基于FRFT和LSTM的变速器齿轮早期故障诊断

Early Fault Diagnosis of Transmission Gears Based on FRFT and LSTM

针对变速器齿轮早期故障诊断中故障特征微弱、难以提取和识别的问题,提出一种基于分数阶傅里叶变换(FRFT)和长短时记忆网络(LSTM)的故障诊断模型。利用FRFT分离出故障齿轮所在挡位的啮合分量,以该分量的时间序列作为特征向量输入到LSTM网络中训练和识别。试验验证了该模型的有效性,能实现齿轮故障的识别,相比BP神经网络和支持向量机(SVM)可提高故障诊断的准确率。

In order to solve the problems that the fault features are weak and difficult to be extracted and identified in early fault diagnosis of transmission gears,a fault diagnosis model based on fractional Fourier transform(FRFT)and long short time memory network(LSTM)is proposed.FRFT is used to separate the meshing component of the gear where the fault gear is located,and the time series of the component is used as the eigenvector to input into the LSTM network for training and recognition.The experiment verified the validity of the model and realized the identification of gear faults.Compared with BP neural network and support vector machine(SVM),the accuracy of fault diagnosis is improved.