Lab Model Reveals How Burning Coal Seams Crack Their Way to the Surface

Beneath the windswept coalfields of Shenmu City in northern Shaanxi, China, seams of Jurassic coal have been smoldering for years, quietly devouring a valuable energy resource while poisoning the air and destabilizing the ground above. These coalfield fires are not simply burning in place; they are engineering their own oxygen supply. As the coal is consumed, the rock layers overhead fracture, sag, and collapse, eventually sending cracks all the way to the surface. Those surface-penetrating fractures act like chimneys, funneling fresh air down to the combustion front and allowing the fire to spread laterally for decades. A team of researchers led by Li-Feng Ren of Xi’an University of Science and Technology has now reproduced this process in the laboratory, capturing for the first time in such detail the exact moment when a fracture network breaks through to the surface and becomes a sustained air leakage pathway.

The study, published in Environmental Earth Sciences, combined three complementary approaches: high-temperature triaxial compression tests on siltstone, a large-scale physical similarity model built to replicate the geology of the Bainita Coal Mine fire zone, and discrete element numerical simulations using the UDEC software. The engineering background is the shallow-buried 1−2 coal seam of the Bainita fire zone in Zhongji Town, an area that has already undergone emergency hazard elimination. The seam averages 5.9 meters in thickness, dips at a gentle 1 to 3 degrees, and lies at an average burial depth of about 85 meters, conditions that make it particularly vulnerable to the coupled effects of mining disturbance and thermal damage.

The physical model was constructed in the Physical Similarity Simulation Laboratory at Xi’an University of Science and Technology on a test bench measuring 300 centimeters long, 20 centimeters wide, and 200 centimeters high. Working from the similarity principle, the team selected geometry, unit weight, and stress as scaling factors and derived similarity constants of 50 for length, 1.5 for unit weight, and 225 for stress. Each rock layer was mixed, compacted into 1-centimeter increments, and separated by mica powder to mimic natural bedding planes. After 20 days of drying and ventilation to stabilize the material properties, the researchers installed ten vertical monitoring lines on the model face, each carrying 30 measurement points, and tracked displacement in real time with a LEICA TS16 total station as the simulated coal seam was progressively destroyed.

Before the model was built, the team needed to understand how heat weakens the rock itself. Cylindrical siltstone specimens, 50 millimeters in diameter and 100 millimeters tall, were thermally treated at temperatures ranging from ambient conditions up to 700 degrees Celsius and then crushed on a SAM-2000 servo-controlled triaxial testing machine under a confining pressure of 5 megapascals. The results were strikingly non-monotonic. Rather than degrading smoothly with temperature, both the elastic modulus and the peak stress fluctuated, remaining comparatively high and stable between 200 and 400 degrees Celsius but dipping at 100 and 500 degrees. Poisson’s ratio hovered between 0.1 and 0.3, peaking at 0.279 at 300 degrees, when evaporating bound water and thermally induced microcracks between mineral grains allowed maximum lateral deformation. Cohesion stabilized near 20 megapascals above 100 degrees, while the internal friction angle swung between roughly 26 and 45 degrees. Beyond 700 degrees, strength and stiffness declined sharply, marking a critical thermal threshold for the rock.

Armed with these temperature-dependent mechanical parameters, the researchers watched the physical model fail in stages. When the simulated coal seam damage reached an advancement distance of 68 centimeters, the first weighting of the overlying strata occurred: the floor siltstone reached its ultimate strength and collapsed locally, opening a bed-separation space beneath the sandy mudstone above. Periodic collapses followed at intervals of 12 and then 15 centimeters as the destruction front advanced to 80 and 95 centimeters, with separation spaces migrating upward and fractures propagating obliquely along weak bedding planes. At 150 centimeters, the collapse interval jumped to 65 centimeters, and fracture clusters extended the equivalent of 35 meters above the coal seam, yet the overburden still held together as an intact load-bearing arch, its tensile stresses short of the failure threshold.

