L.M. Waghmare, S.C. Saxena, and V. Kumar (India)
Inferential control, Model reference adaptive control, Neural
control, Process control, CST Process.
This paper deals with model reference adaptive
interferential controller (MRAC) using neural network,
which has been proposed for the temperature control of CST
process. The controller learns continuously, even while
operating for control action so that the changes in the system
are immediately reflected in control signal, and there is no
need of explicit learning separately for dynamic adaptation.
In this work, the feed-forward neural network has been used
for the forward modeling of the plant. The network is trained
using identification error that is the error between the plant
output and output of the neural network model. The trained
network parameters and tracking error have been used to
construct the control law.
The performance of the controller has been
evaluated on the experimental setup of a continuously stirrer
tank (CST) process. In the CST process, the controller has
been used to control the temperature of water in the kettle by
controlling the flow of coolant flowing in the jacket. The
robust ness of the system has been confirmed for the set
point tracking and also for under the influence of
disturbances. The performance has been compared interms
of integral square error (ISE) with the direct model reference
neural adaptive controller.