Index of /skoge/publications/1998/larsson_miami

	Name	Last modified	Size	Description
	Parent Directory		-
	README.html	2010-07-15 10:56	1.9K
	larsson_miami.pdf	2010-07-15 10:56	119K
	larsson_miami.ps	2010-07-15 10:56	117K

Abstract.
A hierarchical control system arises when design is done in a sequential manner. Typically, a chemical plant has at least three layers; a regulatory (base) control layer, a supervisory control layer, and an optimizing layer. In this paper we will look at possibility for imposing fundamental limitations by improper design of the lower layer control. One tool that is useful for analyzing this problem is the concept of partial control. This tool is well suited because it shows how the plant will look as seen from the higher layer controller. Analysis of these transfer-functions shows that we cannot introduce new limitations, provided that we have access to the measurements of the already closed loops and can adjust the setpoints to the loops. We find that if both y2 and r2 are available at the next level, then no new fundamental limitation are introduced as we can cancel the lower layer controller by a positive feedback. However if either y2 or r2 are unavailable then it is possible to introduce new limitations for the higher layer:

• If r2 and y2 are unavailable, then it is possible to introduce new RHP-zeros in the transfer function from u1 to y1. However, the zeros are generally located far into the RHP.
• If y2 is unavailable then it is possible to introduce new RHP-zeros at the location of RHP-poles.
• If y2 is unavailable then it is possible to introduce a sensitivity to uncertainty and disturbances into the remaining plant.
• One may introduce ill-conditioning into the problem. We have shown an example where we do the opposite (the DV-configuration).