【正文】 routine or innovation design. Design of injection mould mainly involves consideration of design of the following elements or subsystems: (1) mould type (2) number of cavities (3) cavity layout (4) runner system (5) ejector system (6) cooling system (7) venting (8) mounting mechanism Most of the elements are interdependent such that itis virtually impossible to produce a meaningful flowchart covering the whole mould design process. Someof the design activities form a plicated design work as shown in Fig. , in injection mould design, it is difficult for designer to monitor all design parameters. Cavity design and layout directly affects most of other activities. The application of advanced knowledge based techniques to assist designer in cavity layout design at concept design stage will greatly assist in the development of a prehensive puteraided injection mould design and manufacturing system. It is noted from Fig. 1 that a number of different layout patterns are possible with multiple cavities inside a mould. Higher the number of cavities of mould, higher the productivity of the injection mould. But this may lead to difficulties with issues such as balancing the runners or products with the plicated cavity shapes, which in turn may lead to problems of mould manufacturability. It is also possible that the number of cavities and the pattern of cavity layout will influence the determination of parting line, type of gate, 12 position of gate, runner system and cooling system. Most of the main activities of mould design are therefore linked to cavity layout design. Fig. 3 shows the relations between cavity layout design and other design activities. The cavity layout design problem therefore depends upon a number of functionalities of the overall mould design system, which includes: (1) definition of design specifications including analysis and description of characteristics of design problem (2) determination of mould type (3) determination of number of cavities (4) determination of orientation of product (5) determination of runner type and runner configuration (6) determination of type and position of gate (7) cavity layout conceptual design (8) evaluation of ejection ability, manufacturingability and economic performances (9) determination of cooling system (10) graphic results display and output 3. Structure of ICLDS and the Design Process The structure of the Intelligent Cavity Layout Design System (ICLDS) is based on casebased reasoning and ruledbased reasoning designed around the RETE++software system. Fig. 4 shows the overall structure of ICLDS schematically. Fig. 5 shows the general design process of ICLDS. The design process starts with the definition of design specifications. The ICLDS system retrieves similar cases from case base by puting the similarity between the cases and the new case. If the solution is satisfactory, then results are displayed graphically. If the solution is not satisfactory, then ICLDS will use rulebased reasoning with forward or backward chaining or a mixture of both to arrive at a solution. If the solution is still unsatisfactory, then the user has to modify some of the initial design specifications. The use of casebased technology in the design process in ICLDS allows the user to obtain the solution(s) of design problem more quickly and flexibly. The structure of knowledge base and database used in the development of ICLDS is 13 based on the underlying knowledge base and database structure from the RETE++ software system, which is a mercially available software development platform. 4. Development of ICLDS . Classifications of Knowledge For various logic and steps involved in layout design, there are different kinds of knowledge that needs to be described and represented in cavity layout design. The types of knowledge can be classified into five kinds based on object oriented (OO) concept as described below: (1) Design instance/case: previous design cases and current design instances (2) Relation: superclassclasssubclass relation, classin stance relation (3) Attribute: design variables, features, attributes of design problem (4) Rule: general design rules, design experiences (5) Procedure and/or model: numeric calculation, mathematical modeling, analysis, evaluation and procedures. . Knowledge Representations To describe each of these types of knowledge, the internal data structures of the ECLIPSE language, included in RETE++ inherently, can be used to make the object orientated representation of the design process as explained earlier. . Casebased Reasoning CaseBased Reasoning (CBR) is dependent firstly on case retrieved. Casebased retrieval is based on “Similarity Metric”. Therefore, how to calculate the similarity is obviously the key technique in CBR, and it is described in detail as below. which, since dij must range between 0 and 1, must