基于广义回归神经网络的金属矿山水害危险性预测

Prediction of water hazard in metal mines based on general regression neural network

水害是金属矿山重大灾害之一,正确预测水害危险性及变化趋势,对矿山安全生产具有重要的指导意义。在分析水害影响因素和广泛参考其他预测模型的基础上,以矿山地质构造、裂隙发育状况、降水量、采空区面积、地下水水位、回采工艺、开采深度、有无预注浆、含水层层数、围岩岩性等10个影响因素作为预测输入,以有水害危险、无水害危险作为预测输出,建立了GRNN预测模型,并应用MATLAB软件编程,对某金属矿山样本进行了训练和预测。研究结果表明,预测结果最大误差仅为7.62%。GRNN模型预测精度高,对金属矿山水害危险性预测和防治具有借鉴意义。

Water hazard is one of the major disasters in metal mines. Therefore,proper prediction of water hazard and its trends bear great guiding significance for mining operations. Based on the analysis of influence factors on water hazard and comprehensive reference to other prediction models,the paper inputs 10 influence factors including geologic structures of mines,fissure development conditions,precipitation,goaf areas,ground water levels,extraction processes,mining depth,whether grouting or not,aquifer number and lithology of surrounding rocks,while outputs water hazard as ＂ positive＂ or ＂ negative＂. Thus a GRNN prediction model is established,and with MATLAB programming,training and prediction of samples from a metal mine are conducted. The results show that the maximum error is only7. 62 %. The GRNN model has high prediction precision and can provide reference for the prediction and prevention of water hazard in metal mines.