More and more focus is going into the establishment of more sustainable approaches for wastewater treatment (WWT) in South Africa, as well as around the world. Governments are beginning to enforce more economical solutions for WWT, which will have less impact on costs as well as land area requirements. Effective solid-liquid separation in biological wastewater treatment is an important step in the process as it has a major impact on effluent quality. Traditionally this has been achieved using Secondary settling tanks (SSTs) for liquid/solid separation in combination with a biological reactor (for biological degradation of organic matter). SSTs, however, require a large space, which becomes onerous on land requirements. In an immersed membrane bioreactor (iMBR), solid-liquid separation takes place by the wastewater passing through membranes. As the WW flows through, at the same time solids are rejected by the membranes. These membranes are immersed in the bioreactor. iMBR thus eliminates the requirement for SSTs and are becoming more widely used to treat various types of wastewater, due to the decreasing cost of membranes and the resultant reduced plant footprint. MBR is thus becoming an attractive solution to clients due to its sustainable approach. As part of this investigation, 2 types of MBR technology were included, the Kubota FS MBR system and the Zeeweed HF MBR system. As the design of a CAS is sensitive to sludge settleability, various DSVI values were looked at as part of the CAS system. Each system was configured in an MLE and UCT process. In summary, the following systems were included in this investigation: • CAS in an MLE configuration with DSVI of 100,150 and 200 • CAS in a UCT configuration with DSVI of 100,150 and 200 • iMBR using FS membranes in an MLE configuration • iMBR using FS membranes in a UCT configuration • iMBR using HF membranes in an MLE configuration • iMBR using HF membranes in a UCT configuration Each process configuration was designed and sized using the steady state models. Each configuration was then fully costed using actual construction prices from past and current projects. Costing of the MBR systems were done in conjunction with the membrane suppliers who also provided valuable design input. The selection of design MLSS in an MBR and CAS has a significant impact on the reactor and SST size. The MLSS concentration also has an impact on the alpha factor which influences aeration efficiency. As part of this investigation, an optimum MLSS concentration (MLSSopt) cost optimization was done taking into account the effect on reactor size, SST area, membrane area, and aeration CAPEX and OPEX. This resulted in an MLSSopt of 5 500 mg/l and 6 000 mg/l for the CAS MLE and CAS UCT respectively, and 10 000 mg/l for the Zeeweed MBR and Kubota MBR system. The CAS system had the lowest total cost (CAPEX+OPEX) of the 3 systems over a lifespan of 10 years, with the Zeeweed MBR having the 2 nd lowest cost coming in at 61% higher than the CAS system. The Kubota MBR had the highest total cost with a 203% higher cost than the CAS system. In terms of land area requirement, the Kubota MBR required the least amount of land area, followed by the Zeeweed MBR which required 12% more land space. The CAS system required 127-514% more land space at the various DSVI values than the Kubota system. This was due to the additional SST area and a larger reactor requirement.

