Browse
Recent Submissions
Item Platform architecture in the financial sector: A customer perspective(2025-10) Singh, D; Goncalves, Duarte PThe financial industry has witnessed widespread adoption of digital banking services and platform enterprise architectures. The literature highlighted gaps in linking enterprise architecture and platforms and facilitated the synthesis of a conceptual framework. A mixed methods study validated enterprise architectural elements and characteristics critical to the success of banking platforms in South Africa from a customer perspective. Two data collection methodologies were applied to five anonymized South African banks: banking application content analysis and observation, and quantitative analysis of social media customer reviews of customer’s perceptions towards the banking platforms. Vital enterprise architectural elements are strategies, products, services, technical design, privacy, security, governance, communication and trust. Social media review analysis found that the predominant complaints were inefficient services and long service times. While banks invest in their digital platforms, high operational costs and operational inefficiencies within current processes must be resolved to reduce customer complaints.Item The numerical aerodynamic evaluation of geometrical configurations of a vaporizer tube micro-gas combustor(2019) Meyers, Bronwyn C; Grobler, Jan-HendrikA combustor was designed for a 200N micro-gas turbine for the Model aircraft industry using the NREC design method. During the design process, there are various aspects where there are no definitive methodologies for specifying the design detail, such as the design of the hole-sets, and multiple options can be derived that can satisfy the required mass flow split and pressure drop for a particular hole-set. For this study, the various solutions for hole-set configurations were tested using CFD before experimental development will be pursued. The three design parameters tested were 1) annular area split configuration, 2) Hole area splits and 3) Relative hole positions. CFD simulations for a chosen 9 designs were run and the data were processed, analysed and interpreted. Some of the results such as mass flow splits and pressure drop are already quantitative in nature, however, the evaluation of the quality of the recirculation zone, mixing and outlet plane flow are of a more qualitative nature. In order to apply a more quantitative method for choosing a preferable design, a scoring system was devised in order to apply a quantitative value to the aspects of the flow which are initially analysed subjectively. The evaluation was broken up into the relevant aspects/features, namely, Recirculation zone (Rz), Outlet and Mixing. For each feature, the designs were subjectively evaluated relative to each other and given a rating/score. The scoring methodology for ranking combustors proved to be an effective method for evaluating the large mass of data that is generated using CFD and allowed for the use of this data to inform choices when narrowing down the mass of combustor designs that are generated in the preliminary design phase.Item Design and energy modelling of selected off-grid microgrids in South Africa using HOMER Pro®(2025-10) Oloo, Fiona RA; Nombakuse, Z; Mdhluli, Sipho D; Rampokanyo, MApproximately 94 % of South Africa’s population has electricity access, but there are still several remote communities that do not have electricity access. These communities use energy sources such as wood and charcoal, but these can pose a significant health risk. Thus, a least-cost microgrid electricity plan and associated optimization studies are proposed for five select remote unelectrified rural communities in South Africa using HOMER Pro® software. The energy sources considered for optimization are solar, wind, hydropower and battery storage. Diesel were not considered due to security concerns in the remote communities and challenges that are associated with fuel management. Two main scenarios were considered for the optimization. This first scenario was based on meeting the current estimated load demand of the microgrids. The peak demand in the first scenario of each of the villages was 18 kW, 43.46 kW, 36.2 kW, 26.08 kW, 76.5 kW respectively. The second scenario was based on meeting the future demand growth that considers oversizing the demand by 30 % to account for economic development and expected community growth. Both scenarios considered the impact of maximum allowed capacity shortages between 0% - 50 % to evaluate the impact on investment cost and Levelized Cost of Energy (LCOE). Reducing the cost of energy with an increasing capacity shortage had to be balanced with the fact that communities prefer consistent power supply. Therefore, for all the villages, the results show that a maximum allowed capacity shortage of 10 % appeared be the most cost-effective option for all the scenarios. From the optimization studies, solar PV and battery storage is the most optimal combination in both scenarios, for four of the five village locations. The most optimal combination for fifth village is hydropower and battery storage. The LCOE values range from R 4.77/kWh - R 5.5/kWh for the villages in the first scenario and R 4.8/kWh - R 5.48/kWh for the