摘要
渗流的不良影响会提高溃坝风险,利用机器学习精确预测渗流情况对于确保大坝的安全和稳定至关重要。综述了机器学习在大坝渗流预测中的应用、挑战和出路。机器学习不仅可以预测大坝的渗流性态,而且能够识别渗流预测中大坝渗透系数、地下水位等关键参数。人工神经网络、支持向量机、决策树等机器学习算法已经被广泛应用于大坝的渗流预测,集成算法通过整合多种算法的优势,大大提高了预测准确性。机器学习模型在数据数量和质量、模型可解释性、复杂性、可扩展性以及部署和实现等方面仍存在诸多不足。未来的研究方向包括开发先进的机器学习算法、创建物理-数据双驱动模型和可解释模型、加强试验测试和验证等。相关成果可为基于机器学习模型的大坝渗流预测研究提供参考。
The adverse effects of seepage will increase the dam failure risk,and applying machine learning to accurate seepage prediction is crucial to dam safety and stability.This paper reviews the application,challenges,and solutions of machine learning in dam seepage prediction.Machine learning can not only predict the seepage behavior of dams but also identify some key parameters such as dam permeability coefficient and groundwater level in seepage prediction.Artificial neural networks,support vector machines,and decision trees have been widely employed in seepage prediction of dams.Integrated algorithms greatly improve prediction accuracy by integrating the advantages of multiple algorithms.Machine learning models still have many shortcomings in data quantity and quality,model interpretability,complexity,scalability,deployment,and implementation.Future research directions include developing advanced machine learning algorithms,creating physics data dual-drive models and interpretable models,and enhancing experimental testing and validation.The relevant achievements can provide references for studying dam seepage prediction based on machine learning models.
作者
汪卫
廖杰林
朱少坤
WANG Wei;LIAO Jielin;ZHU Shaokun(Guangzhou PRHRI Engineering Survey&Design Co.,Ltd.,Guangzhou 510610,China)
出处
《人民珠江》
2024年第4期1-10,共10页
Pearl River
基金
广东省自然科学基金(2023A1515010754)。
关键词
大坝
渗流预测
机器学习
参数识别
dam
seepage prediction
machine learning
parameter identification
