A LS-SVM modeling approach for nonlinear distributed parameter processes

The distributed parameter system modeling from the input and output data is investigated. The spatio-temporal output of the system is measured at a finite number of spatial locations, while the input is assumed to be a finite-dimensional temporal variable. Firstly, Karhunen-Love (KL) decomposition is used for the time/space separation and the dimension reduction. Subsequently the spatio-temporal output is expanded in terms of a low dimensional Karhunen-Love spatial basis functions. Finally its temporal dynamic model is learned from the temporal coefficients by using least squares support vector machines (LS-SVM). The simulations are presented to show the effectiveness of this spatio-temporal modeling method.