Multi-layer fuzzy system modeling a new approach: Theory and application

In this paper, the concept of multi-layer fuzzy system modeling methodology is presented to deal with rule uncertainties, to cover the different aspect of the modeled system, to reduce data acquisition process and to improve the accuracy and robustness of the systematic fuzzy modeling method proposed in references [1-3]. This method is based on partitioning of the output space using fuzzy c-means clustering and the projection of output space clusters onto input space to construct generic Zadeh fuzzy rules from input-output data. First, several fuzzy models of the system which are called "primary fuzzy model" in this paper, are constructed corresponding to different values of membership grade, the fuzziness parameter, and number of clusters. Then the final output of fuzzy model is calculated based on the weighted sum of the primary fuzzy models output. An approximation of two benchmark nonlinear functions are used to illustrate and describe the new modeling concept. Finally, the developed method is being employed to construct the multi-layer fuzzy dynamic model of a two degrees of freedom robot for control applications.