سال انتشار: ۱۳۹۱
محل انتشار: بیستمین کنفرانس سالانه مهندسی مکانیک
تعداد صفحات: ۴
Reza Kazemi – Department of Mechanical Engineering, K.N.Toosi University of Technology
Majid Abdollahzade – Department of Mechanical Engineering, K.N.Toosi University of Technology
Arash Miranian – School of Electrical & Computer Engineering, University of Tehran
System identification attempts to construct mathematical models of systems, using experimental measurements and observations. In this papercombination of local linear neuro fuzzy model (LLNF) and discrete wavelet transform (DWT), called WLLNF, is proposed for nonlinear system identification. LLNF model trained by local linear model tree (LOLIMOT) learning algorithm exhibits remarkable modeling performance, owing to strategy of dividing complex nonlinear systems into a set of local linear sub-models. Exploitation of good signal processing properties of DWT for denoising of experimental data enhances the modeling capability of the LLNF model. Identification of two nonlinear systems, namely flexible robot arm and hydraulic actuator, and comparison with other methods, demonstrate the promising performance of the proposed approach.