基于BP神经网络的凿岩爆破参数优选

为了得到合理的凿岩爆破参数,使用BP神经网络进行优选.经过简化,以炮孔间距和排距作为输入因子,以总炸药单耗作为综合输出因子;通过有限次的爆破正交试验,建立网络学习、训练样本,优选出最佳的网络模型;增加各输入因子水平,组合成预测、优选样本,从而搜索出最优的凿岩爆破参数.以新桥硫铁矿为例,优选出炮孔间距为1.30 m,排距为1.10 m;预测总炸药单耗为0.459 7 kg/t,比原炸药单耗(0.828 8 kg/t)降低44.53%.

Optimization of drilling and blasting parameters based on back-propagation neural network

Back-Propagation neural network was used to optimize the drilling and blasting parameters. In the process of simplification, the interval and row-space of holes were used as the input data and the sum of unit explosive consumption was confirmed to be the synthesized output data. Some learning and training samples were established by the numbered orthogonal blasting tests to get the best network mode. The best parameters were gotten using the selected network, according to the forecasted and optimized samples formed by combining the more levels of the parameters. BP neural network mode was used in Xinqiao Pyrite Mine. The results show that the best interval and row-space of holes are 1.30 m and 1.10 m, respectively, the sum of forecasted unit explosive consumption is 0.459 7 kg/t, which is 44.53% lower than the former (0.828 8 kg/t).