In open-pit bench blasting operations, the excessive forward throw of the blast pile often leads to large-area coverage of haul roads, irregular pile contours, and increased safety risks. Traditional measurement methods such as RTK survey or total station are insufficient for capturing the full three-dimensional morphology of the muckpile efficiently. To address this challenge, we developed a synergistic approach combining drone technology with LS-DYNA numerical simulation to optimize the initiation location and regulate blast energy distribution. This paper presents our methodology and results obtained from a case study at an open-pit coal mine in Xinjiang, China.
Introduction
Our research team identified that the standard practice of bottom initiation resulted in blast piles extending up to 43.3 m on average, frequently covering the mining road and requiring additional clearing efforts. The primary cause was the uneven distribution of explosive energy along the bench face, particularly the excessive stress transmitted to the upper soft rock layers. To overcome this, we proposed to shift the initiation point to the center of the hard rock layer, expecting to concentrate energy in the lower portion and reduce the forward throw of the upper rock. However, the design required accurate knowledge of the actual bench geometry and rock layer occurrence, which could only be obtained through high-resolution spatial data. Therefore, we employed drone technology equipped with LiDAR to capture the pre-blast topography with millimeter accuracy. The combination of drone-based scanning and dynamic numerical analysis allowed us to systematically evaluate the effect of initiation position before field implementation.
Methodology
Drone LiDAR Data Acquisition
We used a DJI M350 drone carrying a Zenmuse L2 LiDAR sensor to scan the entire blasting area. The flight parameters were set to achieve a point density of approximately 500 points per square meter. After scanning, the raw point cloud was processed in Lidar360 software to remove noise using statistical filtering and then resampled at a uniform density of 1 mm. The resulting dataset had an absolute accuracy of ±3 mm after georeferencing with ground control points. From the cleaned point cloud, we extracted the bench slope angle (62°) and the spatial distribution of rock layers. The hard fine sandstone layer was found to be 1.5 m thick, located 6.5 m above the bench toe. A typical view of our drone scanning system is shown below.
The use of drone technology enabled us to rapidly obtain accurate three-dimensional information of the bench, which was essential for building a representative numerical model.
Characterization of Rock Mechanical Properties
We collected rock samples from the same blasting zone and performed uniaxial compressive strength (UCS) tests on standard cylindrical specimens (50 mm diameter, 100 mm height). The results are summarized in Table 1. The fine sandstone exhibited a UCS of 65.43 MPa, classifying it as hard rock, while the yellow mudstone and sandy mudstone had UCS values below 10 MPa, indicating soft, easily blastable rock. For simplicity, we grouped the two soft rocks into one soft rock category in the numerical model.
| Rock Type | Density (kg/m³) | UCS (MPa) | Shear Strength (MPa) | Tensile Strength (MPa) |
|---|---|---|---|---|
| Hard (fine sandstone) | 2389.7 | 65.43 | 51.85 | 1.62 |
| Soft (mudstone/sandy mudstone) | 2033.4 | 7.0 | 5.75 | 0.12 |
Numerical Model Construction in LS-DYNA
Based on the point cloud data, we built a two-dimensional plane strain model in LS-DYNA representing a vertical cross-section along the direction of maximum throw (i.e., perpendicular to the bench face). The model had a bench height of 18 m (including 1.5 m sub-drilling) and a face angle of 62°. A vertical blast hole of 16.5 m depth and 152 mm diameter was positioned according to the actual drilling pattern (8 m burden, 6 m spacing in the field; the model used 7.2 m spacing between two adjacent holes). A 1.5 m thick hard rock layer was embedded 6.5 m above the bench toe, exactly matching the field observation. The model boundaries, except the bench face and top, were set as non-reflecting boundaries to simulate an infinite medium.
The rock material was modeled using the MAT_272 (RHT) constitutive model, which accounts for strain rate sensitivity, damage evolution, and compression/tension softening. The RHT model parameters were calibrated from the laboratory UCS data. The explosive was modeled as MAT_HIGH_EXPLOSIVE_BURN (MAT_008) with the Jones-Wilkins-Lee (JWL) equation of state:
$$ P = A\left(1 – \frac{\omega}{R_1 V}\right)e^{-R_1 V} + B\left(1 – \frac{\omega}{R_2 V}\right)e^{-R_2 V} + \frac{\omega E_0}{V} $$
where P is the detonation pressure, A, B, R1, R2, ω are explosive constants, E0 is the internal energy per unit volume, and V is the relative volume. The explosive density was 900 kg/m³ and the in-hole detonation velocity was 3803 m/s, measured on site. The delay between the rear hole and the front hole was set to 1000 μs.
Simulation Scheme and Monitoring Points
We compared two initiation schemes: (1) bottom initiation (traditional) and (2) hard rock center initiation, where the detonator was placed at the centroid of the hard rock layer. To quantify the effect on the free face, we defined six stress monitoring points along the bench face: (1) crest, (2) soft rock, (3) upper structural plane, (4) hard rock center, (5) lower structural plane, and (6) toe. The positions of these points were determined from the point cloud layer data. The explicit dynamics analysis was run for 10 ms to capture the full stress wave propagation.
