Zhen Guo , Wenliao Du , Chuan Li , Yannan Yu , Tao Hu , Shuai Wang , Zhiping Liu
{"title":"基于多尺度小波分解和特征融合的旋转机械多层次类不平衡故障诊断","authors":"Zhen Guo , Wenliao Du , Chuan Li , Yannan Yu , Tao Hu , Shuai Wang , Zhiping Liu","doi":"10.1016/j.ymssp.2025.113427","DOIUrl":null,"url":null,"abstract":"<div><div>As the core equipment of industrial systems, rotating machinery, the health status monitoring of which is of vital importance to ensuring production safety and efficiency. However, in actual working conditions, fault samples are scarce and unevenly distributed, resulting in a significant decline in the ability of traditional diagnostic models to identify a few types of faults. To address this issue, this paper proposes a fault diagnosis model based on multi-scale wavelet decomposition and a feature fusion network (MWDFN). Firstly, the low-frequency approximate coefficients of the signal are extracted through the three-level discrete wavelet transform (DWT) to obtain the statistical characteristics. Based on this, the wavelet packet transform (WPT) conducts a secondary decomposition on the statistically significant high-frequency coefficients, generating multi-scale energy-entropy joint features. Secondly, low-dimensional fusion feature vectors are constructed through feature stitching and standardized processing. Finally, a hierarchical adaptive sampling strategy is adopted, combined with the class weights of the random forest (RF) classifier, to adjust and optimize the classification boundaries. The experimental results show that the average accuracy rate on the four datasets reached 88.84 %. The source code of MWDFN is available at <span><span>http://github.com/MR-ach/MWDFN</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":51124,"journal":{"name":"Mechanical Systems and Signal Processing","volume":"240 ","pages":"Article 113427"},"PeriodicalIF":8.9000,"publicationDate":"2025-09-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Multi-scale wavelet decomposition and feature fusion for rotating machinery fault diagnosis under multi-level class imbalance\",\"authors\":\"Zhen Guo , Wenliao Du , Chuan Li , Yannan Yu , Tao Hu , Shuai Wang , Zhiping Liu\",\"doi\":\"10.1016/j.ymssp.2025.113427\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"<div><div>As the core equipment of industrial systems, rotating machinery, the health status monitoring of which is of vital importance to ensuring production safety and efficiency. However, in actual working conditions, fault samples are scarce and unevenly distributed, resulting in a significant decline in the ability of traditional diagnostic models to identify a few types of faults. To address this issue, this paper proposes a fault diagnosis model based on multi-scale wavelet decomposition and a feature fusion network (MWDFN). Firstly, the low-frequency approximate coefficients of the signal are extracted through the three-level discrete wavelet transform (DWT) to obtain the statistical characteristics. Based on this, the wavelet packet transform (WPT) conducts a secondary decomposition on the statistically significant high-frequency coefficients, generating multi-scale energy-entropy joint features. Secondly, low-dimensional fusion feature vectors are constructed through feature stitching and standardized processing. Finally, a hierarchical adaptive sampling strategy is adopted, combined with the class weights of the random forest (RF) classifier, to adjust and optimize the classification boundaries. The experimental results show that the average accuracy rate on the four datasets reached 88.84 %. The source code of MWDFN is available at <span><span>http://github.com/MR-ach/MWDFN</span><svg><path></path></svg></span>.</div></div>\",\"PeriodicalId\":51124,\"journal\":{\"name\":\"Mechanical Systems and Signal Processing\",\"volume\":\"240 \",\"pages\":\"Article 113427\"},\"PeriodicalIF\":8.9000,\"publicationDate\":\"2025-09-29\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Mechanical Systems and Signal Processing\",\"FirstCategoryId\":\"5\",\"ListUrlMain\":\"https://www.sciencedirect.com/science/article/pii/S0888327025011288\",\"RegionNum\":1,\"RegionCategory\":\"工程技术\",\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q1\",\"JCRName\":\"ENGINEERING, MECHANICAL\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Mechanical Systems and Signal Processing","FirstCategoryId":"5","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S0888327025011288","RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"ENGINEERING, MECHANICAL","Score":null,"Total":0}
Multi-scale wavelet decomposition and feature fusion for rotating machinery fault diagnosis under multi-level class imbalance
As the core equipment of industrial systems, rotating machinery, the health status monitoring of which is of vital importance to ensuring production safety and efficiency. However, in actual working conditions, fault samples are scarce and unevenly distributed, resulting in a significant decline in the ability of traditional diagnostic models to identify a few types of faults. To address this issue, this paper proposes a fault diagnosis model based on multi-scale wavelet decomposition and a feature fusion network (MWDFN). Firstly, the low-frequency approximate coefficients of the signal are extracted through the three-level discrete wavelet transform (DWT) to obtain the statistical characteristics. Based on this, the wavelet packet transform (WPT) conducts a secondary decomposition on the statistically significant high-frequency coefficients, generating multi-scale energy-entropy joint features. Secondly, low-dimensional fusion feature vectors are constructed through feature stitching and standardized processing. Finally, a hierarchical adaptive sampling strategy is adopted, combined with the class weights of the random forest (RF) classifier, to adjust and optimize the classification boundaries. The experimental results show that the average accuracy rate on the four datasets reached 88.84 %. The source code of MWDFN is available at http://github.com/MR-ach/MWDFN.
期刊介绍:
Journal Name: Mechanical Systems and Signal Processing (MSSP)
Interdisciplinary Focus:
Mechanical, Aerospace, and Civil Engineering
Purpose:Reporting scientific advancements of the highest quality
Arising from new techniques in sensing, instrumentation, signal processing, modelling, and control of dynamic systems