From ashes to answers: decoding acoustically agglomerated soot particle signatures

Abstract

This study investigated the possibility of extending the soot morphology analyses to acoustically agglomerated soot deposited on the surface of smoke alarms and of applying the validity of soot analysis for unique chemical signatures in the field of fire investigations. Through collecting soot samples, including agglomerated soot acquired from smoke alarms, this research presents a pioneering stride in soot morphology data analyses conducted by leveraging advanced deep learning methodologies. Preliminary outcomes underline that the proposed convolutional neural network model has the potential to decode intricate soot characteristics and to distinguish soot particle images between diverse fuel types and burning conditions. In particular, for the acoustically agglomerated soot collected by smoke alarms, it was also found possible to decode their intricate morphology by applying the proposed data-driven approach.