Browse
Recent Submissions
Item Risk of microplastics in South African freshwater: Case study(2025-09) Komane, Wesley K; Moalusi-Mathye, Salamina M; Moloi, M; Tancu, Yolanda; Lehutso, Raisibe FMicroplastics (MPs) are polymeric materials of size below five millimeters (<5 mm). Anthropogenic activities have led to the abundance of MPs detected in all environmental compartments, indicating their ubiquity nature. Their small size and persistent nature allow them to disperse easily, leading to a growing environmental concern. Of particular, the interaction and/or accumulation of MPs with aquatic organisms has raised concerns since some organisms ingest MPs and can induce adverse effects such as disrupting digestive systems and reducing organisms’ growth and reproduction (Ding et al., 2021). While the potential hazard exposure posed by MPs are somewhat known, their pollution and risk extent in African water systems are unknown. Due to limited experimental data, models are used to predict the risk potential factoring in various MPs properties and exposure extent. Using a suite of complimentary models, the study estimate the MP pollution extent and risk.Item Fine-tuning a machine learning model to detect COVID-19 misinformation in Xitsonga(2025-05) Baloyi, MG; Rananga, S; Mbooi, Mahlatse SThis paper explores the effectiveness of models, specifically Multilingual Bidirectional Encoder Representations from Transformers (mBERT) and Random Forest, for detecting misinformation about Coronavirus Disease 2019 (COVID-19) in Xitsonga, a South African language. The focus is on evaluating how synonym replacement and translation techniques can enhance model performance for misinformation detection by comparing the mBERT and Random Forest models in terms of accuracy, precision, recall, and F1 score. The continuous spread of misinformation during the pandemic COVID-19 has highlighted the need for accurate and efficient classification systems, especially in low-resource languages. To address this challenge, we enhanced an English news dataset using two data augmentation techniques: synonym replacement and translation into Xitsonga using the Marian Machine Translation (MarianMT) model. We fine-tuned mBERT for sequence classification and compare its performance with Random Forest classification model using Term Frequency-Inverse Document Frequency (TF-IDF) features. The results demonstrate that mBERT achieves good accuracy, precision, recall, and F1 score compared to Random Forest. This paper contributes to the field of misinformation detection by introducing a multilingual approach and improving classification performance in low-resource languages.Item Biomass concrete for low carbon buildings: A circular economy case study in South Africa(2025-06) Menardo, A; Van Reenen, Coralie A; Lamb, A; Nduna, G; Mafuwe, P; Van Mele, T; Block, PThis paper presents a case study of a low carbon affordable building in South Africa, utilizing innovative biomass concrete for bricks, funicular floor and roof modules that are optimized in terms of geometry and mix design to reduce their global warming potential (GWP). The objective is to address the urgent need for more sustainable construction practices and to demonstrate circular economy principles by incorporating chipped biomass from invasive alien plants (IAPs) as a supplementary ingredient in the concrete mix, leading to what has been labelled nonCrete (NC). This approach not only mitigates the environmental impact of traditional concrete production but also contributes to the management and eradication of IAPs in the region, which affect the natural water cycle, compete with local flora and fuel frequent wildfires. The focus of this paper is to demonstrate the potential environmental and social impact of NC use in construction in real-world applications through a case study. The design of the floor system, supported by rigorous material tests leading to its accreditation on the local market, indicates that the use of NC has the potential to lower greenhouse gas emissions (GHG) associated with construction, while still fulfilling all the necessary engineering standards and facilitates local job creation. This work is intended to serve as a pivotal step towards more sustainable building methodologies in South Africa and demonstrate circular economy principles in practice.Item Micro-doppler classification of humans and animals using FMCW Radar(2025) Manga, Amisha; Pain, S; Taylor, J-PThis research attempts to address the issue of animal poaching by exploring human and animal classification by making use of micro-Doppler data generated by a Frequency-Modulated Continuous Wave (FMCW) radar. Raw data was collected of human and animal species including dogs, horses and cows. Signal processing techniques