Internal model control using neural network for shell and tube heat exchangers

In this research paper, a nonlinear internal model using neural network (NNIMC) is proposed to the shell and tube heat exchanger system. In past studies, PID controller is implemented in shell and tube heat exchanger, however it is exhibited high overshoot and long settling time. Therefore, NNIMC is introduced to improve the performances of PID controller. The manipulated variable of the controller is the flowrate of the hot fluid in the shell and the controlled variable is the outlet temperature of the cold fluid in the tubes. The addition of the neural network is to compensate time delay and ensures the offset performances. The control structure uses both a forward and an inverse neural network process model. The forward model is placed in parallel to the process model. The inverse neural model (INN) has two input which are previous flowrate and present temperature and one output which is present flowrate. After training for multiple times, one hidden layer INN model with 5 neurons is considered. The forward neural network (FNN) has two inputs which are previous flowrate and previous temperature and one output which is present temperature. After training for multiple times, one hidden layer with 7 neurons is considered. From simulation result, NNIMC outperforms PID controller as it exhibits no overshoot and less settling.