基于主元分析与支持向量机的制冷系统故障诊断方法

上海交通大学学报（自然版） ›› 2011, Vol. 45 ›› Issue (09): 1355-1361.

基于主元分析与支持向量机的制冷系统故障诊断方法

韩华1, 谷波1, 任能2

（1.上海交通大学制冷与低温工程研究所，上海 200240；2.江森自控楼宇设备科技（无锡）有限公司，江苏无锡 214028）

收稿日期:2010-04-12 出版日期:2011-09-30 发布日期:2011-09-30
基金资助:
国家自然科学基金资助项目（50876059）

Fault Diagnosis for Refrigeration Systems Based on Principal Component Analysis and Support Vector Machine

HAN Hua-1, GU Bo-1, REN Neng-2

(1.Institute of Refrigeration and Cryogenics Engineering, Shanghai Jiaotong University, Shanghai 200240, China； 2.Asia Engineering Center of Building Efficiency Business, Johnson Controls, Wuxi 214028, Jiangsu, China)

Received:2010-04-12 Online:2011-09-30 Published:2011-09-30

摘要/Abstract

摘要： 利用主元分析(PCA)法提取制冷系统特征向量，对典型人工智能方法所建故障诊断模型的性能进行理论研究与应用分析，确定了以支持向量机(SVM)算法为基础的故障诊断模型；针对SVM直接解决多种分类问题的困难，分析了3种多类SVM算法，确定了基于“一对其他（One vs others）”多类SVM算法的故障诊断模型，并提出基于PCA与SVM组合的PCASVM故障诊断模型，同时，利用实验数据加以验证.结果表明：PCASVM模型可将16个原始变量转化为相互独立的主元，并可提取前4个主元用于故障诊断而将正常与故障的模式分离，对故障的诊断率不低于98.57%，优于单纯SVM模型，且PCASVM模型的训练速度比SVM模型快约130~350倍；PCASVM模型对小样本的处理能力优于BP神经网络模型，其诊断正确率较高，训练耗时较少（约1/240）.

关键词: 制冷系统, 故障诊断, 主元分析, 支持向量机

Abstract: Principal component analysis (PCA) was employed to make feature extraction from the vast data pool, and the PCA+SVM (support vector machine) model was established for the fault detection and diagnosis (FDD) of refrigeration systems. Considering that SVM can not be used to solve multiclass classification problems directly, several classical multiSVMs algorithms were analyzed and compared, and the “One vs others” algorithm was adopted. The hybrid PCASVM FDD model was presented and validated by the historical data from specially designed experiments. The results show that it can isolate normal from faulty modes (detection) and has a diagnostic rate of no less than 98.57%, which is better than the SVM model without PCA; model training is about 130—350 times faster than the latter; it also has better performance on dealing with small sample problem than BP neural network with higher diagnostic rate and much less training time (about 1/240).

Key words: refrigeration system, fault diagnosis (FD), principal component analysis (PCA), support vector machine (SVM)

中图分类号:

TB66

韩华1, 谷波1, 任能2. 基于主元分析与支持向量机的制冷系统故障诊断方法[J]. 上海交通大学学报（自然版）, 2011, 45(09): 1355-1361.

HAN Hua-1, GU Bo-1, REN Neng-2. Fault Diagnosis for Refrigeration Systems Based on Principal Component Analysis and Support Vector Machine[J]. Journal of Shanghai Jiaotong University, 2011, 45(09): 1355-1361.

参考文献

［1］Qin S J, Yue H Y, Dunia R. Selfvalidating inferential sensors with application to air emission monitoring ［J］. Ind Eng Chen Res, 1997, 36(5): 16751685.

［2］邱天，丁艳军，吴占松. 基于主元分析的故障可检测性的统计指标比较［J］. 上海交通大学学报，2006, 40(8): 14471450.

QIU Tian, DING Yanjun, WU Zhansong. Sensor fault detection statistics based on principal component analysis ［J］. Journal of Shanghai Jiaotong University， 2006, 40(8): 14471450.

［3］高隽. 人工神经网络原理及仿真实例［M］. 北京: 机械工业出版社, 2003.

［4］Vapnik V N. 统计学习理论［M］. 许建华译.北京: 电子工业出版社, 2004: 293323.

［5］Choi K, Namburu S M, Azam M, et al. Fault diagnosis in HVAC chillers: Adaptability of a datadriven fault detection and isolation approach ［J］. IEEE Instrum Meas Mag, 2005, 8(3): 2432.

［6］Cherkassky V, Ma Y Q. Practical selection of SVM parameters and noise estimation for SVM regression ［J］. Neural Networks, 2004, 17(1): 113126.

［7］Lee J M, Yoo C, Choi S W, et al. Nonlinear process monitoring using kernel principle component analysis ［J］. Chemical Engineering Science, 2004, 59(1):223234.