Primary sedimentation involves the separation of solids and liquid in primary settling tanks (PSTs) of wastewater treatment systems. These physical processes are described by various settling conditions such as discrete and flocculent settling, along with other phenomena such as flocculation, coagulation, ammonification or hydrolysis. The modelling of primary sedimentation has often been overlooked because (i) it involves various intricacies that are difficult to replicate and (ii) primary sedimentation has been assumed to be an input to most of the main unit process models, including the activated sludge (AS) system and the anaerobic digestion (AD) models. Though there has been a wide range of proposed mathematical models to describe how PSTs function, the need to correctly disaggregate the total suspended solids (TSS) into realistic fractions of unbiodegradable particulate organics (UPO), biodegradable particulate organics (BPO) and inorganic settleable solids (ISS), remains. This is because PST models that are unable to correctly split the TSS into its characteristic components make incorrect assumptions. These assumptions lead to inconsistencies in predicting the compositions of the primary sludge (PS) that is fed to the AD unit and the settled wastewater (settled WW) that is treated in the AS system. Hence, it becomes difficult to correctly simulate the entire system (plant-wide) towards a holistic evaluation of system strategies. In this study, a realistic PST model was developed, with characterized settling velocity groups, within a plant-wide setting, for municipal wastewater. This involved the improvement of a current TSS-based model into a more accurate and realistic model that could account for the settling of raw wastewater particles. The model was therefore expected to predict the composition of the PS that is treated in the AD system and the composition of the settled WW that is going to the AS unit processes. This could be achieved by splitting the TSS into UPO, BPO and ISS fractions. In developing preparation of such a realistic PST model, the following objectives were established: 1. Disaggregate the TSS into realistic UPO, BPO and ISS fractions, by means of discrete particle settling modelling (Kowlesser, 2014) and the particle settling velocity distribution (PSVD) approach of Bachis et al. (2015). 2. Verify that the model is internally consistent with wastewater treatment plant (WWTP) data, by means of mathematical material mass balances and other specific scenarios. 3. Demonstrate the application and impact of such a model by performing steady state plant-wide simulations. Using the discrete particle settling approach of Kowlesser (2014), a discrete particle settling model was developed in Microsoft Excel and implemented into a dynamic PST framework in WEST® (Vanhooren et al., 2003). The discrete particle settling model was described using steady state and dynamic calculations and the insights obtained from these calculations were implemented in the current TSS-based PST model of Bachis et al. (2015). This was performed towards developing the University of Cape Town Primary Sedimentation Unit (UCTPSU). The influent raw wastewater TSS was fractionated into UPO, BPO and ISS fractions and settling proportions of these fractions were assigned to five settling velocity groups. In addition, a distinct settling velocity was assigned to each settling velocity group. Previous studies data from WRC (1984) and Ekama (2017), were used in the discrete particle settling model, which was able to reproduce PS and settled WW outputs, through steady state and dynamic calculations and under strict material mass balances. As a result, UPO, BPO and ISS settling proportions as well as settling velocities, were extracted from these calculations and used as input parameters into the UCTPSU model. This dynamic model was rigorously verified to be internally consistent with regards to strict material mass balances. The verification scenarios also included variations of high and low settling velocities as well as a combination of both high and low velocities and checking that the model was behaving as expected. The application and impact of the UCTPSU model were demonstrated using plant-wide scenarios in proposing a preliminary integration, under steady state conditions. It showed how incorrect disaggregation of the TSS into UPO, BPO and ISS fractions can lead to incorrect predictions in terms of the settled WW composition, the AS system capacity, the effluent quality, as well as the energy consumption and generation in the AS system and AD unit respectively. The investigation also revealed the need to measure key wastewater parameters such as particle settling velocities and the unbiodegradable particulate COD fraction, when it comes to realistically modelling of primary sedimentation of municipal wastewater, with the view of optimizing plant operations and tactical decision making. The study thereafter recommended the need to conduct an extensive experimental campaign to measure in-situ diurnal data, mainly in terms of settling velocities and settling proportions of UPO, BPO and ISS. It was also suggested to use the settleometer as a tool to extract these settling velocities and settling proportions, after performing biodegradability tests. As such, the data collected from the experimental campaign and the biodegradability tests could be used in calibrating the UCTPSU model and validation could be undertaken by means of full plant scale data.

