Amirkabir University of Technology Amirkabir Journal of Civil Engineering 2588-297X 54 5 2022 07 23 Detection of two simultaneous leakages in water distribution network using hybrid feedforward artificial neural networks Detection of two simultaneous leakages in water distribution network using hybrid feedforward artificial neural networks 1965 1982 4534 10.22060/ceej.2021.20017.7316 FA Hamideh Fallahi Ph.D. Student, Department of Civil Engineering, Roudehen Branch, Islamic Azad University, Roudehen, Iran 0000-0001-6317-8162 Mohammadreza Jalili Ghazizadeh Associate Professor, Department of Civil Engineering, Faculty of Water and Environmental Engineering, Shahid Beheshti University, Tehran, Iran Babak Aminnejad Assistant Professor, Department of Civil Engineering, Roudehen Branch, Islamic Azad University, Roudehen, Iran Jafar Yazdi Assistant Professor, Department of Civil Engineering, Faculty of Water and Environmental Engineering, Shahid Beheshti University, Tehran, Iran Journal Article 2021 05 08 <span style="letter-spacing: .05pt;">Leakage is one of the main challenges in the operation of water distribution networks. In the present study, leakage is detected using Feedforward Artificial Neural Networks (ANNs). For this purpose, two scenarios are considered for training the ANNs. In the first scenario, two simultaneous leakages with equal values, and in the second scenario, two simultaneous unequal leakages are applied to each pair-node of a network. The training data are analyzed by EPANET2.0 hydraulic simulation software linked with the MATLAB programming language. In both scenarios, first, ANNs are trained using flow rates of total pipes number. Then, sensitivity analysis is performed by Hybrid ANNs for the flow rates of different percent of pipes numbers. The results of the proposed Hybrid ANNs indicate that in the first scenario, by having the flow rates of 10% of the total pipes, the locations of two simultaneous leakages are successfully determined. However, for the second scenario, while the difference between the two leakages is less than 80% of the maximum leakage (up to ratio value of 10 and 90 % leakages), by having 10% of the total pipes flow rates, the locations of the two leakages are still successfully determined. However, for larger differences, only the location of the bigger leak could be detected. Despite the complexities of the second scenario, the proposed ANNs could successfully detect the location of the bigger leakage. </span> <span style="letter-spacing: .05pt;">Leakage is one of the main challenges in the operation of water distribution networks. In the present study, leakage is detected using Feedforward Artificial Neural Networks (ANNs). For this purpose, two scenarios are considered for training the ANNs. In the first scenario, two simultaneous leakages with equal values, and in the second scenario, two simultaneous unequal leakages are applied to each pair-node of a network. The training data are analyzed by EPANET2.0 hydraulic simulation software linked with the MATLAB programming language. In both scenarios, first, ANNs are trained using flow rates of total pipes number. Then, sensitivity analysis is performed by Hybrid ANNs for the flow rates of different percent of pipes numbers. The results of the proposed Hybrid ANNs indicate that in the first scenario, by having the flow rates of 10% of the total pipes, the locations of two simultaneous leakages are successfully determined. However, for the second scenario, while the difference between the two leakages is less than 80% of the maximum leakage (up to ratio value of 10 and 90 % leakages), by having 10% of the total pipes flow rates, the locations of the two leakages are still successfully determined. However, for larger differences, only the location of the bigger leak could be detected. Despite the complexities of the second scenario, the proposed ANNs could successfully detect the location of the bigger leakage. </span> https://ceej.aut.ac.ir/article_4534_72cad9e1f9ae79872b8d6ac34fc2851c.pdf