基于信息融合与CNN的齿轮箱故障诊断方法

齿轮箱在实际生产中面临复杂多变的工况,其部件的故障特征随工况发生改变,常规方法在变工况下难以有效识别故障。针对该问题,提出一种基于信息融合和卷积神经网络(IFCNN)的故障诊断方法。IFCNN使用多传感器信息融合和多域特征融合改进卷积神经网络(CNN),首先将不同位置的加速度传感器采集到的振动信号转换成频域、时频域信息,将来自不同传感器的信息融合,然后用CNN对故障信号的频域、时频域信息分别进行特征提取和多域特征融合,结合注意力机制选择重要特征进行故障分类。多组实验结果表明,IFCNN在变工况场景下,可有效提取齿轮箱振动信号的故障特征,12组变工况实验平均识别准确率为98.38%,明显高于所提出的对比方法。

Gearbox Fault Diagnosis Method Based on Information Fusion and CNN

The gearbox is faced with complex and changeable working conditions in actual production,and the fault characteristics of its components change with the working conditions.The conventional methods are difficult to effectively identify faults under variable working conditions.To solve this problem,a fault diagnosis method based on information fusion and convolution neural network (IFCNN) is proposed.IFCNN uses multi-sensor information fusion and multi-domain feature fusion to improve the convolutional neural network(CNN).First,the vibration signals collected by acceleration sensors at different locations are converted into frequency domain and time-frequency domain information,and the information from different sensors is fused.Secondly,CNN is used to extract the frequency domain and time-frequency domain information of the fault signal and fuse the multi domain features respectively.Combined with the attention mechanism,the important features are selected for fault classification.Multiple sets of experimental results show that IFCNN can effectively extract the fault characteristics of gearbox vibration signals in variable working conditions.The average recognition accuracy of 12 sets of variable working conditions experiments is 98.38%,which is significantly higher than the comparison method.