Nitrification can be seen as the weakness of a conventional activated sludge (CAS) process employing biological nutrient removal (BNR). Suspended nitrifiers only grow in the aerobic zone of the biological reactor but are subjected to anaerobic and anoxic conditions where no nitrifier growth takes place. To establish a nitrifier population that consistently produces low effluent ammonia concentrations, long sludge ages are required (about 15 to 25 days) in South African BNR wastewater treatment plants. This results in relatively large biological reactors. Integrated Fixed-Film Activated Sludge (IFAS) systems have been used extensively in European and Scandinavian countries. This process entails the addition of moving-bed biofilm carriers in certain zones of an activated sludge system to establish biofilm growth. The most successful application has been the addition of these carriers in the aerobic zones of activated sludge plants to facilitate the growth of nitrifiers on the biofilm. This allows nitrifiers to grow independently from the suspended sludge age since it remains stationary on the biofilm in the aerobic tank. The system is thereby relieved from the requirement of a long suspended sludge age. For the University of Cape Town (UCT) process commonly employed in South Africa, it is shown that a suspended sludge age of 5 to 7 days is adequate to meet final effluent standards when converted to an IFAS process. As a result, an UCT-IFAS process can treat 50% to 70% more wastewater in an existing process volume or reduce the size required for a new installation by 30% to 40% when compared to a conventional UCT process with a minimum wastewater temperature of 14°C. The intricacies and challenges associated with designing an IFAS process are unpacked in this thesis to gain a better understanding of what is required to harvest the potential benefits.

Fertilizer production is a massive global industry with the global consumption of the three main fertilizer nutrients, nitrogen, phosphate and potassium estimated at 187 million tonnes in 2016 with an anticipated annual growth of approximately 2% for the foreseeable future. In 2016 the global fertilizer market was estimated to have an overall market value of 141 billion US dollars. Fertilizer production produces significant liquid waste as process water used for the various separations, cleaning, emulsifying and dilution processes absorbs various nutrients and contaminants from these production processes. This liquid waste has characteristically high concentrations of nutrients derived from the base fertilizer, such as various dissolved phosphate compounds for phosphate-based fertilizer production or dissolved nitrogenous compounds for nitrogen based fertilizer production. These contaminants are inherently nutrients that could be recovered for beneficial re-use. The phosphate and potassium minerals used in fertilizer production are obtained from ores mined from the earth, thus the re-use of these mineral present particular significance when taking into accounting the declining global supply of these ores. Furthermore, if these liquid wastes are not disposed of correctly they can lead to detrimental environmental impacts such as eutrophication and ecological degradation in water courses. This study addresses this problem by presenting three novel treatment techniques to treat the liquid waste produced from a fertilizer production plant. A liquid waste sample obtained from a particular fertilizer production plant producing primarily nitrogen-based fertilizer is used as a design basis to evaluate the three presented treatment techniques. The techniques are evaluated based on their economic feasibility, technical feasibility and resource recovery ability. The three treatment techniques studied were the Sharon-Anammox bioreaction process, electrodialysis with struvite recovery process and combined forward-reverse osmosis process. The technical feasibility of the processes was primarily evaluated based on the effluent water quality from the treatment systems. The effluent quality index (EQI) was used as a comparative measure of the effluent quality of the processes. All three processes were found to perform inadequately from a technical feasibility perspective as demonstrated by the negative EQI values obtained for the processes. The Sharon-Anammox bioreaction process was found to perform poorly because its application is limited to treatment of waste streams containing high ammonia concentrations such as in conventional domestic waste. Therefore, the Sharon-Anammox process was not suited to the fertilizer effluent which also contained high nitrates, phosphates and total dissolved solids. The electrodialysis process performed poorly as it was unable to effectively remove the ammonium cations from the process water. The combined forward-reverse osmosis process performed poorly because a resource recovery step was not included to treat the concentrated waste stream discharged from the forward osmosis step of the process. It was identified that a similar struvite recovery step should be added to the combined forward-reverse osmosis process to improve the technical feasibility of the process and to provide the process with resource recovery capabilities. From an economic feasibility perspective, it was found that the addition of the struvite recovery setup to the electrodialysis process increased the capital costs of the process to between 300% and 500% of the other two options. However, with the omission of the struvite recovery setup the capital costs of all three processes were in a similar range.

