Leak detection in water distribution networks: a robust hydraulic calibration strategy incorporating modified simulated annealing (SA) and addressing uncertainties

Pipe leakage in water distribution networks (WDNs) has been an emerging concern for water utilities worldwide due to its public health and economic significance. Not only does it cause significant water losses, but it also deteriorates the quality of the treated water in WDNs. Hence, a prompt response is required to avoid or minimize the eventual consequences. This raises the necessity of exploring the possible approaches for detecting and locating leaks in WDNs promptly. Currently, various leak detection methods exist, but they are not accurate and reliable in detecting leaks. This paper presents a novel GIS-based spatial machine learning technique that utilizes currently installed pressure, flow, and water quality monitoring sensors in WDNs, specifically employing the Geographically Weighted Regression (GWR) and Local Outlier Factor (LOF) models, based on a WDN dataset provided by our partner utility authority. In addition to its ability as a regression model for predicting a dependent variable based on input variables, GWR was selected to help identify locations on the WDN where coefficients deviate the most from the overall coefficients. To corroborate the GWR results, the Local Outlier Factor (LOF) is used as an unsupervised machine learning model to predict leak locations based on spatial local density, where locality is given by k-nearest neighbours. The sample WDN dataset provided by our utility partner was split into 70:30 for training and testing of the GWR model. The GWR model was able to predict leaks (detection and location) with a coefficient of determination (R2) of 0.909. The LOF model was able to predict the leaks with a matching of 80% with the GWR results. Then, a customized GIS interface was developed to automate the detection process in real-time as the sensor’s readings were recorded and spatial machine learning was used to process the readings. The results obtained demonstrate the ability of the proposed method to robustly detect and locate leaks in WDNs.

Leakage detection and calibration of hydraulic models are important issues for the management of water and other distribution networks. An inverse transient model based on a hybrid search technique is presented here. The inverse model is developed mainly for the detection of leaks in water distribution networks. The inverse transient procedure is formulated as a constrained optimisation problem of weighted least-squares type. Two optimisation techniques are tested: the genetic algorithm (GA) and the Levenberg-Marquardt (LM) method. After examining their performance, a new hybrid genetic algorithm (HGA) is developed to exploit the advantages of combining the two methods. The resulting HGA-based inverse transient model is compared with the GA and LM-based inverse transient models using two case studies. The HGA-based inverse transient model proved to be more stable than the LM-based model and it is more accurate and much faster than the GA-based inverse transient model.

Objective: This study aimed to implement a model for detecting and locating leaks in water distribution networks, using artificial intelligence (artificial neural networks). Background: Traditional leak detection methods, such as visual inspections and flow and pressure analysis, are being progressively complemented by solutions based on smart sensors and machine learning techniques. Method: Using the method of characteristics to solve the momentum and continuity equations, along with nodal equations, a hydraulic transient simulator was created for water distribution networks with leaks. For the leak simulation, the orifice law (Torricelli's Law) was used. Results and Discussion: The developed model successfully located the leak with an error of approximately 5.65%, corresponding to a distance of approximately 61.22 m for the studied network. Research Implications: Deteriorated infrastructure in water distribution systems can lead to leakage losses, reduced water transport capacity, component failures, increased maintenance and operating costs, frequent system interruptions, and decreased reliability. Thus, the exploration of rapid leak detection and location methods has increased, aiming to mitigate environmental impacts and the losses of natural and financial resources. Originality/Value: The methodological distinction lies in using pressure wave travel times as input data, rather than absolute pressure values. This approach eliminates the need for detailed hydraulic network calibration, reduces sensitivity to operational noise, and broadens the practical applicability of the model.

