Study on the kernel-based IFMS scheduling

This study aims to investigate the scheduling performance for the flexible manufacturing system (FIWS) based on two advanced attribute extraction methods. One is the kernel principal component analysis (KPCA) and the other is the generalized discriminant analysis (GDA). By using nonlinear mapping functions, both methods first map the attributes from the input space into higher dimensional feature space where the PCA and linear discriminant analysis (LDA) are performed to find the eigenvectors associated with the largest eigenvalues and the optimal transform matrix, respectively. The nonlinear mapping operation is done by using the kernel function which performs the inner dot product of input vectors in the input space. With such a manner, the input attributes are transformed into reduced dimensional features that have powerful discriminabilities in classifying various dispatching rules. Also, the task of the attribute selection is automatically done. Experimental results indicate that the KPCA and GDA are able to achieve better scheduling performance for a FMS under several predefined conditions such as different part ratios and part routes.