基于FastICA的遗传径向基神经网络轴承故障诊断研究

Research on Bearing Fault Diagnosis of Genetic Radial Basis Function Neural Network Based on FastICA

针对电机轴承故障诊断效率低和诊断结果准确率不高的问题,提出一种基于FastICA的遗传径向基神经网络的优化算法。利用独立分量分析算法,将信号分离成多个独立的信号源;根据独立信号源构建独立特征向量;将分离所得的独立信号源作为样本,输入到遗传算法优化后的径向基神经网络中进行故障识别,并与其他分类算法比较。实验结果表明,对于电机轴承多信号的故障诊断,该算法具有更好的故障诊断能力。

According to the low efficiency and low accuracy of motor bearings fault diagnosis,a genetic radial basis function neural network optimization algorithm based on FastICA was proposed.The signal was divided into multiple independent signal sources by using independent component analysis algorithm;the independent eigenvectors were constructed by using the independent signal source;the separated independent signal source was taken as a sample and input to the radial neural network optimized by genetic algorithm for fault identification,and compared with other classification algorithms.The experimental results show that this algorithm has better fault diagnosis ability for multi-signal fault diagnosis of motor bearings.