中文摘要：

为解决新桥矿大块率高、炸药单耗高及爆破效率低等问题,在对爆破工艺改进的基础上设计有限的爆破试验（13组试验）获取样本,并建立BP神经网络预测模型（隐含层节点数取9）,以最小抵抗线W、孔间距a、周边孔距Z作为输入因子,以炸药单耗、大块率作为输出因子预测、优选爆破参数。优化推荐W=0.8 m、a=1 m、Z=0.8 m,对应的炸药单耗为0.2001 kg/t,仅为原工艺的50%;大块率为5.2091%,仅为原工艺的20%;生产效率提高了约65%。该方法采用有限的试验与智能预测相结合,实现低成本获取真实样本,并提高了预测精度。

英文摘要：

In order to solve the problem of high boulder yield, high explosive specific charge and low blasting efficiency of Xin-qiao Mine, 13 samples were obtained from limited blasting tests on the basis of improved blasting craft. The blasting parameters were predicted by 9 hidden layer nodes＇ BP neural networks with the minimum burden W, hole spacing a, peripheral hole distance Z as the input factors and with the explosive cost, block rate as the output factor. The recommended parameters were W = 0.8 m, a = 1 m, Z = 0. 8 m, and the explosive specific charge was 0. 2001 kg/t,only 50% of the original process; the block rate was 5. 2091% ,only 20% of the original process; the production efficiency was promoted about 65%. Combined with finite test and intelligent prediction, the method a- chieved low cost and real samples, in addition, the prediction accuracy was improved.