This study presents an innovative data-driven approach to optimally design water distribution networks (WDNs). The methodology comprises five key stages: Generation of 600 synthetic WDNs with diverse properties, optimized to determine optimal component diameters; Extraction of 80 topological and hydraulic features from the optimized WDNs using graph theory; preprocessing and preparing the extracted features using established data science methods; Application of six feature selection methods (Variance Threshold, k-best, chi-squared, Light Gradient-Boosting Machine, Permutation, and Extreme Gradient Boosting) to identify the most relevant features for describing optimal diameters; and Integration of the selected features with four machine learning models (Random Forest, Support Vector Machine, Bootstrap Aggregating, and Light Gradient-Boosting Machine), resulting in 24 ensemble models. The Extreme Gradient Boosting-Light Gradient-Boosting Machine (Xg-LGB) model emerged as the optimal choice, achieving R², MAE, and RMSE values of 0.98, 0.017, and 0.02, respectively. When applied to a benchmark WDN, this model accurately predicted optimal diameters, with R², MAE, and RMSE values of 0.94, 0.054, and 0.06, respectively. These results highlight the developed model’s potential for the accurate and efficient optimal design of WDNs.

Pressure data collection is essential to increase insight into the current condition of water distribution networks (WDNs). To this end, several methods have been proposed over the last decades for measurement site design (MSD). This research presents a novel method for designing measurement sites by using the k-means clustering algorithm as a pre-processing step, followed by utilizing a new optimization algorithm coupled with the fuzzy TOPSIS method as a processing step. The k-means clustering algorithm is employed to narrow down the search space and identify the most suitable candidate nodes. These candidate nodes are then fed into the new optimization algorithm, called the binary genetic-differential evolutionary algorithm (BGDE), to find the optimal nodes, which are then sorted using the fuzzy TOPSIS method. The BGDE considers sensitivity and entropy as objective functions, while the investment cost is taken into account as a constraint. Furthermore, the Bayesian model averaging (BMA) is employed to mitigate the uncertainties in pipe roughness and nodal demands in the hydraulic simulation model. To evaluate the efficiency of the novel method, two WDNs are tested— one from the literature and the other from a real-world case study. Results show that the proposed method reduces the search space, leading to a 70% faster execution, although the accuracy in finding optimal nodes is reduced by roughly 15% compared to the benchmark method.

This study presents an innovative data-driven approach for the optimal design of water distribution networks (WDNs). The methodology comprises five key stages: Generation of 600 synthetic WDNs with diverse properties, optimized to determine optimal component diameters; Extraction of 80 topological and hydraulic features from the optimized WDNs using graph theory; Preprocessing and preparation of the extracted features using established data science methods; Application of six feature selection methods (Variance Threshold, k-best, chi-squared, Light Gradient-Boosting Machine, Permutation, and Extreme Gradient Boosting) to identify the most relevant features for describing optimal diameters; and Integration of the selected features with four machine learning models (Random Forest, Support Vector Machine, Bootstrap Aggregating, and Light Gradient-Boosting Machine), resulting in 24 ensemble models. The Extreme Gradient Boosting-Light Gradient-Boosting Machine (Xg-LGB) model emerged as the optimal choice, achieving R² and RMSE values of 0.98 and 0.02, respectively. When applied to a benchmark WDN, this model demonstrated high accuracy in predicting optimal diameters, with R² and RMSE values of 0.94 and 0.06, respectively. These results underscore the potential of the developed model for accurate and efficient optimal design of WDNs.

The transient flow analyzing by the frequency domain method (FDM) is computationally much faster than the method of characteristic (MOC) in the time domain. FDM needs no discretization in time and space, but requires the linearization of governing equations and boundary conditions. Hence, the FDM is only valid for small perturbations in which the system’s hydraulics is almost linear. In this study, the linearization errors of the FDM applied to a reservoir-pipe-valve system (RPV) are discussed and by using the Genetic Programming (GP), some correction coefficients are defined to reduce them. By applying the correction coefficient at the opening size of the valve, the first frequency of the frequency domain method is modified. Moreover, the responses at higher-order frequencies are evaluated by some new correction factors obtained by the GP. Solving an illustrative example shows that the error of the system can be significantly reduced by using the applied correction factors.

This study introduces an analytical solution for the hydraulic analysis of looped water distribution networks (WDNs). Conventional approaches to solving ∆Q equations for looped water discharge correction entail iterative hydraulic analysis to compute the system pipe flows, velocities, and nodal pressures. In contrast, using the proposed analytical approach, the ∆Q equation is solved with the exact flow directions determined, consolidating known flow directions into a single unknown variable (∆Q) for each loop. Comparative analyses prove that this approach can precisely compute the hydraulic properties of WDNs. Finally, a Z-test hypothesis test is applied to assess the efficacy of the modified shortest-path algorithm. The results show that this algorithm attains an average accuracy of 90% in predicting exact flow directions, with a confidence level of 99%.