Title :
Hydraulic System Faults Diagnosis Based on Multi-class Support Vector Machine
Author :
Sheng, Li ; Peilin, Zhang ; Guode, Wang
Author_Institution :
Dept. 1st, Ordnance Eng. Coll., Shijiazhuang, China
Abstract :
There is no sufficient evidence on classification, because of lack of hydraulic system fault samples. The classification results with definite guess are not exactly right. Meanwhile, there are many types of hydraulic system faults, but present classifiers can only classify two-class problems, which are not fit for hydraulic system faults diagnosis. In order to solve the preceding problems, a method for hydraulic system faults diagnosis based on multi-class support vector machine (MSVM) is proposed. A support vector machine (SVM) has strong classification ability with fewer samples taker. For k -class problem of hydraulic system, it combines k(k-1)/2 two-class SVM classifiers, one for each pair of classes. The experimental results indicate that this method is a more effective and feasible tool for hydraulic system faults diagnosis than Neural Net.
Keywords :
fault diagnosis; hydraulic systems; mechanical engineering computing; support vector machines; hydraulic system faults diagnosis; k-class problem; multiclass support vector machine; two-class SVM classifiers; fault diagnosis; hydraulic system; support vector machine;
Conference_Titel :
Digital Manufacturing and Automation (ICDMA), 2010 International Conference on
Conference_Location :
ChangSha
Print_ISBN :
978-0-7695-4286-7
DOI :
10.1109/ICDMA.2010.126