Condition Residual Life Evaluation by Support Vector Machine

It is more and more important to predict the condition residual life according to the condition information of the equipment in order to make scientific and exact maintenance decision in modern production and defense construction. Currently the adopted models in the on-condition maintenance field are proportional hazards model (PHM) and filtering model. Both of them have complex forms and complex calculation process, the estimation of parameters is established on an amount of sample. However support vector machine (SVM) is a new machine learning method. It has the characters of simple structure, excellent learning capability and fitting in small sample. It also can transfer the problem to the square regression problem. So it can get the best resolution in the public area. SVM is extended to the application of the regression estimation of system. Therefore, SVM is adopted in the condition residual life regression. And the algorithm of the realizing this method is proposed. Finally the predicted result of the example adopted SVM show that SVM can better solve the same kind of problem.