ADAPTIVE CONTROL OF CHEMICAL REACTOR USING NEURAL NETWORKS

ffg areed, HANY ABDELFATTAH

Abstract


Control strategies based on nonlinear

process       models      can     provide      the

potential for significant improvement

over 1 inear c ontroliers f or n onlinear

processes.         An        adaptive        neural

network is applied to a rauitivariable

chemical reactor. The first stage of

tlie project, simulation study, has

been investigated and is presented in

tliis paper. A radial basis function

network is developed to model the

real process, and its weights are on¬

line updated using a self organizing-

map (kohonen algorithm). Design of

a         Proportional-Integral-Derivative

(PID) linear controller for a chemical

process      is presented.       Comparison

between PID controller and a neural

network-based       kohonen       algorithni

controller       are      illustrated       by the

simulation results. Results �howed

tlie proposed technique controller that

.induce a linear response, in input-

output sense and that the nonlinear

controller can be easily tuned.


Refbacks

  • There are currently no refbacks.