mold base standard components. DME
provide six different mold types to choose from UG NX MOLDWIZARD, shown in Figure 2.
Each mold type has its own specification and number of plates (of type TCP,AP, BP, CP ,..
etc) which depends on different plastic product parameters. Product parameters are also input
5 features in pattern recognition technique.
Type 1 Type 2 Type 3
Type 4 Type 5 Type 6
Figure 2. Types of mold constructions: TCP, BCP – clamping plates, AP, BP – cavity plates, CP – risers, SPP –
support plate, NP1 – intermediate plate number 1, NP2 – intermediate plate number 2.
The fundamental objective for pattern recognition is to provide a meaningful categorization of
the input. It can be considered to be a two stage process: feature extraction and pattern
classification. A feature is a measurement taken on an input pattern that is to be classified. A
suitable feature is one which will provide a definite characteristic of the input. The pattern
classifier is then supplied with a set of features which are then mapped onto a classification
state. Given the input features, the classifier must decide which type of class category match
most closely. Classifiers typically rely on distance metrics and probability theory to do this
To transform every non numerical value to numerical value, a following coding and
description rules were developed:
1. Product overall dimensions - parameter values are length(L), width(W) and height(H)
Because these parameters are numerical no coding is needed.
2. Product shape - Plastic parts can be approximated as either rectangular or circular.
Rectangular shape has parameter value 1; circular shape has parameter value 2.
3. Number of cavities – coding occurs as follows:
Parameter value 1 2 3 4 5 6 7 8 9
Number of cavities 1 2 3-5 6 8 10 12 14 16
4. Cavity layout - cavity layout types can be divided in five different categories: single
cavity layout has parameter value 1, circular layout has parameter value 2, series
layout has parameter value 3, inline layout has parameter value 4 and symmetric
layout has parameter value 5. Different cavity layouts are shown in Figure 3.
Figure 3. Types of cavity layouts
5. Plastic material – following coding system are used to give each material grade a
numerical value. Plastic material grade and corresponding parameter values are as
Plastic Material PMMA PC POM FL PA PVC ABS PS PE
Parameter value 1 2 3 4 5 6 7 8 9
As pattern classifier we choose Bayesian classifiers theory. Based on this theory, decision
calculation uses following equation:
gi(x) = - (x – µi)T Σ-1i(x - µi) + ln P(ωi) + ci , (1)
where x is a given feature vector (parameters of product, cavity layout and number), ωi is
classes (types of mold construction), µi is the mean value of the ωi class and Σi is covariance
matrix defined as Σi = E[(x – µi)( x – µi)T].
Also the appended parameters or indexes are used in different Excel spreadsheet files in this
paper. Design parameters can capture and represent the information which reflects the
usability of certain types of elements in mold design process. Accordingly these parameters,
the mold designers can evaluate which elements are the most suitable in certain mold design
Set of product
mold type 6
of mold base
of product and
Figure 4. A pattern recognition technique example in MS Excel spreadsheet file
The formula (1) of pattern recognition technique is calculating different numerical values
(figure 4) for various mold base types. After that numerical values are sorted increasingly.
Resorted list is reflected in mold base selection type area (figure 6).
4 Integration the design knowledge into CAD system
Computer-aided engineering techniques have been used to represent the injection molding
process to assist the mould design by simulation analysis. MOULDFLOW from Mouldflow
Australia, FLOW ANALYSIS from Plastics & Computer Inc., C-FLOW from Advanced
CAE Technology Inc,. POLYCOOL from Structural Dynamics Research Corporation, and
MOULDCOOL from Application Engineering Corp., are some typical commercially
available packages for modeling the mould filling process and/or cooling analysis . Also
the researchers have started to adopt a knowledge-based approach to solving the injection
molding and mold design problems in recent years. IMPARD  is an expert system
developed for injection molded part design. IMES  solved injection molding part-quality
problems. DTMOULD-1  is a knowledge based system for injection mould cost estimation.
MOULDX  and EIMPPLAN-1  incorporated moldability considerations into part
designs and addressed the conceptual design development of injection molded parts. ICAD
, the knowledge based system of Drexel University , etc, were developed for injection
The commercially available package Unigraphics NX MOLDWIZARD is convenient
program to customize mold design scenario. The normal NX MOLDWIZARD mold design
scenario in Unigraphics CAD environment itself could be described shown on figure 5.
Create suitable solid Initialize: Define parting sheets:
model • project name
Use molded part Add the moldbase
validation to verify
• mold csys Add:
• shrinkage • ejectors
• workpiece • slides,
• molding lifters
No corrections • sub inserts
molded part Complete the design
Yes • gates
Plan: No Is solid model • electrodes
• side actions acceptable • pockets
• layout • Bill of
• sub inserts Mat’s
• ejection Yes
• cooling No
• electrodes Is solid model
Figure 5. Overall MOLDWIZARD design process in Unigraphics CAD system.
