356 AI-DRIVEN DETECTION OF LOW SLOPE MOVEMENTS: A TRANSDISCIPLINARY APPROACH TO SAFETY IN OPEN PIT MINING *R. Rimmelin1, J. Calderon2, Rachael Barnes3 1BHP, Resource Centre of Excellence, Australia, (Rigoberto.rp.rimmelin@bhp.com) 2BHP, Resource Centre of Excellence, Australia 3BHP, Decision Science, Australia ABSTRACT This study presents a transdisciplinary collaboration between geotechnical and technology teams to enhance safety in open pit mining through artificial intelligence and data driven decision making. Using ground-based radar data, slow movement analysis was applied to detect low velocity slope displacements that may precede failure. Advanced feature engineering and hierarchical clustering techniques were used to identify movement patterns, which were validated by geotechnical experts and categorised into failure, pre failure, and stable classes. This approach enabled the development of predictive models and interpretive rules that distinguish between benign and concerning movements. By integrating data science into geotechnical workflows, the study supports earlier hazard recognition and proactive risk mitigation. The methodology demonstrates how combining domain expertise with machine learning can extend the available warning lead time, improve slope hazard recognition, and support proactive risk mitigation. The work highlights the critical role of effective transdisciplinary collaboration in translating analytical capability into trusted engineering decision support. KEYWORDS Transdisciplinary collaboration; geotechnical monitoring; slow movement analysis; ground-based radar; hierarchical clustering; early warning systems; interpretability; human in the loop; World Mining Congress 1. CONTEXT AND PROBLEM STATEMENT Open pit slope stability represents a persistent and high-consequence risk in surface mining operations. Advances in ground-based radar technology have significantly improved the detection of rapid slope deformation and imminent failure. However, these systems are intrinsically optimised for short time scale behaviour and are less sensitive to very low displacement rates that evolve over weeks to months. Research across multiple industries reveals that analytics initiatives often fail not due to a lack of technical capability, but rather due to a misalignment between analytical outputs and decision-making contexts. Studies in team science and organisational analytics consistently identify linear handover models, opaque algorithms, and weak integration with domain expertise as primary causes of failure (National Research Council, 2015; Bergmann et al., 2021). In engineering domains, this disconnect is amplified when analytical outputs cannot be reconciled with physical understanding or operational constraints. Low velocity slope movement exemplifies this challenge. Such behaviour is often ambiguous, spatially diffuse, and easily rationalised as benign deformation. Without structured analytical support and expert interpretation, early precursors are frequently recognised only in hindsight. This creates a narrow
RkJQdWJsaXNoZXIy MTM0Mzk2