Computer aided tools for process design space identification and flexibility quantification
File(s)
Author(s)
Sachio, Steven
Type
Thesis
Abstract
Classical process development approaches often centre themselves on the optimisation of economic Key Performance Indicators. Such approaches lead to a design that is constrained around a single, cost-optimal operating point and therefore is inherently less flexible and prone to uncertainties. This lack of flexibility poses a significant challenge for the design of chemical processes across various fields as uncertainty is commonly present in many forms (e.g., composition disturbances, flowrate variations). To design flexible processes that can operate under uncertainty, a systematic approach for examining process flexibility is required.
The identification of design spaces is a promising approach towards examining process flexibility. A design space is defined as a region of candidate operating conditions that ensures satisfaction of target specifications/constraints. Quantifying the size of this region results in a measure of process flexibility. Hence, the development of computer aided tools for the identification of design spaces to quantify process flexibility becomes eminent. Such tools allow for the: (i) quantification of flexibility-performance trade-offs for informed process design and (ii) reduction of experimentation effort needed by examining extreme conditions near the design space boundary. This thesis proposes two new frameworks towards the identification of low- and high-dimensional design spaces. The low-dimensional method relies on quasi-random sampling and alpha shapes. On the other hand, the high-dimensional method involves a novel multi-parametric programming problem formulation.
The frameworks are benchmarked on problems available from literature to evaluate the accuracy of the design spaces and computational cost required. To exhibit the transferability of the frameworks, two industrial problems are considered in this thesis: biopharmaceuticals and carbon capture. The frameworks proposed successfully identify design spaces of the problems and provide valuable insight for the design of well-performing flexible processes.
The identification of design spaces is a promising approach towards examining process flexibility. A design space is defined as a region of candidate operating conditions that ensures satisfaction of target specifications/constraints. Quantifying the size of this region results in a measure of process flexibility. Hence, the development of computer aided tools for the identification of design spaces to quantify process flexibility becomes eminent. Such tools allow for the: (i) quantification of flexibility-performance trade-offs for informed process design and (ii) reduction of experimentation effort needed by examining extreme conditions near the design space boundary. This thesis proposes two new frameworks towards the identification of low- and high-dimensional design spaces. The low-dimensional method relies on quasi-random sampling and alpha shapes. On the other hand, the high-dimensional method involves a novel multi-parametric programming problem formulation.
The frameworks are benchmarked on problems available from literature to evaluate the accuracy of the design spaces and computational cost required. To exhibit the transferability of the frameworks, two industrial problems are considered in this thesis: biopharmaceuticals and carbon capture. The frameworks proposed successfully identify design spaces of the problems and provide valuable insight for the design of well-performing flexible processes.
Version
Open Access
Date Issued
2024-09-02
Date Awarded
01/11/2024
License URL
Advisor
Papathanasiou, Maria M.
Kontoravdi, Cleo
Publisher Department
Chemical Engineering
Publisher Institution
Imperial College London
Qualification Level
Doctoral
Qualification Name
Doctor of Philosophy (PhD)