多项式核植入特征分布适配的滚动轴承故障迁移诊断方法,首先获得源域滚动轴承含健康标记数据集与目标域滚动轴承的监测数据集,输入到深度残差网络后,逐层提取源域与目标域迁移故障特征;通过多项式核植入特征适配最小化分布差异;将目标域故障特征通过Softmax分类器,获得目标域样本特定健康状态的概率分布,然后将概率分布转换为目标域样本的伪标记;通过获得的分布差异与目标域伪标记训练迁移诊断模型后,将目标域轴承的监测数据输入训练完成的诊断模型,输出数据样本对应的标签概率分布,则最大概率所对应的样本标签即为滚动轴承的健康状态;本发明提高迁移诊断模型的性能和训练效率,降低调参难度。
The fault migration diagnosis method of rolling bearing with polynomial implanted feature distribution adaptation includes the following steps: firstly, obtaining the data set of source domain rolling bearing containing health mark and the monitoring data set of target domain rolling bearing, and then inputting them into depth residual network, and extracting the migration fault features of source domain and target domain layer by layer; Minimize the distribution difference through polynomial kernel embedding feature adaptation; Pass the fault features of the target domain through the Softmax classifier to obtain the probability distribution of the specific health state of the target domain sample, and then convert the probability distribution into the pseudo-mark of the target domain sample; After the obtained distribution difference and the false mark of the target domain are used to train the migration diagnosis model, the monitoring data of the target domain bearing is input into the trained diagnosis model, and the label probability distribution corresponding to the data sample is output, so that the sample label corresponding to the maximum probability is the healthy state of the rolling bearing; The method improves the performance and training efficiency of the migration diagnosis model, and reduces the difficulty of parameter adjustment.