villages in the second scenario. In all the villages, considerable excess electricity was produced with one of the villages resulting in excess energy up to about 80 % This significantly contributes the LCOE values obtained. Therefore, the study has shown that considerations must be made for designing off grid microgrids with the intent to fully utilise the excess electricity productively and decrease the resulting LCOE.Item The landscape of quantum computing education in South Africa: Lessons from global practices(2025-11) Featherstone, Coral; Lourens, Roger LThis paper presents a comprehensive literature review ex ploring the landscape of quantum computing (QC) education in South Africa. The review aims to address the scarcity of published research on the provided content and educational experiences of African QC stu dents. While global QC education efforts are expanding, there is a sig nificant gap in published research detailing local educational experiences and offerings. By analysing both limited local and international liter ature, this study identifies key challenges, including the digital divide and the predominant focus on university-level instruction. This contrasts with international trends that incorporate gamification, visual tools, and outreach to primary and high school students. Findings reveal that inter nationally there are efforts to introduce teaching materials more suited to non-physics learners – and since there is interest from industry – teaching aligned to industry’s practical requirements. Future work should focus on understanding where students get stuck in their understanding within current curricula, exploring curricula adaptions for non-physics students, and leveraging off some of the known international use cases.Item Hate speech detection in isiZulu: A semantic approach(2025-05) Sethosa, MR; Rananga, S; Mbooi, Mahlatse SThis research will fill the gap in the availability of hate speech detection models that are inclusive of low-resource languages with a focus on isiZulu. Recent Natural Language Processing studies have concentrated more on high-resource languages, and as a result, languages like isiZulu have been under-represented in this field of research. To bridge this gap, an annotated English dataset has been utilized, leveraging Google translation API to translate the data into isiZulu. We made use of semantic analysis to look for patterns within the labeled categories of data. Without using the categories as target variables in the training process, However, this classification by domain allowed this study to delve into offensive terms with respect to prevalence and context. Machine learning models, such as Support Vector Machines and Random Forests, were trained using TF-IDF (Term Frequency – Inverse Document Frequency) vectorization to achieve state-of-the-art accuracy improvements in Zulu hate speech detection. These results show that a targeted semantically enriched approach that we used in this study enhances the precision of the model, which holds great potential for fine-tuned hate speech detection across multilingual contexts. This research contributes to expanding effective NLP tools for low-resource languages, promoting safer and more inclusive digital spaces.Item Assessment of material parameters of curly hair fiber subjected to axial loading(2024-09) Mathebela, Lebogang; Sigwadi, R; Lekwana, T; Ngwangwa, H; Pandelani, T; Nemavhola, F; Modungwa, Dithoto MThe mechanical properties of hair fiber vary with anatomical variation and for different ethnic groups. The development of material parameters is essential for accurate development of computational models for detailed studying of hair fiber behaviour. Therefore, the aim of this study is to determine the material parameters in hair mechanics using selected hyperelastic constitutive models. The nonlinear hyperelastic constitutive models were used to estimate the material parameters of the single curly hair fiber from the stress-strain nonlinear curve obtained from the tensile raw data subjected to axial loading. The three best fitting nonlinear hyperelastic model were selected, namely Polynomial Model, Mooney-Rivlin Model and Yeoh Model based. Several parameters and statistical measures including C10, R, R2, NE, NRMSE, SA and SE were utilised to compare the performance of each hyperelastic model. Each model was assessed using different strain rate of 100 and 10-2. s-1 strain rate. The best model that captures the axial tensile behaviour is Polynomial model with the R2 parameter higher than 91 % followed by MooneyRivlin model 85 % then Yeoh model at 83 %.The material parameters could be useful in the development of computational models for studying the effect of environment on human. The model with the best fit is polynomial model followed by Mooney-Rivlin then Yeoh model. The best model was selected based on the Evaluation Index [EI] and the study suggest that the new model can be introduced into future studies to check the overall performance of the model not only by assessing the model based on the R2 parameter but also to ensure that the objective function of the model is within the required limit.Item Assessment of modern training innovations for supervisors and trainers in the South African Mining Sector(2025-06) Van Schoor, Abraham