基于深度学习的轴承故障辅助诊断

Bearing Fault Diagnosis Based on Deep Learning

轴承是机械设备中最重要的零件之一,轴承故障的发生将影响机械工作效率,并威胁人员生命安全,因此对轴承故障进行自动化诊断具有重要意义。振动信号是典型的非线性和非平稳信号,并且包含环境噪声等干扰信号,使得故障特征难以被有效提取出来,且适应性较差。将信号处理技术与深度学习技术相结合,采用短时傅里叶变换方式,可将振动信号的一维信号转换为二维图片,并通过一种新的卷积神经网络模型进行训练,从而实现对轴承故障的自动化诊断。实验结果表明,该方法能够拟合非线性和非平稳信号,从而有效地进行轴承故障诊断,相比传统阈值法,诊断准确率提高至96.36%。

Bearings are one of the most important parts in mechanical equipment.The occurrence of bearing faults affects the efficiency of mechanical work and threatens the safety of people’s lives.Therefore,it is of great significance to automatically diagnose bearing faults.Vibration signals are typical non-linear and non-stationary signals,which usually contain multiple low-frequency harmonics,high-frequency shocks,interference noise and other multiple signal classifications,making it difficult to extract the fault features ef⁃fectively with poor adaptability.Compared with the threshold method commonly used in industry,this paper proposes a new method that uses a combination of signal processing technology and deep learning technology to apply a short-time Fourier transform to the vi⁃bration signal to convert a one-dimensional signal into a two-dimensional image,and achieve automatic diagnosis of bearing faults through a new convolutional neural network model training.Experiments show that this method can effectively fit nonlinear and non-sta⁃tionary signals,and can effectively complete the task of bearing fault diagnosis.Compared with the traditional threshold method,the accuracy of this method is improved to 96.36%.