Fracture characteristics and failure mechanisms of coal-rock composites under complex true triaxial stress paths Authors: Chenyang Zhang, Yang Hao, Hongyuan Ma, Yiqun Zhang, Farun Li, Quanfu Liu Published: 2026 ABSTRACT Coal-rock composite structures are common in deep roadway roofs and floors, and their instability is strongly affected by excavation-induced unloading and coal-rock relative thicknesses. To clarify their fracture characteristics and failure mechanisms under realistic stress adjustment paths, laboratory true triaxial loading-unloading tests were conducted on 100 mm cubic specimens with coal-rock ratios of 0:1, 1:2, 1:1, 2:1, and 1:0, combined with acoustic emission (AE) monitoring and PFC3D simulations to investigate their mechanical response, damage evolution and energy characteristics. The results show that with increasing coal-rock ratio, the failure mode gradually transitions from relatively stable splitting-shear failure in rock-dominated specimens to abrupt coal-dominated instability, and composite specimens with intermediate ratios exhibit the most significant interface-controlled X-shaped or semi-X-shaped conjugate shear damage. Due to the mismatch in elastic modulus and Poisson’s ratio between coal and mudstone, distinct AE precursor peaks appear during the unloading and stress readjustment stages, which are stronger in composite specimens than in pure coal or pure rock specimens. The core quantitative findings are that the cumulative absorbed energy first increases and then decreases with increasing coal-rock ratio, reaching a maximum at a coal-rock ratio of 67%, whereas the peak strength decreases monotonically. This indicates that rock burst proneness is governed more directly by energy accumulation and release than by strength alone. Numerical simulations well reproduce the fracture process of specimens, and demonstrate that interface bond breakage promotes the conversion of stored strain energy into kinetic energy during final instability. These findings provide a mechanistic basis for evaluating dynamic instability and optimizing support strategies in deep composite strata. Introduction As coal resources are increasingly developed at greater depths and intensities worldwide, the occurrence of rock burst in coal mines, particularly in China, has increased significantly1–4. Due to its difficulty in prediction, short duration, and strong destructive potential, rock bursts have become one of the most serious dynamic disasters threatening safe coal mine production5–7. Existing theories suggest that rock bursts are nonlinear dynamic phenomena, in which energy stored in a stable state is suddenly transformed into released energy once the coal–rock mass system undergoes dynamic instability under stress8–16. Relevant reaches indicate that nearly 90% of rock bursts occur in roadways with hard roofs17–20. Consequently, studying the dynamic instability mechanism of roadway surrounding rock, as well as the intrinsic rock burst proneness of coal-rock composite structures and their impact tendency under engineering stress paths, from the perspectives of strength and energy has become a research hotspot in this field. In recent years, researchers have conducted extensive rock mechanics experiments to study the stress and energy evolution of roof and floor rocks during deformation and failure under loading conditions. Through field investigations, Pan21 proposed energy-based criteria for rock bursts induced by combined dynamic and static loading, as well as for spontaneous rock bursts under purely static loading. Xu et al.22 established a mechanical model describing the occurrence and influencing factors of floor rock bursts. Zhang23 studied the influence of bottom coal thickness on the impact initiation of roadway floor. In terms of laboratory research, Dou et al.24 carried out uniaxial compression tests on coal–rock composite specimens and analyzed their rock burst proneness during the loading process. Yang et al.25 studied the energy evolution and damage characteristics of coals with different rock burst proneness under uniaxial compression and proposed the concepts and formulations of the strain energy release ratio and the kinetic energy growth ratio. Zhang et al.26 studied the influence of combination modes on the mechanical properties and failure characteristics of coal–rock composite specimens and identified the compressive strength and failure modes of three types of coal–rock composites under uniaxial and triaxial compression. The above studies provide important insights into the stress and energy evolution of rocks under loading, and lay a foundation