基于深度特征提取神经网络的滚动轴承故障诊断

滚动轴承作为旋转机械的重要组成部分，其运行安全性受到大量关注，但传统的基于信号处理的时频分析故障诊断方法较为依赖专家知识从而难以广泛应用。结合应用较广的卷积神经网络和长短时记忆网络模型的优点-自动提取振动信号的深层特征信息以及可识别所提取的长时连续的振动信号时序特征信息，提出一种深度特征提取神经网络模型，将原始的振动信号作为模型输入，进而通过多层卷积与长短时记忆网络对振动信号进行故障特征信息提取，可以有效提取滚动轴承振动信号中的深层时序故障特征信息，进而准确辨识滚动轴承不同的故障模式，并且避免了复杂的信号预处理与人工进行信号特征提取的过程。通过凯斯西储大学滚动轴承故障实验的10类健康状态数据验证了所提方法的有效性，并对实验结果进行分析，解释了在迭代过程中出现精度波动的可能原因。

Fault diagnosis of rolling bearing based on depth feature extraction neural network

Rolling bearings are an important component of rotating machinery, and their operational safety has attracted widespread attention. Traditional time-frequency analysis fault diagnosis methods based on signal processing rely on expert knowledge and are difficult to use. Combining the advantages of the widely used convolutional neural network and the long-short-term memory network model allows the deep feature information of the vibration signal to be extracted and the time sequence feature information of the extracted long-term continuous vibration signal to be recognized automatically. In this paper, we propose a deep feature extraction neural network model, which takes the original vibration signal as the model input, and then extracts the fault characteristic information of the vibration signal through a multi-layer convolution and long-short-term memory network. This can effectively extract the characteristic deep time sequence fault information in the raw rolling bearing vibration signal, and can accurately identify the different failure modes of rolling bearings. This method does not require complicated signal preprocessing or manual signal feature extraction processes. The effectiveness of the proposed method has been verified using ten types of data sets from the Case Western Reserve University rolling bearing failure center. In addition, the experimental results have been analyzed in order to explain the possible reasons for the accuracy fluctuations in the iterative process.