Impact of Intelligent Controller in a Multiprocess System using Artificial Neural Network - BPN

This paper presents a new design perception for slave controller in Cascade process control System. It ensures enhanced robustness by minimizing the physical presence of slave controller in the cascaded loops. The desired outputs are trained by ANN using BPN. A Level and flow measurement system is taken as an experimental set up. The level controller is the master controller and the flow is the slave controller in a conventional PID controller. Using ANN, we minimize the presence of the flow controller. ANN with three inputs and one output with single hidden layer with five hidden neurons are used as a slave controller. The training set attained from the knowledge of the conventional PID controller is used to train the network. The performance of the process is evaluated MSE criterion. The output response of the cascade process control system using suggested control methods is compared with the traditional PID controller. Error versus epoch analysis gives us a clear picture of convergence of output nearer to the real time values attained from the conventional PID controller. The control results clearly state that this soft slave controller shows better performance than the physical slave controller.