基于权重自适应特征融合的轴承故障诊断方法

Bearing Fault Diagnosis Method Based on Adaptive Weight Feature Fusion

传统的轴承故障诊断方法大多直接从原始振动信号中降维,仅利用时域特征诊断故障,存在故障特征单一的问题。针对上述问题,提出基于权重自适应特征融合的轴承故障诊断方法。首先对原始故障信号进行时频分析,得到故障的频域和时频域信息,然后建立双通道特征提取网络模型,分别对频域特征和时频域特征进行提取,最后提出特征自适应加权算法对不同维度特征动态加权,实现特征加权融合诊断。采用凯斯西储大学和帕德博恩大学轴承故障数据集进行实验,在测试集上最佳准确率为99.75%和98.57%,且能保持较好的收敛速度,有效提高了轴承的故障诊断效率。

Most of the traditional bearing fault diagnosis methods directly reduce the dimension from the original vibration signal,and use the time-domain characteristics to diagnose the fault,which has the problem of single fault characteristics.Aiming at the above problems,a bearing fault diagnosis method based on adaptive weight feature fusion was proposed.Firstly,the time-frequency analysis of the original fault signal was carried out to obtain the frequency-domain information and time-frequency-domain information of the fault.Then,a dual channel feature extraction network model was established to extract the frequency-domain features and time-frequency-domain features respectively.Finally,a feature adaptive weighting algorithm was proposed to weight the features of different dimensions to realize the feature weighted fusion diagnosis.Experiments were carried out on the bearing fault data sets of Case Western Reserve University and Paderborn University.The best accuracies were 99.75%and 98.57%on the test sets,maintaining a good convergence speed,and the efficiency of bearing fault diagnosis was effectively improved.