Application of support vector machine to reliability analysis of engine systems

Reliability analysis plays a very important role in assessing product performances and making maintenance plans for maintainable production. To find effective ways of forecasting the reliability in engine systems, the paper presents a comparative study on the prediction performances using support vector machine (SVM), least square support vector machine (LSSVM) and neural network. The reliability indexes of engine systems are computed using the Weibull probability paper programmed with Matlab. Illustrative examples show that probability distributions of forecasting outcomes using different methods are consistent to the actual probability distribution. And the two methods of SVM and LSSVM can provide the accurate predictions of the characteristic life, so SVM and LSSVM are both effective prediction methods for reliability analysis in engine systems. Moreover, the predictive precision based on LSSVM is higher than that based on SVM, especially in small samples. Because of its lower computation costs and higher precision, the reliability prediction using LSSVM is more popular. © 2013 Universitas Ahmad Dahlan.