M; De Kock, N; Khan, Sumaya; Müller, R; Van Rensburg, R; Govindasamy, K; Botha, W; Maphalala, Busisiwe V; Mpofu, Mvikel; Pelders, Jodi L; Ramparsad, SAs the mining industry modernises, skills development needs to be a priority. The aim of this paper was to develop guidelines that consider the assessment of modern training solutions for supervisors and trainers in modern mining. A literature review was conducted on best-practice criteria for the evaluation of modern mining upskilling and reskilling solutions. A draft evaluation matrix was developed based on the literature review insights and incorporated 48 best-practice criteria for the assessment of training solutions. The assessment instrument was applied to training curriculums for supervisors and trainers of two participant entities. Data gathering included assessments of the training solutions, and an industry panel review process. Strengths, weaknesses, opportunities, threats, and gap analyses were undertaken. Key insights were revealed for the respective training solutions. Recommendations included continuous review and improvement of the curriculum for alignment to mining modernisation skills needs, including consideration of modern training methodologies and facilitation; revised content and assessments; skills and training required for modernisation; tracking of graduates and learner feedback; better alignment with modernisation objectives and industry skills needs, increased focus on safety and risk assessment and control; and more immersive learning experiences. While sample training innovations were selected for evaluation, the recommendations remain relevant for training entities looking to align to their curriculum to mining modernisation skills needs and industry skills demands for modernisation.Item Exploratory analysis of modified deep learning model for potholes data augmentation(2025-04) Adebiyi, RF; Bello-Salau, H; Onumanyi, Adeiza J; Adebiyi, BH; Adekale, AD; Bello-Salahuddeen, RA major factor contributing factor resulting to a large proportion of vehicular-related traffic accidents in developing nations is the poor condition of road networks, characterized by potholes, bumps, and other anomalies. Despite efforts by authorities to address these issues, they persist. A new approach involves equipping vehicles with sensors to detect road anomalies, enabling drivers to make informed decisions. Various models using road surface images to detect and classify these anomalies have been proposed, with recent methods leveraging deep learning. The effectiveness of these models depends on the presence of abundant and well-labelled training datasets. To address this need, a modified Deep Denoising Diffusion Probabilistic Model (mDDPM) was proposed, enhancing the U-Net backbone architecture to improve the original DDPM's performance in augmenting pothole images. The mDDPM generates more diverse augmented images, evaluated through subjective and objective assessments, including the Fréchet Inception Distance (FID) score. Experimental results showed that 98% of participants could not distinguish between real and synthetic images, classifying the augmented images as real. Additionally, an FID score of 0.52 indicated that the augmented images closely resemble real pothole images. This demonstrates the model's effectiveness in generating training data for deep learning models aimed at road anomaly detection and classification, contributing to the development of robust models for detecting and classifying potholes and other road anomalies.Item Requirements management of the Auckland City Rail Link Project - An update(2025-08) Dennehy, S; McMullan, R; Meyer, Isabella AThe Auckland City Rail Link (CRL) project was the first major transport infrastructure project in New Zealand to adopt a requirements-led approach. This paper reflects on lessons learned, with the aim of identifying learnings that may inform approaches in other complex, real-world transport infrastructure projects. The recommendations do not cover the full scope of requirements management or systems engineering, rather they are selected aspects that presented particular challenges or successes, including project planning, tool choice and configuration, establishing the requirements management database as the single, trusted source of truth, change control, stage gates, levels of abstraction, interface management, reliability/availability/maintainability/safety, and post-construction verification. This work is limited to the analysis of a single case, with recommendations made from the perspective of the requirements management team rather than from a wider range of disciplines. Notwithstanding these limitations, the insights and recommendations should be considered for future infrastructure projects, in conjunction with experience and lessons learned from other projects, and with appropriate tailoring for project context – which is always unique.Item Correlating defects in electroluminescence images to photovoltaic module power loss using DeepLabV3 for semantic segmentation(2025-07) Pratt, Lawrence E; May, Siyasanga I; Mkasi, Hlaluku WThis article investigates a multi-class semantic segmentation model for correlating defects in electroluminescence (EL) images and PV module power degradation in fielded modules. The