Determination of the Approach to the Construction of the Neural Network Algorithm for Detection of Emergency and Pre-Emergency Situations by Multicriterial Electro-Optical System

Abstract

Determination of the approach to the construction of the neural network algorithm for detection of emergency and pre-emergency situations by multicriterial electro-optical system was performed. The analysis for forecasting and recognition of situations in the coal mine is carried out. Two approaches on forecasting of situations – autoregression and neural networks are analyzed. The advantages of neural networks to forecasting of situations were determined. Two approaches on recognition of situations: on dynamics of change concentration of one gas and on ratios of the several gases concentration are analyzed. Classification of situations in the coal mine is formulated. Admissible concentration was define for methane – up to 2%, carbon oxide – up to 0.0017%, carbon dioxide – 1%, coal dust of 150 mg/m3. The limiting concentrations for situations in coal mine are calculated. It is proposed to use MatLab for modeling neural network. Approach to creation of neural network algorithm for forecasting situation in coal mine is proposed. Structure of the neural network for forecasting situation in the coal mine is developed.