Water distribution networks are often susceptible to pipeline leaks caused by mechanical damages, natural hazards, corrosion, and other factors. This paper focuses on the detection of leaks in water distribution networks (WDN) using a data-driven approach based on machine learning. A hybrid autoencoder neural network (AE) is developed, which utilizes unsupervised learning to address the issue of unbalanced data (as anomalies are rare events). The AE consists of a 3DCNN encoder, a ConvLSTM decoder, and a ConvLSTM future predictor, making the anomaly detection robust. Additionally, spatial and temporal attention mechanisms are employed to enhance leak localization. The AE first learns the expected behavior and subsequently detects leaks by identifying deviations from this expected behavior. To evaluate the performance of the proposed method, the Water Network Tool for Resilience (WNTR) simulator is utilized to generate water pressure and flow rate data in a water supply network. Various conditions, such as fluctuating water demands, data noise, and the presence of leaks, are considered using the pressure-driven demand (PDD) method. Datasets with and without pipe leaks are obtained, where the AE is trained using the dataset without leaks and tested using the dataset with simulated pipe leaks. The results, based on a benchmark WDN and a confusion matrix analysis, demonstrate that the proposed method successfully identifies leaks in 96% of cases and a false positive rate of 4% compared to two baselines: a multichannel CNN encoder with LSTM decoder (MC-CNN-LSTM) and a random forest and model based on supervised learning with a false positive rate of 8% and 15%, respectively. Furthermore, a real case study demonstrates the applicability of the developed model for leak detection in the operational conditions of water supply networks using inline sensor data.

Objective: The objective of this study is to develop a model for locating leaks in small-scale water distribution networks, based on the integration of a hydraulic simulator with multicriteria analysis. Theoretical Framework: The water supply system for human consumption is defined as an installation consisting of a set of civil works, materials and equipment, from the collection to the building connections, intended for the production and collective supply of drinking water, through the distribution network. However, real losses, also known as physical losses, denote the entire volume of water made available in the distribution that does not reach the end consumer, requiring a tool to assist professionals in locating these leaks. Method: The methodology adopted for this research consisted of a literature review and the development of the Leak Location Model (LLM) that was built for the hypothetical network and for the discretized hypothetical network. The PSO algorithm was adopted for optimization. Later, in the model validation stage, for all simulated cases containing one, two or three leaks, the observed sections were compared with the three best alternatives indicated by the LLM to calculate the functions f1 and f2, and subsequently the objective function (fo). Results and Discussion: The results obtained revealed that the application of the multicriteria method as a tool for locating leaks in small water distribution networks. It was found that the evaluation of the model satisfactorily described the decision-making problem (average error less than 30%) when it comes to the occurrence of a leak in the network. Research Implications: The leak location model (LLM) is promising for the sector, encompassing the detection of possible leaks in the distribution network that reflect environmental effects. LLM is especially relevant given the lack of resources to assist both regulatory bodies and companies responsible for water distribution. Originality/Value: The study contributes to the area, mainly due to the lack of archived data as a previous history for comparing leak location models for small networks.

Water Distribution Networks (WDNs) suffer substantial water losses due to pipeline leaks, resulting in economic ramifications and exacerbating global water scarcity concerns. This paper seeks to improve the precision of leak detection and the identification of leak locations within WDNs. The pervasive issue of leaks in WDNs poses significant challenges with economic and environmental implications for water utilities. Traditional leak detection methods are time-consuming, resource-intensive, and susceptible to inaccuracies and false alarms due to the random placement of sensors. The detection of concealed background leaks, invisible to the naked eye and situated beneath the surface, presents a particular challenge. This situation complicates efforts for their real-time identification and subsequent repairs. To address these challenges, this paper introduces the SVM-CNN-GT algorithm, an advanced ensemble supervised Machine Learning (ML) approach that incorporates Support Vector Machines (SVM), Convolutional Neural Network (CNN), and Graph Theory (GT). By combining multiple ML algorithms, the SVM-CNN-GT model takes into account various factors that influence leak detection and localization, resulting in more precise and reliable assessments of leak presence and location. The algorithm leverages automatic feature extraction and heterogeneous dual classifiers to accurately assess leaks based on data related to flow rate, pressure, and temperature. Furthermore, a combination probability scheme enhances leak detection efficiency by integrating diverse classifier models with distinct prediction outputs. Through the EPANET performance evaluations, the SVM-CNN-GT algorithm outperforms CNN and SVM algorithms, demonstrating remarkable proficiency with the highest average leak detection accuracy of 98%, followed by CNN at 82% and SVM at 78%.