case study uses machine learning and statistical methods to identify a potential root cause for power degradation in 680 PV modules sampled from a multi-MW PV plant in South Africa. The EL image and the electrical performance were measured and recorded for each module at the CSIR PV Module Quality and Reliability Lab. A deep-learning model previously trained for multi-class defect detection in EL images of solar cells was re-trained to include 24 samples of a new ‘brightening’ defect. The updated model was used to classify each pixel on 48,960 solar cells into one of 29 classes. Based on an exponential decay model, the degree of ‘brightening’ defect averaged across all cells in a module correlated to the module fill factor (r2 = 0.42), indicating that 42% of the variability in Fill Factor (FF) measured for this set of modules could be explained by the brightening defect. While the strength of the correlation is moderate, it may provide some insight into the root cause of the degradation in FF. The authors speculate that the brightening defect is a signature of solar cells with non-uniform current distribution resulting from ageing. Updates to the multi-class prediction model were relatively simple and fast, enabling a similar analysis for new defects in PV modules as they emerge.Item Exploring the role of dew computing in enhancing cybersecurity and digital forensics: A systematic literature review(2025-06) Nelufule, NthatheniThe Fourth Industrial Revolution (4IR) has precipitated the emergence of sophisticated cyberthreats, which evolve with time. The evolution of these threats demands innovative solutions applied in cybersecurity and digital forensics to improve data security and incidence response. This paper presents a systematic literature survey conducted on dew computing architectures, which integrates local computing capabilities with cloud resources to create a hybrid framework that enhances data accessibility, data security, and digital forensic analysis. By allowing devices to operate independently of constant internet connectivity, dew computing addresses significant challenges faced by traditional cloud computing architectures, such as latency and data vulnerability during transmission. The main objective of this paper was to study how dew computing architectures can enhance real-time data processing, data protection, data integrity, and enhance digital evidence acquisition. Furthermore, this paper discusses the challenges and limitations associated with implementing dew computing, including technical barriers and privacy concerns. The research survey findings suggest that dew computing offers a promising approach to mitigating cybersecurity risks and provides digital forensic investigators with the tools necessary for effective investigations.Item Towards a circular economy in mobile communications technology: A systematic review(2025) Ebrahim, Rozeena; Burger, Chris R; Masonta, Moshe T; Sikrenya, Siyanda KS; Hlatshwayo, Ronald OThis paper presents the findings of a systematic literature review examining the relationship between the circular economy (CE) and terrestrial mobile networks. This review identifies key approaches to address a CE from 14 highquality academic papers and classifies them according to four categories: infrastructure reduction and hardware optimisation; reduced energy demand; reduced reliance on fossil fuel-based power grids; and recycling. These approaches were also mapped to the “3R” framework which reveals a strong emphasis on reducing resource consumption and reusing resources, with minimal attention to recycling. The study emphasises the need for a holistic CE approach addressing all stages of the product lifecycle and calls for expanded research and development efforts, particularly in developing economies. Collaborative initiatives between academia, industry, and policymakers are recommended to promote sustainable and comprehensive CE practices in mobile telecommunications.Item Power Consumption Analysis of a 5G NR Base Transceiver Station Using a Custom Measurement Setup(2025-07) Ebrahim, Rozeena; Vilakazi, Mlamuli C; Mabunda, Ntsako D; Makaleng, Koketso F; Sikrenya, Siyanda KS; Lysko, Albert A; Mfupe, Luzango PAs Fifth Generation (5G) and future mobile networks continue to expand, understanding the power consumption of the base transceiver station (BTS) is necessary for improving the energy efficiency of the Radio Access Network (RAN). The power consumption of the BTS is influenced by factors such as its operating mode, transmission power and traffic levels. This work has explored the power consumption of an outdoor commercial 5G New Radio (NR) BTS using an inexpensive and custom-built power measurement setup. Indoor testing was done at a reduced transmit power level to analyse the power consumption under different operation modes. The measurements presented a clear correlation between activity levels and energy usage. Insights have been provided into the power consumption requirements of outdoor BTSs compared to 5G setups deployed on an in-house, indoor mobile network testbed, including the trade-off between energy efficiency and operational stability. Further, these findings have highlighted areas for future research, such as the impact of deployment environments, transmit power levels