DECOMPOSITION-BASED DATA ANALYSIS WITH HYDRAULIC MODEL CALIBRATION FOR LEAKAGE DETECTION AND ISOLATION

This paper proposes an integrated approach for leak detection and localization. The leak detection is undertaken by data analysis in multiple steps including time series data decomposition to ensure stationarity, leakage event detection by statistical process control methods using the correlation analysis of multiple sensor events of monitored flows and pressures. A detected leak event is localized via hydraulic model calibration with the snapshots of flow and pressure data collected during the leakage event. The approach is applied to L-Town with two-year monitoring data and the hydraulic model. The results show that the proposed method is effective at detecting and localizing both incipient leaks and pipe bursts.