ACCELERE: a case-based design assistant for closed cell rubber industry

https://doi.org/10.1016/S0950-7051(99)00025-8Get rights and content

Abstract

The paper describes a knowledge-based system called ACCELERE, which was developed in order to aid in designing closed cell rubber. Designing a closed cell rubber is a complex activity. The complexity comes from the high number of design variables and the interaction between them. Consequently, the available knowledge is incomplete and does not permit to solve a problem at the first attempt. Then a design episode is an iterative process that can last up to 24 months. ACCELERE was developed to reduce this time, and to help the designers in synthesis as well as in evaluation phases of the design cycle. As a result of the lack of knowledge, ACCELERE uses Case-Based Reasoning as main problem solving method.

Introduction

This paper presents ACCELERE, a knowledge-based system that aids in designing closed cell rubber.1 Designing a closed cell rubber is a complex activity. Because of the high number of design variables and the interactions between them, there is no model, no strong theory, allowing designers to solve a problem at the first attempt. Consequently, designing new products can often last up to 24 months. The primary industrial objective of ACCELERE is to reduce the design cycle. Closed cell rubber design in an industrial setting involve teams of designers, each of whom having and using a certain amount of knowledge. The second application objective of ACCELERE is to capitalize on the expert's knowledge, in order to distribute it among all the designers, especially the novices. There is at least two ways of acquiring and capitalizing on the knowledge. The first one consists in interviewing the experts of the domain, and storing their knowledge, for example, in the form of production rules. The limits of this method are now well known (difficulties to extract knowledge, difficulties to maintain the knowledge base, etc.). The second way consists in storing design problem solving experiences, carried out by the experts. In fact, to solve a problem expert's use their knowledge. Hence storing design episodes permit to capitalize on expert knowledge. Case-Based Reasoning (CBR) is a technology that deals with storing, retrieving and reusing past experiences. This observation and the fact that designers often make use of previous design experiences for new design problems, lead us to integrate CBR in ACCELERE.

The paper is organized as follows. In Section 2, we present the problem of designing closed cell rubber. The concepts of needs, functions, artefact, structure and specifications will be clarified. In Section 3, after a presentation of the CBR principles, we present the ways ACCELERE can be used by designers. Section 4 deals with knowledge representation, and Section 5 treats the retrieval of cases from the memory.

Section snippets

Closed cell rubber design

According to Rosenman and Gero [1] design proceeds from a conceptual description of a problem (a need) to a concrete description of an artefact as a solution to the problem. The description of the artefact (i.e. artefact's elements and their relationships), is called a structure. Mapping a need to a structure is not straightforward, and the design process is often divided in phases. Maher [2] divides the design process into three phases: formulation, synthesis and evaluation:

  • 1.

    Design formulation

Principles of CBR

The principle of CBR consists in solving a problem by using the solution of a solved problem, similar to the present problem (for an introduction to CBR see Ref. [4], contributions to CBR in design can be found in the book by Maher [5]). A design case is often made of a pair (problem description, problem solution). A problem to be solved is called a target case. Solved problems are stored in cases called source cases. The source cases are stored in a case-base. The two main tasks a CBR system

Knowledge representation

ACCELERE uses an object-oriented knowledge representation. The knowledge used can be divided into domain objects, domain knowledge which express relations between domain objects and cases that represent specific knowledge of a design episode.

Case retrieval

The case-base is interrogated thanks to a query, noted req, representing the specifications. In the following sections, the concept of query will be defined, as well as the way that the similarity between a case (or more generally an instance) and a query is evaluated. The similarity computation must reflect the constraints stated in the specifications, in the most pertinent way. For example, if specifications express that results for the oil resistance test should be greater than 50%, the

Conclusion

ACCELERE is still under development but the prototype is being used by a team of designers for the past one year. The first results are the following:

  • Designers use the system in their every days works. To date the case-base contains over 1000 design cases (trials).

  • The ability to browse among previous design episodes, for synthesis as well as for evaluation, is appreciated by the designers and seems to stimulate their creativity. The designers appreciate too that ACCELERE does not try to propose

References (8)

  • M.A. Rosenman et al.

    The what, the how and the why in design

    Applied Artificial Intelligence

    (1994)
  • M.L. Maher

    Process models for design synthesis

    AI Magazine

    (1990)
  • R. Tassinari

    Pratique de l'Analyse Fonctionnelle

    (1992)
  • A. Aamodt et al.

    Case-based reasoning: foundational issues, methodological variations, and system approaches

    AI Communications

    (1994)
There are more references available in the full text version of this article.

Cited by (0)

View full text