光面爆破效果参量预测的投影寻踪回归模型及应用
来源期刊：中南大学学报(自然科学版)2012年第10期
论文作者：李启月 徐敏 董陇军 钱守一
文章页码：3976 - 3981
关键词：光面爆破；爆破效果评价；投影寻踪回归；参量预测
Key words：smooth blasting; blasting effect evaluation; projection pursuit regression; parameter prediction
摘 要：基于投影寻踪回归理论，选取炮眼密集系数、最小抵抗线、装药量、炮眼深度、抗压强度和泊松比6个因素为判别指标，建立光面爆破炮眼利用率、超挖量和半眼率的预测模型。将该方法应用于某矿巷道掘进光面爆破效果预测问题中，对现场实测的24组数据进行训练和检验，用另外4组现场数据作为预测样本进行测试。预测结果与实测情况较吻合。比较BP神经网络和投影寻踪回归2种方法对光面爆破炮孔利用率的预测结果，发现后者比前者预测结果更接近实际。以上研究表明：该方法回判估计性能良好，判别精度高，是一种预测光面爆破效果的有效方法，可以在实际工程中推广应用。
Abstract: Based on the projection pursuit regression theory, six factors i.e. borehole intensive coefficient, the minimum resistance line, charge amount, depth of blast hole, compressive strength and Poisson ratio, were selected as discriminating index to construct the model of predicting 3 factors i.e. borehole utilization ratio, excessive rock breakage and half hole ratio. The established model was applied to predict the effect of smooth blasting in one mine. 24 groups of data were trained and tested; the other 4 groups of data were tested as forecast samples. The predicted results were consistent with the measured ones. Compare the BP neural network and projection pursuit regression to predict bore hole utilization ratio, it can be found that the latter is more accurate. The result shows that the method of projection pursuit regression has high discriminating ability and more accurate prediction, so it is a new method to predict smooth blasting effect and can be applied to practical engineering.
李启月，徐敏，董陇军，钱守一
(中南大学 资源与安全工程学院，湖南 长沙，410083)
摘 要：基于投影寻踪回归理论，选取炮眼密集系数、最小抵抗线、装药量、炮眼深度、抗压强度和泊松比6个因素为判别指标，建立光面爆破炮眼利用率、超挖量和半眼率的预测模型。将该方法应用于某矿巷道掘进光面爆破效果预测问题中，对现场实测的24组数据进行训练和检验，用另外4组现场数据作为预测样本进行测试。预测结果与实测情况较吻合。比较BP神经网络和投影寻踪回归2种方法对光面爆破炮孔利用率的预测结果，发现后者比前者预测结果更接近实际。以上研究表明：该方法回判估计性能良好，判别精度高，是一种预测光面爆破效果的有效方法，可以在实际工程中推广应用。
关键词：光面爆破；爆破效果评价；投影寻踪回归；参量预测
LI Qi-yue, XU Min, DONG Long-jun, QIAN Shou-yi
(School of Resources and Safety Engineering, Central South University, Changsha 410083, China)
Abstract:Based on the projection pursuit regression theory, six factors i.e. borehole intensive coefficient, the minimum resistance line, charge amount, depth of blast hole, compressive strength and Poisson ratio, were selected as discriminating index to construct the model of predicting 3 factors i.e. borehole utilization ratio, excessive rock breakage and half hole ratio. The established model was applied to predict the effect of smooth blasting in one mine. 24 groups of data were trained and tested; the other 4 groups of data were tested as forecast samples. The predicted results were consistent with the measured ones. Compare the BP neural network and projection pursuit regression to predict bore hole utilization ratio, it can be found that the latter is more accurate. The result shows that the method of projection pursuit regression has high discriminating ability and more accurate prediction, so it is a new method to predict smooth blasting effect and can be applied to practical engineering.
Key words:smooth blasting; blasting effect evaluation; projection pursuit regression; parameter prediction
