Data-driven leak localization in WDN using pressure sensor and hydraulic information

Current data-driven methods for leak localization (LL) in water distribution networks (WDNs) rely on two unrealistic assumptions: they frame LL as a node-classification task, requiring leak examples for every node—which rarely exists in practice—and they validate models using random data splits, ignoring the temporal structure inherent in hydraulic time-series data. To address these limitations, we propose a temporal, regression-based alternative that directly predicts the leak coordinates, training exclusively on past observations and evaluating performance strictly on future data. By comparing five machine-learning techniques—k-nearest neighbors, linear regression, decision trees, support vector machines, and multilayer perceptrons—in both classification and regression modes, and using both random and temporal splits, we show that conventional evaluation methods can misleadingly inflate model accuracy by up to four-fold. Our results highlight the importance and suitability of a temporally consistent, regression-based approach for realistic and reliable leak localization in WDNs.

This paper presents a new data-driven method for leak localization in water distribution networks. The proposed method relies on the use of available pressure measurements in some selected internal network nodes and on the estimation of the pressure at the remaining nodes using Kriging spatial interpolation. Online leak localization is attained by comparing current pressure values with their reference values. Supported by Kriging; this comparison can be performed for all the network nodes, not only for those equipped with pressure sensors. On the one hand, reference pressure values in all nodes are obtained by applying Kriging to measurement data previously recorded under network operation without leaks. On the other hand, current pressure values at all nodes are obtained by applying Kriging to the current measured pressure values. The node that presents the maximum difference (residual) between current and reference pressure values is proposed as a leaky node candidate. Thereafter, a time horizon computation based on Bayesian reasoning is applied to consider the residual time evolution, resulting in an improved leak localization accuracy. As a data-driven approach, the proposed method does not need a hydraulic model; only historical data from normal operation is required. This is an advantage with respect to most data-driven methods that need historical data for the considered leak scenarios. Since, in practice, the obtained leak localization results will strongly depend on the number of available pressure measurements and their location, an optimal sensor placement procedure is also proposed in the paper. Three different case studies illustrate the performance of the proposed methodologies.

In the first place, this paper presents an examination of the effects of missing cadastral information on the modeling and calibration of water distribution networks (WDN). For this purpose, cadastral errors were artificially introduced to the hydraulic models of the WDN of three cities of the province of Valle del Cauca (Colombia), with population in the range of 15,000–30,000 inhabitants. The cadastral errors were introduced by eliminating some pipes from the unaltered models using the resilience index as selection criterion. From this examination, typical responses observed during the calibration process were associated to specific topological and hydraulic errors and it was also noted that there is a direct relation between the resilience index and the effect of the removal of a certain pipe on the hydraulic performance of the model. In the second place, after identifying the effects of missing and erroneous cadastral and hydraulic information on the calibration of WDN models, the inverse procedure was carried out: the calibration process was used as a tool for detecting missing and erroneous cadastral and hydraulic information. This methodology was successfully applied to the WDN of Sector Estadio (250,000 inhabitants), which is a complex area of the city of Bucaramanga (Colombia); through the calibration of this WDN, errors such as failing water meters, pipes and valves that existed in the real system but were not in the model, significant errors in pipe diameters in the model, among others, were identified.

The optimal placement and operation of pressure control valves in water distribution networks is a challenging engineering problem. When formulated in a mathematical optimization framework, this problem results in a nonconvex mixed integer nonlinear program (MINLP), which has combinatorial computational complexity. As a result, the considered MINLP becomes particularly difficult to solve for large-scale looped operational networks. We extend and combine network model reduction techniques with the proposed optimization framework in order to lower the computational burden and enable the optimal placement and operation of control valves in these complex water distribution networks. An outer approximation algorithm is used to solve the considered MINLPs on reduced hydraulic models. We demonstrate that the restriction of the considered optimization problem on a reduced hydraulic model is not equivalent to solving the original larger MINLP, and its solution is therefore sub-optimal. Consequently, we investigate the trade-off between reducing computational complexity and the potential sub-optimality of the solutions that can be controlled with a parameter of the model reduction routine. The efficacy of the proposed method is evaluated using two large scale water distribution network models.

