摘要：构建有效的故障诊断方法对提高矿山安全生产具有重要意义。尽管深度学习取得了相当大的成功，但设备故障数据的缺乏限制了模型的性能。 针对故障数据缺乏的问题，我们提出了一种故障诊断方法，该方法从模型的角度出发，针对实际数据与实验室数据分布的混淆，提出迁移学习策略，解决故障数据的缺乏问题。 从刮板输送机上采集实际工况数据， 并在试验台上进了故障诊断实验。 提出的方法得到对实际轴承数据集的故障诊断准确率达90%，泛化性能优于传统方法。  

TOPIC：A deep transfer learning fault diagnosis method for mine rotating machinery
ABSTRCT：An effective fault diagnosis method is of great significance to improve the safety of mine production. Despite the considerable success of the deep learning the absence of components failure data limits the performance of the model. In this paper, a fault diagnosis method is proposed to solve the problem of lacking in fault data. In the method, a transfer learning strategy is presented to confusion actual and labratory data distrbution from the perspective of model to solve fault data lacking. Fault diagnosis experiments conducted on test bed are carried out. And actual working condition data are collected from scraper converyor. The proposed method obtains % fault diagnosis accuracy in actual bearing data sets, which the generalization performance outperforms tranditional methods.

 

设备故障数据,工况数据,故障诊断