CONTROL STRATEGIES FOR CYCLIC POLYMER PROCESSES

This paper presents the simulation of a typical cyclic polymer process ing operation with respect to the performance of different control str ategies for part quality when subjected to disturbances with realistic load-profiles. Disturbances enter the simulation in a complex, probab ilistic manner creating a unique flexibility in load representation wh ich helps the simulation more accurately capture the performance of re al processing operations. Realistic simulations can provide necessary information and be less costly than benchmark studies on actual proces ses. The objective of the research is to examine the effect of differe nt control strategies on the overall control of part-to-part quality, not the control of the continuous processes that occur in part fabrica tion. The performance of six different control strategies is compared; two strategies are conventional feedback, and four strategies are sta tistically based. The statistically based algorithms effect closed loo p control, but only when a true load exists. Results are analyzed usin g statistical analysis of variance (ANOVA) via a completely randomized block experimental design, and Duncan's multiple range test is used t o rank the control strategies. The results indicate that as more distu rbances enter the system, the conventional controllers perform better than the other strategies. When fewer disturbances are present, howeve r, the statistically based controllers perform better. The most notabl e result is that one statistically based controller, the Western Elect ric runs rules controller, performs well over the entire range of dist urbances.