The impact of hydraulic model calibration on model-based leak localization accuracy: Conclusions drawn from a real-world case study

This study examines the influence of the quality of hydraulic model calibration on the accuracy of model-based leak localization. It is based on a real-world case study where pressures were measured at 12 positions for more than 4 months. Fire flow tests were performed to obtain calibration data—artificially generated leaks with different sizes between 0.25 l/s to 1 l/s at 4 different positions served as dataset for model-based leak localization. Three different calibration approaches were applied on an uncalibrated model generated from GIS data: (i) height adjustments of measurement positions through detailed analysis of long term pressure signals in the leak-free system; as well as (ii) manual trial and error calibration; and (iii) automatic calibration with fire flow test data. The height-adjusted model lead to increased leak localization accuracy compared to the uncalibrated GIS model. However, the most significant improvement was achieved through fire flow test based calibration—revealing a partially closed valve in the system. The auto-calibrated model, which simultaneously calibrated roughness and minor loss coefficients, provided the highest leak localization accuracy.