基于EMD和RBFNN的地铁辅助逆变器故障检测

Fault Detection for Subway Auxiliary Inverter Based on EMD and RBFNN

针对地铁辅助逆变器故障信号非平稳的特征,提出了一种基于经验模态分解方法和径向基神经网络的地铁辅助逆变器的故障诊断方法,并应用经验模态分解方法对采集的非平稳的原始信号进行处理,将原始信号分解成多个平稳的本征模函数(intrinsic mode function,IMF),同时,采用K-均值聚类算法确定RBF神经网络的模型参数,并借助径向基神经网络的分类能力对特征向量进行故障检测。仿真结果表明,基于K-均值聚类算法的RBF神经网络,在48个测试样本中有46个正确,准确率为95.8%,高于标准RBF神经网络77.0%的准确率,说明其准确性明显高于标准的径向基神经网络。该研究能够满足地铁辅助逆变器故障检测对准确性的要求,可高效识别地铁辅助逆变器的故障。

Focusing on the non-stationary characteristics of the fault signal of subway auxiliary inverter, this paper proposes a method, by combining empirical mode decomposition （EMD） with radical basis function （RBF） neural network to diagnose fault for metro auxiliary inverter. Empirical mode decomposition method is applied to analyze original non-stationary signal. The original signal is decomposed into several smooth intrinsic mode functions （IMF）. The k-means clustering algorithm is used to determine the param- eters of RBF neural network model. It can detect the faults of feature vector according to the classified a- bility of RBF neural network. According to the analysis results of the fault signal of subway auxiliary inverter, the accuracy of this algorithm is higher than the foundational RBF neural network. The results satisfy the requirement of fault diagnosis of subway auxiliary inverter, and identify the fault efficiently.