Hydraulic System Fault Diagnosis Based on Support Vector Machine Decision Tree

The algorithm of Support Vector Machine Decision Tree (SVMDT) and the sort application in hydraulic system multi-category faults diagnosis are researched. Considering the conventional algorithms of "one-versus-one" and "one-versus-rest", Support Vector Machine (SVM) and Decision Tree (DT) are combined, which uses the advantages of the two algorithms for structuring a multi-category classifier. The between-class distance of each two classes is descending sorted, in order that easily classified category is firstly separated, and a two-class SVM classifier is established in every node of DT for next classification. The experiments show that, compared with "one-versus-one" and "one-versus-rest", the algorithm of SVMDT guarantees better classification accuracy in hydraulic system faults.