such as creating range-Doppler and Constant False Alarm Rate (CFAR) maps for detection and using Short-Time Fourier Transform (STFT) for spectrogram generation were applied. Principal Component Analysis (PCA) was used for data reduction. The dataset was classified and evaluated across various target class configurations, comparing the full dataset and its PCA-reduced versions, using Convolutional Neural Network (CNN), k-Nearest Neighbor (kNN), Random Forest (RF) and Support Vector Machine (SVM) models. Following this, a two-stage classification process was implemented. In the first stage, the 4 classifiers were used to distinguish between human and animal. In the second stage, these classifiers differentiated among the specific animal species. The SVM-SVM achieved the highest accuracy at 97.66%, closely matching the 97.50% from the multi-class classification.Item PS-Aware OSD – An energy-efficient paging approach for 5G-enabled industrial IoT devices(2025-09) Ogbodo, EU; Jasperneite, J; Mendes, LL; Neumann, A; Kurien, A; Abu-Mahfouz, Adnan MIWith the increasing adoption of 5G networks in smart cities and Industry 4.0 applications, energy efficiency (EE) has become a critical concern, particularly for industrial Internet of Things (IIoT) devices that operate continuously in latency-sensitive environments. This paper presents a novel Paging Signal-Aware Ordered Statistical Decoding (PS Aware OSD) algorithm to optimize energy use during paging operations. By intelligently predicting paging occasions (PO) based on historical patterns, the algorithm reduces unnecessary wake-ups in extended discontinuous reception (eDRX) cycles. Simulations demonstrate that this approach improves energy efficiency by up to 40%, reduces block error rate (BLER), and improves throughput and latency compared to standard third generation partnership project (3GPP) paging techniques. The paper discusses practical implementation aspects, acknowledges system limitations, and proposes future work, including the integration of machine learning for dynamic paging optimization and the development of security measures against spoofing or unnecessary battery drain.Item Designing a resilient and sustainable science centre for rural education: A case study of Cofimvaba, South Africa(2025-07) Van Reenen, Coralie A; Van Reenen, Tobias H; Bole, SheldonThis paper highlights lessons learnt in the design, construction and operation of a science centre using innovative design principles and technologies to achieve a resilient and sustainable building in a rural community. This case study research compares the design intent of the building with the operational data and practices. The innovative technologies and design are found to successfully achieve indoor comfort levels and provide off-grid services, however, limited operational expertise result in sub-optimal performance in some aspects. The research is a case study and the findings may not be generalisable to all types of buildings and contexts. Further case studies of similar projects in future can help develop a better understanding of the benefits and limitation of innovative design and technologies for different building types and in different contexts, such as in urban areas. Innovative technologies can be successfully implement, however an extended hand-over period and in-depth training is recommended to be included in the project scope. Continuous data collection is recommended for monitoring and evaluation purposes.Item Effectiveness of the Rosahl Micro–Dehumidifier for humidity management in camera housings(2025-12) Seletani, Rofhiwa; Baloyi, Andrew AManaging humidity is crucial in the operation of electronic and electro-optical products. Excessive moisture buildup inside the enclosures of electro-optical devices, such as cameras and telescopes, can lead to system malfunctions and eventual mission failures. Thus, incorporating humidity control measures is crucial throughout the design, manufacturing, and operational phases of electro-optical systems. There is increasing interest in utilising micro-dehumidifiers for humidity control due to their compact size, energy efficiency, environmentally friendly nature, cost-effectiveness, and silent operation. However, most recent studies have focused on developing prototype devices. As a result, there is a lack of research examining the performance of commercially available microdehumidifiers. In this paper, the effectiveness of commercially available Rosahl MicroDehumidifiers in managing humidity in enclosed camera housings was investigated. The paper further investigated the leak rate of the Dehumidifier’s Solid Polymer Electrolyte (SPE) membrane to Nitrogen and Argon gases under 5%, 10% and 15% above atmospheric pressures. The latter investigation