Abstract
An adaptive law and a learning method are proposed and evaluated through simulation study on controlling a PVC (Poly-Vinyl Chloride) reactor process which has highly nonlinear and time varing characteristics.
In identification of the process parameters, there are some cases where persistently excited condition is not satisfied because of constant input-output signal. As a result, the identification is not correct and the control performance degrade. On the one hand, there are some requirements Whether any methods can have learning function which the more process operation repeat, the more control effect improve.
First, in order to cope with the constant input-output signal, the proposed adaptive law prepared the forgetting factor based on restricting the trace value of adaptive gain matrix. It is obtained that the proposed adaptive law is superior to a constant trace algorithm and a variable adaptive gain algorithm.
Next, to compensate the time delay included in estimated parameters, we compose the learning method using estimated parameters in last operation. It is demonstrated that the control effect of proposed learning method attains improvements in the early temperature heating stage whenever process operation repeat by computer simulation.
