The use of deep neural networks to detect alarms in mines

Abstract

Abstract The article discusses the problem of detecting signals against a background of noise for alarms for miners in underground mine workings in case of emergency. The theoretical aspects of linear and non-linear filtering of alarms are given, the solution to the problem of constructing a non-linear filter based on deep neural network (DNN) is described. The simulation results are presented and a comparative analysis of the performance of an individual miner receiver using the methods of linear coherent reception and a DNN filter is made. Neural network training was carried out on model and experimental data obtained at an existing underground mine.