基于切换系统理论的机电伺服转台多工况优化控制

There are several different working stations in the electromechanical servo turntable, such as even running, zero-crossing distortion, rapid acceleration/deceleration, etc. Thus the input-output relationship changes between different modes, as a result, the effect of traditional single-structure controller is limited. In our former work, it is found that the separate using of linear and NARMAX model on even and uneven states led to outstanding imitative effect. Therefore, the switched system theory is considered because of its ability to deal with the multi modes of working station. However there are still researches need to do on the aspects of modeling and optimal control, during its introduction to the electromechanical turntable. Based on the existing research, the main contents of this program are as follows: 1. Design and realize the identification experiments to the servo turntables. The clustering algorithm is applied to operating space division. Then the switched model parameters identification is converted into the Constrained Multi-objective Optimization Problem (CMOP). Simultaneously, both of the signal objective and Pareto heuristic multi-objective algorithms are proposed to get parameters and adjust the switching rules. Finally, the cross validation will be applied to prove the effectiveness of the switched model. 2. For propose of the errorless control result, the model predictive controller is designed for the linear subsystem, in the same time, nonlinear predictive functional controller is constructed for the nonlinear subsystem. The next step, the multi-models observer is applied to realize the undisturbed switching, and the opposite optimization is carried out for the modeling method and control algorithm as well. In conclusion, though the proposal of switching optimal control strategy, this project aims to make the theoretical basis of the performance improving for the electromechanical servo turntable in multi modes.

机电伺服转台存在平稳运行、过零畸变、急加/减速等多种工作状态，系统呈现出不同的输入输出关系，传统单一结构控制策略作用受限。申请者在前期研究中发现，分别用线性和NARMAX模型描述转台不同工况，可得到良好拟合效果。因此将切换系统理论引入机电伺服转台控制策略中应对多工况问题，但尚需在模型建立和优化控制方面深入研究。本项目研究内容包括:1.设计机电伺服转台的切换模型辨识实验，应用K-means聚类算法划分数据的操作空间，再将切换模型参数估计问题归结为有约束多目标优化问题，并研究单目标全面学习和多目标Pareto求解算法，最后交叉验证输出切换模型。2.分别对线性和非线性子系统构造模型预测控制和非线性预测函数控制算法，实现子系统无差控制；再由多模型观测器实现无扰动切换，并对切换模型和控制策略进行逆向优化。本研究提出的切换优化控制策略期望为机电伺服转台在多工况下性能指标的提高奠定理论基础。

电机驱动伺服转台是精密跟踪雷达、射电天文望远镜、惯导平台等装备的关键组成部分，其技术水平高低直接影响整体设备性能优劣，在军事和空间探索等领域有重要应用。机电伺服转台存在平稳运行、过零畸变、急加/减速等多种工作状态，系统呈现出不同的输入输出关系，传统单一结构控制策略作用受限。. 本研究主要内容包括：1. 设计机电伺服转台的切换模型辨识实验，应用K-means聚类算法划分数据的操作空间，再将切换模型参数估计问题归结为有约束多目标优化问题，并研究单目标全面学习和多目标Pareto求解算法，最后交叉验证输出切换模型；2.设计并实现基于非线性预测函数控制（Nonlinear Predictive Functional Control, NPFC）的切换控制策略；3.建立基于神经网络的温度扰动模型和LuGre摩擦模型，并设计和实现前馈补偿器，提高系统的跟踪精度。. 设计的机电伺服转台辨识实验和应用的辨识算法，得到了非线性切换模型，数据拟合精度可由传统的90.5%提高到99.94%。应用该模型设计的NPFC控制策略具有跟踪精度高、无超调、调节时间短、鲁棒性好的优点。设计的神经网络温度前馈补偿减小了1.82%的跟踪误差，基于模型的摩擦前馈补偿减小了7.32%的前期跟踪误差。. 本研究提出的机电伺服转台整体建模、扰动建模方法和切换优化控制策略期望可多工况下为性能指标的提高奠定理论基础，并可应用于其他类型的电机驱动伺服系统中。