Novel Optimisation Framework for Process Synthesis, Design and Intensification Using Rigorous Models

Abstract

Facing the challenges of global warming, the appearance of new resources, the emergence of technologies, and the transitions of production modes, more efficient and novel chemical processes are expected to be developed in these years. This calls for a reliable and efficient optimisation-based process design and synthesis framework. However, most of the existing works on process synthesis are based on short-cut or surrogate models, which cannot guarantee the obtained solutions are optimal or even feasible when validated by rigorous unit operation models. Although process synthesis using rigorous unit operation models is desired for the optimal design, it usually generates large-scale strongly nonlinear and nonconvex mixed-integer nonlinear programming (MINLP) problems that are difficult to solve by the existing algorithms. Furthermore, process intensification (PI) technologies, such as dividing-wall columns (DWC) have shown significant energy and capital cost saving potential, but their optimal design is still challenging. In this work, we first propose a computationally efficient MINLP model for the synthesis of reaction-separation-recycle processes. However, suboptimal or infeasible solutions may be generated when solving the model with existing algorithms. To address that issue, we proposed two homotopy continuation enhanced branch and bound algorithms. The solutions of four process synthesis problems using rigorous unit operation models show that the algorithms can always get nearly the same solutions from different initial points, while the existing algorithms cannot. On the other hand, novel feasible path algorithms using steady-state and pseudo-transient continuation simulations are proposed to provide initial points for the MINLP algorithms and to conduct the optimal design of intensified chemical processes. The optimisation of several DWC and DWC-based processes validates the convergence and efficiency of the proposed feasible path algorithms. Finally, to further improve the convergence and efficiency of the feasible path algorithm, an improved sequential quadratic programming (SQP) and an improved sequential least squares programming (SLSQP) algorithm are proposed. The case studies show that compared with existing algorithms, the improved SLSQP algorithm is the most robust, whilst the improved SQP algorithm is much more efficient in some cases.

Date of Award	31 Dec 2022
Original language	English
Awarding Institution	The University of Manchester
Supervisor	Nan Zhang (Supervisor) & Jie Li (Supervisor)