Natural safety prediction of non-coal mine accident based on BP neural network

Mine disaster system has the typical non-linear features. The traditional, previously function-setting evaluation methods and prediction methods have appeared their limitations. The BP neural network, with the nonlinear dynamic characteristics, eliminated the drift value brought about by man-made factors during the weight determination using the previous method. It is a promising natural safe-forecasting method. First, obtain the network weight parameters meets the convergence conditions through studying the known samples. Then using them as foundation to calculate mine forecast indicator system parameters, made safety prediction of forecast mines. The error between BP calculated predictive value and the actual value range from 2.22 to 5.54 percent, which showed that the training model is more accurate and reliable to forecast. The study contents have important guiding significance to mine safety management and scientific decision-making.