Scholarlink

<p>Vertical ground motions during earthquakes refer to the upward and downward movements of the ground triggered by seismic waves, influenced by factors such as fault type, earthquake depth, and surface soil composition. Despite their potential to significantly impact buildings, especially tall structures, vertical ground motions have historically been overlooked in earthquake engineering practices. The duration of ground motion plays a critical role in determining structural damage levels. The growing recognition of the vertical component’s significance and duration of ground motion in earthquake damage has led to the application of machine learning to analyze extensive data and improve our understanding of these significant forces. This study investigates 630 vertical ground motions collected from the Himalayan region to analyze statistical correlations between intensity measures and duration metrics using various machine learning models. Linear regression models were commonly used in earthquake engineering, often struggling to capture the complexity of real-world data. Machine learning methods, however, offer a robust alternative capable of uncovering hidden patterns and unique ground motion characteristics. Various Machine learning models were investigated, and the best-performing model was then optimized using hyperparameters. The findings reveal that the Optimizable Ensemble model demonstrates superior accuracy in predicting Significant Duration, while the Optimizable Gaussian Process Regression (GPR) model excels in Bracketed Duration predictions, as validated through the holdout technique.</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

Machine learning based prediction of vertical ground motion duration using hyperparameter optimization

  • 期刊论文

  • 发表日期：

    2026年2月1日

查看全文
  • Chaitanya Bhargav Nerella,
  • Chaturya Ganne,
  • Swati Negi,
  • Jayaprakash Vemuri

摘要

Vertical ground motions during earthquakes refer to the upward and downward movements of the ground triggered by seismic waves, influenced by factors such as fault type, earthquake depth, and surface soil composition. Despite their potential to significantly impact buildings, especially tall structures, vertical ground motions have historically been overlooked in earthquake engineering practices. The duration of ground motion plays a critical role in determining structural damage levels. The growing recognition of the vertical component’s significance and duration of ground motion in earthquake damage has led to the application of machine learning to analyze extensive data and improve our understanding of these significant forces. This study investigates 630 vertical ground motions collected from the Himalayan region to analyze statistical correlations between intensity measures and duration metrics using various machine learning models. Linear regression models were commonly used in earthquake engineering, often struggling to capture the complexity of real-world data. Machine learning methods, however, offer a robust alternative capable of uncovering hidden patterns and unique ground motion characteristics. Various Machine learning models were investigated, and the best-performing model was then optimized using hyperparameters. The findings reveal that the Optimizable Ensemble model demonstrates superior accuracy in predicting Significant Duration, while the Optimizable Gaussian Process Regression (GPR) model excels in Bracketed Duration predictions, as validated through the holdout technique.