Control of Acoustic Extinguisher with Deep Neural Networks for Fire Detection

Abstract

The search for fast and environmentally safe methods of fighting fires has been a particularly important topic in recent years. Many academic centres are conducting research on the use of Deep Neural Networks to detect flames. One of the most promising is the acoustic method of extinguishing flames. In theory, an acoustic extinguisher can be applied to extinguish fires of different classes because acoustic waves pass through solids, liquids, and gases. In principle, the technology described in the article can be used to extinguish B- and C-class fires when gases or liquids are burning. Until now, the known studies have been conducted only for low-power acoustic extinguishers. Therefore, there is a need to fill a theoretical and practical gap in this respect (scientific novelty). The result of the activities is the development of new techniques for extinguishing flames with the use of Deep Neural Networks, and then extinguishing flames using a high and very high power loudspeaker applied to the acoustic extinguisher. The main aim of this paper is to present the possibilities of using Deep Neural Networks to detect fires, as well as the results of research on the extinguishing of flames with the use of square waveforms with Amplitude Modulation (AM) for several frequencies, which is also a scientific novelty, including the minimum acoustic power and sound pressure level as a function of a distance from the output of the acoustic system. On this basis, it became possible to determine the minimum power delivered to the extinguisher and the minimum sound pressure level that causes the extinguishing effect at given input parameters.