and the number of connected devices on energy efficiency. Understanding these factors is considered essential for designing, building and deploying more sustainable mobile networks for the future.Item A satellite-based decision support service for the marine fisheries and aquaculture industry of southern Africa(2025-09) Smith, Marié E; Molapo, Nkadimeng R; Ngulube, Mabuela; Sibolla, Bolelang, H; Vhengani, Lufuno MMarine Aquaculture in South Africa includes abalone, mussels, oysters, and finfish. Depending on the cultured organisms these could include land-based pump-ashore operations in-water cages or rafts. Each method of operation has different environmental risk.Item Future trends in AI for cybersecurity and digital forensics: A systematic literature survey(2025-06) Nelufule, Nthatheni NThe Fourth Industrial Revolution has brought many opportunities, including the integration of artificial intelligence technologies into cybersecurity and digital forensics. These integrations represent a transformative change in how organizations protect their digital assets and investigate their cybersecurity incidents. As cyber threats become increasingly sophisticated, traditional cybersecurity measures often fail, necessitating the adoption of advanced AIdriven solutions. This paper presents a systematic literature survey that explored future trends in artificial intelligence applications in these critical domains, focusing on their potential to improve threat detection, automate incident response, and improve the efficiency of forensic investigations. The survey has identified some key challenges associated with the deployment of Artificial Intelligence technologies, including ethical considerations, data privacy issues, and the complexities of integration into existing systems. The findings from this survey paper have revealed a growing reliance on artificial intelligence for real-time threat detection and response, highlighting its effectiveness in identifying anomalies and predicting potential breaches before they escalate into significant incidents.Item Competency-based training for mine emergency response(2025-06) Lange, Pieter; Bergh, Adriaan V; Pelders, Jodi L; Khan, SumayaMine workers are exposed to hazards that can cause injuries or fatalities, including fires, underground explosions, irrespirable, falls of ground and mobile machinery. Emergency preparedness is important for improved safety outcomes and includes the deployment of self-contained self-rescuer (SCSR devices, effective escape routes, and adequately located refuge bays. The need has been identified for improved training solutions or mine worker escape situations, which should provide some exposure to the stressors that would be experienced. The CSIR mining Cluster has developed innovative multimodal competency-based training modules to improve the emergency response of mine employess. The modules include interactive e-learning, virtual reality training, SCSR donning and breathing simulation, and competency-based assessments. A pilot study was successfully completed with participation from a major coal mining operation. The competency-based approach improves the overall efficiency of the training and is especially well-suited to training for high-consequence, low-frequency scenarios.Item Understanding the risk: Mapping deepfake cyberattacks to a temporal attack model(2025-07) Pieterse, HeloiseThe advancement of artificial intelligence (AI) technologies has become a trending topic in the cybersecurity domain. These technologies, however, present cybersecurity with a double-edged sword as AI offers enhanced threat detection and protection, but also enables cybercriminals to craft sophisticated cyberattacks. Deepfakes, which are a form of digitally manipulated synthesised media created using deep learning techniques, have garnered widespread attention due to the use of deepfakes in cyberattacks to cause influence, spread disinformation, or conduct fraudulent activities. While extensive research efforts have been undertaken to develop defences against deepfakes, the solutions are technical and not easily accessible. Innovative strategies are required to equip personnel from government, academia, and the business sector with the fundamental knowledge to detect and defend against cyberattacks employing deepfake technology. This paper evaluates the most significant events involving deepfakes since the emergence of the technology in November 2017. Key trends and characteristics are identified and mapped to a temporal attack model to separate the different stages of a cyberattack involving deepfakes. The outcome is a Deepfake Attack Framework that offers valuable insights essential to understanding the risks associated with deepfakes. The Deepfake Attack Framework presents a theoretical solution that can be applied practically to minimise risk and enable personnel to be better prepared to defend against deepfake-driven cyberattacks.Item A case for high capacity coal trucks to reduce costs and emissions at Eskom(2021-07) De Saxe, Christopher; Van Eeden, J; Steenkamp, A; Mokone, OlwethuSouth Africa’s national power utility, Eskom, is under heavy strain to maintain an undisrupted electricity supply and contain costs, while at the same time reducing its environmental impact. In 2018/19, Eskom