The breakthrough came at 160 centimeters. The fifth collapse, with an interval of only about 10 centimeters, triggered rapid propagation of fractures on the right side of the model, and an interconnected network punched through to the simulated ground surface. This first surface-penetrating fracture, the initial air leakage and oxygen supply channel, measured 62.1 centimeters long and inclined at 59.2 degrees, while a companion fracture on the left, 59.4 centimeters long, fell just short of full connection. The collapsed zone took on a regular trapezoidal shape, with fracture density concentrated in the lower overburden. Crucially, once the first through-going fracture released the accumulated stress, the collapse intervals after penetration grew markedly larger than those before it, and the length of the dominant fracture actually decreased with further advance as new fractures formed at steeper angles.

Subsidence data told the same story from a different angle. The overburden underwent continuous dynamic subsidence throughout the destruction process, with monitoring lines farther from the simulated surface recording larger displacements and smoother profiles. Displacement at point A06 reached 57.4 millimeters at the first collapse and climbed to 65.5 millimeters at point A09 by 95 centimeters. After the first surface breakthrough at 160 centimeters, displacement spread across all monitoring points in nonlinear, asymmetric patterns, and by 280 centimeters the subsidence trough had migrated to point A26, which recorded a maximum of 73.2 millimeters. The stabilization of earlier peak points reflected the shift from stepwise caving to wholesale integral collapse once the load-bearing structure was lost.

To test how the breakthrough timing depends on geological and thermal conditions, the team built a two-dimensional UDEC model 300 meters long and 74 meters high, layered from siltstone down through sandstones, mudstones, coal, and floor mudstone, using the measured Mohr-Coulomb parameters. The numerical model validated well against the physical test: the first surface-penetrating fracture emerged at a damage advancement distance of 168 meters, corresponding closely to the 160-centimeter physical result, with matching fracture morphology and trapezoidal collapse geometry. Varying the coal seam thickness to 3, 5, and 7 meters showed that thicker seams produced breakthrough earlier, at 280, 224, and 168 meters respectively, because thicker destruction zones create larger bed-separation spaces that favor fracture growth. Regression fits relating fracture length to advance distance yielded coefficients of determination above 0.93 in every case.

The rate of combustion propagation proved to be the decisive variable. With a 7-meter seam held constant, incremental destruction steps of 7, 14, and 28 meters per cycle produced surface breakthrough at 118, 168, and 224 meters respectively. A slower burn gives the overburden more time to accumulate stress and develop fractures, so the fire reaches the surface sooner and with greater subsidence, whereas a fast-moving front outruns fracture development and delays breakthrough. Temperature played a supporting role: at 400 degrees Celsius the first through-going fracture appeared at 140 meters instead of 168 meters, because heat degrades interparticle cementation and weakens the rock, while 200 degrees made little difference. Grey relational analysis quantified the hierarchy, assigning degrees of 0.72 to the damage advancement rate, 0.67 to seam thickness, and 0.51 to temperature, confirming that under a given overburden height the speed of the burning front most strongly controls when air leakage channels form.

The practical implications are significant. Current fire zone control relies on stripping the burning area, injecting cooling agents, and sealing air leakage pathways, yet incomplete sealing leaves fires prone to reignition. By identifying the dominant factors and the critical distances at which surface-penetrating fractures form, the study gives engineers a quantitative basis for predicting where and when to seal leakage channels in shallow-buried fire zones like Bainita. For a region where coal fires consume resources, release pollutants, and threaten communities, knowing that a slower-moving fire is paradoxically the more dangerous one for fracture breakthrough could reshape how monitoring and suppression campaigns are prioritized across China’s northern coalfields.

Subject of Research: Evolution of overburden fractures and air leakage channels in shallow-buried coal fire zones

Article Title: Experimental study on similar simulation of overburden fracture evolution in shallow-buried coal fire zones

Article References: Experimental study on similar simulation of overburden fracture evolution in shallow-buried coal fire zones. (n.d.). https://doi.org/10.1007/s12665-026-13137-3

Image Credits: AI Generated

DOI: 10.1007/s12665-026-13137-3

Keywords: coal fires, overburden fractures, similarity simulation, siltstone, triaxial compression, thermal damage, UDEC, air leakage channels, surface subsidence, grey relational analysis, Shenmu City, Environmental Earth Sciences