Leak detection in water supply networks using inverse transient analysis in time and frequency domain: a comparative investigation

This paper compares the inverse transient analysis (ITA) in the frequency domain (FITA) with its counterpart in the time domain (TITA) for leak detection and calibration of looped water supply networks (WSNs). The leak detection results demonstrate that both FITA and TITA achieve 100% accuracy in predicting leak location and size for the case study. However, FITA exhibits superior computational efficiency, converging twice as fast as TITA. Regarding the calibration performance, TITA demonstrates higher precision in estimating pipe friction factors, with an overall average error of 0.019%. In comparison, FITA achieves an average error of 4.14%, due to the linearization of the friction loss formula. It is also noteworthy that the high sensitivity of FITA to leak parameters allows for reliable leak detection even without simultaneous calibration of the friction factor. In fact, the impacts of leaks and frictional losses are nearly separable in the frequency domain. Furthermore, the novel dominant frequencies (DFs) method provides direct leak detection through frequency response analysis, while the Decision Table Method (DTM) optimizes measurement site placement with reduced computational overhead. Uncertainty analysis confirms FITA’s robustness to friction factor variations within reasonable ranges, making it particularly suitable for large or complex WSNs where computational efficiency is paramount.