was conducted to establish whether the Rosahl MicroDehumidifiers can be utilised in pressurised electro-optical housings. An Experimental Test Rig was designed by modifying a commercially off-the-shelf housing to integrate a Rosahl Micro-Dehumidifier and fit temperature, humidity, and pressure sensors. The findings indicated that Rosahl Micro-Dehumidifiers effectively decrease humidity within enclosed housings. However, the rate at which humidity is removed depends on the humidity gradient and the time of day. At higher humidity levels, the humidity removal rate was found to be strongly dependent on humidity concentration between the interior and exterior of the housing, i.e., the higher the humidity inside the housing, the higher the removal rate. It was also observed that the humidity removal rate varied based on the time of day. The humidity removal rate was higher between Late Morning and Late Afternoon periods and lower between Late Afternoon and Early Morning. These findings could be attributed to the differences in vapour pressures of the moisture within the housing between those time intervals, i.e. the vapour pressure is higher between Late Morning and Late Afternoon, whereas it is lower between Late Afternoon and Early Morning. The results from the leak rate investigation suggested that the SPE membrane of the Rosahl Micro-Dehumidifier has a high leak rate to both Nitrogen and Argon under pressurised conditions. However, the leak rate of Nitrogen was slightly higher than that of Argon under the same pressure conditions. The higher leak rate of Nitrogen could be attributed to its smaller molecular size when compared to Argon. The Rosahl Micro-Dehumidifiers proved to be effective in removing humidity from sealed housings. However, they are not suitable for use in pressurised housings where pressure shall be maintained above atmospheric pressure.Item Shooter bearing estimation for airborne hostile fire indication using networked low-complexity radar sensors and machine learning(2025-10) De Witt, Josias J; Nel, Willem ALow-flying helicopters are vulnerable to small arms fire, highlighting the need for Hostile Fire Indication (HFI) systems that can detect bullet threats and estimate shooter bearing to support evasive action or counter measures. Acoustic sensors have traditionally been used for shot detection but are limited in airborne environments due to noise, atmospheric variability, and weather sensitivity. Radar offers an attractive, all-weather alternative, but conventional angle-of-arrival (AoA) estimation requires complex, multi-channel sensors. This paper proposes a novel alternative approach using only low-complexity, single-channel FMCW radar sensors mounted around the airborne platform. Although these sensors cannot measure AoA individually, combining their bullet detection data with machine learning enables shooter bearing estimation. A Random Forest regressor, trained on simulated bullet trajectories from AK47 and AR15 rifles, achieves shooter bearing estimation RMSE below 4° for miss distances under 20 m, with useful accuracy (RMSE<15°) up to 40 m miss distance. The approach enables a low-complexity, cost-effective, radar-based HFI solution for airborne platforms.Item Combined approaches for cybersecurity awareness training in South Africa(2025-08) Veerasamy, Namosha; Khan, Z; Mahlasela, Oyena N; Badenhorst, Danielle P; Mashali, Mamello L; Ntshangase, Ntomfuthi LTraditional cybersecurity awareness training typically consists of classroom environments with lecture style learning. However, a more transformative approach is required as to how we perceive and implement cybersecurity awareness. Seamless integration into digital and real-world scenarios can help learners shift their behaviour and knowledge base. This paper presents a paradigm shift to more innovative and reimagined styles towards cybersecurity awareness. The authors present the demand for skills development, the state of digital literacy, unique local needs and engagement strategies to facilitate a more novel approach to cybersecurity awareness within South Africa by incorporating a combination of approaches like simulated content, different forms of learning, micro-learning, gamification, intelligent classrooms, interactive role plays, long reads, podcasts and screencasts.Item Acceptance of synthetic speech in South African languages: A comparative study of Afrikaans, isiZulu, and Sepedi in healthcare contexts(2025-11) Louw, Johannes A; Wilken, IlanaWhile text-to-speech technologies have made significant advances in recent years, ques tions remain about how synthesised speech is accepted in culturally and linguistically di verse settings such as South Africa. This study explores how South Africans perceive synthetic speech in