acquired 118 Mt of coal, at a purchase cost of approximately R 47 billion, of which around R7 billion (15%) can be attributed to the transport of coal via conveyor, rail and road. Eskom has been unable to meet its road-torail modal shift targets, and so road haulage still accounts for around 30% of coal deliveries. The “Smart Truck” or “PBS” demonstration project in South Africa has shown how an innovative approach to truck design and regulation can drastically improve the efficiency of road haulage, reducing the cost per tonne-km, while reducing emissions and improving safety. An existing Smart Truck trial in coal transport has demonstrated a 15% reduction in fuel use and associated carbon emissions per tonne-km, which translates into an approximate 6% reduction in total road transport costs. This was achieved through the introduction of innovative 74-tonne tridem interlink truck combinations, which has resulted in fewer truck trips and reduced costs for the same haulage task. At the same time, the trucks are more road friendly due to additional axles and fewer truck trips, and the trucks are designed to be inherently safer than the conventional coal interlinks currently in use. In this paper, we benchmark the costs and emissions of Eskom’s current road haulage coal supply operations in South Africa, and calculate the potential savings from migrating to 74tonne interlink PBS truck combinations. We demonstrate potential savings of R 120 million and 35 000 tonnes of CO2 per year, while removing 300 000 truck trips from the roads.Item Classification of trucks using camera data(2021-07) Mokone, Olwethu; De Saxe, ChristopherUnderstanding the precise movements of different commodities on South African roads can help in not only describing the logistics sector more accurately, but also in the planning of road infrastructure maintenance and investment. Truck combinations can be classified into several classes broadly associated with different commodity groups, including tautliners, tankers, flatbeds (general freight) and flatbed (containerised freight). Current truck classification systems in South Africa can classify trucks by number of axles and vehicle mass but are unable to determine the combination type and hence commodity group. Video data allows for truck combinations to be classified in more detail using image-based classifiers. The latest developments in deep learning algorithms have made it possible for accurate classification of vehicle types using camera data. A CCTV camera feed of a section of the N3 was provided by the South African National Roads Agency Limited (SANRAL) and was used as a case study to develop a proof-of-concept classifier for tautliner and tanker truck combinations, using a transfer learning approach and the pretrained ResNet50 classifier. The results indicate good accuracy based on relatively small datasets. Future work will focus on further optimisation and investigating the training dataset requirements in more detail.Item Accelerated Pavement Testing (APT) of a test section surfaced with an asphalt wearing coarse containing plastic waste incorporated using the ‘wet method’(2025-07) Smit, Michelle A; Rust, FC; Mturi, Goerges; Mokoena, Refiloe; Ntombela, R; Marais, HermanThe incorporation of plastic waste in road pavement materials presents a promising opportunity for sustainable infrastructure development. In South Africa, introducing any innovation requires compliance with national performance criteria and guidance from mechanistic-empirical design methods. This study evaluated the rutting resistance performance of a road pavement section surfaced with plastic waste modified asphalt (PWMA) produced via the wet method – where plastic waste is integrated into the bituminous binder before mixing. An Accelerated Pavement Testing (APT) program was adopted for the permanent deformation testing of a coarse continuously-graded asphalt wearing course modified with plastic waste. The PWMA was produced with post-consumer recycled plastic waste and also incorporated a Reactive Elastomeric Terpolymer (RET). Test sections were constructed in Gauteng, South Africa, comprising a reference asphalt (based on a standard unmodified bitumen used in South Africa) structure and a PWMA layer. Both sections were subjected to simulated traffic loading using a Heavy Vehicle Simulator (HVS) at speeds of 12km/h, varying wheel loads (40, 60 and 80 kN dual wheel load), and controlled temperatures reflective of local pavement conditions. Performance monitoring involved surface and embedded measurement tools, including Road Surface Deflectometer (RSD), Multi Depth Deflectometer (MDD), standard straight edge, thermocouples and temperature buttons. After 2.9 million equivalent standard axles (ESALs) of HVS loading, the PWMA section demonstrated enhanced rutting resistance, with an average rut depth of 7.2 mm, compared to 10.4 mm for the reference section, which reached a maximum rut of 12mm.These results align with laboratory findings, confirming that the addition of plastic waste increases the structural integrity of asphalt layers by enhancing resistance to permanent deformation. This study supports the potential for adopting PWMA in South African road infrastructure to meet national performance standards and sustainability goals.