PhD Thesis Defence Presentations - CHRISTOS PATILAS

In the preliminary design stage of an arbitrary process, the designer must be able to express the process characteristics through mathematical formulations, and study the dynamic behavior of such systems to fully comprehend the over-all operation under any conditions. Simultaneously, they must be able to operate the process in optimal or near-optimal conditions based on an appropriate objective function. Following this trend, researchers over the years followed different approaches and techniques in order to address the design and control problem and produce plants that operate at optimal conditions in different time scales. These formulations may utilize different methodologies based on optimization techniques or heuristics but overall they can be divided in two main categories based on the addressed problem: (a) Sequential Design and Control and (b) Integrated Design and Control. This Thesis contributes to both of these categories with a variety of formulations that are based on the concept of back-off.

For almost three decades, the back-off methodology has been extensively developed and refined for addressing the control structure selection problem. Previous work has been based on a linear and a non-linear formulation. The linear formulation ensures quick determination of the optimal solution but with a potentially significant loss of accuracy in complex non-linear processes. The non-linear formulation offers higher accuracy and the opportunity of simultaneous consideration of design and control. This improved accuracy usually comes with an increase in computational cost and complexity. A new methodology based on linear models and a quadratic approximation of economics is proposed in this work which demonstrates improved accuracy when compared with the linear approximation and reasonable increase in computational effort than the non-linear counterpart.

The importance of the simultaneous consideration of process design and control at the early design stages has been identified in the past and a lot of effort has been expended towards the development of systematic and practically applicable methodologies to address it. The complex nature of chemical processes combined with the stochastic nature of uncertainty are the main reason for the slow development in this field. This Thesis introduces a novel methodology for the simultaneous solution of the design and control problems. Use is made of the back-off methodology in order to avoid solving complex dynamic optimization problems. The method is based on the solution of a static optimization problem where the information related to process dynamics is captured in the necessary back-off from the constraints to ensure feasibility. A novel algorithmic approach is also presented in this Thesis which offers improvements and overall advantages compared to previous efforts of addressing the integrated design and control problem. This new approach utilizes the back-off concept and methodology to address dynamic optimization issues. Furthermore, this new procedure can be used in large-scale or even plant-wide applications. The practical usefulness and effectiveness of both methodologies are exhibited in several examined applications.