NEURAL NETWORK CONTROL FOR HEAT EXCHANGER

This and individual Final Year Research Project entitle 'Neural Network Control in
Heat Exchanger' which carries four credit hours.
The main objective of this project is to develop the Neural Network system from the
specified industrial scale heat exchanger and implementing it in a simple feedback
control system. The four input parameters involved for the heat exchanger which are
the temperatures and flow rate for both shell and tube. There are only two outlet
parameters studied here in this project, the shell outlet temperature and the tube
outlet temperature. There are 3 main stages of this project. The data was taken from
PETRONAS Penapsian Melaka, Sdn. Bhd. First with the available data, the network
was designed and tested for its performance. Then, the dynamic model of the heat
exchanger was designed using ARX method. Finally, both the network and the
dynamic model are integrated in a simple Internal Model Control System.
This report will highlight the data analysis used on the data obtained for this project.
This report will also consist of the Neural Network architecture used for the heat
exchanger as well as its performance data. The dynamic model of the heat
exchanger also was constructed based on ARX Method. Full detail on the ARX
method will be explained in the report. With the dynamic model and the Neural
Network, a simple feedback system based on the concept of IMC system was
designed and studied. This project will be an extension to the existing controlling
strategies used in the industries. This will be the new age in heat exchanger control
system which will eventually propagate to other applications in the production
areas.