Scholarlink

<p>This study presents a novel hybrid ensemble learning farmwork of improved grey wolf optimiser (IGWO) and gradient boosting regressor (GBR) to perform reliability analysis (RA) of pile foundations in cohesive soils at different design loads and coefficient of variation (COV) levels. Two well-established methods namely Monte Carlo and Subset simulations were used followed by automation of the process of RA using the proposed hybrid ensemble model, i.e., GBR-IGWO. The performance of the GBR-IGWO model was&#xa0;also compared with other ensemble learning algorithms (ELAs). A comparative analysis of Monte Carlo, Subset, and first-order second-moment method (FOSM) and FOSM-based ELAs was conducted to perform the RA of piles at different COV levels. The prediction outcomes demonstrated that the GBR-IGWO outperformed other ELAs with a determination coefficient of&#xa0;1 in the testing phase. The probability of failure was subsequently determined using Monte Carlo, Subset, FOSM, and GBR-IGWO approaches, considering the effect of uncertainty in soil properties. Moreover, a graphical user interface was developed and attached as supplementary material, allowing users to input parameters such as pile length, sample size, and design load to perform RA of piles with varying diameters and lengths, estimating failure probability across COV levels.</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

Risk Analysis of Pile Foundations Using an Improved Hybrid Ensemble Paradigm Coupled with Monte Carlo and Subset Simulations

  • 期刊论文

  • 发表日期：

    2025年3月18日

查看全文
  • Subodh Kumar Suman,
  • Shiva Shankar Choudhary,
  • Avijit Burman

摘要

This study presents a novel hybrid ensemble learning farmwork of improved grey wolf optimiser (IGWO) and gradient boosting regressor (GBR) to perform reliability analysis (RA) of pile foundations in cohesive soils at different design loads and coefficient of variation (COV) levels. Two well-established methods namely Monte Carlo and Subset simulations were used followed by automation of the process of RA using the proposed hybrid ensemble model, i.e., GBR-IGWO. The performance of the GBR-IGWO model was also compared with other ensemble learning algorithms (ELAs). A comparative analysis of Monte Carlo, Subset, and first-order second-moment method (FOSM) and FOSM-based ELAs was conducted to perform the RA of piles at different COV levels. The prediction outcomes demonstrated that the GBR-IGWO outperformed other ELAs with a determination coefficient of 1 in the testing phase. The probability of failure was subsequently determined using Monte Carlo, Subset, FOSM, and GBR-IGWO approaches, considering the effect of uncertainty in soil properties. Moreover, a graphical user interface was developed and attached as supplementary material, allowing users to input parameters such as pile length, sample size, and design load to perform RA of piles with varying diameters and lengths, estimating failure probability across COV levels.