Document Type : Research Paper

Authors

• Amin Ahmadi ¹
• Reza Etesami ²

¹ Islamic Azad University Kerman Branch

² Shahid Bahonar University of Kerman, Kerman, Iran

Abstract

Blast-induced ground vibrations threaten structures and operations in mining environments, making accurate prediction essential for safety. This study develops a hybrid Particle Swarm Optimization (PSO)-Artificial Neural Network (ANN) model to predict ground vibrations at the Sarcheshmeh Copper Mine. Using 29 field records, the model relates four inputs—charge weight per delay, distance, stemming height, and number of hole-rows—to the Peak Particle Velocity (PPV). A compact single-hidden-layer network (3 neurons, 19 parameters) was adopted to limit overfitting, with PSO optimizing the weights and biases. The model achieved a training R² = 0.983 and, on representative test samples, R² ≈ 0.984 with an average absolute relative error of about 4.5%. While limited by dataset size, the results indicate that the hybrid PSO-ANN approach is a promising tool for vibration control in open-pit mining.

Keywords

• Artificial Neural Networks
• Particle Swarm Optimization
• Blast-induced Vibrations
• Prediction Modeling
• Hybrid Optimization
• Open-pit Mining