{"title":"RCSGAN:残差特征循环半生成对抗网络在极稀缺标记数据下的冷水机组故障诊断","authors":"Xuejin Gao , Zhiyuan Zhang , Huayun Han , Huihui Gao , Yongsheng Qi","doi":"10.1016/j.enbuild.2025.116085","DOIUrl":null,"url":null,"abstract":"<div><div>The fault diagnosis of chillers is of significant importance for equipment maintenance and energy saving. Semi-supervised learning based methods alleviate the model’s rigid dependence on labeled data by learning from the information contained in a large amount of unlabeled data. However, the dynamic coupling characteristics and multi-fault severity levels of chillers result in poor inter-class separability of their data in low dimensional space, making existing models struggle to extract sufficient information from unlabeled data, and their performance tends to rely heavily on the number of labeled samples. Therefore, a fault diagnosis method based on residual feature cycle semi-generative adversarial network (RCSGAN) is proposed. This method enhances the inter-class separability by extracting residual features, allowing the model to mine more effective information from unlabeled data. Additionally, a cyclic training strategy combined with a novel pseudo-label selection method is proposed to further reduce the model’s reliance on the quantity of labeled data. Experimental results on the ASHRAE Research Project 1043 (RP-1043) dataset and real datasets show that the proposed method still achieves good fault diagnosis performance even in scenarios with extremely scarce labeled data. With only one labeled sample per category, RCSGAN improves the fault diagnosis accuracy by 26.16% compared with the current advanced methods on the RP-1043 dataset.</div></div>","PeriodicalId":11641,"journal":{"name":"Energy and Buildings","volume":"345 ","pages":"Article 116085"},"PeriodicalIF":6.6000,"publicationDate":"2025-06-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"RCSGAN: Residual feature cycle semi-generative adversarial network for chillers fault diagnosis under extremely scarce labeled data\",\"authors\":\"Xuejin Gao , Zhiyuan Zhang , Huayun Han , Huihui Gao , Yongsheng Qi\",\"doi\":\"10.1016/j.enbuild.2025.116085\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"<div><div>The fault diagnosis of chillers is of significant importance for equipment maintenance and energy saving. Semi-supervised learning based methods alleviate the model’s rigid dependence on labeled data by learning from the information contained in a large amount of unlabeled data. However, the dynamic coupling characteristics and multi-fault severity levels of chillers result in poor inter-class separability of their data in low dimensional space, making existing models struggle to extract sufficient information from unlabeled data, and their performance tends to rely heavily on the number of labeled samples. Therefore, a fault diagnosis method based on residual feature cycle semi-generative adversarial network (RCSGAN) is proposed. This method enhances the inter-class separability by extracting residual features, allowing the model to mine more effective information from unlabeled data. Additionally, a cyclic training strategy combined with a novel pseudo-label selection method is proposed to further reduce the model’s reliance on the quantity of labeled data. Experimental results on the ASHRAE Research Project 1043 (RP-1043) dataset and real datasets show that the proposed method still achieves good fault diagnosis performance even in scenarios with extremely scarce labeled data. With only one labeled sample per category, RCSGAN improves the fault diagnosis accuracy by 26.16% compared with the current advanced methods on the RP-1043 dataset.</div></div>\",\"PeriodicalId\":11641,\"journal\":{\"name\":\"Energy and Buildings\",\"volume\":\"345 \",\"pages\":\"Article 116085\"},\"PeriodicalIF\":6.6000,\"publicationDate\":\"2025-06-30\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Energy and Buildings\",\"FirstCategoryId\":\"5\",\"ListUrlMain\":\"https://www.sciencedirect.com/science/article/pii/S0378778825008151\",\"RegionNum\":2,\"RegionCategory\":\"工程技术\",\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q1\",\"JCRName\":\"CONSTRUCTION & BUILDING TECHNOLOGY\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Energy and Buildings","FirstCategoryId":"5","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S0378778825008151","RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"CONSTRUCTION & BUILDING TECHNOLOGY","Score":null,"Total":0}
RCSGAN: Residual feature cycle semi-generative adversarial network for chillers fault diagnosis under extremely scarce labeled data
The fault diagnosis of chillers is of significant importance for equipment maintenance and energy saving. Semi-supervised learning based methods alleviate the model’s rigid dependence on labeled data by learning from the information contained in a large amount of unlabeled data. However, the dynamic coupling characteristics and multi-fault severity levels of chillers result in poor inter-class separability of their data in low dimensional space, making existing models struggle to extract sufficient information from unlabeled data, and their performance tends to rely heavily on the number of labeled samples. Therefore, a fault diagnosis method based on residual feature cycle semi-generative adversarial network (RCSGAN) is proposed. This method enhances the inter-class separability by extracting residual features, allowing the model to mine more effective information from unlabeled data. Additionally, a cyclic training strategy combined with a novel pseudo-label selection method is proposed to further reduce the model’s reliance on the quantity of labeled data. Experimental results on the ASHRAE Research Project 1043 (RP-1043) dataset and real datasets show that the proposed method still achieves good fault diagnosis performance even in scenarios with extremely scarce labeled data. With only one labeled sample per category, RCSGAN improves the fault diagnosis accuracy by 26.16% compared with the current advanced methods on the RP-1043 dataset.
期刊介绍:
An international journal devoted to investigations of energy use and efficiency in buildings
Energy and Buildings is an international journal publishing articles with explicit links to energy use in buildings. The aim is to present new research results, and new proven practice aimed at reducing the energy needs of a building and improving indoor environment quality.