基于深度自编码网络的轴承故障诊断

Bearing Diagnosis based on Deep Neural Network of Auto-encoder

在故障诊断领域,通常采用信号处理技术提取特征,然后将特征输入到故障分类器中进行故障识别。对于提取特征部分,采用信号处理技术可以使故障诊断取得较好的效果,但是仍然存在不足之处:一是人为提取的特征很大程度上依靠专业的诊断知识;二是绝大多数方法都需要使用标签数据来进行故障特征分类,其中标签数据必须通过大量的实验才可以得到。提出一种基于深度编码网络的轴承故障新型智能诊断方法,可以克服上述故障诊断中存在的缺陷。为了验证该方法的有效性,利用具有不同健康状况的大量滚动轴承测量振动信号数据进行测试,实验结果表明效果良好。

In the field of fault diagnosis,signal processing technology is usually used to extract features,and then the features will be transferred to the fault classifier for fault diagnosis.For the feature extraction,using signal processing technology can cause the fault diagnosis to yield good results.But this method has some shortcomings:(1)the manual feature extraction strongly depends on professional diagnosis knowledge;(2)this method usually needs to use the tag data to extract the fault characteristics and classify them,but the tag data can be got only through a lot of experiments.In this paper,the bearing fault diagnosis based on deep neural network of auto-encoder is proposed as a new intelligent fault diagnosis method,which can overcome the defects existing in the traditional intelligent fault diagnosis methods.In order to verify the effectiveness of this method,the bearing data with a large number of signals and different health conditions is used for testing.The experimental results show that the proposed fault diagnosis method can adaptively extract the fault features and get excellent diagnostic results.