The main aspects of an expert system for process identification are presented. The knowledge base acts as a supervisor, alternating its reasoning concerning the model structure selection, with numerical parameter-estimation. The knowledge representation relies on a mixed structure: an object-oriented representation and production rules. Fuzzy compatibility functions are used for the numerical-symbolic conversion of values supplied by the parameter estimation module. The strategy for an "optimal" model structure search has been implemented for SISO as well as MISO processes. The performance of the expert system is illustrated on a real MISO case study.