基于GADF-TL-ResNeXt的滚动轴承故障诊断方法

Based on the GADF-TL-ResNeXt Rolling Bearing Fault Diagnosis Method

针对传统诊断方法难以有效提取故障特征的问题,提出了一种基于格拉姆角场(GAF)与TL-ResNeXt相结合的故障诊断方法。首先利用GAF对原始振动信号编码为时间序列相关的二维特征图;再将这些特征图输入到层级更深的分组残差网络ResNeXt中进行自动的识别、分类;模型训练的同时,在网络的最后一层结合了迁移学习(TL)模块以加快模型特征提取能力、快速的进行学习。为了验证该方法的有效性,利用凯斯西储大学轴承数据对比了其他方法,结果表明该方法表现最优。且在轧机模拟实验平台上收集的轴承故障数据表明,该方法在改变工况时同样具有好的泛化性与识别能力。

To solve the problem that traditional diagnosis methods are difficult to extract fault features effectively,a fault diagnosis method based on Gramian angular field(GAF)and TL-ResNeXt is proposed.Firstly,GAF is used to encode the original vibration signal into a two-dimensional feature map of time series correlation.Then these feature maps are input into a deeper level of packet residual network ResNeXt for automatic recognition and classification.At the same time of model training,transfer learning(TL)module is combined in the last layer of the network to accelerate the feature extraction ability of the model and fast learning.In order to verify the effectiveness of the method,the bearing data of Case Western Reserve University are compared with other methods,and the results show that the method performed best.The bearing fault data collect on the rolling mill simulation test platform show that the method also has good generalization and recognition ability under different working conditions.