Artificial Neural Network Control of an Experimental Heat Exchanger Facility

Abstract

The artificial neural networks technique is applied to control the dynamic behavior of a fin-tube single-row compact heat exchanger. The experimental setup consists of a variable-speed wind-tunnel facility built specifically for heat exchanger analysis. Two different control methodologies were studied. The first one corresponds to adaptive control in which the weights and biases of the artificial neural network that acts as a controller are modified depending on the error obtained between the desired outlet air temperature and its measured value. Experimental results show that the stability of the system varies depending on the different ways of performing the adaptation of the controller. The second control strategy tested corresponds to internal model control. We added a filter and an integral control structure to obtain an offset-free steady state prediction. The control methodology was extensively tested and the results compared to those of conventional PID control. The results were very favorable for the neural controller.