Smart leak detection breakthrough boosts water management

Dr Giresse Komba, has developed a real-time AI system that detects and locates water leaks with 96% accuracy, providing a scalable solution to reduce losses and improve infrastructure efficiency.

Dr Giresse Komba Dr Komba, a lecturer in Computer Systems Engineering, recently graduated with a Doctorate in Computer Systems Engineering from the Tshwane University of Technology (TUT) with research that advances smart water leak detection. His study introduces a real-time machine-learning system designed to transform water distribution management and address a major challenge facing utilities.

Water leak detection remains a critical issue in Water Distribution Networks, where undetected leaks result in significant water losses and costly infrastructure damage. Traditional methods rely on manual inspections, which are slow, inefficient and often miss hidden leaks. This leads to wasted resources, delayed maintenance and rising operational costs.

To address this, Dr Komba developed a hybrid supervised machine learning algorithm, SVM-ANN-GT. The model integrates Support Vector Machine (SVM), Artificial Neural Network (ANN) and Graph Theory to improve leak detection and localisation. The system enables faster leak identification and more effective responses.

Simulations conducted in EPANET and MATLAB show strong performance. The system achieved 96% detection accuracy, outperforming traditional SVM and ANN models, which achieved 85% and 81%, respectively. It also achieved 95% precision, 90% recall and an F1-score of 87.5%, indicating reliable performance with minimal false positives.

The research also identified 14 distinct leakage zones, improving sensor placement and system efficiency. This enables real-time monitoring, rapid intervention and precise localisation in complex networks.

The impact is significant for South Africa, where nearly 50% of treated water is lost annually, amounting to about R10 billion. Dr Komba’s solution offers a practical, scalable approach to conserve water, strengthen infrastructure resilience and reduce economic losses. It positions the country at the forefront of smart water management and supports sustainable resource use.

The thesis titled “A Hybrid SVM-ANN-GT Algorithm for Real-Time Water Leak Detection and Localisation in Water Distribution Networks” was supervised by TUT’s Prof TE Mathonsi and Prof PA Owolawi. The study explored advanced machine learning techniques to improve detection accuracy and optimise sensor placement.

The research contributes a high-performance algorithm that reduces false alarms, improves detection accuracy, optimises sensor deployment and lowers energy use. The work has gained international recognition through presentations at four international conferences and the publication of two articles in peer-reviewed journals.

Dr Komba’s achievement reflects the Faculty of ICT’s commitment to impactful research and innovation. His work advances computer engineering and supports sustainable water management, smart infrastructure and the use of artificial intelligence to solve real-world challenges.