The Western Cape has suffered severe droughts over the past decade which has placed severe strain on raw water resources for both agriculture and municipal use. The crisis was due to many factors including climate change, increasing urbanisation and ageing infrastructure to name a few. The water scarcity problems will persist in the future globally unless water management authorities are able to augment existing raw water resources with a mix of desalination, groundwater and reclamation of treated effluent. The town of Stellenbosch was selected as part of a case study to determine the feasibility of implementing a direct potable reuse (DPR) plant to augment the future water resource mix from a technical, social, environmental and economic standpoint. Over the past two decades there has been a global shift towards direct and indirect potable reuse schemes to augment existing surface and groundwater resources. The shift has been accelerated by advances in treatment technology, water quality monitoring and research which have reduced the costs of potable reuse when compared to conventional water resources. The effluent from the Stellenbosch Wastewater Treatment Works was investigated as a reliable raw feed water source for the Stellenbosch DPR Plant. The Stellenbosch DPR Plant treatment train followed the multiple barrier approach to ensure high quality product water and mitigate potential risks to human health. The process design favoured granular activated carbon filtration instead of reverse osmosis due to the lower costs, inland location and brine disposal issues along with the acceptable total dissolved salt levels within the source water. The process design was developed further to determine the energy consumption, chemical consumption and process monitoring and control framework for the plant. A technical feasibility was done on three scenarios which were selected based on mix of reclaimed water and current surface water resources to supply the town of Stellenbosch with potable water. Scenario A – ‘do-nothing’ approach whereby the Stellenbosch Municipality would continue to be supplied with bulk raw water from the Theewaterskloof Dam treat it at the Paradyskloof WTW Scenario B – DPR Plant which produced potable water and injected it upstream of the Paradyskloof WTW Scenario C - DPR Plant which produced potable water and injected it downstream of the Paradyskloof WTW The research found that it would be feasible to implement a DPR scheme in Stellenbosch to improve the towns’ water security to meet future demands. The technical, social and environmental issues introduced in this research would need to be considered and developed further once a decision was made to pursue DPR. The unit costs of DPR would be higher than expanding the current raw surface water allocation and conventional water treatment works, which would have a knock-on effect on consumer tariffs. These economic costs would need to be compared to the towns risk exposure to climate change and water demands from surrounding areas within the Western Cape should they continue to abstract water from surface water resource not under their control.

UCTOLD, IAWPRC and UCTPHO are activated sludge system diurnal simulation software programs written and compiled in TurboPascal 3.1 by the Water Research Group at the University of Cape Town in the late 1980s and early 1990s. IAWPRC is identical to International Water Association (IWA) Activated sludge Model No 1 (ASM1, Henze et al., 1987) and UCTOLD is an earlier variation of ASM1 that makes no material difference to simulation results when each are run with their default parameters – the difference is explained by Dold and Marais (1985). Details on the use of UCTOLD and IAWPRC was published by Dold et al. (1991) and this guide remains essential reading for the responsible use of the recoded UCTOLD and UCTPHO presented in this thesis. In the early 1990s, when the kinetics of phosphorus accumulating organisms (PAO) were sufficiently well understood, biological excess P removal (BEPR) was added to UCTOLD to yield UCTPHO. This software has the same look and feel as UCTOLD/IAWPRC and so a guide for it was never published. Details of the BEPR research that was codified into it are given by Wentzel et al. (1991) and Clayton et al. (1991). Later, much of the Wentzel BEPR model was integrated into ASM1 to form ASM2 (Henze et al., 1995). So UCTPHO and ASM2 differ in the same way as UCTOLD and ASM1/IAWPRC differ, essentially only in the fate of the product of the hydrolysis of slowly biodegradable particulate organics (BPO) – in ASM1/2, this product enters the bulk liquid as readily biodegradable soluble organics (BSO) with the ordinary heterotrophic organisms (OHO) utilizing only BSO, whereas in UCTOLD/UCTPHO the BPO product is utilized directly by the OHO and so does not enter the bulk liquid (Dold and Marais, 1985). One aspect is very important to note about the