Acoustic sensors are widely used for monitoring urbanized water distribution networks (WDNs) to detect and localize pipe leakages. Since their inception, few research studies have focused on developing a generic, effective, and practical methodology to analyse complex acoustics signals for leakage detection and localization in large-scale WDNs. In collaboration with PUB, Singapore’s National Water Agency, a generic acoustic data analysis approach has been developed to facilitate PUB’s present Smart Water Grid (SWG) management. The proposed approach encompasses multi-stage systematic analyses, namely: (1) data quality assessment; (2) data pre-processing; (3) near real-time leakage event detection and classification; and finally (4) near real-time leakage localization. Our proposed approach is then tested in major WDNs in Singapore having more than 1100km of underground water pipelines and 82 permanently installed hydrophone acoustic sensors between 1 Aug 2019 and 31 Aug 2020, where multiple historical leakage events were reported to within 600m, or less, from neighbouring hydrophones across the large complex networks. By emulating the near real-time detection and localization analyses daily, our proposed methodology could localize reported leakage events to an error range of around 150m on average, while demonstrating significant and stable acoustic leakage power rate over the temporal size of the leakage event cluster(s).

Leakages in the water distribution networks (WDNs) are real problems for utilities and other governmental agencies. Timely leak detection and location identification have been challenges. In this paper, an integrated approach to geospatial and infrared image processing was used for robust leak detection. The method combines drops in flow, pressure, and chlorine residuals to determine potential water leakage locations in the WDN using Geographic Information System (GIS) techniques. GIS layers were created from the hourly values of these three parameters for the city of Sharjah provided by the Sharjah Electricity, Water, and Gas Authority (SEWA). These layers are then analyzed for locations with dropped values of each of the parameters and are overlaid with each other. In the case where there were no overlaying locations between flow and pressure, further water quality analysis was avoided, assuming no potential leak. In the case where there are locations with drops in flow and pressure layers, these overlaying locations are then examined for drops in chlorine values. If overlaying locations are found, then these regions are considered potential leak locations. Once potential leak locations are identified, a specialized remote sensing technique can be used to pinpoint the leak location. This study also demonstrated the suitability of using an infrared camera for leak detection in a laboratory-based setup. This paper concludes that the following methodology can help water utility companies in the timely detection of leaks, saving money, time, and effort.

The water distribution industry is facing challenges of water leakages in drinking water distribution networks. Mainly high pressure and the deterioration of ageing pipes and fittings cause the increase of the leak, and the consuming water and the number of bursts in the network. This study based on hydraulic and geospatial information system model has been suggested to estimate the probability of leaking for each pipe segment in the selected water distribution network. In this model, the likelihood of leaking is illustrated using pressure, age, length, diameter and slope of the pipe. The system is implemented with GIS tools, WaterGEMS, Google Earth Pro and Minitab software. Google Earth Pro is used to design the pipe network. The hydraulic model creates with WaterGEMS is integrated with ArcGIS. Minitab is used in elaborating the linear regression model, which is proposed to gain the leakage probability. The formed methodology is applied in identifying vulnerability ranges that have been identified for the pipe segments in the water distribution network. For the adducting results, a Geographic Information System (GIS) is used. Using the capabilities of this GIS model, the colour coded map represents the vulnerability ranges of the areas. The gained results can be used in detecting the most leakage areas and control pipeline leakage in the water distribution network. This methodology is developed to provide maximum protection against water leaks in the selected region.KeywordsGeographic information system (GIS)WaterGEMSGoogle earth proMinitab

