Screening plant designs and control structures for uncertain systems

Abstract

Screening tools are developed which provide nonconservative estimates of the achievable performance in the presence of general structured model uncertainty. These tools allow the rational selection among plant designs, or can be applied to provide recommedations on how to modify a plant design to improve the closed loop performance. Many of these tools are applicable to the pairing or partitioning of inputs and outputs of decentralized controllers. It is shown through examples that ignoring or improperly characterizing plant/model mismatch while selecting among plant designs can lead to erroneous results.