The single standard elements or subassemblies could be described in Excel spreadsheet
format. To customize these available spreadsheets was emerged. To analyze decision based
models and make design decisions based on best praxis of companies, the pattern recognition
technique described afore applied to Excel spreadsheet.
The goal was to extend existing possibilities in Unigraphics NX MOLDWIZARD. It helps to
support the mold design process in tool making companies.
For example, if Moldbase Management dialog window in NX MOLDWIZARD was selected
(figure 6), it is clearly noticeable that two Excel spreadsheet file are used:
1. Mold base register file and
2. Corresponding mold base database file.
In mold base register file “moldbase_reg_mm.xls” pattern recognition technique (figure 4)
was used, to find what type of mold base components are most suitable which correspond to
different specific criteria’s (for instance DME metrical mold base components). The goal is to
sort different mold base types so what the most reasonable mold base type is the first in
selection in “Mold base Management” dialog window (figure 6). It helps the mold designer,
to select the most reasonable mold base type to his specific mold design project. Also, as
more specific data which correspond to input classifier are entered to mold base register file,
the overall mold base type selection is become more accurate.
Type of mold The resorted mold types
base catalogue. according to pattern recognition
technique in mold base register
The indexes which The shortcuts to two
reflect the different Excel spreadsheet
dimensions of file (mold base
selected mold register file and
type. specific database
variable “Index of
Figure 6. Mold base Management dialog window
Additionally, new parameter called “Index of usability” was entered. The meaning of “Index
of usability” is to record every single selection of mold base variant with specific dimensions
in “Moldbase Management” dialog window and copy that information back to the mold base
database file. It gives to mold designer possibility to evaluate which mold base variants are
mostly used in same kind of projects. Additional variable “Index of usability” is visible in
MOLDWIZARD “Moldbase Management” dialog window, in variable sector (figure 6).
The suitability of this methodology was tested on practical examples. The results indicates
that pattern recognition technique as additive element in NX MOLDWIZARD’s mold base
register file and additional variables are convenient to take further research consideration.
This paper has presented an approach and methodology of supportive technique of CAD-
based mould design procedure based on existing CAD resources (Unigraphics NX
MOLDWIZARD) and collected experimental data.
The proposed methodology does not aim to replace existing systems in companies but rather
to be a supportive tool for communicating and sharing specific heuristical and empirical data.
In the implementation of the proposed methodology in CAD could facilitate to utilize specific
knowledge in tool making company.
The future direction of this research is to implement that approach to other similar decision
making problems in mold design process. Combination of different knowledge representation
methods are also considered to find best solution for decision making.
The research was supported by the Estonian Science Foundation (Grant 5620)
 Mok C. K., Chin K. S. and John K. L. Ho., “An Interactive Knowledge-Based CAD
System For Mould Design in Injection Moulding Processes”, The International Journal
of Advanced Manufacturing Technology, Vol. 17, 2001, pp. 27 - 38
 Hui K., C. and Tan S., T., “Mould design with sweep operations – a heuristic search
approach”, Computer Aided Design, Vol. 24, 1992, pp. 81 - 91
 Lye S., W. and Yeong H., Y., “Computer assisted mould design for Styrofoam
products”, Computers in Industry, Vol. 18, 1992, pp. 117 - 126
 Huang T., S., “IMD – an aided design software for the design of injection moulds”,
Pasific Conference on manufacturing, 1994, pp. 377 - 385
 Xiang Q., Bell D., McGinnity M., “Multiknowledge in decision making”, Knowledge
and Information Systems, 2004
 Rodriguez K., Al-Ashaab A., “Knowledge web-based system architecture for
collaborative product development”, Computers in Industry, Vol. 56, 2005, pp.125-140
 Lardeur E., Longueville B., “Mutual enhancement of systems engineering and
decision-making through process modeling: toward an integrated framework”, Vol. 55,
2005, pp. 269 - 282
 Russell S., Norvig P., “Artificial Intelligence: A modern Approach”, 2003, pp. 1081
 Theodoridis S., Koutroumbas K., “Pattern Recognition. Second Edition”, 2003, pp. 689
M.Sc. Aigar Hermaste
Department of Machinery
Faculty of Mechanical Engineering
Tallinn University of Technology, TUT
Ehitajate tee 5, 19086 Tallinn, Estonia
Tel: Int +372 620 3269
Fax: Int +372 620 3250