Patel, Pushpendra and Verma, Harsh Kumar (2026) Evaluation of Blast-Induced Ground Vibration Using Machine Learning for Safe Construction of Hydropower Tunnels: A Case Study. Journal of Rock Mechanics and Tunnelling Technology , 32 (2). pp. 161-175. ISSN 0971-9059

Full text not available from this repository.

Abstract

Large hydropower projects require extensive tunnel excavation, where the drill-and-blast method (DBM) is widely adopted for excavating hard rock formations. However, DBM generates blast induced ground vibrations (BIGV), which may adversely affect surrounding rock masses and nearby structures. Peak Particle Velocity (PPV) is widely used to evaluate vibration intensity and assess structural safety. This study develops machine learning (ML)-based PPV prediction models for tunnel blasting at the Tapovan-Vishnugad Hydroelectric Power Project, Uttarakhand. A dataset comprising 332 field-monitored blasting events was used with a 70:30 training–testing split. Three tree-based regression algorithms, namely Decision Tree (CART), Random Forest (RF), and Extreme Gradient Boosting (XGB), were developed using six input parameters: tunnel cross-sectional area (a), rock mass quality index (Q), blasthole length (L), charge per delay (W), specific charge (SC), and monitoring distance (R). Model performance was evaluated using the coefficient of determination (R²), root mean square error (RMSE), and mean absolute error (MAE), and compared with the conventional USBM empirical predictor. SHapley Additive exPlanations (SHAP) analysis was performed to evaluate parameter influence, while a parametric optimization framework was developed for safe charge estimation under varying blasting conditions. The results show that ML models significantly outperform the empirical approach. Among the developed models, RF achieved the best performance with R² values of 0.98 and 0.90 for training and testing datasets, respectively, along with the lowest RMSE and MAE values. SHAP analysis indicated that the influence of input parameters follows the order: R > W > Q > a > SC > L. The proposed framework provides a reliable approach for estimating safe charge per delay for different geological conditions and supports safer tunnel blasting practices.

Item Type:	Article
Uncontrolled Keywords:	Tunnel blasting; PPV prediction; Machine learning; Random forest; SHAP analysis.
Subjects:	Blasting
Divisions:	UNSPECIFIED
Depositing User:	Mr. B. R. Panduranga
Date Deposited:	08 Jul 2026 04:56
Last Modified:	08 Jul 2026 04:56
URI:	https://cimfr.csircentral.net/id/eprint/3047

Actions (login required)

View Item