为了在露天爆破中更准确地预测出飞石的抛掷距离,研究引入多科得分的概念,利用多科得分思维进化算法(Multidisciplinary Score Mind Evolutionary Algorithm,MSMEA)对 BP 神经网络(Back-Propagation Neural Network)进行优化并建立模型来预测飞石距离.通过分析隐含层神经元个数、种群规模、子种群规模、优胜及临时子种群个数建立了 64个多科得分思维进化算法优化BP神经网络模型(Back-Propagation Neural Network Optimized by Multidisciplinary Score Mind Evolutionary Algorithm,MSMEA-BP),并选取了其中最优的MSMEA-BP模型.为了验证预测模型的有效性,分别用MSMEA-BP模型、思维进化算法优化BP神经网络模型(Back-Propagation Neural Network Optimized by Mind Evolutionary Algorithm,MEA-BP)和 BP 神经网络模型对 10组爆破飞石距离进行预测.结果显示,MSMEA-BP模型得到的预测结果与真实值之间的平均相对误差、决定系数、均方根误差、均方根百分比误差分别达到3.67％、0.980 8、7.357 1、1.33％,依次优于MEA-BP模型和BP神经网络模型,表明在相同训练条件下,采用多科得分思维进化算法对BP神经网络模型进行优化,可以克服BP神经网络易陷入局部最优解的问题,进而显著提高模型的预测精度.该方法为预测爆破飞石距离提供了一个新思路.
Intelligent prediction of flyrock distance in open blasting
To enhance the accuracy of predicting open blast flyrock throwing distance,a multidisciplinary score concept was introduced in this study.The proposed Multidisciplinary Score Mind Evolution Algorithm(MSMEA)was utilized to optimize the BP neural network models for flyrock distance prediction.Initially,a dataset of 91 blasting flyrock instances was collected,including parameters such as hole depth,stemming,burden,maximum charge per delay,powder factor,rock rebound value,and flyrock distance.Besides,81 sets were employed to build the prediction models,while the remaining 10 sets were kept aside for model validation.Next,a total of 64 MSMEA-BP models were constructed by considering factors such as the number of neurons in the hidden layer,population,and subpopulation size,as well as the number of superior and temporary subpopulations.The optimal MSMEA-BP model was determined based on the R2 and RMSE values obtained from training and testing sets.Subsequently,the effectiveness of the prediction model was evaluated by comparing the performance of the MSMEA-BP model with the mind evolution algorithm optimization BP neural network model(MEA-BP model)and the standard BP neural network model.The MSMEA-BP model exhibited superior results,with an average relative error of 3.67％,a determination coefficient of 0.980 8,a root mean square error of 7.357 1,and a root mean square percentage error of 1.33％.These metrics outperformed the MEA-BP model and the BP neural network model,respectively.These findings highlight the significant improvement achieved by optimizing the BP neural network using the multidisciplinary score mind evolution algorithm over the single subject score mind evolution algorithm.This approach effectively mitigates the issue of the BP neural network falling into local optimal solutions and substantially enhances the accuracy of the prediction model.Ultimately,this method can be applied to predict the throwing distance of flyrock in open blasting operations.
NETL
NSTL
万方数据
