Identification Of Linear Model For Scanning Mirror System With Nonlinear Distortion

In order to use controller designing method based model in image motioncompensation (IMC) control of aerial cameras, the dissertation gives research aboutidentification method of linear model for IMC system with nonlinear distortion. Theplant to be identificated is the scanning mirror of an aerial camera. The research iscomposed of detection of nonlinear distortion, non parametric model identification,parametric model identification and close loop validation. The best linear approxim-ation of the scanning mirror system is obtained finally.It is introduced the composition of the scanning mirror system, and two mainnonlinear distortion. Then its theoretical model is deduced and simulation model issetup. It is found that the driving system behaves nonlinearly around the dead zone.Two concepts, equivalent input voltage and equivalent start voltage are given forPWM power driving system. The equivalent start voltage is increased by thenonlinearity of driving system.It is given the different method to identifing non-parametric model of thescanning mirror. The Blackman-Tukey spetrum analysis method and Randommultisine signal exitating method are studied for obtaining the non-parametric modelof the scanning mirror system, while the best linear approximation of the scanningmirror system and its uncertanity are given. The research result shows that the revisedpseudorandom binary sequence and the optimized multisine signal decrease the biaserror between non-parametric model and the swept sine measuremnt result at the lowfrequency zone. Different method is used to identify the parametric model of the scanning mirror.The prediction error method based time data is studied, using OE and ARX modestructures, while in frequency domain, the sample maximum likelihood is studied.The research result shows different validation method should be used at the same toget propriate orders of parametric model for the plant with nonlinear distortion.The validation of parametric model of scanning mirror system is carried out in theclosed loop condition. The closest model to real system is given by examing thecloseness between model and real system in closed loop under certain given controller.By comparing the sensitivety funciton and the step response between model and realsystem, it is found that the best linear approximate model is the best model to reflectthe behavior of scanning mirror system in closed loop due to the nonlinear distortion.The research result of this dissertation show that for the scanning mirror systemwith friciton dead zone nonlinearity and the power driving load effect nonlinearsystem, the best linear approximate model is the best model to reflect the behavior ofscanning mirror system in closed loop by comparing the different models obtainedwith different methods.