Block caving is the heavyweight champion of underground mining methods. Engineers undercut an orebody, removing a layer of rock beneath it so that the overlying ore loses its support and fractures progressively under its own weight and gravity, feeding down toward drawpoints like a self-breaking conveyor of stone. The method’s appeal is obvious: enormous production volumes at low operating cost, ideally suited to the deep, low-grade, sprawling orebodies that increasingly dominate the resource base. But the method carries a brutal prerequisite. The rock must actually cave. If the rock mass is too strong, too intact, or too favorably oriented relative to the stresses squeezing it, the cave stalls, hangs up, or never initiates at all, stranding ore and investment underground.

That risk is especially acute in open-pit-to-underground transition mines, the category into which Haoyaoerhudong falls. The mine, located in Urad Middle Banner in Bayannur, Inner Mongolia, covers roughly 10 square kilometers and is planned to produce 8.25 million tonnes of ore per year. Two open pits trace the strike of a steeply dipping gold orebody. The western pit has already been mined out and backfilled with waste rock; the eastern pit is scheduled to cease production in 2028. Beyond the final pit walls, the deep ore will be won by block caving, with the eastern segment divided into 1050-meter and 840-meter levels and the western segment into 1400-meter and 1250-meter levels, mined from the top down.

The research team, led by Peng Liu with colleagues Lianku Xie, Yimeng Song, and Jiong Wang, built their assessment on an unusually rich foundation of field data. Sixteen discontinuity survey points on the exposed pit slopes recorded the orientation, trace length, and spacing of the dominant fracture sets. Seven representative drillholes provided core-based structural data. Borehole ultrasonic televiewer imaging across 13 boreholes, spanning elevations from +840 meters to +1650 meters, resolved the discontinuity patterns at depth. Hydraulic fracturing tests measured the in-situ stress field directly, and laboratory tests characterized the strength of the five main rock types: andalusite schist, limestone, metamorphic sandstone, biotite schist, and slate.

The structural picture that emerged is strikingly consistent. The rock mass is dominated by steeply dipping fracture sets, with dip angles generally between 60 and 85 degrees, oriented mainly northwest-southeast and northeast-southwest. Borehole imaging showed that more than 75 percent of discontinuities dip between 70 and 85 degrees, with the 75-to-80-degree range alone accounting for over 30 percent of the population. Gently dipping fractures below 45 degrees make up less than 5 percent. Joint spacing averages roughly half a meter, and it widens with depth as increasing confinement closes fractures and suppresses brittle cracking. In plan view, the western portions of both pits show systematically wider joint spacing than the eastern portions, flagging zones of more fractured, weaker rock.

From this database, the team constructed three-dimensional block models of rock quality in the 3DMine software, using 99 effective boreholes and a block size of 10 meters, refined to 5 meters where detail mattered. Inverse distance weighting interpolated the borehole attributes into the model, and where the nearest borehole lay more than 150 meters away, a K-nearest-neighbour procedure generated supplementary virtual boreholes to fill the gaps. The team then validated the interpolation with leave-one-borehole-out cross-validation, hiding each complete borehole in turn and predicting its values from the rest. The errors were quantified honestly: root mean square errors of 17.67 percent for RQD, 8.55 for RMR, 25.63 centimeters for joint spacing, and 11.11 for the caving-specific MRMR rating. Crucially, 98 percent of MRMR predictions landed in the same or an adjacent classification class, and uncertainty rose only modestly in data-sparse regions, from an RMSE of 10.93 near boreholes to 12.19 beyond the 150-meter threshold.

The headline result is a caveability verdict for the whole orebody. The mean Mining Rock Mass Rating from the 840-meter level to the surface is 54, placing the deposit firmly in the moderate-caveability class. Volumes with MRMR between 40 and 60 account for 53.52 percent of the evaluated rock, and the more favorable 20-to-40 range covers another 36.52 percent. Only 9.61 percent falls in the difficult 60-to-80 range. Most lithologies, including metamorphic sandstone, andalusite schist, slate, and biotite schist, sit in the moderate class, but limestone stands out with an MRMR of 61, meaning higher integrity, stronger self-supporting capacity, and a real risk of delayed cave initiation or hanging ground after undercutting.

