Hybrid Machine Learning Model for Blast-Induced Peak Particle Velocity Estimation in Surface Mining

Strategic Insights -

A recent study conducted at Kyushu University in Japan has made significant advancements in the field of artificial intelligence by proposing a hybrid machine learning model for estimating peak particle velocity (PPV) in surface mining operations. The research team, led by Kesalopa Gaopale, developed a Sparrow Search Algorithm-optimized Artificial Neural Network (SSA-ANN) that outperformed existing methods, such as Genetic Algorithm-optimized ANN (GA-ANN) and empirical formula (USBM), in predicting PPV. The study highlights the importance of incorporating rock mass characteristics, particularly GSI (Geological Strength Index) and UCS (Unconfined Compressive Strength), to improve the model's accuracy and geotechnical sensitivity.

Key Takeaways:

  • The SSA-ANN model demonstrated superior prediction accuracy, achieving an average R^2 of 0.51 using bootstrap validation, surpassing GA-ANN (0.41) and standard ANN (0.26).
  • The incorporation of GSI enhanced the model's geotechnical sensitivity, allowing for more accurate predictions of PPV.
  • The research concluded that the application of SSA-ANN alongside comprehensive rock mass characteristics can substantially decrease uncertainty in PPV prediction, enhancing safety within the blast area and improving vibration control methods in blasting operations.
  • The study was funded by Jica and conducted by a team of researchers from Kyushu University, including Kesalopa Gaopale, Takashi Sasaoka, Akihiro Hamanaka, and Hideki Shimada.
  • The researchers proposed a new approach to estimating PPV using a hybrid machine learning model, which has the potential to improve safety and efficiency in surface mining operations.

Statistics:

  • Average R^2 of SSA-ANN model: 0.51
  • Average R^2 of GA-ANN model: 0.41
  • Average R^2 of standard ANN model: 0.26
  • Bootstrap validation used to evaluate model performance
  • The study was published in the journal Algorithms, volume 18, issue 9, page 543

Sources:

  • "Hybrid Machine Learning Model for Blast-Induced Peak Particle Velocity Estimation in Surface Mining: Application of Sparrow Search Algorithm in ANN Optimization." Algorithms, 2025, 18(9):543. doi: 10.3390/a18090543.
  • Kesalopa Gaopale, Takashi Sasaoka, Akihiro Hamanaka, and Hideki Shimada. "Hybrid Machine Learning Model for Blast-Induced Peak Particle Velocity Estimation in Surface Mining: Application of Sparrow Search Algorithm in ANN Optimization." Algorithms, 2025, 18(9):543. doi: 10.3390/a18090543.