当前位置: X-MOL 学术Environ. Earth Sci. › 论文详情
Our official English website, www.x-mol.net, welcomes your feedback! (Note: you will need to create a separate account there.)
Soil liquefaction in seismic events: pioneering predictive models using machine learning and advanced regression techniques
Environmental Earth Sciences ( IF 2.8 ) Pub Date : 2024-03-16 , DOI: 10.1007/s12665-024-11480-x
Pouyan Abbasimaedeh

This study goes into the complex area of predicting soil liquefaction potential, a crucial issue in geotechnical engineering. The work presents a novel viewpoint on forecasting soil liquefaction occurrences by combining traditional geotechnical concepts with cutting-edge machine-learning approaches. The research maintains authenticity and credibility in its conclusions by using real-world datasets from actual liquefaction incidents. While Pearson and Spearman correlation analyses clarify the complex interrelationships between the variables, a fundamental descriptive analysis offers a thorough statistical summary. Three machine learning models—logistic regression, random forest, and SVM—that make use of cutting-edge computational methods constitute the basis of the analytical efforts. Twenty different parameter sets were examined, and among them, sets 4, 12, 2, 8, 16, and 19 stood out as being particularly interesting. Set 4 achieved an accuracy of 86 percent. According to the study's definition of the "Gray zone", a zone of caution should be followed when calculated probabilities are close to the 0.5 decision threshold. The model's reliability and the significance of several geotechnical elements are further highlighted through sensitivity analysis. The research highlights complementary approaches' usefulness in merging traditional geotechnical evaluations with cutting-edge data analytics, as recommended by literature and regulations. To sum up, this study combines conventional geotechnical concepts with cutting-edge analytical methods, providing essential insights for upcoming geotechnical projects in the face of seismic difficulties.



中文翻译:

地震事件中的土壤液化:使用机器学习和先进回归技术的开创性预测模型

这项研究涉及预测土壤液化潜力的复杂领域,这是岩土工程中的一个关键问题。这项工作通过将传统岩土概念与尖端机器学习方法相结合,提出了预测土壤液化发生的新颖观点。该研究通过使用来自实际液化事件的真实数据集来保持其结论的真实性和可信度。虽然皮尔逊和斯皮尔曼相关分析阐明了变量之间复杂的相互关系,但基本描述性分析提供了全面的统计摘要。利用尖端计算方法的三种机器学习模型——逻辑回归、随机森林和支持向量机——构成了分析工作的基础。检查了 20 个不同的参数集,其中第 4、12、2、8、16 和 19 组特别有趣。第 4 组的准确率达到 86%。根据该研究对“灰色区域”的定义,当计算的概率接近 0.5 决策阈值时,应遵循谨慎区域。通过敏感性分析进一步强调了模型的可靠性和几个岩土元素的重要性。该研究强调了补充方法在将传统岩土工程评估与尖端数据分析相结合方面的有用性,正如文献和法规所建议的那样。总而言之,这项研究将传统的岩土工程概念与尖端的分析方法相结合,为即将面临地震困难的岩土工程项目提供了重要的见解。

更新日期:2024-03-16
down
wechat
bug