基于灰色关联分析及GA BP模型的岩体爆破块度预测

在土石坝筑坝材料的爆破开采过程中，准确预测岩体爆破块度并进行块度控制，可保证土石坝的填筑质量。结合长河坝工程的过渡料现场爆破试验，采用灰色关联分析法分析影响爆破块度的主要因素，以此选取孔距、不均匀系数等分别作为预测模型的输入、输出参数，并采用遗传算法(GA)优化反向传播(BP)神经网络，建立了预测爆破块度的GA-BP模型。该模型的工程应用结果显示，不均匀系数C_u、曲率系数C_c、分形维数D预测值的平均相对误差分别为5.918%、8.862%、2.867%，且预测级配曲线的线形及走向均与实际结果较为接近，表明预测效果良好。对比GA-BP模型与BP网络的预测结果发现，GA-BP模型预测值的平均相对误差更小，表明总体上GA-BP模型优于BP网络。

Prediction of Rock Blasting Fragmentation Based on Grey Correlation Analysis and GA-BP Model

In the process of blasting mining of rock-fill dam materials, accurately predicting and controlling the blasting fragmentation of rock mass can guarantee the filling quality of rock-fill dam. Combined with the field blasting test of the transition material in Changheba Project, the main influencing factors of blasting fragmentation were analyzed by using grey correlation analysis method, and then hole spacing and non-uniformity coefficient were selected as the input and output parameters of the prediction model respectively. GA-BP model of blasting fragmentation prediction was established by using genetic algorithm (GA) to optimize BP network. Application results of this model show that the average relative errors between the predicted values of the non-uniformity coefficient C_u, curvature coefficient C_c and fractal dimension D?and the actual values are 5.918%, 8.862% and 2.867%, respectively, and the line shape and trend of the predicted grading curve are close to the actual results, indicating that the prediction effect of grading is good. By comparing the predicted results of GA-BP model with those of BP network, the average relative error of GA-BP model is smaller, and the results show that GA-BP model is superior to BP network on the whole.