Formulating polyurethanes using case based reasoning

A large amount of historical knowledge exists in the form of ‘formulation experiences’ across polyurethane manufacturing companies. This knowledge is difficult to formalise, share and use in new formulations. As a part of an effort to support the polyurethane formulating problem, the use of case based reasoning (CBR) has been assessed. Two basic problems in the development of the proposed hybrid tool that uses past formulations to solve new problems are studied. The problems investigated are related to the retrieval of former formulations that are similar to a new problem description by the CBR module, and the adaptation of the retrieved case to meet the problem constraints using an artificial neural network (ANN). Results indicated that the CBR-ANN system is useful for reusing historical data. Although the obtained ANN is unable to generalise well when presented with more data independent from the original data set, results proved that real formulation data can be used as a ‘knowledge repository’ that can guide CBR adaptation without human expert intervention.