It goes without saying that water scarcity is a widespread and increasingly pressing global challenge. One of the methods which can mitigate water shortage is to increase freshwater production via desalination of saline waters. Seawater and saline aquifer sources represent 97.5% of all water on Earth. Hence, treating even a small portion of saline water could significantly reduce water shortage. Although reverse osmosis is one of the state-of-the-art pressure-driven membrane desalination technologies, it is incapable of desalinating high-salinity streams due to the very high osmotic pressure to overcome. Membrane distillation (MD) is one of the emerging methods, which has attracted much attention for desalinating highly saline brines. MD is a thermally driven process in which only vapor molecules pass through the pores of a microporous hydrophobic membrane. This process, however, has not been fully commercialized due to a number of challenges, including “pore wetting”. Pore wetting refers to the presence of liquid, instead of just water vapor, inside the membrane pores, which may cause a decrease in MD flux and/or deterioration of distillate quality. Herein, a comprehensive review on pore wetting is presented, and then this phenomenon is investigated from four aspects. In the first phase of this project, a theoretical model is presented according to which the pore size distribution of membrane, a parameter affecting pore wetting risk, is estimated by employing only a few experimental data points in accordance with the wet/dry method, reducing the number of data required to be recorded largely. In the next phase, an equation is presented for the estimation of liquid entry pressure (LEP), a membrane parameter closely related to pore wetting, using computational fluid dynamics (CFD) tools and genetic programming (GP) as an intelligent technique. This equation can estimate LEP in closer agreement to experimental values in comparison to the Young-Laplace equation. In the third phase, movement of liquid-gas interface inside the membrane pore is tracked using a well-founded model, and consequently, the pressure and velocity at the interface and the required time for replacement are studied. Finally, in the last phase, a model is developed for pore wetting in vacuum MD, considering heat and mass balances at the vapor-liquid interface. This model assumes that heat only enters the pore inlet and is removed due to liquid vaporization at the vapor-liquid interface, with heat transfer through the pore wall neglected. This model shows that partial pore wetting is possible since the vapor-liquid interface might remain within the pore at the steady-state condition. Further, this model can predict the decrease in temperature from the pore inlet to the vapor-liquid interface, a phenomenon that has been reported in the literature without any proof.

In this work polymeric materials were developed to be used as a solvent and heat resistance membrane for different applications. In ultrafiltration, poly (ether imide sulfone) membranes were manufactured by combining phase inversion and functionalization reaction between epoxy groups and amine modified polyether oligomers (Jeffamine®). Polysilsesquioxanes or oligo silsesquioxanes containing epoxy functionalities were in-situ grown in the casting solution and made available for further reaction with amines in the coagulation/annealing baths. Water permeances up to 1500 l m-2 h-1 bar-1 were obtained with sharp pore size distribution and a pore diameter peak at 66 nm, confirmed by porosimetry, allowing 99.2 % rejection of γ-globulin. The membranes were stable in 50:50 dimethylformamide/water, 50:50 N-methyl pyrrolidone/water and 100 % tetrahydrofuran. In pervaporation, Novel hydrophobic Hyflon®/Extem® and Hyflon®/PVDF were developed and investigated for ethylene glycol dehydration and n-butanol dehydration respectively. For ethylene glycol different Extem® concentrations were evaluated with regard to both flux and amount of water in the permeate side. Eighteen (18) wt% gave more than 90 wt% water in the permeate. Increasing feed temperature from 25 to 85°C increased the water flux from 31 to 91 g m-2 h-1 when using 5 wt% water in ethylene glycol as feed. The water flux of 40 wt% water:ethylene glycol at 45°C was found to be 350 g m-2 h-1. And for n-butanol dehydration the coating protocols for thin defect-free Hyflon® selective layer on the PVDF support was optimized. Water and n-butanol transport was measured, analyzing the effect of operating conditions. The water flux through the newly developed membranes was higher than 150 g m-2 h-1 with selectivity for water higher than 99 wt%. The membrane application can be extended to other solvents, supporting an effective and simple method for dehydration with hydrophobic membranes. In membrane distillation, PVDF and Extem® membranes before and after coating with Hyflon® were examined for ionic liquid dehydration on 23.6 mS cm-1 feed concentration. Different feed temperatures and flow rates were evaluated for flux as well as rejection. High flux was obtained at 70°C and increased at high flow rate from 2 Kg m2 h to 10 Kg m2 h.

