Machinery Fault Diagnosis Based on Improved Algorithm of Support Vector Domain Description and SVMs

In order to improve accuracy of fault diagnosis based on SVMs, an improved algorithm of support vector domain description (ISVDD) is proposed, used to pretreat the fault data. ISVDD constructs the recognizer of fault data by introducing an optimal sphere instead of the minimum sphere. The recognizer can sift out the fault data belonging to new unknown fault types and avoid erroneous diagnosis. A new method of fault diagnosis is given based on ISVDD and hierarchy structure SVMs for the multi-fault problem. Numerical experiments are performed on a real dataset. The results show that ISVDD can be used to pretreat the fault data effectively and that the new method of fault diagnosis has higher precision and can be used in practice.