Theoretical and Experimental Results on an Efficient Method for the Design of More Effective Dynamic Decouplers and Feedforward Compensators

In taking the classical approach to the design of dynamic decouplers and feedforward compensators, it is well known that the major obstacle is that of realization: the inversion back, into the time domain, of the usually very complicated inverses of transfer function matrices. The recent External System Restructuring (ESR) technique (Ogunnaike, 1986) efficiently achieves the same decoupling (and feedforward compensation) objectives while completely circumventing the realization problem. Based on a judicious restructuring of the control variable vector in the multivariable system´s state space representation, the ESR structure is very simple to design and is by far more straightforward to implement than its classical counterparts. In this paper, the theoretical results are first summarized; then we present results from real time applications on an inherently non-linear, and severely interacting experimental system. The ease of design and implementation coupled with the rather impressive performance of this technique on the experimental system seem to indicate a real potential for effective industrial process control applications.