基于深度置信网络的轴承故障识别分析与研究

Analysis and Investigationon Discrimination of Bearing Fault Based on Deep Belief Networks

轴承为诸多机械设备的重要零部件,对其故障状态的识别对于设备的稳定运行具有重要的意义。本文首先利用改进的自适应噪声完全集合经验模态分解(ICEEMDAN)与小波阈值相结合的方法去除轴承振动信号中的伪迹,然后分别提取信号的标准差、峭度、样本熵等线性和非线性特征,最后将多域特征作为输入项,利用深度置信网络(DBN)进行训练识别,建立了能够有效识别轴承故障类型的网络模型。试验结果表明:该模型对轴承故障类型识别的正确率可达97.8%。

Bearings are important components of many mechanical equipment,and the discriminationof their healthy operating status is of great significance for the stable and safe operation of theequipment.In the present work,an improved method combining Improved Complete Ensemble Empirical Mode Decomposition with Adaptive Noise(ICEEMDAN)and wavelet threshold wasused to denoise the bearing vibration signals.Then,extracting standard deviation,kurtosis,sample entropy,and other linear and non-linear features from the denoised signal.Finally,multi-domain featureswere used as input and training classificationwas carried out with Deep Belief Network(DBN),and a network model thatcan effectively discriminatethe bearing fault types wasestablished.Theexperimental results show that the accuracy of thisdiscriminant model for bearing fault type is up to 97.8%.