基于PSO-BP神经网络的单位注浆量预测  

PSO-BP Neural Network Based Unit Grouting Volume Prediction

作　　者：陈泓 黄永辉[3] 张智宇 陈成志 CHEN Hong;HUANG Yonghui;ZHANG Zhiyu;CHEN Chengzhi(Faculty of Land Resources Engineering,Kunming University of Science and Technology,Kunming 650093,China;Advanced Blasting Technology Engineering Research Center of Yunnan Province Education Department,Kunming 650093,China;Faculty school of Electrical Engineering,Kunming University of Science and Technology,Kunming 650500,China)

机构地区：[1]明理工大学国土资源工程学院,昆明650093 [2]云南省教育厅爆破新技术工程研究中心,昆明650093 [3]昆明理工大学电力工程学院,昆明650500

出　　处：《有色金属(中英文)》2025年第2期288-297,共10页Nonferrous Metals

基　　金：国家自然科学基金资助项目(52064025,52164009);云南省重大科技项目(202202AG050014)。

摘　　要：帷幕注浆作为矿山控制地下水的重要手段之一,对矿山的安全生产十分重要,单位注浆量作为注浆效果的关键评价指标,具有不确定性。基于尖山磷矿帷幕注浆试验段注浆数据,进行单位注浆量影响因素相关性分析,分别构建单位注浆量卷积神经网络(CNN)、BP神经网络、遗传算法优化神经网络(GA-BP)和粒子群算法优化神经网络(PSO-BP)预测模型进行预测和准确性分析。结果表明:斯皮尔曼相关系数法和肯德尔相关系数法对单位注浆量影响因素分析结果一致,影响因素相关性由强到弱为:注浆持续时间、水灰比、注前透水率、注浆段长度、注浆压力、钻孔深度;PSO-BP神经网络模型预测效果明显优于另外三种预测模型,R^(2)达到0.94527,RMSE值分别降低80%、56%、49%;MAE值分别降低68.3%、48.6%、23.2%,验证了该模型的优越性。该模型能够更准确地对单位注浆量进行预测,对后续注浆工作的实施具有一定参考,可为帷幕注浆效果评价提供重要的指导建议。As one of the important means to control groundwater in mines,curtain grouting is very important for the safe production of mines.As a key evaluation index of grouting effect,the unit grouting volume is uncertain.Based on the grouting data of the curtain grouting test section of Jianshan Phosphate Mine,the correlation analysis of the influencing factors of unit grouting volume was carried out.The prediction models of unit grouting volume convolutional neural network(CNN),BP neural network,genetic algorithm back propagation neural network(GA-BP)and particle swarm optimization back propagation neural network(PSO-BP)were constructed for prediction and accuracy analysis.The results showed that the results of the influencing factors of the unit grouting volume analyzed by Spearman correlation coefficient method were consistent with the results analyzed by Kendall correlation coefficient method.The correlation of the influencing factors from strong to weak wass:grouting duration,water-cement ratio,pre-injection permeability,grouting length,grouting pressure,drilling depth.The prediction effect of PSO-BP neural network model was obviously better than that of the other three prediction models.The R^(2) reached 0.94527,and the RMSE values wer reduced by 80%,56%and 49%respectively.The MAE values were reduced by 68.3%,48.6%and 23.2%,respectively,which verified the superiority of the model.The model can predict the unit grouting volume more accurately,which has a certain reference for the implementation of subsequent grouting work,and can provide important guidance for the evaluation of curtain grouting effect.

关 键 词：帷幕注浆 单位注浆量 相关性分析 BP神经网络 粒子群优化算法 

分 类 号：TD745[矿业工程—矿井通风与安全]