Performance Analysis in Blast Induced Ground Vibrations with Advanced Machine Learning Algorithm Techniques

For mining and construction, it is very important to analyze blast-induced ground vibrations because of the risks they pose, such as structural damage and environmental disturbance. Machine learning provides powerful tools for analyzing these complex, non-linear signals, allowing accurate prediction and classification of vibration patterns. Deep learning and related optimization procedures offer opportunities for improving the vibration impact assessment procedures. In the research work, the SMO algorithm is applied to perform the dimensionality reduction of piezoelectric sensors (shielded and unshielded). The reduced features are then classified using Decision Tree and Random Forest methodologies. Random Forest performs even better at 97.92% accuracy when a shielded sensor is used, as opposed to only 95.84% accuracy when an unshielded sensor is used, thus demonstrating superior overall efficiency.