Inverse artificial neural network control design for a double tube heat exchanger

Abstract

This work focused on a new developed and experimentally tested non-linear control approach to control a double tube heat exchanger's output cold water temperature. The control approach uses an inverse artificial neural network (ANNi) and an integral control law. The ANNi objective is to obtain the cold water flow considering as a reference the desidered output cold water temperature. The ANNi was built from a feedforward ANN with four inputs (input cold water temperature, input hot water temperature, cold water flow, hot water flow), a hidden layer with a neuron, and two outputs (output cold water and output hot water temperature). The integral control law is implemented to reduce the error between the setpoint and the ANNi estimate. As a result, the output cold water temperature regulation is achieved through a simple controller that does not need a complex ANN model and does not depend on the process parameters' exact value. The proposed control scheme (Inverse Artificial Neural Network with Integral Control (ICANNi)) showed good reference tracking, obtaining a root mean square error (RMSE) of 0.2025, standard deviation (SD) of 0.0410, establishment time of 23 s, and overshoots less than 2.7 °C. In addition, the ICANNi control has a settling time that is 1.91 times faster than the PID control and 1.5 times faster than the ANNi control.