Dynamic optimisation and parameter estimation for biochemical process systems
Rodman, Alistair David
A vast number of contemporary production processes rely on living cells or biomolecules preforming chemical transformations as a vital step in the manufacture of valuable products. Providing sustainable supplies of food and drink relies on such biochemical processes which are essentially unchanged for millennia, and will remain quintessential for future communities and societies. Likewise many pharmaceutical products and produced from similar systems, and have formed an essential component of modern civilisation. To ensure high productivity without wasting resources (agricultural feedstock, equipment, time), it is critical to determine optimal dynamic operating profiles toward prescribing implementable control methodologies. Mathematical models have been developed for many important food and drink manufacturing processes operated using suboptimal recipes: these journal publications are quite rigorous and extensive (often describing not only composition, but also how flavour can be tuned as desired), but they frequently require consistent kinetic parameter estimation on the basis of industrial data, which is itself a dynamic optimisation problem (multi-parametric error minimisation). Dynamic optimisation of biochemical processes is of extreme technoeconomic interest and importance in industrial control practice, particularly for biochemical process systems which display steady-state and/or operating regime multiplicity, and require sizeable vectors of time-dependent concentrations and temperature-dependent kinetic parameters. Alcohol fermentation is undergoing continuous development for several millennia via concurrent advances in chemistry and chemical engineering (which greatly affected the art of bringing yeast, barley and hops together); at the same time, the biological evolution of yeast strands by natural selection as well as empirical recipes and procedures have impacted brewing even more. Ensuring high product quality is not a trivial task, particularly when strong market demand dictates process intensification. Producing food and drink in shorter times (more efficiently) with optimised processes (more cost-effectively) requires in-depth knowledge of reactive systems. The problems of consistent kinetic parameter estimation and systematic determination of optimal operating profiles to improve industrial practice are explored in this thesis for several different biochemical systems. For the first time attainable performance in beer fermentation has been exhaustively mapped under a comprehensive family of realistic time dependent temperature manipulations, providing invaluable insight to industrial brewing collaborators. This is expanded upon with the computation of optimal dynamic fermentor temperature profiles subject to a range of realistic threshold constraints on flavour degrading compounds in the product. Herein the influence of each individual by-product level on the achievable process performance can be explicitly quantified and visualised. Furthermore, the inherent trade-off in brewing process targets (batch time vs product quality) has been explored for the first time in this work, mapping the Pareto front via multi-objective dynamic optimisation. These results can be used by decision makers to better inform process decisions with significant economic implications. Following an extensive experimental campaign the first lumped parameter model and associated parameter values for the enzymatic hydrolysis of keratin waste is also proposed in this work. The model is used to formulate a dynamic optimisation problem, demonstrating that treatment of this waste can be accelerated with novel feed strategies. This work highlights the immense value in systematic and rigorous model based simulation and optimisation campaigns for biochemical process systems.