Machine Learning-Assisted Model for Leak Detection in Water Distribution Networks Using Hydraulic Transient Flows

This study introduces a novel hybrid leak detection method based on machine learning (ML) and hydraulic transient modeling for pipe networks. First, the transient hydraulic simulation model is developed in the time domain. Then, the optimum measurement sites for sampling the network's hydraulic responses are determined using a graph-based method. The model exploits the network's high-frequency transient responses at measurement sites to generate data sets. The generated samples are transformed into the frequency domain using the fast Fourier transform (FFT). The neighborhood component analysis (NCA) is used for feature selection and the optimum classifier is selected by comparing the performance of different classification algorithms. The model is finally applied to two case studies: an experimental reservoir-pipe-valve (RPV) system and a complex water distribution network (WDN). The accuracy of leak detection is evaluated considering fast and slow transient excitations concerning various levels of uncertainty in the system parameters. The results indicated that the model could detect leaks accurately and is stable and reliable against high uncertainties in pipe friction factors and nodal demands.