【作者】 王杰；

【导师】 蒋静坪；

【作者基本信息】 浙江大学 ， 工业自动化， 1996， 博士

【摘要】 交流伺服驱动系统具有二阶线性的简化数学模型，但实际上却存在着多变量耦合和非线性，并且系统反应非常迅速。针对这些特点，本文主要研究了适合于这类系统的基于模型和无需模型的控制策略。 利用多项式矩阵互质分解和控制器结构变换技术导出的调节器简化综合方法同时解决了正方形和非正方形系统传递函数矩阵的闭环零、极点任意配置问题且无须求解丢番矩阵方程。混合H₂及H^∞鲁棒最优控制方法可以使控制系统同时满足稳态及动态两类指标。特征向量参与优化的最优频谱配置设计方法不须“返回”且具有全局最优性。在解决了权值的无扰动切换之后，由同一网络实现的变结构神经网络控制器具有单神经元PID和BP网络控制器的各自的优点。由PID型BP算法和修正的Solis&Wets随机最优化方法既具有快速收敛性，又能在有限次的迭代里实现全局最小。带权值修正功能的神经网络—模糊控制方法具有明显优于单纯模糊控制方法的性能。 所研究的控制方法均在AC200交流伺服驱动系统的频率域实验模型上做了仿真实验，结果支持了本文分析所得的结论。更多还原

【Abstract】 The simplified mathematical model of an AC Servo-Drive-System is theoretically of 2-nd and linear, but there exist really nonlinearities and coupling of multi-variables. The response speed of the system is further very fast. For these characteristics, this paper deal with two kinds of control strategies that are model-necessary and model-free respectively.By using coprime decomposition of polynomial matrices and structure variation of controllers, the simplified synthesis method of linear regulators achieved the arbitrary assignment of closed-zeros and poles of square and non-square transfer matrices without necessity of solving complicated Diophantine matrix equations.The hybrid H₂ and H^∞ robust control strategy is able to make thesystem satisfy steady and transient requests simultaneously. The non-recursive optimal design procedure of close loop eigen values with participation of eigenvectors has a global optimality. With the problem of non-disturbance transportation of weightings being solved, the variable structure neuron controller, that is realized in fact by only one network, has all the advantages of single neuron PID- and BP network-controllers. A hybrid algorithm combining PID-BP and revised Solis&Wets random optimal method has a very fast convergence speed and can achieve the global minimum in a limit number of recursions. The neuron network realized fuzzy controller with function of weightings-learning is proved to be superior to normal fuzzy controllers in control precisions.All the concerned control strategies are supported with the simulation results that are made on the experimental frequency domain model of the AC200 servo drive system.更多还原