电机故障诊断的仿真研究

Study on Fault Diagnosis of Electromotor Based on RBF Neural Network

研究电机故障诊断问题。针对电机信号具有非平稳和随机性特点,为保证电机运行的安全性,准确进行故障诊断,传统方法不能有效识别故障信号特征,导致故障识别准确率低,现提出一种基于小波分析和神经网络相结合的电机故障诊断方法。采用小波包变换技术对电机故障振动信号进行去噪处理,然后利用小波包分解系数计算各子频带能量值,根据能量值的变化构建故障特征向量,利用将特征向量作为RBF神经网络的输入进行故障识别,并在Matlab仿真平台上进行仿真。仿真结果表明方法提高电机故障诊断的准确率,有效克服了传统方法存在不足,同时缩短了电机故障诊断的时间。

Study electromotor fault diagnosis. Electromotors have nonstationary and random signal, Fourier trans- form cannot effectively extract fault signal features and filter the weak signals which contain rich fault information, then correct recognition rate is low. To improve the electromotor fault diagnosis accuracy, a fault diagnosis method of electromotor is proposed based on the wavelet analysis and neural network. Firstly, electromotor fault vibration signals are de-noised by wavelet packet transform technology; secondly the decomposition coefficients of wavelet packet are used to calculate the energy of every sub-band, and feature vector is built according to the change of signal energy. Finally, feature vector of fault is used for RBF neural network ＂s input, and studied 4n Matlab simulation platform. Simulation results show that this method can effectively diagnose the electromotor fault and improve the accuracy of fault diagnosis of electromotor.