Application of support vector machine for equipment reliability forecasting

In information age, reliability of digital manufacturing equipment has a large impact on throughput, productivity and executing predictive maintenance. Accurate reliability forecasts can provide a good assessment of machine performance in order to execute predictive maintenance effectively. This paper investigates a methodology of applying support vector machines (SVMs) to predict reliability in computerized numerical control (CNC) machine tool of digital manufacturing system. SVM is capable to solve nonlinear regression and times series problems lie on conducting the structural risk minimization principle seeking to minimize an upper bound of the generalization error rather than minimize the training error. A real reliability data (for 42 suits) of CNC machine tool were employed as the data set. SVM can be trained to learn the relationship between past historical reliability indices and the corresponding targets, and then future reliability or failures can be predicted. The experimental results demonstrate that the SVM prediction model is a valid potential for predicting system reliability and failures.