The research about online monitoring and leakage automatic location of water distribution networks (WDN) has a wide range of applications that include water resource protection, monitoring, and allocation. Variational mode decomposition (VMD) and cross-correlation (CC) based leakage location is a popular and effective method in WDN. However, the value of K intrinsic mode functions (IMFs) based on VMD decomposition needs to be determined artificially, which affects the separation effect of signal frequency band characteristics directly. Hence, this work proposes an adaptive method to determine the parameter K of leakage vibration signal’s IMFs, which will be applied to automatic leakage location in WDN. Firstly, the number of saddle points in the frequency domain envelope of the sampled signal in different step sizes is calculated. The parameter K is determined according to the curvature change of the number of saddle points and the sampled signal. Finally, the selective IMFs are reconstituted into a new signal, which can determine a leak position using CC based time-delay estimation (TDE). To verify the effectiveness of the proposed algorithm, the different methods based on EMD and Fast ICA are compared. The experimental results demonstrate that the proposed parameter K value adaptive VMD (KVA-VMD) decomposition method is more suitable for leakage location in WDN.

This study aims to propose Micro-electromechanical System (MEMS) accelerometers for leak localization in the water distribution network and assess the performance of machine learning models in accurately estimating leak locations. Intensive field experimentation was conducted to collect data for model development. Machine learning algorithms were employed to develop leak localization models, specifically artificial neural network (ANN) and support vector machine (SVM). Seventeen time-domain and frequency-domain features were extracted, and feature selection was performed using the backward elimination method. The results indicate that the ANN and SVM models are suitable classifiers for localizing leak distance. Both models achieved leak location predictions with over 80% accuracy, and the mean absolute errors were measured at 0.858 and 0.95 for the ANN and SVM models, respectively. The validation results demonstrated that the models maintained accuracies close to 80% when the distance between sensors and the leak was less than 15 m. However, the performance of the model deteriorates when leaks occur at distances greater than 15 m. This study demonstrates the applicability of MEMS accelerometers for leak localization in water distribution networks. The findings highlight the promising potential of employing MEMS accelerometers-based ANN and SVM models for accurate leak localization in urban networks, even under real-world, uncontrolled conditions. However, the current model exhibits limited performance in long-distance leak localization, requiring further research to address and resolve this issue.

This paper extends previous research to analytically identify leaks within a water distribution network (WDN), by combining hydraulic simulation and network science based data analysis techniques. The WDN model is used to run several ‘leakage scenarios’, by varying leak location (pipe) and severity, and to build a dataset with corresponding variations in pressure and flow, induced by the leak. All junctions and pipes are considered for potential pressure and flow sensors deployment; a clustering procedure on these locations identifies the most relevant nodes and pipes, and cost-effectiveness was considered. A graph is then generated from the dataset, having scenarios as nodes and edges weighted by the similarity between each pair of nodes (scenarios), in terms of pressure and flow variation due to the leak. Spectral clustering groups together similar scenarios in the eigen-space spanned by the most relevant eigen-vectors of the Laplacian matrix of the graph. This method uses superior traditional techniques. Finally, support vector machines classification learning is used to learn the relation between variations in pressure and flow at the deployed meters and the most probable set of pipes affected by the leak.

An accurately calibrated hydraulic model of a water distribution network (WDN) is crucial for tasks such as leakage detection, resilience assessment, and rehabilitation planning. Since model calibration relies on real data, increasing monitoring locations for pressure and flow raises both accuracy and cost. Therefore, selecting suitable data collection points is crucial for efficient calibration. This research proposes a practical, graph-theory-inspired approach for determining pressure sensor locations based on the network topology and the demand distribution in the WDN. The method employs a customized metric of Node Demand Betweenness Centrality (NDBC) as a criterion to estimate flow distribution, and thereby identifies influential paths within WDNs. Sensor locations are then optimized using the nondominated sorting genetic algorithm II (NSGA-II), which identifies nodes located along the most influential paths in WDNs while maximizing network coverage by increasing the graph-based hydraulic distance between sensors. This methodology is further supported by additional graph analysis methods, which serve as a faster alternative and are compared against a widely utilized hydraulic model-based approach. The proposed approach was applied to both the Modena and L-Town WDNs. In both case studies, the calibrated models based on pressure data from the sensor locations proposed in this research demonstrated satisfactory accuracy.