comparison to human recorded speech across three official languages: Afrikaans, isiZulu, and Sepedi, with healthcare as the application context. Using a blind and randomised listening test, 65 participants rated audio prompts across four acceptance metrics: trust, knowledgeability, lik ability, and relatability. Statistical analysis us ing the Wilcoxon signed-rank test revealed no significant difference between natural and syn thesised speech perception among Afrikaans speakers. However, low participation rates pre vented meaningful analysis of speech percep tion for isiZulu and Sepedi speakers. When combining data from all participants, a medium effect size favouring natural speech was ob served, though this difference was not statisti cally significant. These findings suggest that synthetic speech adapted from natural recordings may be suit able for certain applications in South Africa, though larger and more linguistically represen tative samples are needed to confirm these re sults.Item B31D-09: Phytoplankton community composition in inland waters from airborne and satellite hyperspectral data(2025-12) Sharp, SL; O’Shea, RE; Guild, LS; Cortés, A; Forrest, AL; Kravitz, J; Lain, L; Mpapane, S; Mudzielwana, R; Smith, Marie EThis presentation covers the basis of aquatic food webs, loss of biodiversity and the negative impact on the ecosystem.Item Photonic-biosensing towards drug-resistant tuberculosis diagnosis(2025-07) Chauke, Sipho H; Tjale, Mabotse A; Maphanga, Charles P; Dube, F; Ombinda-Lemboumba, Saturnin; Mthunzi-Kufa, PEarly detection and treatment of tuberculosis (TB) remain key strategies to reduce transmission and disease progression. However, this is hampered by time-consuming, insensitive diagnostic methods, particularly for the detection of drug-resistant forms and in patients with human immunodeficiency virus infection (HIV). Several genes, such as the RpoB and InhA genes, contain mutations that are responsible for drug resistance. This study aimed to use an SPR-based biosensor platform to detect RpoB and InhA genes. DNA probes, specific to RpoB and InhA, were used as biorecognition elements to capture the corresponding target DNA sequences. The RpoB and InhA gene-specific thiolated DNA probes were immobilized on a gold-coated glass substrate before the target DNA was introduced for detection. As a negative control, a non-specific target to both genes was used to confirm the binding of the specific target. The shifts in the resonance angles indicated the binding properties associated with DNA hybridization between the specific target and the capture probe. The results obtained from this study demonstrated the use of a simple SPR setup and its potential for identifying genes associated with drug-resistant TB.Item Assessment of the interventions required to enable participation of South African SMMEs in renewable energy Global Value Chains(2025-10) Rakaibe, Tshwanelo K; Mbam, VuyoIn the second quarter of 2022, the then Small Enterprise Development Agency (SEDA) reported that small, micro, and medium enterprises (SMMEs) accounted for 59% of total employment in South Africa. Some of these enterprises are attempting to, and or aspire to, localise the production and assembly of components in the renewable energy value chains given the increasing demand for renewable energy in South Africa, which can contribute to the reduction of unemployment, poverty and inequality. However, they do so in the context of these components being manufactured in global value chains dominated by east-Asian countries, particularly China. This paper provides a detailed analysis of how South African SMMEs can participate in the renewable energy global value chains (GVCs), highlights the benefits that SMMEs can derive by participating in GVCs and the challenges that they, in general, encounter when trying to participate in GVCs. The insights presented herein are drawn from a review of international literature and a survey questionnaire that was sent out to South African SMMEs involved in the renewable energy industry. Drawing from the literature review and SMME survey responses, a case is made for why it would be beneficial for South African SMMEs to be integrated into renewable energy GVCs and, most importantly, what actions would need to be taken by both the SMMEs and the government - whose responsibilities would include, among others, creating an enabling and supportive environment for SMME integration into GVCs. The interventions recommended in this paper are aligned with the South African Renewable Energy Masterplan (SAREM) vision of industrialising the renewable energy value chain in South Africa to enable participation in the energy transition, serve the needs of society and contribute to economic revival. Furthermore, these