基于机器学习的异步电机故障诊断方法

Machine learning based fault diagnosis method for asynchronous motors

针对以往异步电机故障诊断中特征提取能力不足导致诊断效果较差的问题,提出了一种基于机器学习的异步电机故障诊断方法,该方法使用了注意力机制(AM)的多尺度卷积神经网络(MSCNN)-双向长短时记忆网络(BiLSTM)的异步电机故障诊断模型,通过加入通道注意力机制改进学习机制,使用3种不同尺度提取数据特征,使用BiLSTM对周期故障振动信号进行时序特征的提取,添加自注意力机制关注重点故障特征,引入残差模块减少噪声和冗余数据的影响,最后,通过Softmax分类输出诊断结果。结果表明,该模型能够有效提取数据集中的故障特征,与其他4种常见模型进行对比,体现其稳定性和高诊断性能,针对异步电机故障诊断的准确率达到98.5%。

Aiming at the problem of poor diagnostic effect due to insufficient feature extraction capability in previous asynchronous motor fault diagnosis,a machine learning-based asynchronous motor fault diagnosis method is proposed,which uses a multi-scale convolutional neural network(MSCNN)-bidirectional long and short-term memory network(BiLSTM)asynchronous motor fault diagnosis model of the attention mechanism(AM).We improve the learning mechanism by adding channel attention mechanism,and use three different scales to extract data features,use BiLSTM to extract temporal features of periodic fault vibration signals,add the self-attention mechanism to focus on the key fault features,introduce the residual module to reduce the influence of noise and redundant data,and finally,output the diagnostic results through Softmax classification.The experimental results show that the model can effectively extract the fault features in the dataset,and compared with the other four common models,reflecting its stability and high diagnostic performance,and the accuracy rate for asynchronous motor fault diagnosis reaches 98.5%.