自适应滤波器的新型变步长算法及其应用

Novel variable step size algorithm of adaptive filter and its application

针对变步长算法的选取对自适应滤波器的性能有重大的影响,而常见变步长算法的步长与误差或输入信号有关,在自适应调节过程中步长存在较大的波动,影响自适应滤波器的滤波效果等问题,通过分析和比较常见的变步长算法,提出一种新型变步长算法。该算法先将误差信号取绝对值,然后再求其均值,以该均值决定步长的变化,克服步长波动的不足。采用新型变步长算法的自适应滤波器,不仅具有较快的收敛速度,而且可以获取较小的均方误差。最后,将基于新型变步长算法的自适应滤波器应用于笼型异步电动机的断条故障诊断中。实验证明,新型变步长算法可以增强断条故障的特征,有利于信号特征的提取。

The variable step size algorithm affects the performance of the adaptive filter heavily.The step size of the usual variable step size algorithms has a certain relation with the error signal and the input signal.In the adaptive regulation process,the step size has a big fluctuation.This property severely affects the filtering effect of the adaptive filter.To solve these problems,a novel variable step size algorithm（NVSSA） was put forward by analyzing and comparing the usual variable step size algorithms.In the NVSSA,the absolute value of error signal was taken,and then its mean was found.The mean decided the change of the step size.The NVSSA can overcome the shortage of the step size fluctuation.The adaptive filter based on the NVSSA has a faster convergence rate and a less mean square error.In the end,the adaptive filter based on the NVSSA was used in the fault diagnosis of the broken rotor bar of the squirrel cage induction motor.Experiments show that the NVSSA can highlight the feature of the broken rotor bar fault and is propitious to the signal feature exaction.