The detection and localization of leaks in water distribution networks (WDNs) is one of the major concerns of water utilities, due to the necessity of an efficient operation that satisfies the worldwide growing demand for water. There exists a wide range of methods, from equipment-based techniques that rely only on hardware devices to software-based methods that exploit models and algorithms as well. Model-based approaches provide an effective performance but rely on the availability of an hydraulic model of the WDN, while data-driven techniques only require measurements from the network operation but may produce less accurate results. This paper proposes two methodologies: a model-based approach that uses the hydraulic model of the network, as well as pressure and demand information; and a fully data-driven method based on graph interpolation and a new candidate selection criteria. Their complementary application was successfully applied to the Battle of the Leakage Detection and Isolation Methods (BattLeDIM) 2020 challenge, and the achieved results are presented in this paper to demonstrate the suitability of the methods.

Water is a valuable resource that has to be handled appropriately. However, a significant volume of water is wasted annually due to leaks in Water Distribution Networks (WDNs). This emphasizes the necessity for reliable and effective leak detection and localization systems. Several types of solutions have been proposed during the last few years. Among these solutions, data-driven ones are gaining more traction due to their impressive performance. In this paper, we propose a new method for leak detection and localization. The method is based on water pressure measurements acquired at a series of nodes of a WDN. Our technique is a fully data-driven solution that makes only use of the knowledge of the WDN topology, and a series of pressure data acquisitions obtained in absence of leaks. The proposed solution is based on an autoencoder trained on no-leak data, so that leaks are detected as anomalies. The results achieved on the LeakDB dataset demonstrate that the proposed solution outperforms recent methods for leak detection and localization.

This chapter presents a data-driven based approach for detection of leaks in water distribution networks in which the demand is formed by a known periodic pattern plus a stochastic variable. The leak detection method is based on an adaptation of the dynamic principal component analysis (DPCA), and it is assumed that only pressures at selected consumption nodes are measured. Since the variables of water distribution networks (WDNs), even in normal conditions, are nonstationary and time-correlated the data are preprocessed with a periodic transformation previous to the application of DPCA. The proposed approach is validated with the Hanoi network model. The performance is evaluated with three indexes: the leak detection rate, the false alarm rate, and the delay of the detection with respect to the leak’s occurrence time. All of them are satisfactory for diverse leaks’ scenarios, and the proposed approach presents an improvement in the leak detection rate of approximately \(70\%\) as compared with the traditional PCA and DPCA methods.

One of the major types of water wastage in a distribution network is the leakage that occurs in pipes or other network components such as connections. Leaks in urban water supply networks impose many costs and workforce on governments and related organizations annually. If the cause of leakage is traced to the cracks and slits on various parts of a network, the amount of leakage from the location of the crack is said to be directly associated with the amount of pressure at that point. To trace leakage in water distribution networks, this study introduced a method based on hydraulics modeling and the inverse solution of flow equations to predict the location and amount of leakage in water distribution networks, by considering the values of the measured pressure in some network nodes. For this, a hydraulic model of the network under study was provided in the EPANET Hydraulic Analysis software, and the network analysis of various values and states of hypothetical leaks led to obtaining pressure values in the different nodes of the network via modeling water modeling under a steady state. Then, the artificial neural network, having been trained, was used to provide measured pressures in some network nodes at the testing hour as input data to the neural network to locate possible leaks in the water distribution network and to predict their approximate values. The findings were found to enjoy desirable accuracy. DOI: https://doi.org/10.52783/pst.835

Towards an improvement of GPR-based detection of pipes and leaks in water distribution networks

Leaks represent one of the most relevant faults in water distribution networks (WDN), resulting in severe losses. Despite the growing research interest in critical infrastructure monitoring, most of the solutions present in the literature cannot completely address the specific challenges characterizing WDNs, such as the low spatial resolution of measurements (flow and/or pressure recordings) and the scarcity of annotated data. We present a novel integrated solution that addresses these challenges and successfully detects and localizes leaks in WDNs. In particular, we detect leaks by a sequential monitoring algorithm that analyzes the inlet flow, and then we validate each detection by an ad hoc statistical test. We address leak localization as a classification problem, which we can simplify by a customized clustering scheme that gathers locations of the WDN where, due to the low number of sensors, it is not possible to accurately locate leaks. A relevant advantage of the proposed solution is that it exposes interpretable tuning parameters and can integrate knowledge from domain experts to cope with scarcity of annotated data. Experiments, performed on a real dataset of the Barcelona WDN with both real and simulated leaks, show that the proposed solution can improve the leak detection and localization performance with respect to methods proposed in the literature.

