On-line learning Fuzzy Relational Model based dynamic matrix control of anopenloop unstable process

Fuzzy Relational Models (FRM) are used when implementing a dynamic matrix c ontrol (DMC) algorithm on a nonlinear process which exhibits multiplicity o f steady state solutions. Following Ozkan(1), Ozkan and Camurdan(2), the op enloop unstable equilibrium point of the process is first stabilized by an analog proportional only controller. DMC is then used to control this stabi lized system. Both servo and regulatory control are investigated. Locally a nd globally obtained fixed and globally obtained variable FRMs are used. In some runs the recursive least squares (RLS) algorithm is also used to upda te the relational array so as to provide on-line learning. It is shown by s imulations that FRMs that are identified in a local region performed as wel l as those that are identified in the whole operating region. On the other hand the performance of the system was found to be poor if a variable model is used.