iii Table of Contents
iii Table of Contents
复制标题
DOI：
--
复制
发表时间：
1954
期刊：
影响因子：
--
通讯作者：
J. Huseynov
中科院分区：
文献类型：
--
作者：
J. Huseynov
美国加州大学欧文分校信息与计算机科学哲学博士Javid J.Huseynov于2008年发表的论文《气体泄漏的超声源的分布式定位》中，教授卢博米尔·比克担任主席，及时检测可燃气体泄漏是一项基本的安全措施，有助于挽救任何工业设施的生命。传统的气体检测系统使用红外或电化学传感器来确定周围环境中的气体量，而一种相对较新的方法--使用管道泄漏气体产生的声波--不仅可以检测潜在危险的来源，还可以定位潜在危险的来源。在这项工作中，提出并分析了一套利用MEMS(微型机电系统)麦克风接收的超声信号来定位气体泄漏的分布式算法。虽然声学定位在海洋学、语音识别和雷达跟踪应用中得到了广泛的研究，但使用MEMS传声器的宽带超声源的分布式定位是一个新的应用。所提出的算法旨在与任何工业设施中的分布式传感器网络一起部署。根据宽带声信号的物理特性，将现有的气体泄漏定位算法分为基于能量衰减(ED)和基于到达时延(TDOA)两种方法。统计工具，如最大似然(ML)和最小二乘(LS)估计器，结合XIII迭代梯度下降和牛顿方法，在信号中存在加性高斯白噪声(AWGN)的情况下定位信源。开发了一个基于Java的仿真平台，从准确性、通信开销和响应时间三个方面对算法进行了实现和测试。仿真输入根据传感器和信号源放置拓扑、传感器数量和AWGN的不同级别而不同。为了提高算法的可伸缩性和响应时间，对所提出的一些算法开发了分散式版本。此外，所提出的基于能量衰减的定位算法被成功地测试了一组来自四个分布式MEMS麦克风的输入，该麦克风同时观测到一个小孔中的氮气泄漏。
OF THE DISSERTATION Distributed Localization of Ultrasonic Sources of Gas Leak By Javid J. Huseynov Doctor of Philosophy in Information and Computer Science University of California, Irvine, 2008 Professor Lubomir Bic, Chair Timely detection of combustible gas leaks is a fundamental safety measure which helps to save lives at any industrial facility. While conventional gas detection systems make use of infrared or electrochemical sensors to determine the amount of gas in the surrounding environment, a relatively new approach of using acoustic waves generated by a gas leak from pipes allows to not only detect but also localize the source of potential danger. In this work, a suite of distributed algorithms is proposed and analyzed for localizing gas leaks using ultrasonic signals received by MEMS (Micro-Electro-Mechanical Systems) microphones. While the acoustic localization has been extensively studied in context of oceanography, speech recognition, and radar tracking applications, a distributed localization of broadband ultrasonic sources using MEMS microphones is a novel application. The proposed algorithms are intended for deployment with distributed sensor networks at any industrial facility. In terms of the physical characteristics of broadband acoustic signal, the proposed gas leak localization algorithms were classified into being based either on energy decay (ED) or on time delay of arrival (TDOA). Statistical tools such as the maximum likelihood (ML) and the least-squares (LS) estimators were deployed in combination with xiii iterative gradient descent and Newton’s methods to localize the source in presence of additive white Gaussian noise (AWGN) in the signals. A Java-based simulation platform was developed for implementing and testing the algorithms in terms of accuracy, communication overhead and the response time. Simulation input was varied in terms of sensor and source placement topology, number of sensors and different levels of AWGN. The decentralized versions were developed for some of the proposed algorithms to improve the scalability and the response time. In addition, the proposed energy-decay based localization algorithms were successfully tested with inputs from a set of four distributed MEMS microphones simultaneously observing a nitrogen gas leak from an orifice.