Dynamic simulation programs were created in the Python programming language, to describe a laboratory scale, sub-zero distillation column, that is used to separate mixtures of tetrafluoroethylene (TFE), hexafluoropropylene (HFP) and octafluorocyclobutane (OFCB). Both the equilibrium and rate-based modelling approaches were taken, to generate a comparison between the efficiency and simulation time of both models. A physical properties data bank for the three components had to be created, as the main and many of the sub-models require physical or thermodynamic properties for evaluation. The different physical property models, found in literature, were programmed into functions that could easily return the wanted property, given a set of required inputs. The applicable mixing rules for each property type was also programmed into functions, to allow for easy retrieval. The vapour-liquid equilibrium (VLE) model used, is also one that comes from literature and is based on parameters for the three binary systems. The VLE model consists of the Peng-Robinson equation of state, that utilises the Mathias-Copeman alpha function and the Wong-Sandler mixing rules, to describe the vapour phase. The liquid phase is described by the non-random two liquid (NRTL) activity coefficient model. Furthermore, the γ-Φ VLE formulation was used to put the thermodynamic model together. These models were also written into functions to serve as simulation building blocks. Mass and energy transfer on packed sections in the rate-based model was described by the Maxwell-Stefan diffusion model. The form of this model that was utilised, is the matrix-based, exact solution of the Maxwell-Stefan equations, under the two-film theory. This model was slightly simplified by assuming that the corrective flux matrix reduces to the identity matrix- an assumption that is regularly made in distillation modelling. Emphasis was laid in documenting how the models are put together to build the simulations. Dynamic simulation algorithms rarely accompany distillation models reported in literature, or authors make use of commercial software to order the modelling equations for them. One of the objectives of the research presented here was, therefore, to report on the process developed to solve the problem. Both simulation programs delivered typical responses that can be expected of distillation systems. The actual change in the magnitude of the values, however, proved to be significantly small. The cause of this, being the large liquid molar hold-up values that were produced by the model initialisation. The feed flow rate, in comparison, is too small to bring about a significant effect when suddenly increased. This could mean that the system is not capable of reaching the steady-state produced by the initialisation (as the feed cylinder may be too small to contain the required amount of feed gas) and that the column may have to be run in a continuous dynamic state. To be sure of this, however, the model will first need to be validated against experimental data. Furthermore, the simulation programs proved to progress very slowly, particularly the simulation built around the rate-based model. A time step-size of 0.5 resulted in an integration time around 1 minute and 20 seconds for the equilibrium model, while the rate model ran for over 19 minutes, both for a timespan of 300 s. It is recommended that future research focuses on building start-up simulations for the models, to provide better initial results and to give more insight into the operation of the column. Experimental validation of the models is also important, to establish their accuracies. Finally, work has to be done to improve the simulation speeds, especially if it is required that one of the models are integrated into the column's control system. / Dissertation (MEng)--University of Pretoria, 2019. / Fluorochemical Expansion Initiative / Department of Science and Technology / Department of Trade and Industry / Chemical Engineering / MEng / Unrestricted

No description available.

This thesis presents advancements in the development of hybrid and data-driven models of distillation columns. First, it introduces a hybrid model structure that incorporates a novel multiplicative correction term for inferential monitoring. This model architecture outperforms previous hybrid structures, especially in extrapolation conditions, and can be adapted for different conditions. Second, it presents a methodology for selecting temperature measurement for inferential models. This methodology demonstrates that nonlinear columns can be effectively modeled with linear models requiring two temperature measurements per section (previous works state requiring more) when the measurements are selected systematically. Finally, an iterative Real-Time Optimization (RTO) based on an augmented inferential data-driven model is demonstrated. The accuracy of the model enables estimation of the sensitivity matrix of the plant from the model without the need for additional plant measurements. The proposed RTO framework produces results similar to those achieved by optimizing rigorous tray to tray distillation models. / Thesis / Candidate in Philosophy / This thesis presents advancements in the development of hybrid and data-driven models of distillation columns. First, it introduces a hybrid model structure that incorporates a novel multiplicative correction term for inferential monitoring. This model architecture outperforms previous hybrid structures, especially in extrapolation conditions, and can be adapted for different conditions. Second, it presents a methodology for selecting temperature measurement for inferential models. This methodology demonstrates that nonlinear columns can be effectively modeled with linear models requiring two temperature measurements per section (previous works state requiring more) when the measurements are selected systematically. Finally, an iterative Real-Time Optimization (RTO) based on an augmented inferential data-driven model is demonstrated. The accuracy of the model enables estimation of the sensitivity matrix of the plant from the model without the need for additional plant measurements. The proposed RTO framework produces results similar to those achieved by optimizing rigorous tray to tray distillation models.

No description available.

With Industrial Internet, businesses can pool their resources to acquire large amounts of data that can then be used in machine learning tasks. Despite the potential to speed up training and deployment and improve decision-making through data-sharing, rising privacy concerns are slowing the spread of such technologies. As businesses are naturally protective of their data, this poses a barrier to interoperability. While previous research has focused on privacy-preserving methods, existing works typically consider data that is averaged or randomly sampled by all contributors rather than selecting data that are best suited for a specific downstream learning task. In response to the dearth of efficient data-sharing methods for diverse machine learning tasks in the Industrial Internet, this work presents an end-to end working demonstration of a search engine prototype built on PriED, a task-driven data-sharing approach that enhances the performance of supervised learning by judiciously fusing shared and local participant data. / Master of Science / My work focuses on PriED - a data sharing framework that enhances machine learning performance while also preserving user data privacy. In particular, I have built a working demonstration of a search engine that leverages the PriED framework and allows users to collaborate with their data without compromising their data privacy.

