Mammalian cell cultures are intrinsically heterogeneous at different scales (from the molecular to bioreactor level). The cell cycle is at the centre of capturing heterogeneity since it plays a critical role in the growth, death, productivity and product quality of mammalian cell cultures. Attempts to model the cell cycle heterogeneity are not new and have proved to be challenging both experimentally and computationally. Most current cell cycle models rely on biological variables (mass/volume/age) that are non-mechanistic and difficult to experimentally quantify to describe cell cycle transition and to capture the culture heterogeneity. In this thesis, the development of integrated modelling and experimental approaches that facilitates the study of cell cycle subpopulations in cell cultures is presented. The recently proposed closing the loop framework (Kiparissides et al., 2011) is employed to facilitate the development of cell cycle models with biological relevance and applicable to real life problems (industrial and clinical). The work herein presents a novel experimental-modelling platform whereby experimental quantification of key cell cycle metrics (cell cycle timings, cell cycle fractions, and cyclin expression determined by flow cytometry) is used to develop a cyclin and DNA distributed model for the industrially relevant GS-NS0 cell line. The cell cycle model captured the population heterogeneity, which further enables in silico studies of the complex system. It is envisaged that this modelling approach will pave the way for model-based developments of industrial cell lines and clinical studies. A second cell cycle model was developed to assist the industrially relevant selection of temperature profiles in mammalian cell cultures. The combined experimental-mathematical approach avoided unnecessary experimentation and guided the model development for the temperature selection. The model was successfully validated by predicting different temperature profile scenarios. The presented contributions assist the development of meaningful mathematical models with predictive capabilities accounting for the cell cycle heterogeneity in bioprocesses.

The most significant issue associated with the oxidative methane conversion processes is the use of pure oxygen, which is extremely expensive. By using a dense oxygen permeable membrane reactor, a possible decrease in the air separation cost can be expected due to the elimination of oxygen plants. Besides, higher reaction yields can be attained due to the selective dosing of oxygen into the reaction zone. This thesis focuses on the development and potential application of functional micro-structured catalytic ceramic hollow fibre membrane reactor (CHFMR) in oxidative methane conversion to syngas (known as partial oxidation of methane (POM)) and to ethane and ethylene (known as oxidative coupling of methane (OCM)). As the membrane reactor performance is crucially dependant on the oxygen permeation rate and good contact between oxygen and methane, two types of membrane reactor designs were proposed in this study. The first design involves the development of CHFMR that consists of two layers i.e.: an outer oxygen separation layer and an inner catalytic substrate layer, known as dual-layer catalytic hollow fibre membrane reactor (DL-CHFMR). The DL-CHFMR was fabricated via a novel single-step co-extrusion and co-sintering technique, in which the thickness and the composition of each functional layer can be controlled in order to improve reactor performance. The second design involves the development of CHFMR with an outer dense separation layer integrated with conical-shaped microchannels open at the inner surface, created via a viscous fingering induced phase inversion technique. Besides substantially reducing resistance across the membrane, the microchannels can also act as a structured substrate where catalyst can be deposited for the catalytic reaction to take place, forming a catalytic hollow fibre membrane microreactor (CHFMMR). Although the CHFMRs discussed in this study are designed particularly for POM and OCM, there are general advantages of such membrane structures and reactor designs for improving the overall reaction performance. Therefore, these reactor designs can be transferred to other important catalytic reactions.

