文章摘要
郭崧.基于机器学习的边坡稳定性预测[J].矿产勘查,2026,17(S1):307-315
基于机器学习的边坡稳定性预测
Slope stability prediction using machine learning
投稿时间:2026-01-21  修订日期:2026-02-24
DOI:10.20008/j.kckc.2026S1035
中文关键词: 边坡稳定性  BP神经网络  随机森林  广义可加模型回归  思维进化算法
英文关键词: slope stability  Backpropagation neural network  random forest  generalized additive model  mind evolution algorithm
基金项目:本文受中煤湖北地质勘察基础工程有限公司项目“保康县店垭镇望粮山村山体滑坡治理工程设计”(2023-3034)、“华新骨料(武穴)有限公司一体化二期3000万吨骨料项目边坡设计”(2022-3027)联合资助。
作者单位邮编
郭崧* 湖北煤炭地质物探测量队,湖北 武汉 430200 430200
摘要点击次数: 4
全文下载次数: 1
中文摘要:
      为实现对边坡稳定性的精准预测,本研究系统对比了BP神经网络、随机森林及广义可加模型回归3种机器学习算法的预测性能。初步研究结果表明,BP神经网络在该问题上展现出最优的潜力,其预测误差最小,泛化能力最佳。然而,针对BP神经网络易陷入局部最优解的固有缺陷,本文进一步引入思维进化算法对其初始权重与阈值进行优化,构建了一种MEA-BP混合预测模型。通过与传统BP模型的对比分析证实,优化后的MEA-BP模型各项误差指标均显著降低,预测精度得到实质性提升。结论表明,利用MEA优化BP网络的初始化过程,能有效引导网络收敛至更优解,从而获得更精确的非线性映射关系和更强的泛化性能,为边坡稳定性分析提供了一种更为可靠的工具。
英文摘要:
      To achieve accurate prediction of slope stability, this study systematically compares the predictive performance of three machine learning algorithms: Backpropagation (BP) Neural Network, Random Forest, and Generalized Additive Model regression. Preliminary results indicate that the BP Neural Network demonstrates the highest potential for this specific problem, exhibiting the smallest prediction error and the best generalization capability. However, to address the inherent drawback of the BP Neural Network, namely its tendency to converge to local optima, this paper further introduces the Mind Evolution Algorithm (MEA) to optimize its initial weights and thresholds, thereby constructing a hybrid MEA-BP prediction model. Comparative analysis with the conventional BP model confirms that the optimized MEA-BP model achieves a significant reduction across all error metrics and a substantial improvement in prediction accuracy. The findings demonstrate that utilizing MEA to optimize the initialization process of the BP network effectively guides the network towards a superior solution, resulting in a more accurate nonlinear mapping relationship and enhanced generalization performance. This work provides a more reliable tool for slope stability analysis.
查看全文   查看/发表评论  下载PDF阅读器
关闭