<p><strong>The influence of stochastic water demand on model-based leak localization</strong></p><p>Globally, water demand is rising and resources are diminishing. In the context of climate change and a growing world population, a further increase in water scarcity seems inevitable. Aiming towards a sustainable future, water should be used as efficient as possible by minimizing water losses, which can be higher than 50% in some drinking water networks.<sup>3</sup> To minimize losses it is crucial to detect, localize and repair leaks as soon as possible.</p><p>Leaks cause changes in flow and pressure. By monitoring the network with pressure and flow sensors and coupling these measurements with hydraulic computer models, leaks can be detected and located. The success of this so-called model-based leak localization depends heavily on our knowledge of water demand, since every water consumption affects the pressure and flow in the network as well. Nowadays, demand is modelled based on water billing information and the network’s inflow. This study proposes a new strategy by modelling stochastic demands. Realistic residential demands are generated in high spatial and temporal resolution based on Dutch water use statistics with SIMDEUM<sup>4</sup>. Subsequently, the stochastic demands are used within hydraulic simulations. The influence of demand fluctuations on pressure in the system is analyzed using Monte-Carlo sampling and the corresponding effects on model-based leak detection and localization are investigated.</p><p>The proposed method is applied on a real Dutch water distribution network, containing inflow and six pressure measurements. Statistical information like the number of residents, households and annual billing information in the area is known. The corresponding hydraulic model is calibrated on pipe roughness by minimizing the mean squared error of the modelled and measured pressure at the sensor locations. Pressure driven simulations are performed and the resulting pressure changes at the sensors are simulated. Through the stochastic simulations in combination with Monte-Carlo sampling, confidence intervals for pressure changes at the sensor locations are determined and compared with the real measurements. The performance of leak detectability and localization is subsequently examined.  </p><p>This study shows that stochastic water demand simulations provide a better understanding on the reliability of model-based leak localization. By using these simulations, confidence intervals of demand related pressure changes at the sensor locations can be determined which affect the performance of leak detectability and localization under the variability of water demand. A better grip on the reliability of leak localization yields in a more efficient quest for leaks.</p><p> </p><p><sup>3</sup><sub>EurEau 2017, Europe’s water in figures, an overview of European drinking water and waste water sectors, The European Federation of National Associations of Water Services</sub></p><p><sup>4</sup><sub>Blokker, E. J. M. (2010), Stochastic water demand modelling for a better understanding of hydraulics in water distribution networks, PhD thesis, Delft University of Technology</sub></p><p> </p>

Leaks in water distribution networks (WDNs) are one of the main reasons for water loss during fluid transportation. Considering the worldwide problem of water scarcity, added to the challenges that a growing population brings, minimizing water losses through leak detection and localization, timely and efficiently using advanced techniques is an urgent humanitarian need. There are numerous methods being used to localize water leaks in WDNs through constructing hydraulic models or analyzing flow/pressure deviations between the observed data and the estimated values. However, from the application perspective, it is very practical to implement an approach which does not rely too much on measurements and complex models with reasonable computation demand. Under this context, this paper presents a novel method for leak localization which uses a data-driven approach based on limit pressure measurements in WDNs with two stages included: (1) Two different machine learning classifiers based on linear discriminant analysis (LDA) and neural networks (NNET) are developed to determine the probabilities of each node having a leak inside a WDN; (2) Bayesian temporal reasoning is applied afterwards to rescale the probabilities of each possible leak location at each time step after a leak is detected, with the aim of improving the localization accuracy. As an initial illustration, the hypothetical benchmark Hanoi district metered area (DMA) is used as the case study to test the performance of the proposed approach. Using the fitting accuracy and average topological distance (ATD) as performance indicators, the preliminary results reaches more than 80% accuracy in the best cases.

Locating Leaks in Water Distribution Networks with Simulated Annealing and Graph Theory