CONCLUSION
In this paper, a novel fuzzy identification method that combines
a modified cluster validity criterion index and ACO and
OLS algorithms is proposed. In general, the initial premise
structure and consequent parameters might affect the final
identification result. Therefore, in this paper, the ACO algorithm
is adopted to sift better initial premises and the consequent
parameters. In addition, an insufficient number of
rules will lead to an inappropriate premise structure and unreasonable modeling results. A modified cluster validity
criterion index is thus adopted to determine the appropriate
rule number. In addition, instead of the conventional FCRM
algorithm, an improved FCRM algorithm is utilized, and
the fuzzy membership matrix U and consequent parameters
2 can be refreshed and varied towards the direction
that minimizes the fitness function gradually. Furthermore,
the antecedent parameters can be obtained from U and the
final consequent parameters can be obtained via the OLS
algorithm. Last, three examples are given to show that the
proposed identification method provides better approximation
results and robustness than those obtained using other
methods.