Ultrasound-based air leak detection using a random microphone array and sparse representations

In this paper, we propose a method for the detection and localization of air leaks in pressurized air systems using the broadband ultrasonic sound that is generated by the escaping air. We apply a random, sparse array of microphones in conjunction with an algorithm inspired by compressed sensing (CS) to finely localize air leaks in 2D (that is, we estimate the full direction vector of the leak). To achieve this, an intermediate acoustic image is generated from the raw microphone signals, called the energyscape. A sparse decomposition algorithm based the system's Point Spread Function (PSF) is applied to the energyscape to estimate the real position of the air leaks. The method allows the simultaneous detection and localization of multiple air leaks of varying strengths and can also localize weak air leaks overshadowed by a simultaneously present strong leak.