Generalized open set fault diagnosis method based on deep adversarial migration network

Abstract

The invention discloses a generalized open set fault diagnosis method based on a deep adversarial migration network, and the method comprises the steps: collecting original vibration signals of samples with labels of a rotating machine under a certain working condition, obtaining fault samples without labels from different operation conditions, and constructing a source domain sample set and a target domain sample set; constructing a deep adversarial migration network with a double-weighting mechanism, wherein the deep adversarial migration network comprises a feature extractor, a domain discriminator, a non-adversarial domain discriminator and a multi-classification ensemble learning device; performing joint optimization training on parameters of the feature extractor and parameters of the multi-classification ensemble learning device by using the source domain data set and adopting a gradient descent method; weighted training is carried out through dual weights; and judging whether the data belongs to a new fault category or not through the calculated weight value, and outputting a final diagnosis result. According to the method, through the deep adversarial migration network, the influence of new fault categories of a source domain and a target domain on feature matching is reduced by using a double-weighting mechanism, and generalized open set new fault task diagnosis is realized.

Description

technical field [0001] The invention belongs to the field of intelligent fault diagnosis of rotating machinery, and in particular relates to a generalized open set fault diagnosis method based on a deep anti-transition network. Background technique [0002] The mechanical rotation system fault identification network based on deep adversarial transfer learning has achieved good classification results in different transfer diagnosis tasks, providing an effective solution for intelligent fault diagnosis. In practical applications, there are usually no precursors before equipment damage, which makes mechanical failures occur suddenly. There are few types of fault data collected and new fault types may appear. Therefore, the fault types contained in the target domain are usually unknown. In view of the sporadic nature of faults, it may lead to the intersection of the fault category space of the source domain and the target domain in the domain migration problem, that is, the situ...

AI Technical Summary

Problems solved by technology

However, the above method assumes that the label category of the source domain data is larger than the label category of the target domain, and assumes that the target domain category is a subcategory of the source domain category; and in the actual industrial environment, the target domain category may exist simultaneously with The same shared categories and new fault categories in the source domain, therefore, the source domain and the target domain not only have shared categories, but each has a new fault category, the above invention cannot solve this problem of diagnosing the new fault category

Images