In this work, molecular diffusion coefficients of carbon dioxide (CO2) in liquids relevant to carbon capture, utilisation and sequestration and enhanced oil recovery are reported. These parameters are necessary for the accurate and optimal design and control of such processes. Knowledge of these values is required to fully describe the migration of CO2 away from the injection wells and also for calculating the rate of absorption of CO2 into the formation fluids. However, diffusion coefficients are amongst the least studied of thermophysical properties, especially at high pressure, high temperature conditions. This work extended previous measurements where available, and produced new measurements where not, of diffusion coefficients at infinite dilution of CO2 in H2O, and several relevant brines and hydrocarbons at high temperatures (< 423 K) and high pressures (< 69 MPa). The Taylor dispersion method was used to determine diffusion coefficients for CO2 in water and selected hydrocarbons. The hydrocarbons chosen as representative of major crude oil components were n-heptane, nhexadecane, squalane, cyclohexane, and toluene. A technique based on nuclear magnetic resonance was used to measure effective diffusion coefficients of CO2 in several brines, encompassing monovalent and divalent salts, and a mixed brine. The diffusion coefficients of CO2 in water were correlated using the Stokes-Einstein equation in which the Stokes-Einstein number was assigned a value of 4 and the hydrodynamic radius was treated as a linear function of temperature. No relationship between brine salinity and the hydrodynamic radius was found. The results indicated pressure did not have an observable impact on the diffusivity in aqueous systems. The experimental uncertainty was found to be 2.3% with a coverage factor of 2 for the CO2-water system and 1.5% with a coverage factor of 2 for the CO2-brine systems. In contrast to aqueous systems, the diffusion coefficient of CO2 in hydrocarbons was found to be strongly dependent on pressure. At a given temperature the diffusion coefficient decreased by up to 50% over the pressure range investigated (1 to 69 MPa). A correlation based on the Stokes-Einstein equation and a two-parameter correlation based on the rough hard sphere theory was used to model the experimental results. The experimental uncertainty was found to be 1.5% with a coverage factor of 2.

The focus of this thesis is the development of novel methodologies for systematic identification of optimal solvents for chemical reactions. Two aspects are considered: the integrated solvent and process design using a mixed solvent, and the design of an optimal solvent using ab initio methods that do not rely on experimental data. A methodology is developed for the integrated design of a CO2-expanded solvent in a reaction process. Posing as objective function the cost of the process, for a defined production rate, an optimisation problem is formulated, with decision variables that include the organic co-solvent, the composition and the mass of the mixed solvent. Emphasis is placed on the prediction of the reaction rate, for which the solvatochromic equation combined with a preferential solvation model are used, and on solid-vapour-liquid phase equilibrium, for which the group-contribution volume translated Peng-Robinson equation of state is used. The proposed methodology is applied to the Diels-Alder reaction of anthracene and 4-phenyl-1,2,4-triazoline-3,5-dione (PTAD), and three CO2-expanded solvents are considered (acetone, acetonitrile and methanol). Acetonitrile and acetone are found to offer good performance over a range of CO2 concentrations. The importance of taking into account multiple process performance indicators, when designing gas-expanded liquids, is highlighted. As a further step toward systematic solvent design approaches that are not limited by the availability of experimental data and consider a large number of candidate solvents, an ab initio methodology is developed for the design of optimal solvents for reactions. The developed method combines quantum mechanical calculations with a computer-aided molecular design formulation. In order to limit the number of QM calculations but also retain accuracy and ensure convergence, the Kriging approach is used. Kriging is a response surface approach, which has recently attracted a lot of attention because it is an exact extrapolator with a statistical interpretation which makes it stand out from other methods. The proposed approach is used successfully to identify promising solvents for the Menschutkin reaction of phenacyl bromide and pyridine and the Cope elimination of methylamine oxide. The use of Kriging as the surrogate model is found to lead to improved solvents when compared to the simpler solvatochromic equation used in previous work.

It has long been a goal of the bioprocessing field to be able to produce proteins in a cost-effective and efficient manner. The current method of choosing high-producing mammalian cells is labour-intensive and time-consuming, representing an opportunity to employ new analytical methodologies in an effort to expedite progress. It would be advantageous to measure the intracellular concentration of key metabolites as the cells are growing. This would allow the user to have more information regarding the wellbeing and growth potential of cells. In this thesis the construction, optimisation and use of FRET biosensors for the in vivo measurement of glucose and glutamine in Chinese Hamster Ovary cells will be discussed. Experiments have been conducted in batch and fed-batch cultures as well as small-scale investigations using the BioLector™. The work presented here suggests the use of genetically encoded FRET biosensors allows for quantification of intracellular metabolite concentrations via FRET ratios during cell growth. These FRET biosensors also show that it may be possible to predict intracellular metabolite concentrations based solely on the FRET ratio of these cells. It has also been suggested that in a smaller volume micro fermentation system (BioLector™) these biosensor transfected cell lines can be used to detect changes in intracellular metabolite levels. These biosensor cell lines were then used for cell line selection, results indicate that the FRET ratio may also be used as a tool to discard various cell lines based on time of progression. Using data in the final study, it may be concluded that the early progressed cell lines may not necessarily be the highest protein producers and with, respect to scFv production, no product specific differences exist. The results indicate the usefulness of these FRET biosensors as tools in aiding the process of information gathering at an early stage during cell line selection.

