A Fault Alarm and Diagnosis Method Based on Sensitive Parameters and Support Vector Machine

9th International Conference on Compressors and their Systems

Zhang, Jinjie^1 ; Yao, Ziyun^2 ; Lv, Zhiquan^3 ; Zhu, Qunxiong^1 ; Xu, Fengtian^4 ; Jiang, Zhinong^4

College of Information Science and Technology, Beijing University of Chemical Technology, Beijing

100029, China^1

PetroChina Beijing Gas Pipeline Co., Ltd., Beijing

100101, China^2

Aromatic Plant of PetroChina Liaoyang Petrochemical Company, Liaoyang

111003, China^3

Diagnosis and Self-Recovery Engineering Research Center, Beijing University of Chemical Technology, Beijing

100029, China^4

关键词: Automatic classification;    Diagnostic techniques;    Distance evaluation techniques;    Feature extraction and classification;    Machinery fault diagnosis;    On-line monitoring system;    Practical engineering;    Typical mechanical faults;   

Others  :  https://iopscience.iop.org/article/10.1088/1757-899X/90/1/012081/pdf
DOI  :  10.1088/1757-899X/90/1/012081

来源: IOP

【 摘 要 】

Study on the extraction of fault feature and the diagnostic technique of reciprocating compressor is one of the hot research topics in the field of reciprocating machinery fault diagnosis at present. A large number of feature extraction and classification methods have been widely applied in the related research, but the practical fault alarm and the accuracy of diagnosis have not been effectively improved. Developing feature extraction and classification methods to meet the requirements of typical fault alarm and automatic diagnosis in practical engineering is urgent task. The typical mechanical faults of reciprocating compressor are presented in the paper, and the existing data of online monitoring system is used to extract fault feature parameters within 15 types in total; the inner sensitive connection between faults and the feature parameters has been made clear by using the distance evaluation technique, also sensitive characteristic parameters of different faults have been obtained. On this basis, a method based on fault feature parameters and support vector machine (SVM) is developed, which will be applied to practical fault diagnosis. A better ability of early fault warning has been proved by the experiment and the practical fault cases. Automatic classification by using the SVM to the data of fault alarm has obtained better diagnostic accuracy.