Abstract
This study proposes an intelligent approach for classifying damage and deformation stages in backfill materials using acoustic emission (AE) monitoring and machine learning. Uniaxial compression tests were conducted to capture the evolution of AE parameters during backfill failure, and sensitivity analysis was performed to identify key indices reflecting damage progression. A back propagation (BP) neural network model was developed, with input features optimized to {rise time, count, energy, duration, amplitude} based on statistical evaluation of parameter sensitivity. The model was trained and validated using experimental AE data. The effectiveness of the approach for classifying damage and deformation stages in backfill was confirmed with a classification accuracy of 93.52%. This approach enhances the understanding of failure mechanisms in backfill materials and provides a reliable tool for real-time monitoring and early warning of backfill failure in filling mining.