A Systematic Method For Fuzzy Modeling From Numerical Data

minyou chen,d a linkens

DOI: https://doi.org/10.1109/ICSMC.2001.969783

2001-01-01

Abstract:A systematic fuzzy modeling method that includes the initial fuzzy model self-generation, significant input selection, partition validation, parameter optimisation and rule-base simplification is proposed in this paper. In this framework, the whole procedure of structure identification and parameter optimisation is carried out automatically and efficiently by the combined use of a self-organisation network, fuzzy clustering, adaptive back-propagation learning and similarity analysis. The proposed fuzzy modeling approach has been used to non-linear system identification and mechanical properly prediction in hot rolled steels. Experimental studies demonstrate that the proposed fuzzy models have a good balance between model accuracy and interpretability.