DATA FOR EXPLOSION MONITORING: ACOUSTIC EXPLOSION SIGNATURES COLLECTED ON INFRASOUND MICROPHONES AND SMARTPHONES

Abstract

Explosions produce pressure waves that consist of low (<300 Hz) and infrasonic (<20 Hz) frequencies that can travel vast distances in the atmosphere. These pressure waves are used to characterize explosions and are monitored through various sensor networks of differing scale and density. Although there is a global network (i.e. International Monitoring System) in place that can monitor large explosions (>1 kiloton), regional and local networks that can monitor small explosions are few and far between due to their cost. Additionally, data from these networks are rarely openly available to the public, making research limited to those with special access, potentially stifling progress.This dissertation introduces acoustic explosion datasets collected on infrasound microphones and smartphone sensors to help address the lack of available data and dense sensor networks for explosion monitoring. In Chapter 2, a chemical high explosive (HE) dataset collected on infrasound sensors is presented. For those who are interested or new to the field, it is accompanied by an overview of acoustic explosion features and standardization methods. In Chapter 3, data from two explosions recorded on an infrasound sensor and smartphone sensor network are presented. The audio data from both sensors are compared, showing how the smartphone microphone filters the waveform due to its frequency response. However, due to the similarities in the frequency and time-frequency domains, as well as data from the additional sensors included in the smartphones, it was concluded that smartphones have the potential to enhance traditional explosion monitoring and identification capabilities. This potential is demonstrated with the introduced source localization method. In Chapter 4, a HE dataset collected on smartphones is presented and used to train machine learning (ML) models to detect explosions. The resulting model showed success in classifying smartphone microphone data as either “explosion,” “ambient,” or “other,” highlighting the potential of smartphones as attritable, ubiquitous sensor networks for explosion monitoring.