In this paper, the background and principles of case-based reasoning are described, and we discuss how the use of a CBR(l) approach to data retrieval allows us to provide more intelligent support during the materials selection process as compared to the support offered by more traditional databases or conventional expert systems. This will be illustrated in particular for two case-based reasoning applications we recently developed: (a) for selecting materials corresponding to a given set of required (physical) properties and specifications, and(b) for the more challenging task of selecting materials for corrosion resistance in complex environments (e.g. mixtures of chemicals). These applications represent a first step towards the development of novel information retrieval tools that better support the engineer during the materials selection and specification phases.
In the past 10 years or so, a number of attempts were made to overcome the limitations of conventional materials databases by building knowledge or reasoning capabilities into this type of information-handling systemsl-2. The vast majority of such present-day know/edge-based systems used for selecting materials, or for similar materials engineering problems, are expert systems that encode their knowledge as a large set of domain-specific ''if-then'' rules. These rule-based methods are an effective way of utilizing heuristic knowledge, but they suffer from severe limitations in capturing human expertise, such as the difficulty to represent deep knowledge (i.e., fundamental or theoretical knowledge) and to incorporate exceptional cases in the knowledge base. In the real world, however, many rules have exceptions and there are many situations where the use of if-then rule chains does not accurately model the problem- solving strategies used. In particular, most human experts solve new problems by recalling into their memory past experiences, and by utilizing these previously encountered -similar- cases or prior successfully applied solutions as a guide during their problem-solving behaviour. This approach to problem- solving is called case-based reasoning, and CBR systems are systems which mimic this kind of problem- solving behaviour3-7. In a CBR system, experience is stored as a set of cases in a case-base. New solutions are derived using old solutions to previously solved cases, i.e. the system uses its knowledge of relevant past cases to interpret or solve a new problem.
Due to their vast potential in a number of domains7-11, companies are using CBR applications in industrial design, planning, and explanation systems. Other prominent CBR application areas are call centers and help desks for technical support as well as corporate memories. In this paper, the potential of using a CBR approach to materials data retrieval and (improved) materials selection is illustrated by discussing two CBR applications that have been developed in recent years. The first application (named ?M- BASE?) is designed to facilitate the process of retrieving materials which conform as closely as possible to a given set of specifications, while the second application (named ?C-BASE?) is developed to help the materials engineer in selecting corrosion-resistant materials for complex chemical environments.
-------------------- -------------------
?case-based reasoning, Esteem Software Inc.
A CBR APPROACH TO MATERIALS INFORMATION RETRIEVAL
Information retrieval and materials selection
A first but important step in the creation of new or improved industrial systems and structures is the selection of appropriate materials of construction from the wide variety of advanced materials which are av
