Gas-leak localization using distributed ultrasonic sensors

We propose an ultrasonic gas leak localization system based on a distributed network of sensors. The system deploys highly sensitive miniature Micro-Electro-Mechanical Systems (MEMS) microphones and uses a suite of energy-decay (ED) and time-delay of arrival (TDOA) algorithms for localizing a source of a gas leak. Statistical tools such as the maximum likelihood (ML) and the least squares (LS) estimators are used for approximating the source location when closed-form solutions fail in the presence of ambient background nuisance and inherent electronic noise. The proposed localization algorithms were implemented and tested using a Java-based simulation platform connected to four or more distributed MEMS microphones observing a broadband nitrogen leak from an orifice. The performance of centralized and decentralized algorithms under ED and TDOA schemes is analyzed and compared in terms of communication overhead and accuracy in presence of additive white Gaussian noise (AWGN).