改进二阶统计量算法在风力机主轴承裂纹识别中的应用

Application on crack identification for main bearing of wind turbine according to improved second-order statistics algorithm

针对风力机主轴承裂纹故障的非平稳信号,提出一种改进二阶统计量的盲源分离算法,利用信号的非平稳性,将传感器数据分成不重叠的时间窗,用广义时滞协方差矩阵代替标准协方差矩阵,然后估计每个窗内的时滞协方差矩阵平均值来提高算法的稳健性和精确度。仿真试验选取了适当的时延系数和加权参数,证明了改进算法在信噪比低于12dB的情况下优于传统SOS算法。最后,采集了某风场带有裂纹故障的风力机主轴承振动信号,根据本文提出的改进算法剔除了噪声干扰,从观测信号中分离出裂纹冲击振动源信号,并通过信号处理结果的频率特征来识别轴承裂纹状态。

A blind source separation algorithm based on improved two-order statistics is presented for the non-stationary signals collected from the wind turbine main bearing.The sensor data is divided into the different non-overlapping time windows according to the non-stationary characteristics,using generalized delay covariance matrix instead of the standard covariance matrix,and then the delay covariance matrix of each window is estimated to improve the algorithm＇s robustness and accuracy.The proper time delay coefficient and weighted parameters are chosen through the simulation test,and it is proved that the improved algorithm is superior to the traditional SOS algorithm on the condition that SNR is below 12 dB.Finally,the vibration signals of wind turbine main bearing with crack failure are collected from a wind scene.The noise is eliminated and the crack impact vibration signals are separated from the observation signals according to the proposed improved algorithm in this paper.The bearing crack status can be indentified based on the signal processing results of the frequency characteristics.