NONLINEAR AND DIRECTION-DEPENDENT DYNAMIC PROCESS MODELING USING NEURAL NETWORKS

The paper discusses several methods of modelling complex nonlinear dyn amics using neural networks. Particular reference is made to the probl em of modelling direction-dependent relationships. A typical example o f this would be top product composition control in a distillation colu mn, where it is easier (i.e. faster) to make the product less pure tha n it is to make it more pure by an equivalent amount. Recurrent neural networks are identified as a potential method of modelling this type of relationship. The particular architecture chosen for this example i s referred to as 'semirecurrent', since only past values of the predic tions of the network are fed back to the input layer. This architectur e is successfully used to model direction-dependent relationships in b oth simulated and actual industrial process data.