interventions relate directly to the SAREMs objectives, such as “growing the industrial capacity in the renewable energy and battery storage value chain”; “building the capabilities needed for the industry”, and “contributing to a just energy transition” in which SMMEs could play an integral role in democratising the energy sector, creating employment and contributing to economic growth.Item Bootstrapping Siswati lexical resources from isiZulu(2025) Wierenga, Roné; Marais, Laurette; Wilken, IlanaIsiZulu and Siswati are closely related languages that share significant morphosyntactic characteristics. Systematic differences between these languages have been identified at the phonological and morphosyntactic levels. Due to the resource-scarce status of these languages, this similarity has led to bootstrapping of computational language resources at the morphological and syntactic levels. In this work, we investigate the feasibility of adapting lexical items in a computational lexicon from isiZulu to Siswati. We use Grammatical Framework resource grammars for both languages to analyse and transform lexical items, which are then evaluated against a parallel term list. An iterative process yields a success rate of 70.5%, indicating that this approach is largely viable as a means of significantly reducing the manual effort needed to develop lexicons for computational resources for Siswati.Item Impacts of climate change on long-term wind forecasting(2025-10) Rapotu, Dimakatso R; Masoga, Mandla A; Karamanski, StefanWind forecasting is a critical tool for understanding the long-term capacity of wind energy. As global temperatures rise due to climate change, it is essential to understand how these changes will impact future wind generation. This study presents a physics-based approach to forecasting the effects of climate change on wind energy in South Africa. The analysis leveraged five different global climate models, all using the Representative Concentration Pathway 8.5 (RCP 8.5) scenario, which represents a high-emission future. These climate models provided forecasted wind data from 1960 to 2100 for a specific location. This data was then used in a four-step long-term wind forecasting model. The model calculates wind energy production based on specific turbine characteristics, adjusts for air density and hub height, and then aggregates the total energy output for all turbines available at the site. The model was tested on 35 onshore wind farms in South Africa and validated using power output predictions for [2012,2016,2020,2024] using Sere and Amakhala data acquired from Renewables ninja. The results showed a Pearson correlation of 0.99 for both sites, demonstrating the high reliability of the proposed forecasting method. This approach provides a robust way to assess the long-term viability of wind energy projects under future climate conditions.Item Urban Air Mobility: Regulatory Pathways and Readiness for Integration in South Africa(2025-07) Ndlovu, Hlamulo P; Niken, Adrian; Madonsela, N; Teane, Tshegofatso O; Moodley, TheolanUrban Air Mobility (UAM) refers to the use of small, electric or hybrid-electric aircraft, often highly automated, designed to transport passengers or goods at low altitudes within urban and suburban environments, which has been developed in response to traffic congestion in the cities. This presentation highlights work that forms part of the International Forum for Aviation Research (IFAR) project titled “Navigating the Skies: A Guide to Certification for Urban Air Mobility”.Item Protection motivation theory (PMT) as a driving force for cybersecurity awareness(2025-07) Veerasamy, Namosha; Mashiane, Charmaine TDespite the use of various technical controls and technologies, a pivotal weakness in Information Communication Technology (ICT) security remains the human element. Cybersecurity awareness training can be used to influence and change behaviour towards the adoption of secure cyber practices. It is imperative to reduce cybersecurity risks through awareness and cybersecurity training to influence safer behaviour. Various behaviour and motivation theories exist, which attempt to explain how behaviour can be influenced. This paper focuses on the benefits of Protection Motivation Theory in the delivery of security awareness topics. Examples of stakeholders that can be involved in cyber security awareness training are mapped onto the protection motivation theory. This helps to show how PMT can be used in different user environments to effectively communicate cybersecurity awareness at the correct level. This paper proposes the usefulness of applying PMT to the cyber security awareness domain to create a culture of cybersecurity awareness and promote cybersecurity education and awareness.Item