Design of decentralized controllers by a relay technique: extensions of tuning rules

MIMO autotuning has been identified as an efficient technique for finding reasonable tuning parameters for multivariable systems when no process knowledge is available. A major drawback lies in the reliability of the tuning rule used at each step once the critical parameters have been identified with the relay tests. No technique which uses a constant detuning can guarantee both stability and reasonable performance for a wide class of processes. Different possibilities of refining the Ziegler and Nichols tuning rule (1942) have been outlined. Further relay tests are needed at each step in order to find the information necessary to use the refined tuning rules. Application to a test case, which poses serious problems to the classical autotuning, has shown the potential benefits of the developed procedure, which gives a better understanding of the characteristics of a process with unknown dynamics.