基于ANN耦合遗传算法的爆破方案选择方法  

Selection of Blast Scheme Based on Coupling of Genetic Algorithm and Artificial Neural Network

作　　者：崔铁军[1,2] 马云东[2] 白润才[3] 

机构地区：[1]辽宁工程技术大学安全科学与工程学院,辽宁阜新123000 [2]大连交通大学辽宁省隧道与地下结构工程技术研究中心,辽宁大连116028 [3]辽宁工程技术大学资源与环境工程学院

出　　处：《中国安全科学学报》2013年第2期64-68,共5页China Safety Science Journal

基　　金：辽宁省重点实验室项目(LS2010075)

摘　　要：为预防露天矿爆破引起的事故,基于安全和经济方面的考虑,对露天矿爆破方案进行优化选择,提出将人工神经网络(ANN)模型与遗传算法(GA)相耦合,从而进行爆破方案优化。研究露天矿爆破可能引起的2种主要危害形式:超爆和飞石,进而确定超爆深度和飞石距离为爆破方案的被优化目标参数。另一方面,炮眼深、间距、装药深度、阻塞深度、单位炸药消耗量和钻孔率对超爆深度和飞石距离的影响是复杂的、非线性的,因而将其作为爆破方案的影响参数。分别用影响参数和目标参数作为ANN的输入值和输出值加以训练,训练后的ANN数据作为GA的适应度函数进行方案优化。结果表明:可找到符合工程实例数据的爆破方案集合,借助Pareto图,可得到相关参数的值最小(超瀑深度为0.6999m,飞石距离为27.3386m)的最优爆破方案。In order to prevent the accidents caused by open iron mine blast/ng, taking safety economic factors into consideration, a method based on coupling of ANN and GA was worked out and to optimize and select the blast scheme. The backbreak and flyrock, the two main blasting damaging and used phenomena, were researched, and then backbreak depth and flyrock distance were taken as object parameters; hole length, spacing, burden, stemming, powder factor and specific drilling were taken as impact parameters. The effects of impact parameters on backbreak depth and rlyrock distance are complex nonlin- ear. The object parameters and the impact parameters were taken respectively as ANN input value and out- put value to train it. The data of the artificial neural network trained as the fitness function of GA were used to optimizing the blasting scheme. The results show that the blasting scheme set according with actual data can be found with the method, and that with Pareto, the optimal blasting scheme （flyrock distance 27. 338 6 m and backbreak depth 0. 699 9 m） can be obtained.

关 键 词：人工神经网络(ANN) 遗传算法(GA) 适应度函数选择 采矿爆破 爆破方案选择 

分 类 号：X932[环境科学与工程—安全科学]