The present thesis focuses on amino acids (a.a.) and their metabolism by Chinese hamster ovary cells, the workhorse of the multibillion dollar biopharmaceutical industry. The aim of the research was to explore a.a. transport and metabolism and define optimal operating conditions during fed-batch culture, which is the most common process mode used industrially. A fast and reliable way to calculate a.a. concentration ranges in media and feeds is of vital importance, as a.a. are the monomers of proteins, which account for 70% of dry cell weight. The desired recombinant product of bioprocesses is typically also a protein. The transport of a.a. into the cells was studied at the mRNA level of a.a. transporters for the first time in a bioprocessing context. The presented results demonstrate that a.a. transport is not the limiting step for recombinant protein formation. Also, the study allowed for a staged feeding strategy to be designed, where a.a. were not fed altogether. Following linear projection of an integral of viable cell concentration target and using the specific a.a. consumption rates during batch culture, six feeds were formulated containing a.a. and glucose. Three designs were based on the results of the a.a. transport study; however, they underperformed in comparison to the other feeds. In the latter, all nutrients were fed at the same time, resulting in cell culture performance comparable to that obtained with a commercial feed that was tested in parallel. This renders the presented method the first to define a traceable quantitative way to calculate amount of nutrients in the feeds. Flux balance analysis, a powerful technique that allows for investigation of intracellular dynamics, was used to analyse the metabolic data. An enhanced intracellular network was created by coupling two pre-existing in the literature that also for the first time included the glycosylation of the host proteins in the biomass equation. Finally, a novel methodology was developed and coded in R to calculate specific rates of consumption/production of various metabolites in cell culture. The methodology couples mass balances for fed-batch culture operation with constructed vectors of the sampling and feeding schemes. This can be further developed to a bioprocess relevant software platform for analysing cell culture data.

The Non-Inertive-Feedback Thermofluidic Engine (NIFTE) is a two-phase thermofluidic oscillator which, by means of persistent periodic thermal-fluid oscillations when placed across a steady temperature difference, is capable of utilising low-grade (i.e., low temperature) heat to induce a fluid motion. Innovative devices which comprise no or few moving parts and that can operate utilising low-grade heat for fluid pumping and/or pressurisation are currently under development based on the NIFTE concept to: (i) understand the fundamental principle of operation of this novel technology; (ii) construct reliable, simple models that capture the first-order dominant underlying processes that govern its operation and performance for the purpose of early-stage engineering design; and (iii) to investigate the potential of this technology in specific fluid-pumping applications. Three spatially lumped linearised models of the NIFTE are developed through the use of electrical analogies. The first model (LTP) imposes a static (i.e. steady) linear temperature profile along the surface of the heat exchangers, the second model (CTD) imposes a constant-temperature difference between the surface of the heat exchanger and the working fluid, the third model (DHX) allows the solid heat exchanger blocks to store and release heat dynamically as they interact thermally with the working fluid. Through carrying out a parametric study on the LTP model, with and without inertial effects in the liquid phase it is shown that the inclusion of inertia has a significant effect on the trends and magnitudes of key performance indicators, namely the temperature gradient along the heat exchangers, oscillation frequency and exegetic efficiency. In addition, much improved predictions of the oscillation frequency and temperature gradient are possible when using the inertive LTP model. Following from this, a parametric study on the three models, all including inertia, is used to show that the CTD model predicts unrealistically high exergetic efficiencies, and as such is omitted from any further studies. A dissipative thermal loss parameter that can account for the exergetic losses due to the parasitic, cyclic phase change and heat exchange within the device is included in the LTP and DHX models in an effort to make realistic predictions of the exergetic efficiencies. A parametric study on the LTP and DHX models, including and excluding the thermal loss parameter is carried out and the results are compared to experimental data. It is found that the inclusion of the thermal loss parameter greatly improves the prediction of the exergetic efficiency in both the LTP and DHX models, both in trend and approximate magnitude. From the results it is concluded that, on accounting for thermal losses, the DHX model achieves the best predictions of the key performance indicators of the NIFTE, that is, of the oscillation frequency and exergetic efficiency of the device. An investigation on the applicability of different working fluids for the NIFTE, based on the dynamic heat exchanger model including thermal losses, with emphasis on the effects of key thermodynamic properties on the maximum thermal efficiency of an idealised cycle and the predicted exergetic efficiency of the device is also carried out. The change in specific volume due to vaporisation and the maximum saturation pressure of the working fluid in the cycle are found to have a dominant role in determining these efficiencies. Thirty one pure working fluids are studied, under a given set of scenarios, each representing a different practical application for the NIFTE device. For the scenario where the maximum pressure of the engine is defined by the pumping application, higher efficiencies are predicted for wet and isentropic fluids. For the scenario where the hot and cold heat exchanger temperatures are set by the external heat source and sink, higher efficiencies are predicted by dry and isentropic fluids. In this work, it is estimated that, with optimised designs and well-selected working fluids, the NIFTE may be capable of thermal efficiencies in the range 1 - 5 % when operating with low-grade heat at temperatures from 50 to 100°C, with current best performance of 1.5% at 80°C.

