CIESC Journal ›› 2022, Vol. 73 ›› Issue (9): 3950-3962.DOI: 10.11949/0438-1157.20220568

• Process system engineering • Previous Articles     Next Articles

Fault diagnosis of chillers using central loss conditional generative adversarial network

Xuejin GAO1,2,3,4(), Kun CHENG1,2,3,4, Huayun HAN1,2,3,4, Huihui Gao1,2,3,4, Yongsheng QI5   

  1. 1.Faculty of Information Technology, Beijing University of Technology, Beijing 100124, China
    2.Engineering Research Center of Digital Community, Ministry of Education, Beijing 100124, China
    3.Beijing Laboratory for Urban Mass Transit, Beijing 100124, China
    4.Beijing Key Laboratory of Computational Intelligence and Intelligent System, Beijing 100124, China
    5.School of Electric Power, Inner Mongolia University of Technology, Hohhot 010051, Inner Mongolia, China
  • Received:2022-04-21 Revised:2022-06-02 Online:2022-10-09 Published:2022-09-05
  • Contact: Xuejin GAO

基于中心损失的条件生成式对抗网络的冷水机组故障诊断

高学金1,2,3,4(), 程琨1,2,3,4, 韩华云1,2,3,4, 高慧慧1,2,3,4, 齐咏生5   

  1. 1.北京工业大学信息学部,北京 100124
    2.数字社区教育部工程研究中心,北京 100124
    3.城市轨道交通北京实验室,北京 100124
    4.计算智能与智能系统北京市重点实验室,北京 100124
    5.内蒙古工业大学电力学院,内蒙古 呼和浩特 010051
  • 通讯作者: 高学金
  • 作者简介:高学金(1973—),男,博士,教授,gaoxuejin@bjut.edu.cn
  • 基金资助:
    北京市自然科学基金项目(4222041)

Abstract:

Aiming at the problem that the fault data generated by the chiller is insufficient and the quantity of normal data and fault data in the historical data set is unbalanced, which leads to the decline of the fault diagnosis accuracy. In this paper, a fault diagnosis method based on central loss conditional generative adversarial network (CLCGAN) and support vector machine (SVM) were proposed. Firstly, CLCGAN generates new fault data from a small amount of real fault data. Then, the generated fault data was mixed with the initial data set to balance the amount of normal data and fault data. Finally, the SVM model was constructed by using the balanced data set for fault diagnosis. When GAN generates chiller fault data, the dynamic center loss term was constructed and added into the objective function. Through the dynamic center loss term, the intra-class distance of various fault data generated by the chiller was reduced, so that the overlapping degree of each fault generated data was reduced and the reliability of generated data was increased. Before generating fault data, configure fault labels and input them into CLCGAN to guide the data generation process. In this way, the generated fault data can be evenly distributed among different fault types. The proposed method was validated on ASHRAE 1043-RP project data set, and the results show that the proposed method has higher fault diagnosis accuracy than other fault diagnosis methods that solve the problem of data imbalance.

Key words: chiller, fault diagnosis, generative adversarial networks, neural networks, algorithm, center loss, integrate

摘要:

针对冷水机组产生的故障数据不足,数据集中正常数据和故障数据数量不平衡,进而导致故障诊断精度下降的问题,提出一种基于中心损失的条件生成式对抗网络(central loss conditional generative adversarial network,CLCGAN)和支持向量机(support vector machine,SVM)的故障诊断方法。首先,CLCGAN利用少量真实故障数据生成新的故障数据;然后,将生成的故障数据与初始数据集混合,使正常数据与故障数据的数量达到平衡;最后,利用平衡数据集构建SVM模型进行故障诊断。在GAN生成冷水机组故障数据时,构建动态中心损失项并加入到目标函数中,利用动态的中心损失减少冷水机组生成的各种故障数据的类内距离,从而降低各个故障生成数据之间的重叠程度,增加生成数据的可靠性。在生成故障数据之前配置相应的故障标签,并输入到CLCGAN中指导数据生成过程,使生成的故障数据可以均衡地分布于各个故障类别。在ASHRAE 1043-RP数据集上对所提方法进行了验证,结果表明,相较于其他解决数据不平衡问题的故障诊断方法,所提方法具有更高的故障诊断准确率。

关键词: 冷水机组, 故障诊断, 生成式对抗网络, 神经网络, 算法, 中心损失, 集成

CLC Number: