基于机器学习方法预测地浸过程中铀浸出金属量的变化  

Predicting the Variation of Uranium Leaching Metal Content in Ground-leaching Process Based on Machine Learning Methods

作　　者：余东原 罗跃 梁大业[3] 李立尧 YU Dongyuan;LUO Yue;LIANG Daye;LI Liyao(School of Water Resource and Environmental Engineering,East China University of Technology,Nanchang 330013,China;Key Laboratory of Nuclear Resource and Environment,East China University of Technology,Nanchang 330013,China;Inner Mongolia Mining Co.,Ltd.,CNNC,Huhhot 010010,China)

机构地区：[1]东华理工大学水资源与环境工程学院,南昌330013 [2]东华理工大学核资源与环境重点实验室,南昌330013 [3]中核内蒙古矿业有限公司,呼和浩特010010

出　　处：《有色金属（冶炼部分）》2024年第2期92-98,共7页Nonferrous Metals（Extractive Metallurgy)

基　　金：国家自然科学基金资助项目(42062017,42372295);江西省自然科学基金资助项目(20224BAB203038,20212BAB213006)。

摘　　要：在地浸采铀过程中,准确预测铀浸出金属量具有重要意义。使用多元线性回归、多层感知机(MLP)和随机森林(RF)多种机器学习方法分别建立预测模型。结果表明:1)相比于传统的多元线性回归算法,MLP和RF两种方法能够得到预测性能更好的模型。2)多层感知机(MLP)模型在预测铀浸出金属量变化的上性能表现最佳(R 2=0.91)。3)在相同的预测精度下,RF模型比MLP模型耗时更短,超参数设置更加简单。4)在以总流量和每平米铀量为铀浸出金属量的关键因素进行分析时,总流量对铀浸出金属量的权重占比为81.6%。In the process of in-situ leaching of uranium,it is of great significance to accurately predict the amount of uranium metal extracted.Multiple machine learning methods such as multiple linear regression,Multi-Layer Perceptron(MLP)and Random Forest(RF)were used to build prediction models.The results indicate that:1)Compared with the traditional multiple linear regression algorithms,MLP and RF methods can obtain models with better prediction performance.2)The MLP model has the best performance in predicting the change of uranium leaching metal content(R 2=0.91).3)With the same prediction accuracy,the RF model takes less time than that of the MLP model,and the hyperparameters setting is simpler.4)The weight ratio of the total flow rate to the amount of uranium metal leached is 81.6%when the total flow rate and the amount of uranium leaching metal content per square meter are taken as the key factors.

关 键 词：地浸采铀 线性回归模型 机器学习预测 

分 类 号：TL212.12[核科学技术—核燃料循环与材料]