for the evaluation of coal-rock rock burst proneness. However, three key limitations remain to be addressed. First, most existing research has relied on uniaxial or pseudo-triaxial loading conditions, which differ markedly from the actual in situ stress environment induced by roadway excavation, and thus cannot accurately characterize the impact tendency of coal-rock composites under realistic engineering stress paths. Moreover, energy analyses have primarily focused on elastic–plastic energy distribution ratios at specific stress states, with limited attention to the evolution of energy throughout the entire loading-unloading process, which is critical for revealing the dynamic instability mechanism of coal-rock systems. In addition, coal–rock composite specimens are often prepared with fixed coal–rock ratios, and the influence of coal-rock ratio on the rock burst proneness, impact tendency and dynamic instability characteristics of composites has not been systematically examined, thereby limiting their applicability in guiding on-site roof and floor coal retention. Fig. 1Schematic illustration of stress adjustment around the roadway floor during excavation. Schematic illustration of stress adjustment around the roadway floor during excavation. As shown in Fig. 1, the excavation of coal mine roadways disrupts the original in-situ stress equilibrium of surrounding rock masses and redistributes the stress field. According to the classical solution of the plane strain problem in elastic mechanics, the stress redistribution in the floor strata after rectangular roadway excavation follows these patterns: the vertical stress primarily transfers toward the solid coal on both sides of the roadway, forming stress concentration zones; the shallow regions of the floor experience significant pressure relief, with vertical stress σ1 gradually recovering to the initial in-situ stress level as depth increases; at the floor surface, the vertical stress theoretically equals zero, which constitutes the fundamental mechanical boundary condition of a free surface27–30. Under this stress path, coal-rock composite structures with different coal-rock ratios exhibit distinct mechanical responses, directly altering their impact tendencies and corresponding dynamic instability modes. Therefore, true triaxial loading and unloading tests that reproduce the actual roadway excavation stress path were conducted on coal–rock composite specimens with varying coal-rock ratios in this study. From a macroscopic perspective, the strength, deformation, acoustic emission behavior, and fracture evolution of specimens subjected to disturbance loading under true triaxial pre-stress conditions were analyzed, to clarify the evolution law of their impact tendency under engineering stress paths. The experimental results were further verified through PFC3D numerical simulations, to reveal the intrinsic dynamic instability mechanism of coal-rock composites, and to quantify the influence of coal-rock ratio on their inherent rock burst proneness. Laboratory experiments Experimental setup The coal-rock specimens used in this study were from the Mengcun Coal Mine, Shanxi Province, China. Coal and rock blocks were from the 401,101 working face of a gas drainage entry and were transported to the laboratory. The basic physical and mechanical properties, including burst tendency indices, were tested in strict accordance with the Chinese national standard GB/T 25217.2–2010. The calculation formulas of the key energy indices are as follows:1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$W_{t} = \frac{{A_{e} }}{{A_{p} }},K_{E} = \frac{{U_{e} }}{{U_{p} }},W_{b} = \frac{W}{S}$$\end{document} Where Ae is the elastic deformation energy, Ap is the post-peak plastic deformation energy, Ue is the elastic deformation energy released during unloading, Up is the plastic deformation energy consumed during the loading process, W is the total energy under the load-deflection curve, S is the cross-sectional area of the bending specimen. The detailed test results are listed in Table 1. Table 1Basic physical and mechanical properties of coal and mudstone.Uniaxial compressive strength(MPa)Uniaxial tensile stress(MPa)Young’s modulus/GPaPoisson ‘s ratioDynamic failure time/msImpact energy index WtElastic energy index KEBending energy index Wb/kJOutburst pronenessCoal19.371.232.690.34726.213.35---HighRock28.751.688.510.26---------7.29Low Table 1Basic physical and mechanical properties of coal and mudstone.Uniaxial compressive strength(MPa)Uniaxial tensile stress(MPa)Young’s modulus/GPaPoisson ‘s ratioDynamic failure time/msImpact energy index