| Over the last few decades, controller performance assessment has become one of the most active research areas in process control community. Though most algorithms are based on minimum variance control (MVC) benchmark, other methods, with consideration of time varying dynamics, disturbances, model plant mismatch etc., are gaining ground as more realistic benchmarks for advanced control monitoring. This thesis focuses on the controller performance assessment under disturbance effects. Linear matrix inequalities (LMIs) and covariance analysis methods are used as main mathematical tools for solving problems.; First, the controller performance of a class of linear processes is investigated under linear time invariant (LTI) control subject to linear time varying (LTV) disturbances, abbreviated as the LTVD problem. The structured closed-loop response is introduced to formulate the performance limit problem and performance assessment problem. The problems are solved for both SISO and MIMO processes by using LMI techniques. The regular, weighted and generalized LTVD benchmarks are derived respectively with distinct objective functions which result in different control performance in dealing with different disturbances.; A more general framework based on the structured closed-loop response is proposed for performance assessment subject to a pre-specified variance/covariance upper bound constraint. Its feasibility equivalence is derived with covariance control methods, giving rise to a full or reduced order solutions accordingly. An optional optimization strategy is presented for a practical solution by minimizing the gap between the resultant structured closed-loop response and the existing one in the sense of Hinfinity norm.; A higher level performance assessment for model predictive control (MPC) applications is studied with the consideration of disturbance effects. Both variability and constraint are taken into account for economic benefit potential. They are utilized as two tuning knobs to improve economic performance. The variance performance is shown to be readily transformed to the economic performance. A systematic approach is given to evaluate the performance of existing MPC applications, which includes variance and economic performance assessment, sensitivity analysis and tuning guidelines.; Finally, a practical framework for industrial implementation is suggested. The software package developed in this thesis is plant-oriented with standard DCS interfaces and is readily applied to process industries. |
