TagLeak: Non-Intrusive and Battery-Free Liquid Leakage Detection with Backscattered Signals

Leakage detection is a crucial issue for factories with numerous pipelines and valves. Conventional methods for leakage detection are mainly rely on manual checking, which results in both high delay and low accuracy. In this paper, we propose TagLeak, a real-time and low-cost system for automatic leakage detection with commercial off-the-shelf (COTS) RFID devices. The key intuition behind TagLeak is that the leaked liquid around tags will change their phase and RSSI (Received Signal Strength Indicator) readings. Multiple challenges need to be addressed before we can turn the idea into a functional system, including: i) it is difficult to detect the slight signal variation that caused by the leaked liquid, based on the coarse-grained RSSI sequence; ii) multipath and interferences can undermine the tags signal, making the variation caused by leaked liquid more difficult to detect. We propose solutions to these challenges and evaluate the systems performance in different environments. The experimental results tell that TagLeak achieves a higher than 90.2% true positive rate (TPR) while keeps false positive rate (FPR) below 14.3%. Moreover, as an exploration of the industrial Internet, we have deployed TagLeak in a real-world digital twin system Pavatar for liquid leakage detection in an ultra-high-voltage converter station (UHVCS).