Scholarlink

<p>The development of excess pore water pressure (PWP) in saturated loose sandy soils under cyclic loading is an ongoing subject of research in geotechnical engineering practice. It continues to be an important issue in the field, primarily because of its link to liquefaction and ground instability. In this paper we apply and validate data-driven symbolic regression models to develop predictive models for PWP generation in sandy soils subjected to dynamic loads. An extensive dataset including 1.375 data points from 125 separate cyclic triaxial tests was employed to train the models. Two alternative methods were implemented: (i) a symbolic regression-based empirical model; and (ii) a hybrid logarithmic model which integrated multiple soil and loading parameters, namely cyclic stress ratio (CSR), relative density (<i>D</i><sub><i>r</i></sub>), fines content (FC), and frequency of loading (<i>f</i>). To validate the models, new undrained cyclic triaxial testing was performed on clean Podima sand at two confining stresses (35 and 50&#xa0;kPa) using regular (sin) and irregular waveform inputs. The results indicate that both models are capable of producing very accurate predictions (R<sup>2</sup> &gt; 0.9) and demonstrate that the hybrid model has significant generalization ability when applied to various cyclic loading conditions. Furthermore, analysis indicates that the primary factors affecting PWP development are cycle-related parameters (N/N<sub>liq</sub>) whereas CSR, <i>D</i><sub><i>r</i></sub>, and FC are governing variables affecting the resistance and evolution process. Additionally, the application of symbolic regression provides improved predictive capability compared to established empirical relationships as well as contribute to developing physically meaningful predictive models. Overall, the work represents an advancement in the application of Explainable Artificial Intelligence (XAI) in geotechnical engineering. Additionally, the work provides a solid basis for the development of predictive models for liquefaction under realistic cyclic loading conditions.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Hybrid and symbolic regression-based modeling of pore water pressure generation in sands under cyclic loading

  • 期刊论文

  • 发表日期：

    2026年7月27日

查看全文
  • Berkay Ertekin,
  • Kamil Bekir Afacan

摘要

The development of excess pore water pressure (PWP) in saturated loose sandy soils under cyclic loading is an ongoing subject of research in geotechnical engineering practice. It continues to be an important issue in the field, primarily because of its link to liquefaction and ground instability. In this paper we apply and validate data-driven symbolic regression models to develop predictive models for PWP generation in sandy soils subjected to dynamic loads. An extensive dataset including 1.375 data points from 125 separate cyclic triaxial tests was employed to train the models. Two alternative methods were implemented: (i) a symbolic regression-based empirical model; and (ii) a hybrid logarithmic model which integrated multiple soil and loading parameters, namely cyclic stress ratio (CSR), relative density (Dr), fines content (FC), and frequency of loading (f). To validate the models, new undrained cyclic triaxial testing was performed on clean Podima sand at two confining stresses (35 and 50 kPa) using regular (sin) and irregular waveform inputs. The results indicate that both models are capable of producing very accurate predictions (R2 > 0.9) and demonstrate that the hybrid model has significant generalization ability when applied to various cyclic loading conditions. Furthermore, analysis indicates that the primary factors affecting PWP development are cycle-related parameters (N/Nliq) whereas CSR, Dr, and FC are governing variables affecting the resistance and evolution process. Additionally, the application of symbolic regression provides improved predictive capability compared to established empirical relationships as well as contribute to developing physically meaningful predictive models. Overall, the work represents an advancement in the application of Explainable Artificial Intelligence (XAI) in geotechnical engineering. Additionally, the work provides a solid basis for the development of predictive models for liquefaction under realistic cyclic loading conditions.