Estimating the strength properties of intact rock is a fundamental step in all engineering phases of civil and mining projects/operations. These properties can exhibit high variability and therefore require continuous evaluation to enhance the confidence in their estimation and reduce uncertainty.
This work introduces a spatial analysis methodology to improve the estimation of rock strength properties by accounting for local variations, where different strength envelope fitting parameters can occur in various locations within the study region, highlighting the importance of spatial variability.
The variability of these properties has significant implications for the design and stability of geotechnical projects/operations (open pit, underground and/or infrastructure among others). Therefore, exploring and improving these estimations is imperative.
To tackle this complexity, kernel-based spatial smoothing techniques and integrate artificial intelligence models are applied, specifically Graph Neural Networks (GNN), to model and predict rock strength properties with greater accuracy than traditional methods.
Combining spatial smoothing with advanced machine learning algorithms not only enhances the spatial understanding of rock mechanical properties, but also provides a tool for decision making in civil engineering and mining projects, opening new paths for future research and practical applications in rock mass characterization.
