基于PCA-BP神经网络的岩石爆破平均粒径预测

Prediction of Mean Particle Size of Rock Blast based on Combination of PCA and BP Neural Networks

岩石爆破平均粒径的预测对岩石采装及二次破碎具有重要意义,然而常规的神经网络预测岩石爆破平均粒径存在较大的误差。为更加合理准确预测岩石爆破粒径分布,选取台阶高度与钻孔荷载比(H/B),间距与荷载比(S/B),荷载与孔径比(B/D),炮泥与荷载比(T/B),粉因数(Pf),弹性模量(E)和现场块度大小(XB)7个主要影响岩石爆破粒径的因素,并结合BP神经网络较好的预测性,以及主成分分析(PCA)能消除自变量间的相关性和减少BP神经网络输入数据的特点,建立了基于PCA-BP神经网络的岩石爆破粒径预测模型。以48组实测数据为例,对7个影响因素进行主成分分析,最终降为4个主成分,进而将其作为BP神经网络的输入因子,对岩石爆破粒径进行了预测。结果表明:BP神经网络与最小二乘法预测的平均误差分别为15.71%、27.32%,而PCA-BP神经网络预测平均误差仅为9.21%,实现了对岩石爆破粒径的较准确预测。综上所知,PCA-BP神经网络模型为岩石爆破平均粒径预测提供了一种科学、可靠的方法。

The prediction of mean particle size of rock blast is of great importance in rock transportation and secondary rock breaking. However,the conventional neural networks have relatively large errors in mean particle size prediction. To predict the particle size distribution of rock blast fragment more precisely and accurately,seven main factors influencing the particle size of rock blast fragment were chosen,including the ratio of bench height to drilled burden（ H / B）,ratio of spacing to burden（ S / B）,ratio of burden to hole diameter（ B / D）,ratio of stemming to burden（ T/B）,powder factor（ Pf）,modulus of elasticity（ E） and in-situ block size（ XB）. In addition,the BP neural networks has a good predictability and the principal component analysis（ PCA） can eliminate the correlation between independent variables and reduce the BP neural network input data. Then the particle size of rock blast fragment predict model was built,combines with the PCA and BP neural networks. Furthermore,the PCA-BP method was tested on 48 group data,and a principal component analysis was performed on the seven factors which were eventually reduced into four main factors. Then the four factors were used as BP neural networks input to predict particle size of rock blast fragment. Results show that,the average errors of BP neural networks and the least square method are 15.71% and 27. 32%,while the same reference value of PCA-BP neural networks prediction is only 9. 21%. In conclusion,the PCA and BP neural networks model provides a scientific and reliable method for the prediction of mean particle size of rock blast fragment.