models embedded in these codes – UCTOLD and ASM1 are only for biological N removal systems and UCTPHO and ASM2 are only for biological N and P removal systems. One cannot simulate a N and P removal system with UCTOLD (or ASM1) and one cannot simulate a N removal system only with UCTPHO (or ASM2). The kinetics of denitrification in UCTOLD/ASM1 and UCTPHO/ASM2 are not the same - the details of the differences are explained by Clayton et al. (1991). For two decades up to around 2010, UCTOLD and UCTPHO provided free activated sludge system simulation tools to consultants, municipalities and operators to predict the performance of the activated sludge systems of wastewater treatment plants. Therefore smaller consultancies had the software available for use without requiring expensive commercial software for the modelling of household/municipal wastewater. With the advancement of computer hardware and Windows operating systems over two decades, the UCTOLD and UCTPHO software (not the models coded within them) grew outdated and gradually lost functionality. Then when 64 bit processor personal computers (PC) became available, the old compiled TurboPascal 3.1 software could no longer be run – they can still run on 32 bit processor WinXP PCs. Re-coding the software was therefore required if the continued use of the software were to be ensured. This recoding into Excel/VBA was undertaken in this thesis. Some considerable effort also went into the improvement of the user interface. The new interface allows quick input and configuration of the biological reactors of the activated sludge system, and the recycle streams. Several challenges presented themselves during the recoding stage of the project, most notable of which were the data management structures and the calculation time minimisation. Also, significant time was also spent on simulations to determine the effects of the integration parameters on calculation time, and to determine how to minimise the calculation time. After struggling with these challenges for some months, the recoded programs started to present results similar to the original TurboPascal compiled code, about a year after commencement with the project. This Beta version has the same functionality that the original Turbo Pascal versions had including Adjustment of the Wastage Pattern and Graphical Display of Diurnal Results. The calculation of activated sludge ISS concentration was added as an option by including the influent OHO and PAO ISS to determine reactor TSS concentration for N and N&P removal systems using results from Ekama and Wentzel (2004). So, instead of having to enter an activated sludge VSS/TSS ratio, the option to calculate it from entered influent VSS and TSS data can now be selected to calculate the reactor TSS concentration.

Water is a scarce natural resource, which requires to be treated with much care and importance. It is a finite resource and should be used sparingly. The process of treating domestic wastewater varies from ponds to the more advanced system, namely the activated sludge system. The main purpose of wastewater treatment is the reduction of pathogenic contamination, coliform bacteria, suspended solids, oxygen demand, and nutrient enrichment. The application or use of stabilisation ponds, as a part of the wastewater treatment process, depends on, among other factors, the influent loading and climate conditions. Waste Stabilisation Ponds (WSPs) are used to biologically treat domestic wastewater or industrial wastewater. The present study focuses on the treatment of domestic wastewater by using the WSPs in the absence of mechanical and electrical equipments. Different countries use different methods of pond design or WSP sizing and different parameters to ensure that the effluent discharge guidelines of the Department of Water and Sanitation (DWS) and World Health Organisation (WHO) are met. There are insufficient literature studies focusing on the design models and water quality data that can be used for sizing the WSPs in South Africa. There is a requirement for a study that can compare the existing WSP design models in different countries and check their suitability for South Africa, particularly their applicability to provinces with respect to climate and domestic wastewater quality. The comparison between the WSP design models will assist the process designers in the early stages of projects, particularly in the feasibility study stages (Scenario 1). The objective of the present study is to perform a comprehensive review of the use of WSPs in domestic wastewater treatment, their design and operating requirements for optimal performance, and the existing mathematical models used to virtually replicate the WSP treatment processes. Also considered is the development of a simplified model to demonstrate its application as a tool for the effective design of WSPs, including a case study of a WSP in the Eastern Cape (EC).

