Fault detection and diagnosis of batch process using kernel local FDA

In process monitoring of batch process, Fisher discriminant analysis is a very popular method and has be widely applied. In this paper, a new kernel local Fisher discriminant analysis (KLFDA) algorithm is proposed for fault diagnosis. The main contributions of the presented approach are as follows: 1) the proposed algorithm can simultaneously extract the global European distribution of data and local manifold structure; 2) the nonlinear problem of data is solved by mapping the data into high-dimensional space; 3) the similarity analysis method is utilized to diagnose the fault type, which improves the performance of the fault diagnosis. Experimental results on Beer fermentation process verify the performance of the proposed method.