Application of Support Vector Machine to Reliability Analysis of Engine Systems

Abstract

Reliability analysis plays a very important role for assessing the performance and making maintenance plans of engine systems. This research presents a comparative study of the predictive performances of support vector machines (SVM) , least square support vector machine (LSSVM) and neural network time series models for forecasting failures and reliability in engine systems. Further, the reliability indexes of engine systems are computed by the weibull probability paper programmed with Matlab. The results shows that the probability distribution of the forecasting outcomes is consistent to the distribution of the actual data, which all follow weibull distribution and the predictions by SVM and LSSVM can provide accurate predictions of the characteristic life. So SVM and LSSVM are both another choice of engine system reliability analysis. Moreover, the predictive precise of the method based on LSSVM is higher than that of SVM. In small samples, the prediction by LSSVM will be more popular, because its compution cost is lower and the precise can be more satisfied.

DOI: http://dx.doi.org/10.11591/telkomnika.v11i7.2624