The study then asked a subtler question: how much undercut exposure does each zone need to start caving and, just as importantly, to keep it going? Using the Laubscher caving chart, the predicted initial caving hydraulic radius ranges from 13 to 25 meters by elevation, with sustained caving radii of 26 to 39 meters, corresponding to sustained caving areas between roughly 11,300 and 25,800 square meters. The Mathews stability graph, which explicitly incorporates the measured stress field through its rock stress factor, tells a sharper story. It predicts sustained caving areas from 14,100 up to a remarkable 57,121 square meters, with the maximum demand concentrated in the 1200-to-1250-meter interval of the eastern orebody, where the stability number N is 3.24 and the required hydraulic radius reaches 59 meters. The discrepancy between the two methods is not a failure but a feature: Laubscher’s empirical chart smooths variation by caveability class, while Mathews exposes the local hot spots where stress, joint orientation, and rock quality conspire to resist caving.

The fragmentation analysis adds a third dimension to the picture. Using the BCF program, the team predicted how large the rock pieces would be at detachment and after their journey down the cave. In the Phase I mining area, initial fragment sizes range from 0.65 to 1.67 cubic meters, with the coarsest blocks predicted in the 1300-to-1350-meter interval, where rock integrity is high and structural segmentation weak. Phase II fares better, with initial fragments of 0.44 to 0.82 cubic meters. As blocks descend, caving pressure, collision, and repeated splitting break them down further: average caving fragment sizes shrink to 0.49 to 1.22 cubic meters in Phase I and 0.34 to 0.61 cubic meters in Phase II. The BCF model treats this breakage as a stochastic, cycle-by-cycle process, with splitting probability rising from about 10 percent for equidimensional blocks to certainty for blocks whose aspect ratio exceeds 10, and the pressure factor declining from 1.0 below 1 megapascal to 0.5 above 12 megapascals.

Perhaps the most valuable insight is the coupling the study reveals between caveability and fragmentation. The same geological factors that make rock hard to cave, namely high MRMR, high RQD, and wide joint spacing, also make it prone to producing oversized fragments that jam drawpoints and demand costly secondary breakage. The difficult-caving zones and the coarse-fragmentation zones are largely the same zones, and the 1300-to-1350-meter interval exemplifies both problems at once. The authors point to instructive international precedents: at the Esmeralda sector of Chile’s El Teniente mine, cave breakthrough occurred at a hydraulic radius near 26 meters with an active footprint of about 16,800 square meters, comfortably within the Laubscher predictions for Haoyaoerhudong, while at Northparkes E26 Lift 1 in Australia, caving that was expected at a hydraulic radius of 20 to 25 meters failed to sustain and required induced-caving measures. The lesson is that empirical charts are screening tools, not guarantees. For transition mines like Haoyaoerhudong, the study’s spatially integrated framework, linking borehole data, three-dimensional quality models, stress measurements, and fragmentation prediction in a single workflow, offers a way to see the trouble spots before the first undercut is fired, and to design the draw strategy, induced caving, and oversize management around them rather than discovering them one stalled cave at a time.

Subject of Research: Three-dimensional assessment of caveability and fragmentation for block caving at an open-pit-to-underground transition gold mine

Article Title: Integrated three-dimensional assessment of caveability and fragmentation for block caving in an open-pit-to-underground transition mine

Article References: Liu, P., Xie, L., Song, Y., & Wang, J. (2026). Integrated three-dimensional assessment of caveability and fragmentation for block caving in an open-pit-to-underground transition mine. Results in Engineering, 32, Article 113412. https://doi.org/10.1016/j.rineng.2026.113412

Image Credits: AI Generated

DOI: 10.1016/j.rineng.2026.113412

Keywords: block caving, caveability, fragmentation, open-pit to underground transition, MRMR, Mathews stability graph, Laubscher caving chart, BCF program, rock mass quality, in-situ stress, gold mining, 3D block modeling