The binding free energy is one of the most important and desired thermodynamic properties in simulations of biological systems. The propensity of small molecules binding to macromolecules of human bio-substrates regulates their sub-cellular disposition. This subject is fundamental in transdermal permeation and hair absorption of cosmetic actives. Biomechanical and biophysical properties of hair and skin are related to keratin as their major constituent. A key challenge lies in predicting molecular and thermodynamic basis as the result of small molecules interacting with alpha helical keratin at the molecular level. In addition, elastic properties of human skin which are directly related to the interactions of keratin intermediate filaments remain a challenging subject. Molecular dynamics (MD) simulations provide a possibility of observing biological processes within atomistic resolution providing more detailed insight into experimental results. However, MD simulations are limited in terms of the achievable time scales. Hence, in this thesis MD simulations were employed in order to provide better understanding of the experimental results conducted in parallel and to overcome the main limiting factor of MD - the simulation time. For this purpose, thermodynamic and detailed structural basis have been delivered for small molecules interacting with keratin explaining and validating experimental data. On the top of this the fast free energy prediction tool has been built within all-atom force field by a use of steered molecular dynamics alone. Within the coarse grain approach, the force field was developed for the application of elastic properties of human skin enabling orders of magnitude faster than all-atom force fields simulations. The application of the coarser representation enabled assessing the influence of the natural moisturizing factor composed of small molecules on the elastic properties of the outermost human skin layer. In this work, MD results reached excellent agreement with the experimental data.

CO2 capture, transport and storage (CCTS) is gaining a broad interest as a countermeasure to global warming. The systematic development of CCTS network infrastructure is a non trivial activity that involves choosing the optimum design of the selected CO2 capture plant technology and transportation mode, and identifying the key performance operating parameters and limiting uncertainties that need to be mitigated or optimized to ensure a safe cost-optimal network. This thesis focuses on developing a systematic multiscale modelling and optimization approach that integrates validated sub-process models of the MEA-based CO2 capture plant, compression train and pipelines in which thermodynamic properties were calculated using SAFT-VR with the supply-chain CO2 network model. A number of simulations were performed to analyse and identify the cost-optimal design and operating variables while considering different CO2 prices, flue gas bypass option and uncertainty in transporting flow temperature and composition. A meta model that combines the results of the fine scale model was then used in the supply chain network model to successfully determine the cost-optimal CCTS network for a case study in Abu Dhabi. A key result of the thesis was that the cost- optimal degree of capture is a function of several site-specific factors, including exhaust gas characteristics, proximity to transportation networks, adequate geological storage capacity, CO2 price, and the option to partially bypass flue gas. A higher CO2 price had a clear impact on encouraging higher degree of capture. The flue gas bypass option was seen to be an optimal option for lower than 60% degree of capture. It was also observed that transportation companies should levy a charge to discourage transporting flow from low CO2 content sources. This thesis serves to underscore the need to comprehend the science governing the behaviour at different scales and the importance of a whole-system analysis of potential CCTS networks.

No description available.