In the face of South Africa's escalating water crisis, where nearly 50% of treated water is lost annually through leaks—equating to roughly R10 billion in economic damage—a groundbreaking development from Tshwane University of Technology (TUT) offers real hope. Dr. Giresse Komba, a lecturer in Computer Systems Engineering at TUT's Faculty of Information and Communication Technology (ICT), has pioneered a real-time artificial intelligence (AI) system that detects and pinpoints water leaks with an impressive 96% accuracy. This innovation, born from his doctoral research, promises to revolutionize water distribution networks (WDNs) across the country, turning a persistent challenge into an opportunity for sustainable management.
South Africa grapples with non-revenue water (NRW) losses hovering around 47% nationally, with some provinces like KwaZulu-Natal reaching 60%. Aging infrastructure, manual inspection methods, and complex urban pipe networks exacerbate the issue, leading to undetected leaks that waste precious resources and inflate costs for municipalities. TUT's breakthrough addresses these pain points head-on, leveraging machine learning to enable proactive, precise interventions.
South Africa's Water Woes: A National Emergency
The nation's water infrastructure is under strain, as highlighted in the latest Blue Drop and Green Drop reports released in March 2026. These assessments reveal deteriorating drinking water quality in 60% of systems and critical wastewater management failures in nearly half of facilities. Physical losses from leaks account for 40.8% of NRW, underscoring the need for smarter technologies.
Traditional leak detection relies on labor-intensive patrols and ground microphones, which are slow and prone to missing subtle anomalies in buried pipes. In cities like Johannesburg and Cape Town, this results in billions of liters lost daily, straining supplies amid climate variability and population growth. Dr. Komba's work at TUT positions higher education institutions as key players in national problem-solving, bridging academia and industry for tangible impact.
Dr. Giresse Komba: From PhD to Practical Innovation
Dr. Komba's journey exemplifies the power of university research in addressing real-world issues. As a lecturer at TUT, he completed his PhD in Computer Systems Engineering, focusing on hybrid machine learning models for WDNs. Supervised by Prof. Topside E. Mathonsi, Head of the Department of Information Technology, and Prof. Pius Adewale Owolawi, a leading expert in telecommunications and IT, his thesis—"A Hybrid SVM-ANN-GT Algorithm for Real-Time Water Leak Detection and Localisation in Water Distribution Networks"—marks a milestone.
Prior publications, including works on SVM-CNN-GT and ANN-XGBoost algorithms presented at international conferences, laid the groundwork. These efforts have garnered citations and recognition, affirming TUT's role in fostering AI talent for South Africa's Fourth Industrial Revolution (4IR) agenda.
The Core Technology: Breaking Down SVM-ANN-GT
At the heart of this breakthrough is the SVM-ANN-GT hybrid algorithm, integrating three powerful components:
- Support Vector Machine (SVM): Classifies leak patterns from pressure and flow data with high precision.
- Artificial Neural Network (ANN): Learns complex non-linear relationships in sensor inputs for adaptive detection.
- Graph Theory (GT): Models WDNs as graphs to optimize sensor placement and localize leaks across 14 identified zones.
The process unfolds step-by-step: sensors capture real-time hydraulic data (pressure transients, flow anomalies); the algorithm processes it via EPANET simulations for network modeling and MATLAB for computation; leaks are flagged, sized, and located within minutes. This shift from reactive to predictive maintenance minimizes downtime and resource waste.
Photo by Axel Richter on Unsplash
Superior Performance: Metrics That Matter
Tested rigorously, the system outperforms benchmarks:
| Model | Detection Accuracy | Precision | Recall | F1-Score |
|---|---|---|---|---|
| SVM-ANN-GT (TUT) | 96% | 95% | 90% | 87.5% |
| Traditional SVM | 85% | - | - | - |
| ANN Alone | 81% | - | - | - |
Low false positives ensure utilities trust the alerts, enabling targeted repairs. By optimizing sensor deployment, it cuts costs and energy use, vital for resource-strapped municipalities.
From Lab to Life: Implementation Potential
Scalability is key. The model adapts to varying network sizes, from urban grids to rural supplies. In pilot scenarios, it could slash SA's R10 billion NRW losses by 20-30%, conserving billions of liters amid Day Zero threats. Integration with IoT sensors and municipal SCADA systems paves the way for nationwide rollout. For more on the research, see TUT's detailed announcement here.
Challenges like data scarcity and integration hurdles were overcome through simulations, proving viability without massive upfront infrastructure.
Supervisors' Vision and TUT's Research Ecosystem
Prof. Mathonsi and Prof. Owolawi's guidance blended expertise in IT and telecom, fostering interdisciplinary innovation. TUT's Faculty of ICT, with its AI Hub and 4IR focus, provides fertile ground—home to chairs in AI and smart systems. This environment nurtures PhDs into leaders, as seen in Dr. Komba's conference presentations and journal publications.
Such mentorship highlights higher education's role in South Africa's National Research Foundation (NRF) priorities, including water security.
Economic and Sustainability Gains
- Cost Savings: Reduces repair expenses by prioritizing leaks.
- Water Conservation: Prevents evaporation and contamination.
- Climate Resilience: Supports drought-prone regions per recent Green Drop findings.
- Job Creation: Demands skilled AI engineers, boosting TUT graduates' employability.
By curbing NRW, it eases pressure on dams like Vaal, aligning with government mandates.
Future Horizons: Deployment and Beyond
Next steps include field trials with metros like Tshwane and partnerships via TUT's industry liaison office. Expansion to predictive analytics for bursts could elevate it further. Dr. Komba's work inspires similar AI applications in energy and transport, cementing TUT's legacy.
For researchers eyeing water tech careers, opportunities abound in SA universities. Explore roles at AcademicJobs.com/research-jobs.
Higher Education's Pivotal Role in National Solutions
TUT exemplifies how South African universities drive 4IR solutions for SDGs like clean water (Goal 6). With 41 new NRF chairs announced, including for health disparities, research momentum grows. Dr. Komba's success underscores the value of PhD programs in applied engineering.
Stakeholders—from government to utilities—must invest in tech transfer, ensuring lab breakthroughs reach pipes nationwide.
