{"title":"多工况下基于域自适应的往复式压缩机故障诊断方法","authors":"Lijun Zhang, Lixiang Duatt, Xiaocui Hong, Xinyun Zhang","doi":"10.1109/ICMA52036.2021.9512625","DOIUrl":null,"url":null,"abstract":"The complex structure and changeable working conditions of reciprocating compressor lead to the strong noise interference of collected monitoring data, the poor universality of diagnosis model and so on. A fault diagnosis method of reciprocating compressor based on domain adaptation is proposed in this paper to solve the above-mentioned problems. It breaks away from the assumption of the same distribution of source domain and target domain data in the traditional artificial intelligence algorithm. In addition, it contributes a new idea to the intelligent diagnosis of reciprocating compressor equipment. Firstly, the vibration signal is decomposed and reconstructed by CEEMDAN. Besides, in combination with wavelet transform, one-dimensional signal is converted into two-dimensional time-frequency image. Finally, a MK-MMD layer is added in front of the classifier for adaptation to the source domain and target domain, so as to realize fault diagnosis of multi-working conditions for the reciprocating compressor based on ResNet50. According to the experimental results, the combination of CEEMDAN and WT can be effective in reducing the noise-induced interference, and the time-frequency image contains rich information. In addition, the ResNet50-MK-MMD method is used for fault diagnosis under multi-working condition, with the average accuracy reaching above 97%.","PeriodicalId":339025,"journal":{"name":"2021 IEEE International Conference on Mechatronics and Automation (ICMA)","volume":"2 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2021-08-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Fault Diagnosis Method of Reciprocating Compressor Based on Domain Adaptation under Multi-working Conditions\",\"authors\":\"Lijun Zhang, Lixiang Duatt, Xiaocui Hong, Xinyun Zhang\",\"doi\":\"10.1109/ICMA52036.2021.9512625\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"The complex structure and changeable working conditions of reciprocating compressor lead to the strong noise interference of collected monitoring data, the poor universality of diagnosis model and so on. A fault diagnosis method of reciprocating compressor based on domain adaptation is proposed in this paper to solve the above-mentioned problems. It breaks away from the assumption of the same distribution of source domain and target domain data in the traditional artificial intelligence algorithm. In addition, it contributes a new idea to the intelligent diagnosis of reciprocating compressor equipment. Firstly, the vibration signal is decomposed and reconstructed by CEEMDAN. Besides, in combination with wavelet transform, one-dimensional signal is converted into two-dimensional time-frequency image. Finally, a MK-MMD layer is added in front of the classifier for adaptation to the source domain and target domain, so as to realize fault diagnosis of multi-working conditions for the reciprocating compressor based on ResNet50. According to the experimental results, the combination of CEEMDAN and WT can be effective in reducing the noise-induced interference, and the time-frequency image contains rich information. In addition, the ResNet50-MK-MMD method is used for fault diagnosis under multi-working condition, with the average accuracy reaching above 97%.\",\"PeriodicalId\":339025,\"journal\":{\"name\":\"2021 IEEE International Conference on Mechatronics and Automation (ICMA)\",\"volume\":\"2 1\",\"pages\":\"0\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2021-08-08\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"2021 IEEE International Conference on Mechatronics and Automation (ICMA)\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.1109/ICMA52036.2021.9512625\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"2021 IEEE International Conference on Mechatronics and Automation (ICMA)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/ICMA52036.2021.9512625","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
Fault Diagnosis Method of Reciprocating Compressor Based on Domain Adaptation under Multi-working Conditions
The complex structure and changeable working conditions of reciprocating compressor lead to the strong noise interference of collected monitoring data, the poor universality of diagnosis model and so on. A fault diagnosis method of reciprocating compressor based on domain adaptation is proposed in this paper to solve the above-mentioned problems. It breaks away from the assumption of the same distribution of source domain and target domain data in the traditional artificial intelligence algorithm. In addition, it contributes a new idea to the intelligent diagnosis of reciprocating compressor equipment. Firstly, the vibration signal is decomposed and reconstructed by CEEMDAN. Besides, in combination with wavelet transform, one-dimensional signal is converted into two-dimensional time-frequency image. Finally, a MK-MMD layer is added in front of the classifier for adaptation to the source domain and target domain, so as to realize fault diagnosis of multi-working conditions for the reciprocating compressor based on ResNet50. According to the experimental results, the combination of CEEMDAN and WT can be effective in reducing the noise-induced interference, and the time-frequency image contains rich information. In addition, the ResNet50-MK-MMD method is used for fault diagnosis under multi-working condition, with the average accuracy reaching above 97%.