基于CS-LSTM-attention的轴承故障诊断

作为旋转机械设备的常用部件，轴承容易受到损伤而影响整个机械设备的运行，因此对其进行故障监测和诊断十分重要。轴承振动信号是一种时间序列数据，基于卷积神经网络的故障诊断模型对时序特征的提取具有局限性。针对上述问题，文中提出了一种基于卷积神经网络(CNN)、改进长短时记忆神经网络(LSTM)和注意力机制(Attention)的轴承故障诊断模型。首先，利用卷积神经网络初步提取经小波变换处理后的时频数据的特征，对数据等段均分后输入LSTM，进一步提取时序特征，再加入Attention模块对不同时刻的特征进行权重参数学习，最后结合全连接层与激活函数完成故障诊断。

Fault Diagnosis for Rolling Bearing Based on CS-LSTM-attention

As a common part of rotating mechanical equipment, bearings are easily damaged and affect the operation of the entire mechanical equipment, so it is very important to monitor and diagnose their faults. Bearing vibration signal is a time series data, and the fault diagnosis model based on convolutional neural networks has limitations on the extraction of time series features. Aiming at the above problems, a bearing fault diagnosis model based on convolutional neural networks (CNN), improved long and short-term memory neural networks (LSTM) and attention mechanism (Attention) is proposed in this paper. First, the convolutional neural networks are used to preliminarily extract the features of the time-frequency data processed by wavelet transform, and the data is equally divided into equal segments and then input into LSTM to further extract the timing features, and then add the Attention module to learn the weight parameters of the features at different times. Finally, the fault diagnosis is completed by combining the fully connected layer and the activation function.