Industrial applications of fuzzy system modeling

Abstract

Aggregate industrial system behaviour models can be built with fuzzy data mining provided the historical system behaviour data are available from system databases. Given the input-output data vectors, a unified system modeling approach can be used to extract "hidden rules" of system behaviour using fuzzy technology. In particular, fuzzy cluster analysis could be used with unsupervised learning to extract fuzzy set membership function and the fuzzy rule structures. A parametric reasoning method combined with supervised learning with minimum error criteria could determine combination operators. This eliminates the arbitrary choice of t-norms and t-conorms that are required in the execution of approximate reasoning algorithms. Examples given include continuous caster scheduling in steel making with criteria of minimum tardiness and minimum mixed grade steel production. This methodology can also be applied to pharmacological analysis of experimental data.