Multilinear-model dynamic matrix control: A normalized-eigenvalue-based fuzzy weighting method

Multi-model based scheme provides a valid methodology for the control problem of nonlinear systems, which uses the philosophy of Divide and Conquer. Operation space division determines which local model gets active or how all local models (/controllers) are weighted for control implementing in the current instant. This paper presents a novel fuzzy weighting method embedding in multilinear-model dynamic matrix control (DMC). Fuzzy weights of local models, calculated by Takagi-Sugeno (TS) fuzzy modeling for controlled systems, are relative to a type of nonlinear measures - normalized eigenvalues, derived from state-space representatives of the local models at different operation points. Then, an overall weighted model is integrated into the conventional DMC framework, and disturbances and input/output constraints can be dealt with. The proposed multi-model control method is a combination of DMC and TS modeling, thus it inherits advantages of both techniques. Case studies of two classic nonlinear systems illustrate the effectiveness of the proposed approach.