Multivariable nonlinear dynamic modeling based on DPLS and Hammerstein model and its application

Process data exhibits both nonlinear and dynamic characteristics. A multivariable nonlinear dynamic modeling method is proposed by combining dynamic partial least squares (DPLS) algorithm and Hammerstein model. This method applies Hammerstein model to the DPLS inner regression. The outer PLS algorithm with ARX inputs is used to model the dynamics of the process, and it is also used to reduce the dimensionality and to remove the collinearity. The inner Hammerstein model is used to capture the dynamics and nonlinearity. As illustration, the proposed method is implemented in the alumina production process to build the model of component concentration in sodium aluminate solution. The results show that the proposed approach is capable of modeling the complex chemical process, and much improved prediction performance is achieved over the conventional linear PLS model.