Derivation of decision rules for the evaluation of product performance using genetic algorithms and rough set theory

In the manufacturing of critical components of a product, it is important to ascertain the performance and behaviour of those components being produced before assembly. Frequently, these part components are subject to stringent acceptance tests in order to confirm their conformance to the required specifications. Such acceptance tests are normally monotonous and tedious. At times, they may be costly to carry out and may affect the cycle time of production. This work proposes an approach that is based on genetic algorithms and rough set theory to uncover the characteristics of the part components in relation to their performance using past acceptance test data, that is, the historical data. Such characteristics are described in terms of decision rules. By examining the characteristics exhibited, it may be possible to relax the rigour of acceptance tests. A case study was used to illustrate the proposed approach. It was found that the cost in conducting the acceptance tests and the production cycle time could be reduced remarkably without compromising the overall specifications of the acceptance tests.