Castro, R & Cuello, D 2018, 'Hang-up analysis and modelling for Cadia East PC1-S1 and PC2-S1', in Y Potvin & J Jakubec (eds), Caving 2018: Proceedings of the Fourth International Symposium on Block and Sublevel Caving, Australian Centre for Geomechanics, Perth, pp. 233-246, https://doi.org/10.36487/ACG_rep/1815_15_Castro
(https://papers.acg.uwa.edu.au/p/1815_15_Castro/)
Abstract: Cadia Valley operations is a gold–copper deposit located in Orange, New South Wales, Australia. Currently, two macroblocks are in production – PC1-S1 and PC2-S1 of Cadia East at undercut levels 4670 and 4475, respectively. The production rate during 2016 reached approximately 43 ktonne/day for PC1 and 19 ktonne/day for PC2. These macroblocks have implemented different variants of the block caving method and have different rock mass characteristics and depth. The PC1-S1 block was subjected to intensive preconditioning, hydraulic fracturing (from Gallery 5050) and DDE (blasting), for a column height of 400 m from the production level (4650 mRL). Finally, to propagate the caving effectively to surface, hydraulic fracturing was performed at 500 m below surface (maximum-depth hole). PC2-S1 was preconditioned only by hydraulic fracturing and is located 194 m below PC1. 
These two macroblocks are at different stages of maturity. PC1, a mature cave, has shown a fine fragmentation, which diminished notably as the cave back reached the surface in 2014. Consequently, there have been few observed hang-ups, resulting in a high production rate when compared to caving standards. On the other hand, PC2 has caving in progress and has shown a coarser fragmentation, and a large number of hang-ups. Therefore, prediction capabilities for hang-ups for PC2 is critical for planning purposes.
In order to understand and model hang-ups for PC2, a BCRisk® model was built. BCRisk is a methodology to assess key gravity flow-related risks based on logistic regression. A BCRisk model of hang-ups delivers the probability (P) that the hang-up rate (HUR) would exceed 1 event/1,000 tonnes, that is P(HUR > 1). A univariate statistical analysis indicated that the key variables to be considered were the accumulated draw height (m), the uniformity draw index and the rock mass rating (RMR). The same analysis indicated that the different lithologies observed at PC2 were not a key variable. Each of the key variables has a significant and relative impact on P(HUR > 1). An increase of 10 m on the draw height decreases the probability of hang-ups by 26%, an increase of the RMR in 10 units increases the probability by 30%, while an improvement of draw uniformity index (by 30%) decreases the probability by 13%. 
The P(HUR > 1) was compared to the hang-ups database of PC2 (which consider hang-up events measured during 2016). The model showed a good fit to the data with an 81% accuracy at predicting events in terms of the number of drawpoints and the percentage of active area that could present hang-up issues. 

Keywords: fragmentation, hang-up events, statistical modelling, mine data