Mechatronic System Modeling Based On Input And Output Data

With the increasing improvement of complexity and control requirements ofmechatronic systems, modern control algorithms which will inevitably need to know thesystem model need to be used in controller designing. However, for complex systems, itis difficult to establish accurate mathematical model of the system. But such systems areusually associated with a substantial amount of experimental data, and expertexperience and knowledge. In these premises, how to obtain the system model becamean urgent problem in system control.This paper first proposes a class of linear subspace identification method, which doesnot depend on any a priori knowledge of the system but entirely by input and outputdata to identify system model. This method has been very widely used in practicalapplications. Simulation and experimental result highlight the effectiveness of themethod.Then a high-performance feed-forward control method by inverse dynamics isproposed. The method is based on model and feedback controllers and is a furtherimprovement of the feedback control. First, the desired output is set for the system, thenthe inverse idea is used to solve a corresponding input signal. The system model used inthis method is received by the above identification. Simulation and experimental resultvalidate that the original control performance is greatly improved.Finally, this article concludes a nonlinear identification method based on linearfractional transformation (LFT). The methods first assume that the linear part is known.Then using the information of system interconnections between linear and nonlinearmodules, the overall input and output signal is decomposed into input and output signalsof each module. Thus the non-linear model can be identified. Simulation examplesverify the effectiveness of the method.