岩溶地质下露天深孔爆破飞石飞散距离的预测  

Prediction of Flyrock distance in Open-pit Deep Hole Blasting under Karst Geology

作　　者：王振毅 李静 钱至桥 肖强 WANG zhen-yi;LI Jing;QIAN Zhi-qiao;Xiao-qiang(Zhejiang Gaoneng Blasting Engineering Co.,Ltd.,Hangzhou 311500,China;School of Mining,Guizhou University,Guiyang 550002,China)

机构地区：[1]浙江省高能爆破工程有限公司,杭州311500 [2]贵州大学,贵阳550002

出　　处：《爆破》2022年第3期199-203,共5页Blasting

摘　　要：飞石是露天矿爆破作业中最危险的事件之一,目前爆破飞石的飞散距离主要通过几种经验公式预测,由于不同工程地质情况、爆破类型和参数的差异性很大,此类经验模型准确率不高,各种参数的权重关系未知是经验模型不准确的主要原因。目前使用人工智能等新方法预测爆破飞石飞散距离具有一定可操作性,由于贵阳清镇市樱桃井水泥用石灰岩矿山在爆破作业过程中发生了飞石飞散距离超出预期,造成周边受保护建筑物损坏的情况,研究人员利用人工神经网络、结合岩石性质、爆破方式和爆破设计参数,对该矿山爆破作业中的爆破参数、岩石性质和飞石飞散距离等参数进行了统计和分析。通过122个爆破试验数据训练了7个不同的神经网络建模,其中一个三层前馈-反向传播神经网络均方根误差最小,该网络由16个隐神经元组成,包含9个输入参数和1个输出参数;经对比,该网络模型的预测结果与实测数据平均相对误差在3.31%以内,具有指导意义。对该网络建模的敏感性分析,发现可爆性指数、延期时间、单段装药量、孔径、孔深、溶洞是对爆破飞石飞散距离影响最大的参数。Flyrock is one of the most hazardous events in open-pit mine blasting operations.At present,the flying distance of flyrock is mainly predicted by several empirical formulas.However,due to the great difference between different engineering,blasting types and parameters,the accuracy of this kind of empirical model is not high.Besides,the unknown weight relationship of various parameters is the main reason for the inaccuracy of the empirical model.To solve this problem,artificial intelligence and other new methods can be used to predict the flying distance of flyrock.During the blasting operation of Yingtaojing limestone mine in Qingzhen city,Guiyang,flying stones were scattered far beyond expectations,resulting in the damage of surrounding protected buildings.In this paper,an attempt is made to predict and control flyrock in blasting operation of this mine using artificial neural network method.Based on 122 blasting test data,7 different neural networks are trained to model.Among them,a three-layer feedforward back-propagation neural network has the lowest root mean square error,which was composed of 16 hidden neurons,including 9 input parameters and 1 output parameter.By comparison,the average relative error between the predicted results and the measured data is within 3.31%,which has guiding significance.The sensitivity analysis of the model shows that the parameters that have the greatest influence on the distance of blasting flyrock are explosiveness index,delay time,charge per delay,hole diameter,hole depth and karst cave.

关 键 词：工程爆破 溶洞 飞石飞散距离 人工神经网络 

分 类 号：TD235.3[矿业工程—矿井建设]