Results
Stress Wave Responses on the Free Face
Figure (not shown) illustrates the time histories of effective stress at each monitoring point. Our analysis revealed distinct differences between the two initiation schemes. In bottom initiation, the upper stress monitoring points (crest, soft rock, upper structural plane) experienced higher peak stresses and prolonged duration compared to hard rock center initiation. In contrast, with hard rock center initiation, the toe and hard rock layer experienced the highest stresses, while the upper points were significantly relieved. Table 2 summarizes the peak stress values at each point (in units of 1011 Pa).
| Monitoring Point | Bottom Initiation | Hard Rock Center Initiation |
|---|---|---|
| Crest | 0.00058 | 0.00053 |
| Soft Rock | 0.000286 | 0.000247 |
| Upper Structural Plane | 0.00038 | 0.00030 |
| Hard Rock Center | 0.00058 | 0.00068 |
| Lower Structural Plane | 0.00040 | 0.00027 |
| Toe | 0.0018 | 0.0020 |
The data clearly show that hard rock center initiation redistributes explosive energy downward, reducing the forces acting on the upper bench and thus limiting the forward ejection of rock from the crest region. This mechanism was confirmed by the significant reduction in flyrock distance observed later in field tests.
Field Validation using Drone Technology
We conducted two full-scale blasts in two adjacent blocks with identical drilling patterns (hole diameter 152 mm, depth 16.5 m, burden 8 m, spacing 6 m, electronic detonators with 31 ms in-row delay and 73 ms between rows). The only changed parameter was the initiation location: bottom initiation for the reference block and hard rock center initiation for the optimized block. After each blast, we re-scanned the muckpile using the same drone LiDAR system within 15 minutes to obtain high-resolution point clouds. The point cloud data were processed to extract cross-sectional profiles at multiple buffer distances (60 m to 100 m apart). Several key morphological parameters were measured and are compared in Table 3.
| Parameter | Unoptimized (Bottom Initiation) | Optimized (Hard Rock Center Initiation) |
|---|---|---|
| Average forward throw distance (m) | 43.3 | 34.9 |
| Minimum pile angle (°) | 24 | 28 |
| Loose coefficient k = Vblast/Vin-situ | 1.12 | 1.16 |
| Farthest flyrock distance (m) | 57 | 45 |
The forward throw distance was reduced by 19.5% (from 43.3 m to 34.9 m), while the minimum pile angle increased from 24° to 28°, indicating a more compact and regular muckpile. The loose coefficient k was computed as:
$$ k = \frac{V_2}{V_1} $$
where V1 is the pre-blast rock volume and V2 is the post-blast bulk volume measured from the point cloud mesh model. The increase from 1.12 to 1.16 reflects better fragmentation and a looser pile, which improves digging efficiency. The reduction in farthest flyrock from 57 m to 45 m enhances operational safety. The drone technology enabled us to precisely quantify these improvements without interfering with the mining schedule.
Discussion
The agreement between numerical simulation and field measurements confirms that hard rock center initiation effectively controls blast pile morphology by focusing energy into the lower hard rock and bench toe. The stress redistribution reduces the momentum imparted to the upper soft rock layers, which are the main contributors to excessive forward throw. Our approach demonstrates the power of integrating drone‑based spatial data with high-fidelity numerical modeling. The drone technology provided the detailed layer geometry and bench profile that were essential for building a realistic LS‑DYNA model. Without such high-resolution field data, the numerical model would have relied on assumptions that could lead to inaccurate predictions.
Furthermore, the use of drone technology for post-blast scanning allowed rapid and accurate measurement of muckpile dimensions, enabling a quantitative comparison between schemes. Traditional methods would have required ground-based surveys taking several hours, whereas our drone scans were completed within 15 minutes. This speed is critical for maintaining an efficient blasting cycle in large open‑pit mines. The point cloud data also facilitated volume calculations and extraction of cross‑sectional profiles, which are impossible to obtain manually with sufficient density.
We also note that the loose coefficient improvement from 1.12 to 1.16 is statistically significant and directly translates to reduced secondary blasting or crushing costs. The more concentrated pile reduces the footprint on haul roads, minimizing the need for dozer re‑handling. Our methodology can be extended to other mine sites by repeating the same workflow: drone LiDAR survey, layer extraction, LS‑DYNA simulation with site‑specific rock properties, and validation with post‑blast drone scans. This synergy between drone technology and numerical simulation provides a robust framework for fine‑tuning blast designs.
Conclusion
We have developed and validated a novel method to optimize blast pile morphology in open‑pit bench blasting through the integration of drone‑based LiDAR scanning and LS‑DYNA numerical modeling. The key findings are:
- Hard rock center initiation concentrates explosive energy in the lower part of the bench, reducing stress on the upper soft rock layers and suppressing excessive forward throw.
- Field tests showed that this optimized scheme reduced the average forward throw distance by 19.5% (from 43.3 m to 34.9 m), increased the minimum pile angle from 24° to 28°, raised the loose coefficient from 1.12 to 1.16, and decreased the farthest flyrock distance from 57 m to 45 m.
- The drone technology enabled rapid, high‑resolution acquisition of both pre‑blast bench geometry and post‑blast muckpile morphology, providing the necessary data for building accurate numerical models and for quantitative validation.
- The combined drone‑numerical approach offers a cost‑effective and transferable solution for improving blasting efficiency and safety in large‑scale open‑pit operations. Future work will focus on automating the workflow and applying it to heterogeneous rock masses with multiple hard layers.