基于BP神经网络的矿山排土场滑坡预警模型

Landslide Early Warning Model for Mine Dumping Site Based on BP Neural Network

排土场滑坡是矿山的重大灾害之一,威胁矿山的安全生产。提出了矿山排土场滑坡预警指标体系,针对综合预警指标,建立了基于BP神经网络的排土场滑坡预警模型。将黏聚力、内摩擦角、地表位移、内部位移、降雨量等10个指标作为网络的输入单元,选取合适的激励函数和学习步长,利用实际工程的20组样本数据,采用改进的梯度算法完成了BP神经网络的学习。应用学习好的预警模型对贵州某矿排土场进行了滑坡预警分析,结果与实际情况一致。该预警模型合理可靠,具有推广应用价值。

Dumping site landslide is one of the major disasters for mine,threaten the safety mining of mine. The early warning index system for mine dumping site landslide is proposed,according to the comprehensive early warning indexes,the dumpling site landslide early warning model based on BP neural network is established. The indexes including adhesion,internal friction angle,surface displacement,internal displacement,rainfall etc. are selected as the input units of BP network,appropriate excitation function and learning step are selected,and 20 sets of sample data for actual engineering are also selected for conducting the training and learning of BP neural network by using improved gradient algorithm. The learned and trained early warning model is applied to analyze the landslide of the dumpling site of a mine in Guizhou Province,the early warning analysis results is accordant with the actual situations of the dumping site,which show that the early warning model established in this paper is reasonable and has application value.