Velocity form PID controller for control and modeling of nonlinear processes using NN based approach

The paper presents an approach to model the identifiable nonlinearity of an overall nonlinear process systems online by using “Extended Minimal Resource Allocating Network” with Radial Basis Function(EMRAN-RBF) approach. An important feature of this general technique is the network flexibility to update the parameters (weights, biases) and maintaining only necessary number of neurons by a growing and pruning criterion. A digital PID control has been implemented that avoids anti-windup. The approach has been demonstrated with a case study on a CSTR involving exothermic reaction, that has to be removed to reach steady operating conditions. Both the cases of set point tracking control and disturbance rejection in feed-flow have been studied. It is concluded that for the given settings of a PID controller with this approach, system satisfies both tracking the set-point change and disturbance rejection strategies without offending the limits of other variables and control effort.