A Study on Determination of the Variance Faults of a Process Using Soft Computational Approaches

Because it is able to significantly improve the process, the research issue of determination of process faults has attracted considerable attention. Although some statistical decomposition methods may provide the possible solutions, the mathematical difficulty could confine the applications. As a consequence, this study proposes the soft computing approaches to determine the source of a process fault. In this study, we apply artificial neural network (ANN), support vector machine (SVM) and multivariate adaptive regression splines (MARS) to identify the faults of a multivariate process. The multivariate process is considered to have five quality characteristics and the variance shifts are presented either on 2, 3, 4 or 5 quality characteristics. A series of computer simulations are performed to evaluate the effectiveness of the proposed approaches.