Physical Model Generation in PDE Analysis using Model-based Case-based Reasoning

Abstract

Model generation has emerged as a key task in engineering design and analysis. AI research in this area has focused on model based reasoning emphasising qualitative models in attempting to automate this process. In this paper, we propose that this work on the use of model based reasoning in model generation would benefit from the inclusion of case-based reasoning (CBR) techniques. We argue that the use of cases constrains the reasoning process as cases reflect known good routes in the solution space. Cases also have the advantage of facilitating the integration of heat transfer exemplars, approximations, formulae and correlations. In addition, much of human competence in this area is based on reusing solutions to previously solved problems and CBR emulates this. In the paper, we advance these arguments based on our experience with CoBRA, a CBR system for physical model generation for the domain of heat transfer described by partial differential equations.