A Novel Technique for Identification Control of a Non-linear System

In this paper combination of neural network (NN) and Fast Traversal Filters (FTF) is used for system identification and control. The error signal is divided into two parts-linear and non-linear. The linear part of the error signal is input by the FTF algorithm, whereas the non-linear part is input to the NN. The minimized errors from the two and then added and it finally becomes the input to the system or plant. The proposed hybrid controller requires less number of samples for training of weights, thus making the system fast. The system under study is Box and Jerkins furnace. The output of the system is carbon dioxide concentration and input is gas flow rate. First identification is done using the proposed technique and then a controller is designed for the same. Comparative analysis is done between conventional feedforward neural networks and the proposed technique. Simulated results show the superiority of the proposed hybrid technique.