The Accuracy of blasting vibration prediction using signature hole and site constant analysis (a case study: Blasting near crusher structure, limestone mining in Indonesia)

Ground vibration caused by blasting has become a common concern, especially when blasting is conducted near buildings or structures. Uncontrolled vibrations can potentially damage nearby infrastructure if appropriate control measures are not implemented. PT Tambang Semen SUKABUMI, a limestone mining company, is planning to carry out blasting activities near a crusher building located approximately 80 meters away. This building is classified as a Class 5 structure, with a maximum allowable Peak Particle Velocity (PPV) of 40 mm/s, in accordance with the Indonesian national standard SNI 7571:2023. To ensure compliance with this limit, a Signature Holes Analysis (SHA) was conducted using the Hanwha HEBS II electronic detonator initiation system. This analysis measured ground vibration from single-hole blasts, with the average vibration from three SHA holes recorded at 49 mm/s. A site constant analysis using the scaled distance formula yielded a multiple R value of 0.87. However, the ANOVA table showed a significance-F value of 0.24, which is greater than 0.05 indicating that the regression model used for prediction was not statistically significant. Using the SHA data and assuming a maximum of 30 blast holes, a recommended delay timing of 22 × 15 milliseconds was selected. This produced a predicted PPV of 17.25 mm/s. The comparison between predicted and actual PPV values demonstrates that the delay configuration derived from Signature Hole Analysis provides reliable results. The observed deviations range from -16% to +29%, with an average deviation of only 8%, indicating a high level of accuracy and consistency. Using the actual blast data, three model prediction were applied for predicting the most accurate prediction that close to the actual value. The SHA method, utilizing actual single-hole vibration measurements and optimized delay timing, provided the lowest average deviation of ±8% and the smallest absolute error (2.45 mm/s) from actual measurements, demonstrating high reliability for sites with sufficient signature data. Meanwhile, the log-log regression model yielded the highest coefficient of determination (R² = 0.751) and the smallest prediction deviation (3%) when standard deviation margins were considered, making it highly effective for predictive modeling.