基于神经网络优化遗传算法的爆破参数优化

Optimization of blast parameters using genetic algorithm optimized by artificial neural network

对于露天铁矿开采而言,爆破是一项重要且危险的开采活动。从安全的角度看,爆破的两种主要危害形式为超爆和飞石。影响爆破方案的因素有很多,包括矿石的自然属性和爆破方案等。对于同一矿场主要考虑的爆破方案参数包括:炮眼深(HL)、间距(S)、装药深度(B)、阻塞深度(ST)、单位炸药消耗量(PF)和钻孔率(SD)。它们对超爆深度(BB)和飞石距离(FR)的影响是复杂的,非线性的。为了防止超爆和飞石造成事故,爆炸的参数必须在安全、经济的范围内。本文使用爆破影响因素和BB及FR分别作为神经网络(神经网络)的输入值和输出值加以训练,训练后的神经网络作为遗传算法(GA)的适应度函数。在使用GA寻找到最优的BB和FR,从而对爆破方案参数进行优化。结果表明该方法有效可行。

Selection of blast gy from blast is exhausted in pattern is a very important task in mining activities. A considerable amount of the ener- the form of unwanted damaging phenomena such as backbreak and flyrock. There are many factors influencing it, including ore natural attribute and blasting scheme, etc. For a mining area, the most in- fluential factors are the hole length （ HL）, spacing （ S）, burden （ B）, stemming （ ST）, powder factor （PF） and spe- cific drilling （SD）. The effect on backbreak depth （BB） and flyrock distance （FR）is complex nonlinear from these parameters. In order to prevent the backbreak and flyrock caused accidents, the parameters must be in security and economic range. This paper uses blasting pattern parameters and BB ＆ FR as neural network （神经网络） input and output to training respectively. Make the neural network after training be a genetic algorithm （GA） of the fit- ness function and search for the optimal BB and FR with GA, so as to optimize the parameters of blasting pattern.