BP神经网络预测外加剂对充填体强度的影响  

Influence Prediction of Admixture on the Strength of Backfill Body by BP Neural Network

作　　者：魏微[1,2] 高谦[1,2] 何环莎[1,2] 

机构地区：[1]北京科技大学土木与环境工程学院,北京100083 [2]金属矿山高效开采与安全教育部重点实验室,北京100083

出　　处：《矿业研究与开发》2014年第1期18-21,共4页Mining Research and Development

基　　金：国家重点基础研究发展计划项目(2010CB731501);长江学者和创新团队发展计划项目(IRT0950);河北省钢铁产业技术升级专项资金项目(SJGS-KJ-12-03)

摘　　要：外加剂对充填体强度影响复杂,具有非线性特性,用数理统计的方法建立充填体强度与外加剂之间的关系模型很困难。因此,首先开展了全尾砂充填体强度与外加剂掺量的正交试验;然后采用BP神经网络进行试验样本的学习训练,建立充填体强度与外加剂种类及掺量之间的关系模型。结果表明,采用BP神经网络建立的预测模型,不仅对学习样本的预测精度高,更重要的是对测试样本的预测精度同样高,预测的最大误差仅为4.16%。实践证明,BP神经网络预测模型可以提高实验工作效率,节省人力、物力,为充填体添加外加剂的研究提供了一条有应用前景的理论设计途径。The influence of the admixture on the backfill strength is complex, with nonlinear characteristics. It is very difficult to establish the relationship model between the backfill body strength and the admixture by the method of mathematical statistics. Therefore, orthogonal test with the backfilling strength and admixture content was carried out firstly. Then BP neural network was used for learning and training of test samples in order to establish the relationship model between the backfill body strength with the type and dosage of the admixture. The results showed that high pre- diction precision was obtained in this model not only fromtraining samples hut also from test samples, whose maxi- mum predicted error was only 4.16%. This model was proved to improve the efficiency of experimental work and save manpower and materials, which provided a promising theoretic design approach in studying admixtures of backfill body.

关 键 词：BP神经网络 外加剂 充填体 抗压强度 

分 类 号：TD679[矿业工程—矿山机电]