超声波电机二阶线性时变模型的迭代学习辨识建模  

Iterative learning identification modeling of second-order linear time-variant models for ultrasonic motors

作　　者：周星龙 史敬灼 ZHOU Xinglong;SHI Jingzhuo(College of Electrical Engineering,Henan University of Science and Technology,Luoyang 471000,China)

机构地区：[1]河南科技大学电气工程学院,河南洛阳471000

出　　处：《电机与控制学报》2024年第8期174-184,共11页Electric Machines and Control

基　　金：国家自然科学基金(U1304501)。

摘　　要：为了提高超声波电机(USM)转速控制模型的精度,提出一种基于迭代学习辨识的二阶线性时变模型建模方法。针对USM的非线性与时变特性,首先基于超声波电机驱动电压的频率作为输入变量,转速作为输出变量,建立一个描述其动态行为的线性时变模型。为提高模型辨识的精度与鲁棒性,改进迭代学习辨识(ILI)算法,通过优化误差加权范数和参数变化率范数相结合的目标函数,并且设计新的学习律,利用归一化柯西滤波窗口进一步地构造学习律的权重矩阵。仿真与实验结果表明,所提的二阶线性时变模型在描述USM转速运动特性方面具有较高的精度,迭代学习辨识算法展现了优良的收敛性和鲁棒性,能够有效应对时变扰动并改进USM的运动控制性能。To improve accuracy of the speed control model for ultrasonic motors(USM),a second-order linear time-varying model based on iterative learning identification(ILI)was proposed.A linear time-varying model was established with the driving voltage frequency as the input variable and the rotational speed as the output variable to capture the dynamic behavior of the USM,addressing its nonlinear and time-varying characteristics.An enhanced ILI algorithm was developed by optimizing an objective function that combines error-weighted norms and parameter variation rate norms,and a new learning law was designed with a weight matrix constructed using a normalized Cauchy filtering window.Simulation and experimental results demonstrate that the proposed second-order linear time-varying model accurately describes the speed dynamics of the USM,while the ILI algorithm exhibits excellent convergence and robustness,effectively addressing time-varying disturbances and improving the motion control performance of the USM.

关 键 词：超声波电机 线性时变系统 迭代学习辨识 学习律 双范数最优 转速控制模型 

分 类 号：TM359.9[电气工程—电机]