Tackling South Africa’s literacy crisis with local technologies(2025-06) Wilken, Ilana; Marais, LauretteThere are 12 official languages in South Africa, and a recent international study sent shockwaves through the country’s education community by revealing that more than eight out of ten Grade 4 learners cannot read for basic meaning in their home language. Without this essential skill, South African learners are deprived of the opportunity to fulfil their true potential, with the impact being the most devastating for those from disadvantaged communities. With many complex factors contributing to this result, the need for a wide range of solutions is paramount to addressing the various dimensions of this crisis. With this in mind, we embarked on a research and development project called Ngiyaqonda (isiZulu for “I understand”), in which speech and text technology for South African languages are harnessed to enhance home language literacy. These technologies were integrated into an Android application designed for Grade 3 learners, who will transition to English as their language of learning and teaching (LOLT) in Grade 4 after receiving instruction in their home language from ages six to nine during the foundation phase. In this work, we will describe the design and development of the application and the pilots we conducted. Hereby, we aim to provide insights into the literacy crisis South Africa faces and how a small intervention can make a difference to the lives of learners and educators.Item IAM-based zero trust architecture for IoT: Securing non-human identities in a connected world(2025-11) Mthethwa, S; Jembere, E; Dlamini, Thandokuhle MThe widespread adoption of IoT devices has fundamentally altered digital connectivity, facilitating real-time data exchange and autonomous interactions worldwide. While this transformation offers significant operational advantages, it also introduces critical security challenges, particularly concerning the Identity and Access Management (IAM) of non-human identities such as sensors, devices, machine agents, and service accounts. Traditional perimeter-based security models, which rely on static trust boundaries and implicit trust for internal actors, have been used over the years for human identities. However, they are inadequate for non-human identities. Their limitations have led to a growing interest in Zero Trust Architecture (ZTA), a cuttingedge security concept, built upon the foundational rule of "never trust, always verify". This paper explores the application of ZTA in securing IoT ecosystems, with a specific focus on managing nonhuman identities. It investigates ZTA’s core tenets, like least privilege, micro-segmentation, continuous monitoring, and identity-centric access control—and analyses how these can be effectively implemented in resource-constrained IoT environments. The study identifies key implementation challenges and considerations for the use of ZTA in IoT. The findings of this paper highlight that ZTA, when properly implemented, offers a robust framework for mitigating cyber risks inherent in IoT ecosystems. Finally, the paper outlines future research directions, aimed at integrating ZTA into IoT environments. Ultimately, this work contributes to the growing body of knowledge advocating for Zero Trust as a foundational approach to modern IoT security.Item Investigating gender bias using artificial intelligence classification models on RAVDESS dataset(2025-08) Sefara,Tshephisho J; Khosa, Marshal V; Kisten, MelvinArtificial intelligence (AI) classification models are increasingly being deployed across a wide array of sectors, becoming fundamental tools in decision-making processes that impact individuals and society. These models are utilised in critical applications such as healthcare diagnostics, financial risk assessment, criminal justice systems, and educational admissions, demonstrating their widespread influence. However, a significant challenge arises from the susceptibility of these models to biases, which can lead to outcomes that are unfair, discriminatory, and ultimately harmful to individuals and specific demographic groups. Artificial intelligence bias refers to systematic errors that occur within decision-making processes, ultimately leading to outcomes that are unfair or inequitable. This can manifest as skewed results stemming from human biases that have influenced the original data used to train the AI model, resulting in distorted outputs with potentially harmful results. In this paper, we mitigate the gender bias that occurred during data selection for a classification model. This research experiment was conducted on RAVDESS emotion recognition dataset. The experiments showed improvement in model accuracy by 6% after bias mitigation.