Scholarlink

<p>Actual prediction of blast induced ground vibrations (BIGV) is still a major challenge of the industry because of the interaction of several blast design parameters. In this study, an effort has been made to improve the prediction of peak particle velocity (PPV) by an integrated approach of statistical and machine learning techniques with field data of a mining site in Jharkhand, India. It considers data on the following parameters: distance (DT), maximum charge per delay (MCPD), total explosives (TE), number of holes (NoH), and hole depth (HD) and measures their effect on PPV. Artificial neural network (ANN) model is developed to incorporate the underlying complex and non-linear interdependencies between the input parameters. ANN performance has been validated using the coefficient of determination (R²) and the root mean square error (RMSE). The results show that DT is the most dominant factor, followed by MCPD and TE, while others also have a significant effect when taken together. The neural network (NN) exhibits higher prediction accuracy with R² = 0.960 and RMSE = 0.092, though the regression models provide simple equations suitable for easy field application. Finally, through multivariate regressions (MVR) the result of the ANN was revalidated in order to develop a site-specific empirical equation based on the significant parameters. This research implies that combining empirical and machine learning approaches significantly improves prediction accuracy. It also helps in understanding the combined impact of multiple variables on the ground vibration for the optimum blast designing.</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

Predictive modelling of blast-induced ground vibration: integrating neural networks and multivariate regression

  • 期刊论文

  • 发表日期：

    2026年7月27日

查看全文
  • Amar Prakash Kaushik,
  • Vivek Kumar Himanshu,
  • Ashish Kumar Vishwakarma,
  • C. Sawmliana,
  • A. K. Mishra

摘要

Actual prediction of blast induced ground vibrations (BIGV) is still a major challenge of the industry because of the interaction of several blast design parameters. In this study, an effort has been made to improve the prediction of peak particle velocity (PPV) by an integrated approach of statistical and machine learning techniques with field data of a mining site in Jharkhand, India. It considers data on the following parameters: distance (DT), maximum charge per delay (MCPD), total explosives (TE), number of holes (NoH), and hole depth (HD) and measures their effect on PPV. Artificial neural network (ANN) model is developed to incorporate the underlying complex and non-linear interdependencies between the input parameters. ANN performance has been validated using the coefficient of determination (R²) and the root mean square error (RMSE). The results show that DT is the most dominant factor, followed by MCPD and TE, while others also have a significant effect when taken together. The neural network (NN) exhibits higher prediction accuracy with R² = 0.960 and RMSE = 0.092, though the regression models provide simple equations suitable for easy field application. Finally, through multivariate regressions (MVR) the result of the ANN was revalidated in order to develop a site-specific empirical equation based on the significant parameters. This research implies that combining empirical and machine learning approaches significantly improves prediction accuracy. It also helps in understanding the combined impact of multiple variables on the ground vibration for the optimum blast designing.