부산대학교 도서관

This paper compares two approaches to tune PI controllers in a multivariable process. A modified Direct Synthesis method extends the technique to a multivariable system. Results show that the modified Direct Synthesis technique relies on model accuracy and suitability of the desired transfer function and process model. The effectiveness of this technique may break down when these conditions are not met. Genetic algorithms provide a method based on the evolution of species via natural selection. The GA method allows the use of a specific structure, such as PI controller, and then gets the best possible results for that structure. The concept has been adapted into an optimisation technique that can be used to find the best controller parameters to meet user defined transient specifications for a complex process. Thus, the robustness of the method allows it to outperform the former technique. Simulation results are shown to illustrate the comparisons between these two methods and the effectiveness of GAs.