Scholarlink

<p>The Himalayan region has experienced major earthquakes in the past, necessitating a robust characterization of ground motion for accurate seismic hazard assessment and infrastructure resilience. However, the scarcity of available records in the seismically active regions poses a significant challenge in developing region-specific ground motion models (GMMs). This study employs a transfer learning framework to develop a GMM for multiple ground motion parameters (GMPs) in the Himalayan region. These GMPs collectively define the amplitude, duration, or frequency characteristics of the ground motions. Initially, an artificial neural network (ANN)-based global GMM is developed using an extensive dataset of recorded ground motions from the PEER NGA-West2 database. The global GMM is subsequently adapted and calibrated using Himalayan ground motion records through transfer learning. The developed ANN-based model accounts for random effects by employing the mixed-effect regression with a nonparametric approach. Model performance is rigorously evaluated through several statistical measures to ensure robustness and predictive accuracy. This global model is adjusted utilizing sparse recorded data in the Himalayan region of India using a transfer learning framework. A thorough residual analysis and performance evaluation through recorded data, as well as regional GMM, are conducted on this model, and its predictive reliability for the Himalayan region is confirmed. The resulting regionally adjusted GMM represents the first comprehensive model developed for multiple GMPs for the Himalayan region of India.</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

Ground motion parameters of Himalayan earthquakes using transfer learning and simulation of time history

  • 期刊论文

  • 发表日期：

    2026年2月18日

查看全文
  • Ajay Kumar Pathak,
  • Ravi Kanth Sriwastav,
  • S. T. G. Raghukanth

摘要

The Himalayan region has experienced major earthquakes in the past, necessitating a robust characterization of ground motion for accurate seismic hazard assessment and infrastructure resilience. However, the scarcity of available records in the seismically active regions poses a significant challenge in developing region-specific ground motion models (GMMs). This study employs a transfer learning framework to develop a GMM for multiple ground motion parameters (GMPs) in the Himalayan region. These GMPs collectively define the amplitude, duration, or frequency characteristics of the ground motions. Initially, an artificial neural network (ANN)-based global GMM is developed using an extensive dataset of recorded ground motions from the PEER NGA-West2 database. The global GMM is subsequently adapted and calibrated using Himalayan ground motion records through transfer learning. The developed ANN-based model accounts for random effects by employing the mixed-effect regression with a nonparametric approach. Model performance is rigorously evaluated through several statistical measures to ensure robustness and predictive accuracy. This global model is adjusted utilizing sparse recorded data in the Himalayan region of India using a transfer learning framework. A thorough residual analysis and performance evaluation through recorded data, as well as regional GMM, are conducted on this model, and its predictive reliability for the Himalayan region is confirmed. The resulting regionally adjusted GMM represents the first comprehensive model developed for multiple GMPs for the Himalayan region of India.