A New Process Industry Fault Diagnosis Algorithm Based on Ensemble Improved Binary-Tree SVM

Support vector machine (SVM) is an effective tool in deal with small sample, nonlinear and high dimension classification problems. In this paper, an improved pre-treatment binary-tree SVM is proposed to solve fault diagnosis. Furthermore an ensemble method is presented to establish ensemble SVM. Here the improved SVM is used as weak learning machine. The new ensemble SVM can improve the performance of single binary-tree SVM. At the end, the new algorithm is applied to fault diagnosis of blast furnace faults and the Tennessee Eastman process (TEP). The experiments results show that the improved binary-tree SVM algorithm has an excellent performance on diagnosis speed and accuracy.