超声电机的EM核鲁棒建模研究

Research on Ultrasonic Motor Modeling Based on EM Kernel Robust Method

针对在超声电机的物理机理建模存在的问题,提出了基于数据驱动的建模方法。运用贝叶斯理论进行模型参数估计。使用EM(参数期望最大化)的性能指标,得到基本理论公式及计算迭代方法。提出了模型适配度作为超声电机模型鲁棒性的评价指标,并进行研究。通过实验和模型计算,参数鲁棒性和建模精度均有提高。经实验证明,所述的建模方法在超声电机的建模工作中具有较好的精度和参数鲁棒性。

Aiming at the ultrasonic motor modeling problems existing in the physical mechanism method,a datadriven modeling method was proposed.Bayesian theory was used to estimate model parameters.The basic theoretical formula and iterative algorithm were obtained by using the performance index of EM(expectation maximization).The model fitness was proposed as an evaluation index of the ultrasonic motor model robustness of,and the relative research has been carried out.The parameter robustness and modeling accuracy have been improved based on experiment and model calculating.Experiments show that the modeling method described has better accuracy and parameter robustness in the modeling of ultrasonic motor.