高速电主轴动力学模型参数多新息随机梯度辨识

Multi-innovation Stochastic Gradient Identification of Dynamic Model Parameters of High-speed Motorized Spindle

针对高速电主轴转子转速、磁链和电流中存在复杂的强耦合、时变非线性因素造成其系统模型难以精确建立的问题,结合多新息辨识理论,提出了一种高速电主轴动力模型参数的多新息辨识方法。根据高速电主轴的结构和特点,建立其动力学模型;通过对高速电主轴动力模型的离散化,估计参数项由当前误差扩展为包含当前误差和历史误差的向量,实现了高速电主轴的多新息模型参数辨识。通过与传统随机梯度辨识方法进行仿真对比,表明了多新息长度p的引入可以有效提高模型参数辨识的速度和精度,并且随着信息长度的增加收敛速度逐步提高,验证了该文方法的有效性和正确性。

In order to solve the problem that the system model of high-speed motorized spindle is difficult to establish precisely because of the complex strong coupling and time-varying nonlinear factors in rotor speed,flux linkage and current,a multi-innovation identification method for dynamic model parameters of high-speed motorized spindle is proposed according to multi-innovation identification theory.According to the structure and characteristics,the dynamic model of the high-speed motorized spindle is established.By discretizing the dynamic model the estimated parameter term of the high-speed motorized spindlesystem is expanded from the current error to a vector containing the current error and the historical error,and the parameter identification is realized.Compared with the traditional stochastic gradient identification method,the simulation results show that the multi-innovation length p of the proposed method can effectively improve the identification accuracy and convergence speed,and the convergence speed will increase with the increase of the information length.Simulation results verify the effectiveness and correctness of the proposed method.