Scholarlink

<p>This research evaluated an artificial neural network (ANN) model to predict the peak particle velocity (PPV) induced by underground metro operations. Numerical simulations of the train-track-tunnel-soil system were conducted under diverse parametric scenarios, generating a dataset of 300 PPV responses. A three-layered neural network architecture was developed and trained using three algorithms: Levenberg-Marquardt (LM), Bayesian regularization (BR), and scaled conjugate gradient (SC). A comparative study and conventional multiple linear regression (MLR) analysis were performed to identify the best-performing ANN algorithm based on the coefficient of correlation (R) and mean squared error (MSE). The findings revealed a significant concordance between the on-site recorded PPV and the numerically simulated PPV. In contrast, the multiple linear regression model exhibited a noticeable deviation in the predicted outcomes. Among the ANN algorithms, Bayesian regularization and Levenberg-Marquardt yielded the most appropriate results compared to the on-site measurements, while the scaled conjugate gradient algorithm displayed an error exceeding 30%. The ANN model presented in this study can be easily applied to predict ground-induced PPV from metro movement, offering a practical engineering solution to quantify ground-borne vibrations.</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

ANN-Based Approach for the Prediction of Peak Particle Velocity of Ground Induced by Underground Metro Operations

  • 期刊论文

  • 发表日期：

    2025年2月12日

查看全文
  • Namrata Bhattacharjee,
  • Arnab Sur,
  • Bappaditya Manna,
  • Arnab Banerjee,
  • J. T. Shahu

摘要

This research evaluated an artificial neural network (ANN) model to predict the peak particle velocity (PPV) induced by underground metro operations. Numerical simulations of the train-track-tunnel-soil system were conducted under diverse parametric scenarios, generating a dataset of 300 PPV responses. A three-layered neural network architecture was developed and trained using three algorithms: Levenberg-Marquardt (LM), Bayesian regularization (BR), and scaled conjugate gradient (SC). A comparative study and conventional multiple linear regression (MLR) analysis were performed to identify the best-performing ANN algorithm based on the coefficient of correlation (R) and mean squared error (MSE). The findings revealed a significant concordance between the on-site recorded PPV and the numerically simulated PPV. In contrast, the multiple linear regression model exhibited a noticeable deviation in the predicted outcomes. Among the ANN algorithms, Bayesian regularization and Levenberg-Marquardt yielded the most appropriate results compared to the on-site measurements, while the scaled conjugate gradient algorithm displayed an error exceeding 30%. The ANN model presented in this study can be easily applied to predict ground-induced PPV from metro movement, offering a practical engineering solution to quantify ground-borne vibrations.