Fuzzy-logic-based process modeling using limited experimental data

Process modeling with limited experimental data is always a dicult task. It becomes even more dicult if the process is highly nonlinear and is characterized by multiple inputs and outputs. Under these circumstances, fuzzy logic may show its capabilities for model development. In this paper, an ecient fuzzy modeling methodology is introduced. The resulting fuzzy model consists of a number of fuzzy implications, each of which is of an IF–THEN form. The IF part consists of a set of logically related antecedents, while the THEN part contains a consequent expressed as a set of linear models. To ensure model simplicity and to accelerate the modeling process, an eective model-development route has been developed. To guarantee the model’s reliability, a t-test-based non-linearity analysis is proposed when each fuzzy implication is developed. The ecacy of the methodology is demonstrated by modeling two nonlinear industrial processes. # 2000 Elsevier Science Ltd. All rights reserved. 1