Application of Neural Networks and Genetic Algorithm on Risk Assessment of Coalmine Fire Safety

摘要

The coalmine fires not only result in great economic loss and casualties, but also influence economy run and people' life seriously. Because of the nonlinear and great risk of coalmine fire risk. now our country has no explicit assessment model of it. According to fire control safety engineering and systematic safety engineering theory, risk evaluation index system for coalmine fire are established based on the comprehensive research and analysis of the fire causing factors at coalmine. Aiming at the irrational distribution of weight value of evaluation index caused by neural network's liability to local minimum, a new model for risk assessment of urban fire is established based on neural network and genetic algorithms. In this model, the likelihood of fire occurring and the severity caused by fire are regarded as input parameters and fire risk grade as output parameter. The paper uses grey clustering which is suitable insufficient samples to create the training GA-BP samples. By adopting error inverse arithmetic to train BP network, the risk grade range of fire is obtained, which effectively solves the dynamic and nonlinear characteristics of coalmine fire. The model is validated by an example, and the results indicate it has better application value in coalmine risk assessment. The model can give a good reference to safety management of coalmine fire control.