AN ABSTRACT OF THE DISSERTATION OFSherri Houmadi, for the Doctor of Philosophy degree in Engineering Science, presented on March 27, 2020, at Southern Illinois University Carbondale. TITLE: THE APPLICATION OF CONVOLUTIONAL NEURAL NETWORKS TO CLASSIFY PAINT DEFECTSMAJOR PROFESSOR: Dr. Julie DunstonDespite all of the technological advancements in computer vision, many companies still utilize human visual inspection to determine whether parts are good or bad. It is particularly challenging for humans to inspect parts in a fast-moving manufacturing environment. Such is the case at Aisin Manufacturing Illinois where this study will be testing the use of convolutional neural networks (CNNs) to classify paint defects on painted outside door handles and caps for automobiles. Widespread implementation of vision systems has resulted in advancements in machine learning. As the field of artificial intelligence (AI) evolves and improvement are made, diverse industries are adopting AI models for use in their applications. Medical imaging classification using neural networks has exploded in recent years. Convolutional neural networks have proven to scale very well for image classification models by extracting various features from the images. A goal of this study is to create a low-cost machine learning model that is able to quickly classify paint defects in order to identify rework parts that can be repaired and shipped. The central thesis of this doctoral work is to test a machine learning model that can classify the paint defects based on a very small dataset of images, where the images are taken with a smartphone camera in a manufacturing setting. The end goal is to train the model for an overall accuracy rate of at least 80%. By using transfer learning and balancing the class datasets, the model was trained to achieve an overall accuracy rate of 82%.

For this thesis, I created a guideline for socially assistive robots (SARs), and used it to evaluate a reading comprehension based social robot. To create the guideline I extracted relevant details from published standards about toy safety, radio equipment, electromagnetic compatibility, internet of things security, ethical considerations for human-computer interaction, and data privacy. I then sent a summarized version to experts in the field for feedback. I received seven responses, five of whom were from researchers in academia. The sample size was too small for statistical analysis. Survey responses identified additional areas, such as interactivity and aesthetics, for the guideline. I evaluated a reading comprehension based social robot called HAPI the Librarian with my newly created guidelines. Using HAPI, I found that the guidelines worked, but needed improvement. Improvements suggested for the next iteration of the guide- line are to provide better directives for intangible concepts such as ethics and data privacy. Additionally, the guidelines should help to identify characteristics that raise ethical or data privacy concerns. Overall, these guidelines can be applied to socially assistive robots designed from scratch or purchased off-the-shelf.

Doctor of Philosophy / Department of Industrial and Manufacturing Systems Engineering / Shing I. Chang / Profile analysis has drawn attention in quality engineering applications due to the growing use of sensors and information technologies. Unlike the conventional quality characteristics of interest, a profile is formed functionally dependent on one or more explanatory variables. A single profile may contain hundred or thousand data points. The conventional charting tools cannot handle such high dimensional datasets. In this dissertation, six unsolved issues are investigated. First, Chang and Yadama’s method (2010) shows competitive results in nonlinear profile monitoring. However, the effectiveness of removing noise from given nonlinear profile by using B-splines fitting with and without wavelet transformation is unclear. Second, many researches dealt with profile analysis problem considering whether profile shape change only or variance change only. Those methods cannot identify whether the process is out-of-control due to mean or variance shift. Third, methods dealing with detecting profile shape change always assume that a gold standard profile exists. The existing profile shape change detecting methods are hard to be implemented directly. Fourth, multiple nonlinear profiles situation may exist in real world applications, so that conventional single profile analysis methods may result in high false alarm rate when dealing multiple profile scenario. Fifth, Multiple nonlinear profiles situation may be also happened in designs of experiment. In a conventional experimental design, the response variable is usually considered a single value or a vector. The conventional approach cannot deal with when the format of the response factor as multiple nonlinear profiles. Finally, profile fault diagnosis is an important step after detecting out-of-control signal. However, current approaches will lead to large number of combinations if the number of sections is too large. The organization of this dissertation is as following. Chapter 1 introduce the profile analysis, current solutions, and challenges; Chapter 2 to Chapter 4 explore the unsolved challenges in single profile analysis; Chapter 5 and Chapter 6 investigate multiple profiles issues in profile monitoring analysis and experimental design method. Chapter 7 proposed a novel high-dimensional diagnosis control chart to diagnose the cause of out-of-control signal via visualization aid. Finally, Chapter 8 summarizes the achievements and contributions of this research.