小波-神经网络在故障诊断中的应用

Wavelet Neural Network for Fault Diagnosis

根据旋转机械振动信号特点,提出了小波分析和概率神经网络相结合的故障诊断方法。该诊断方法利用小波分析进行预处理-获取机械故障特征向量,概率神经网络应用该特征及对应的故障类型建立非线性映射,实现故障诊断。通过计算机仿真和试验的结果,表明该方法运算速度快、对样本噪声有较强的鲁棒形,结构简单,工程上易于实现,为旋转机械故障诊断提供了实践方法。

According to the characteristic of vibration signal in rotating machinery, a faultdiagnosis method based on wavelet probabilistic neural network is proposed, in which the wavelet packet transform is applied to the vibration analysis of rotating machinery and the feature vectors are extracted. By using the relationship between the fault models and fault features, the wavelet probabilistic neural network are capable of mapping the feature vectors to the corresponding fault models and certain fault is recognized. The result of simulation and test indicates that the method can remarkably reduce the training time. It is feasible for machinery fault diagnosis due to its simple structure and the strong robust characteristic.