Proposta de identificação recursiva de sistemas dinâmicos não lineares baseada em modelo nebuloso Takagi-sugeno inverso: uma abordagem direta no contexto do espaço de estados com observador de Estados

Abstract

In the areas of control and system identification, inverse modeling from experimental data is one of the major challenges for nonlinear dynamic systems based control strategies. Therefore, in this dissertation, a proposal for recursive inverse identification of nonlinear dynamic systems based on Takagi-Sugeno fuzzy model and structured in state space with state observer is presented, resulting in a inverse Takagi-Sugeno fuzzy model. To obtain the <antecedent proposition> of this model, this proposal uses the Gustafson-Kessel (GK) fuzzy clustering methodology in a time sliding data window, achieving static and dynamic estimation aspects. In order to obtain the <consequent proposition>, this proposal works with a Kalman observer state-space structure, which has been defining for system inverse mapping at each operation point, resulting in inverse fuzzy local submodels. To estimate the parameters of these submodels, the Kalman filter / Observer IDentification (OKID) methodology is used. Furthermore, to demonstrate the proposed methodology applicability, experimental results, which are concerning to a continuous stirring tank reactor, a Hammerstain model benchmark and a helicopter with two degree of freedom, are presented as case studies. Then, analyzes and final considerations highlight aspects of the optimal experiment, which is resulting from each case study.