［8］Valle S, Li W, Qin S J. Selection of the number of principal components: The variance of the reconstruction error criterion with a comparison to other methods［J］. Industrial and Engineering Chemistry Research, 1999, 38(11): 43894401.

［9］Cui J T. A robust fault detection and diagnosis strategy for centrifugal chillers ［D］. Hong Kong: Department of Building Services Engineering, Hong Kong Polytechnic University, 2005.

［10］Fletcher R. Practical methods of optimization ［M］. New York: John Wiley and Sons, 1987.

［11］Hsu C W, Lin C J. A comparison of methods for multiclass support vector machines ［J］. IEEE Transaction on Neural Networks, 2002, 13(2): 415425.

［12］Liu Y G, You Z S, Cao L P. A novel and quick SVMbased multiclass classifier ［J］. Pattern Recognition, 2006, 39(11): 22582264.

［13］Weston J, Watkins C. Support vector machines for multiclass pattern recognition ［C］// Proc ESANN’99. Brussels: Facto Press, 1999: 219224.

［14］Platt J C, Cristianini N, Shawe T J. Large margin DAGs for multiclass classification ［C］//Advances in Neural Information Processing Systems. Cambridge, MA: MIT Press, 2000: 547553.

［15］Moreira M, Mayoraz E. Improved pairwise coupling classification with correcting classifiers ［C］// Euro Conf Machine Learning. London: SpringerVerlag, 1998: 160171.

［16］Yuan S F, Chu F L. Support vector machinesbased fault diagnosis for turbopump rotor ［J］. Mechanical Systems and Signal Processing, 2006, 20(4): 939952.

［17］Lin H T, Lin C J. A practical guide to support vector classification ［EB/OL］.［20091030］.http://www.csie.ntu.edu.tw/~cjlin/papers/tanh.pdf.

[1]	许勇, 蔡云泽, 宋林. 基于数据驱动的核电设备状态评估研究综述[J]. 上海交通大学学报, 2022, 56(3): 267-278.
[2]	刘秀丽, 徐小力. 基于特征金字塔卷积循环神经网络的故障诊断方法[J]. 上海交通大学学报, 2022, 56(2): 182-190.
[3]	郭金玉, 李文涛, 李元. 在线压缩核主元分析的自适应过程监控[J]. 上海交通大学学报, 2022, 56(10): 1397-1408.
[4]	马航宇, 周笛, 卫宇杰, 吴伟, 潘尔顺. 变工况下基于自适应深度置信网络的轴承智能故障诊断[J]. 上海交通大学学报, 2022, 56(10): 1368-1377.
[5]	施裕升, 王晓科, 周宇泰, 蒋国韬, 徐天洋. 基于卡方检验与SVM的多雷达抗欺骗干扰方法[J]. 空天防御, 2022, 5(1): 108-114.
[6]	席剑辉, 姜瀚, 陈博, 傅莉. 基于PCA-ELM的红外多光谱辐射测温[J]. 上海交通大学学报, 2021, 55(7): 891-898.
[7]	肖冉, 魏子清, 翟晓强. 基于支持向量机的办公建筑逐时能耗预测[J]. 上海交通大学学报, 2021, 55(3): 331-336.
[8]	朱东, 姜萍萍, 颜国正, 王志武, 韩玎, 赵凯, 华芳芳, 姚盛健, 丁紫凡, 周泽润. 人工肛门括约肌系统便意感知重建[J]. 上海交通大学学报, 2020, 54(8): 771-777.
[9]	陶正瑞, 党嘉强, 徐锦泱, 安庆龙, 陈明, 王力, 任斐. 基于支持向量机回归的曲面零件涡流测距标定方法[J]. 上海交通大学学报, 2020, 54(7): 674-681.
[10]	徐彬彬, 洪榛, 赵磊, 俞立. 网络化倒立摆系统的偏差攻击及其检测方法[J]. 上海交通大学学报, 2020, 54(7): 697-704.
[11]	胡晓强，仲训昱，张霄力，彭侠夫，何荧. 基于支持向量机辅助的四轴陀螺两级故障诊断方法[J]. 上海交通大学学报, 2020, 54(11): 1151-1156.
[12]	王江卓,徐文聪,李建勋,贺丰收,曹兰英,缪礼锋. 基于支持向量机的雷达电子支援措施系统点迹-航迹关联算法[J]. 上海交通大学学报, 2019, 53(9): 1091-1099.
[13]	李冬辉，高峰. 基于扰动观测器的压缩式制冷系统改进Smith预估解耦控制[J]. 上海交通大学学报, 2019, 53(5): 593-599.
[14]	马志宏，贡亮，林可，毛雨晗，吴伟，刘成良. 基于稻穗几何形态模式识别的在穗籽粒数估测[J]. 上海交通大学学报（自然版）, 2019, 53(2): 239-246.
[15]	李春祥，裴杨从琪，殷潇. 基于Hermite组合核EMD-WT-LSSVM的非平稳非高斯风压预测[J]. 上海交通大学学报, 2019, 53(10): 1249-1258.