Integrating KPCA with an improved evolutionary algorithm for knowledge discovery in fault diagnosis

Abstract

In this paper, a novel approach to knowledge discovery is proposed based on the integration of kernel principal component analysis (KPCA) with an improved evolutionary algorithm. KPCA is utilized to first transform the original sample space to a nonlinear feature space via the appropriate kernel function, and then perform principal component analysis (PCA). However, it remains an untouched problem to select the optimal kernel function. This paper addresses it by an improved evolutionary algorithm incorporated with Gauss mutation. The application in fault diagnosis shows that the integration of KPCA with evolutionary computation is effective and efficient to discover the optimal nonlinear feature transformation corresponding to the real-world operational data.