Scholarlink

<p>Accurate load sharing in tie beam foundations represents one of those challenges a structural engineer is bound to encounter regarding the transpiring infrastructure’s structural safety and reliability. Conventional approaches generally remain confined to empirical correlations alone, which are found unable to appropriately address high interaction among the individual properties of the soil and configuration and dimensions of the supporting units and environmental variations. The ML models used in this paper bridge these gaps by combining them with optimization algorithms, increasing the efficiency and accuracy of the predictive performance. Data Description The study dataset contained 21 feature variables representing characteristics related to soil, structural parameters, and environmental conditions. In contrast, one variable is the target, which accounts for the load distribution factor. Three machine learning models are developed for the analysis: Random Forest, Gradient Boosting, and ANNs. Furthermore, optimization algorithms such as SWO, FFOA, EVO, and SAO were implemented to select the features and optimize the hyperparameters to improve performance. The results illustrated that the performance of the models improved much after optimization; ANN outperformed others with the best accuracy of R2 = 0.941 with a minimum error metric of RMSE and MAE. Gradient Boosting and Random Forest also showed enhancements that again evidence the transformation after its optimization. Contributions are threefold: proposing two new optimization techniques and developing a robust predictive framework for the structural engineering domain. In this way, this work underlines the potential currently offered by the combination of ML and optimization in solving complex challenges in engineering. These results set the stage for further research by expanding data sets, considering advanced algorithms, and applying this framework within a wide range of geological contexts that will improve safety and efficiency for engineering practices.</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

Predicting load distribution in tie beam-foundation systems using machine learning and nature-inspired optimization algorithms

  • 期刊论文

  • 发表日期：

    2025年3月5日

查看全文
  • Ahmad S. Alfraihat

摘要

Accurate load sharing in tie beam foundations represents one of those challenges a structural engineer is bound to encounter regarding the transpiring infrastructure’s structural safety and reliability. Conventional approaches generally remain confined to empirical correlations alone, which are found unable to appropriately address high interaction among the individual properties of the soil and configuration and dimensions of the supporting units and environmental variations. The ML models used in this paper bridge these gaps by combining them with optimization algorithms, increasing the efficiency and accuracy of the predictive performance. Data Description The study dataset contained 21 feature variables representing characteristics related to soil, structural parameters, and environmental conditions. In contrast, one variable is the target, which accounts for the load distribution factor. Three machine learning models are developed for the analysis: Random Forest, Gradient Boosting, and ANNs. Furthermore, optimization algorithms such as SWO, FFOA, EVO, and SAO were implemented to select the features and optimize the hyperparameters to improve performance. The results illustrated that the performance of the models improved much after optimization; ANN outperformed others with the best accuracy of R2 = 0.941 with a minimum error metric of RMSE and MAE. Gradient Boosting and Random Forest also showed enhancements that again evidence the transformation after its optimization. Contributions are threefold: proposing two new optimization techniques and developing a robust predictive framework for the structural engineering domain. In this way, this work underlines the potential currently offered by the combination of ML and optimization in solving complex challenges in engineering. These results set the stage for further research by expanding data sets, considering advanced algorithms, and applying this framework within a wide range of geological contexts that will improve safety and efficiency for engineering practices.