Water loss poses a significant problem for water utilities and has received a lot of attention. To fulfill the increasing global demand for water, water supply system operations must be streamlined, making leak detection and location crucial. Water utilities have developed a number of techniques over time for finding leaks in water distribution networks (WDNs). These methodologies range from simple visual inspection to the use of hardware systems and now software using models and algorithms. Data from flow or pressure measurements, which are required for the analysis of water networks, is becoming more readily available with the introduction of intelligent sensor devices. Along with the introduction of geographic information systems (GIS) and supervisory control and data acquisition (SCADA) in the water sector, the deployment of model-driven methodologies for leak detection and localization has found extensive use. This paper aims to provide a concise introductory reference for early researchers in the development of a model-based approach for leak detection in WDNs. Thus, a survey of model-based approaches is presented, along with current research trends and applications of model-driven methodologies for leak detection in water supply networks. Several model-driven approaches and research studies for each case are discussed. Some challenges and research gaps are also discussed.

Improving Analytics in Urban Water Management: A Spectral Clustering-based Approach for Leakage Localization

Water Distribution Networks (WDNs) are large-scale, spatially irregular systems composed of interconnected nodes and pipes shaped by urban infrastructure. Despite their critical role in public health and urban resilience, WDNs often lag behind other utilities in terms of digitization and integration into smart city frameworks. This work presents a generative, data-driven method for reconstructing the full hydraulic state—pressures at nodes and flows in pipes—using a feedforward neural network with radial basis function (RBF) activations. The model is trained on synthetic data generated via hydraulic simulations and uses sparse real-time measurements from a limited number of strategically placed pressure and flow sensors. Unlike classical RBF Neural Networks, which rely on fixed spatial centers and local interpolation, the proposed architecture enables full-field inference through a single forward pass. The RBF activations capture spatial dependencies while allowing the network to generalize across the entire topology of the WDN. The approach achieves high-precision reconstruction performance, with Mean Squared Errors on the order of $$10^{-13}$$ <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"> <mml:msup> <mml:mn>10</mml:mn> <mml:mrow> <mml:mo>-</mml:mo> <mml:mn>13</mml:mn> </mml:mrow> </mml:msup> </mml:math> and Mean Absolute Errors around $$10^{-7}$$ <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"> <mml:msup> <mml:mn>10</mml:mn> <mml:mrow> <mml:mo>-</mml:mo> <mml:mn>7</mml:mn> </mml:mrow> </mml:msup> </mml:math> , confirming the method’s accuracy and stability. So, RBF-NNs emerge as a particularly effective solution, as they efficiently capture local-to-global spatial dependencies inherent in the dynamic behavior of WDNs, contributing to the broader integration of AI-driven solutions in infrastructure management.

Water losses in watersupply and distribution systems have been a serious problem in Turkey and allover the world causing economic costs and environmental drawbacks. Total waterlosses is the summation of real (physical) and apparent (commercial) losses. Mainactivities to reduce water losses include digitizing the components of network usingGeographical Information Systems (GIS), dividing large water distributionnetworks (WDNs) into District Metered Areas (DMAs), monitoring, carrying outyearly water balance, hydraulic modelling, pressure management, active leakagecontrol, recording water consumption of the users, reducing meter errors andillegal water consumption. The Ministry of Forestry and Water Affairs in Turkeyissued a new regulation for water losses reduction in 2014 based on the methodsused to determine and reduce the losses. This is followed by issuing practicaldescription of the technical procedure to reduce the losses in 2015. The newlegislation aims to reduce total water losses in all water authorities below 25%within certain time periods. All the water authorities are required to carryout yearly water balance and report them to the ministry together with forms tofollow up the efforts towards water losses reduction. In order to apply theprescribed techniques in a successful way, first of all water losses problemshould be understood well, technical and human resource capacity should beimproved, and sufficient financial resources should be allocated. Also, thewater utilities should have a strategic plan and a strong commitment to reducewater losses.