Batch distillation with chemical reaction when takes place in the same unit is referred to as batch reactive distillation process. The combination reduces the capital and operating costs considerably. Among many different types of batch reactive distillation column configurations, (a) conventional (b) inverted (c) semi-batch columns are considered here. Three reaction schemes such as (a) esterification of methanol (b) esterification of ethanol (c) hydrolysis of methyl lactate are studied here. Four different types of dynamic optimisation problems such as (a) maximum conversion (b) maximum productivity (c) maximum profit and (d) minimum time are formulated in this work. Optimal design and or operation policies are obtained for all the reaction schemes. A detailed rigorous dynamic model consisting of mass, energy balances, chemical reaction and thermodynamic properties is considered for the process. The model was incorporated within the dynamic optimisation problems. Control Vector Parameterisation (CVP) technique was used to convert the dynamic optimisation problem into a nonlinear programming problem which was solved using efficient SQP (Successive Quadratic Programming) method available within the gPROMS (general PROcess Modelling System) software. It is observed that multi-reflux ratio or linear reflux operation always led to better performance in terms of conversion, productivity for all reaction schemes compared to that obtained using single reflux operation. Feed dilution (in the case of ethanol esterification) led to more profit even though productivity was found to be lower. This was due to reduction in feed price because of feed dilution. Semi-batch reactive distillation opertation (for ethanol esterification) led to better conversion compared to conventional batch distillation, however, the total amount of acetic acid (reactant) was greater in semi-batch operation. Optimisation of design and operation (for ethanol esterification) clearly showed that a single cloumn will not lead to profitable operation for all possible product demand profile. Also change in feed and /or product price may lead to adjust the production target to maximise the profitability. In batch distillation, total reflux operation is recommended or observed at the begining of the operation (as is the case for methnaol or ethanol esterification). However, in the case of hydrolysis, total reflux operation was obseved at the end of the operation. This was due to lactic acid (being the heaviest) was withrawn as the final bottom product. / Libyan Petroleum Institute

The process of fractional distillation has been performed for many years but has only within the past half century become a true science. However, because of the many variables involved and the uncertainty of their effect upon fractionating column efficiency, the design of columns has long been a major engineering problem. Plate and column efficiencies, as given in the literature, vary widely and it was thought possible that the inconsistencies of the results reported might be due to poor equilibrium within the column, caused in part by faulty sampling. In order to determine the operating characteristics of a fifteen plate bubble-cap column (8-3/8" I.D., one 3-7/8" bubble-cap per plate, and plate spacing of 5-7/8"), the effects of the operating variables, and the distillation characteristics of two different types of binary mixtures, the column assembly was redesigned for continuous distillation and provisions made for the introduction of feed at its boiling point to any one of the bottom eight plates. The binary — mixtures distilled were isopropyl alcohol — water (distillation rates 157 - 376 gms./min., feed concentrations 3.1 - 10.8 mol per cent isopropyl alcohol, reflux ratio 3:1, and feed rate of 330 gms./min.), and toluene - ethylene dichloride (distillation rates 203 and 196 gms./min., feed concentrations 41.4 and 39.7 mol percent ethylene dichloride, reflux ratio 4:1, and feed rate of 138 gms./min.) In order to reduce the possibility of disturbances within the column by removal of large plate samples, the refractive index method of analysis was used which required maximum samples of only 4 ml. Results of the experiments made indicated that the assumption of the McCabe - Thiele operating line is probably incorrect. In the case of the isopropyl alcohol — water fractionation, a considerable divergence (largest between 10 and 40 mol per cent isopropyl alcohol) between the McCabe - Thiele and the actual operating line was noted, the actual operating line being a curve approximately the shape of the equilibrium curve. Murphree Plate Efficiencies for the mixture varied from 0 to 100 per cent, depending primarily on the relative deviations of the two operating lines from the equilibrium curve. The McCabe - Thiele operating line for the system toluene — ethylene dichloride closely approximated the actual operating line. Murphree Plate Efficiencies varied from 44.5 to 121 percent, but the individual plate efficiencies were more consistent with the average efficiency. The rate of distillation over the range of 157 to 376 gms./min. and feed concentration over the range of 3.1 to 10.8 mol percent isopropyl alc1ohol, at a reflux ratio of 3:1 had negligible effect on product purity which ranged from approximately 62 to 67 mol per cent isopropyl alcohol. / Master of Science

Master of Science