含非线性因素的扫描镜系统线性模型辨识

发布时间：2018-04-20 17:00

本文选题：扫描镜系统 + 系统辨识　；参考：《中国科学院研究生院（长春光学精密机械与物理研究所）》2012年博士论文

【摘要】：为将基于模型的控制器设计方法应用到航空相机像移补偿控制中，本论文研究含有非线性因素的像移补偿系统的线性模型辨识。以某航空相机扫描镜系统为研究对象，从非线性影响程度评估、非参数模型辨识、参数模型辨识和模型的闭环校验四个方面对扫描镜系统进行了研究，获得了扫描镜系统最佳线性近似模型。介绍了扫描镜系统构成，分析了系统中存在的两种非线性因素并建立了扫描镜系统理论模型和仿真模型。通过实际测量获得了扫描镜系统的死区及功率级负载非线性效应，针对脉宽调制（PWM）功率驱动方法引入了等效输入电压和等效启动电压概念，由于功率级负载非线性效应使得扫描镜系统等效启动电压增大。使用不同方法对扫描镜系统非参数模型进行辨识。分别使用Blackman-Tukey谱分析方法和随机相位多正弦信号（Random Phase MultiSine）对扫描镜系统的非参数模型进行辨识，并获得了扫描镜系统最佳线性近似的非参数模型及不确定性的估计。研究结果表明：改进伪随机二进制序列的功率谱分布及优化随机正弦组合信号幅值因子都能有效改善模型低频段与正弦扫描测试结果间的偏差，扫描镜系统最佳线性近似的频率响应函数在低频段与正弦扫描测试结果差别较大。使用不同方法对扫描镜系统线性参数模型进行辨识，给出存在非线性因素影响时系统线性模型阶数选择方法。基于时域数据，使用预测误差辨识方法辨识扫描镜系统的参数模型，，对不同的模型阶数OE模型和ARX模型分别进行研究；基于频域数据，使用采样极大似然方法辨识扫描镜系统的参数模型。研究结果表明，对于存在非线性因素影响的系统，综合使用多种模型校验方法对不同阶数的线性模型进行检验，可以获得适合的辨识结果。扫描镜系统参数模型的闭环校验。选定一个控制器，比较模型与真实系统间在闭环条件下接近程度，给出与系统真实行为特性最接近的模型。通过对模型与实际系统间灵敏度函数和阶跃响应的比较看到，扫描镜系统存在非线性因素的影响，使得最佳线性近似参数模型能更好的反映系统闭环行为特征。本论文研究结果表明，对存在摩擦死区非线性和功率级负载非线性系统因素影响扫描镜系统，通过不同方法获得的线性模型中，系统最佳线性近似模型最能反映真实对象的行为特性。
[Abstract]:In order to apply the model-based controller design method to the image shift compensation control of aerial camera, the linear model identification of the image shift compensation system with nonlinear factors is studied in this paper. The scanning mirror system of a certain aerial camera is studied from four aspects: nonlinear influence degree evaluation, nonparametric model identification, parameter model identification and model closed-loop verification. The optimal linear approximation model of scanning mirror system is obtained. This paper introduces the structure of scanning mirror system, analyzes two kinds of nonlinear factors in the system, and establishes the theoretical model and simulation model of scanning mirror system. The dead-time and power level load nonlinear effects of scanning mirror system are obtained by practical measurement. The concepts of equivalent input voltage and equivalent start voltage are introduced to the PWM PWM-based power drive method. The equivalent start-up voltage of the scanning mirror system is increased due to the nonlinear effect of power level load. Different methods are used to identify the non-parametric model of scanning mirror system. The nonparametric model of scanning mirror system is identified by Blackman-Tukey spectrum analysis method and random Phase MultiSine method of random phase multi-sinusoidal signal, and the optimal linear approximation nonparametric model and uncertainty estimation of scanning mirror system are obtained. The results show that improving the power spectrum distribution of pseudorandom binary sequences and optimizing the amplitude factor of random sinusoidal combination signals can effectively improve the deviation between the low frequency band of the model and the results of sinusoidal scanning. The frequency response function of the optimal linear approximation of scanning mirror system is different from that of sinusoidal scanning in low frequency range. Different methods are used to identify the linear parameter model of the scanning mirror system, and the order selection method of the linear model for the system with the influence of nonlinear factors is given. Based on time domain data, the parameter model of scanning mirror system is identified by using predictive error identification method, and the different model order OE model and ARX model are studied separately, based on frequency domain data, The parameter model of scanning mirror system is identified by sampling maximum likelihood method. The results show that for the systems with nonlinear factors, the linear models of different orders can be tested by using a variety of model checking methods, and the suitable identification results can be obtained. Closed-loop calibration of scanning mirror system parameter model. A controller is selected to compare the close-loop approximation between the model and the real system, and the model that is the closest to the real behavior of the system is given. By comparing the sensitivity function and step response between the model and the real system, it is found that there are nonlinear factors in the scanning mirror system, which makes the optimal linear approximate parameter model better reflect the closed-loop behavior characteristics of the system. The results of this paper show that the linear model of the scanning mirror system is affected by the factors of the nonlinear system with friction dead-time and the nonlinear load of power level, and the linear model is obtained by different methods. The optimal linear approximation model of the system can best reflect the behavior of the real object.
【学位授予单位】：中国科学院研究生院（长春光学精密机械与物理研究所）
【学位级别】：博士
【学位授予年份】：2012
【分类号】：TP273;TP334.22

【参考文献】