Scholarlink

The Northeast part of India has a record of historic and devastating earthquakes which have left their mark in this region. This study has made an attempt to collect and analyse seismic data, which includes ground motion data from past earthquakes, to model the typical attenuation behaviour observed in the North-East Indian terrain. Unlike Spectral Acceleration (Sa), Peak Ground Acceleration (PGA) does not offer insights into how different frequencies of shaking affect structures. This paper throws light on the supremacy of Spectral Acceleration over PGA and sees how a structure or a building could behave when subjected to different time periods with different storeys of a building. The dataset of Ground motion includes data from COSMOS (Consortium of Organizations for Strong Motion Observation System) and PESMOS (Program for Excellence in Strong-Motion Studies). Strong Motion analysis has been carried out using Seismo-Signal, a strong motion processing software. An attempt has been made to use an Artificial Neural Network (ANN), which provides predicted values when observed data is fed into the Network and a comparison is made to see how Sa/g values vary with an increase in epicentral distance. The three input parameters taken were Magnitude, Epicentral Distance and Focal Depth and the output parameter was Sa/g. Here, Neural Network Tool (nntool) from MATLAB (R016a) is used, and the algorithm considered is Feed Forward back propagation (FFBP) in which output nodes are back-propagated through the connections to re-configure the network. The data is trained a number of times by changing the number of neurons in the hidden layer and seeing which network model could best suit Northeast India by giving the least possible error. The general architecture of the ANN model used is illustrated in this paper. In the dataset, Strong Motion data for the North-East part of India is available till 2013, and an effort has been made to predict the attenuation behaviour as correctly as possible. This paper aims to serve as an outlook on future research avenues, the establishment of protocols and standards for employing Artificial Intelligence (AI) in geotechnical engineering practices.

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

AI Application in Geotechnical Engineering Using Artificial Neural Network for Seismic Data Analysis in Northeast India

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Md. Ijaj Ahmed,
  • Sasanka Borah,
  • Jayanta Pathak

摘要

The Northeast part of India has a record of historic and devastating earthquakes which have left their mark in this region. This study has made an attempt to collect and analyse seismic data, which includes ground motion data from past earthquakes, to model the typical attenuation behaviour observed in the North-East Indian terrain. Unlike Spectral Acceleration (Sa), Peak Ground Acceleration (PGA) does not offer insights into how different frequencies of shaking affect structures. This paper throws light on the supremacy of Spectral Acceleration over PGA and sees how a structure or a building could behave when subjected to different time periods with different storeys of a building. The dataset of Ground motion includes data from COSMOS (Consortium of Organizations for Strong Motion Observation System) and PESMOS (Program for Excellence in Strong-Motion Studies). Strong Motion analysis has been carried out using Seismo-Signal, a strong motion processing software. An attempt has been made to use an Artificial Neural Network (ANN), which provides predicted values when observed data is fed into the Network and a comparison is made to see how Sa/g values vary with an increase in epicentral distance. The three input parameters taken were Magnitude, Epicentral Distance and Focal Depth and the output parameter was Sa/g. Here, Neural Network Tool (nntool) from MATLAB (R016a) is used, and the algorithm considered is Feed Forward back propagation (FFBP) in which output nodes are back-propagated through the connections to re-configure the network. The data is trained a number of times by changing the number of neurons in the hidden layer and seeing which network model could best suit Northeast India by giving the least possible error. The general architecture of the ANN model used is illustrated in this paper. In the dataset, Strong Motion data for the North-East part of India is available till 2013, and an effort has been made to predict the attenuation behaviour as correctly as possible. This paper aims to serve as an outlook on future research avenues, the establishment of protocols and standards for employing Artificial Intelligence (AI) in geotechnical engineering practices.