AI-Based Rockfall Prediction and Alert System for Open-Pit Mines

Abstract

Rockfalls threaten personnel and production in open-pit mines, yet radar-based monitoring is too costly for many small and public-sector operations. This paper proposes the Rockfall Prediction and Alert System (RPAS), an artificial intelligence framework that fuses digital elevation models, drone imagery, geotechnical sensor readings, and environmental data into a Probability-of-Failure (PoF) score for each slope sector. The design combines a convolutional branch for imagery and terrain, a recurrent (LSTM) branch for sensor trends, and a tree-ensemble branch for tabular risk factors, joined by a stacking meta-learner. Scores feed a risk-map dashboard and a four-tier SMS and email alert workflow running on low-cost, open hardware. Expected performance is framed against published benchmarks, and a phased field-validation plan is described.