WtElastic energy index KEBending energy index Wb/kJOutburst pronenessCoal19.371.232.690.34726.213.35---HighRock28.751.688.510.26---------7.29Low Basic physical and mechanical properties of coal and mudstone. As shown in Fig. 2, five groups of composite specimens were prepared with different coal-rock ratios, defined as the height ratio of coal to rock in the vertical direction of the cubic specimen. The coal-rock ratios were set to 0:1, 1:2, 1:1, 2:1, and 1:0 respectively. Each group comprises 8 samples. The test results are the average values of 8 groups of tests to ensure the reliability of the test. The coal and rock blocks were firmly bonded by a high-strength structural adhesive, with the bonding thickness controlled to approximately 0.5 mm. The detailed composition and dimensions of each group of specimens are listed in Table 2. Although this adhesive layer cannot fully reproduce the roughness, mineral filling, and long-term geological cementation of a natural coal-rock interface, its thickness was kept very small relative to the 100 mm specimen size and was controlled consistently for all groups. Therefore, the artificial interface mainly provides a repeatable weak bonding plane for comparing the influence of coal-rock stiffness mismatch and interface damage evolution, rather than representing a site-specific natural interface in every detail. Fig. 2Geometries and photographs of coal-rock composite specimens with different height ratios. Geometries and photographs of coal-rock composite specimens with different height ratios. Table 2Composition and dimensions of coal-rock composite specimens.NumberCoal-rock compositeCoal specimens height/mmRock specimens height/mmCoal-rock ratioA R 01000:1B R \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\tfrac{2}{3}$$\end{document} C \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\tfrac{1}{3}$$\end{document} 34661:2C R \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\tfrac{1}{2}$$\end{document} C \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\tfrac{1}{2}$$\end{document} 50501:1D R \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\tfrac{1}{3}$$\end{document} C \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\tfrac{2}{3}$$\end{document} 66342:1E C 10001:0 Table 2Composition and dimensions of coal-rock composite specimens.NumberCoal-rock compositeCoal specimens height/mmRock specimens height/mmCoal-rock ratioA R 01000:1B R \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\tfrac{2}{3}$$\end{document} C \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\tfrac{1}{3}$$\end{document} 34661:2C R \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\tfrac{1}{2}$$\end{document} C \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\tfrac{1}{2}$$\end{document} 50501:1D R \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\tfrac{1}{3}$$\end{document} C \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\tfrac{2}{3}$$\end{document} 66342:1E C 10001:0 Composition and dimensions of coal-rock composite specimens. The test systems used in the research mainly include two parts: true triaxial electro-hydraulic servo test system and PCI-Ⅱ acoustic emission testing system. To ensure full contact between the specimens and the pressure plates of the true triaxial loading device, a loading mold matching the specimen dimensions was prepared in advance. The mold provided sufficient space for the placement of pressure blocks in contact with the specimens. The pressure blocks were slightly smaller than the coal specimens to allow adequate deformation during loading. Three extensometers were installed on the pressure block to measure deformation in the X, Y, and Z directions, each with a measuring range of ± 3 mm. Deformation data of the specimens were collected in real time at a sampling interval of 0.001 h. In addition, Vaseline was used to attach six acoustic emission (AE) probes (No. 1–No. 6) to the pressure blocks in six directions for monitoring acoustic emission signals. The specimen loading mold and the loading system are shown in Fig. 3. Fig. 3Loading mold and sensor arrangement for the true triaxial test. Loading mold and sensor arrangement for the true triaxial test. The test procedure was as follows. First, coal–rock composite specimens with different coal-rock ratios were loaded to three initial in situ stress states. Then, the loading path was adjusted to unload the minimum principal stress in the X direction. After the stress state stabilized, the intermediate principal stress in the Y direction was maintained constant, while the maximum principal stress in the Z direction was increased at a constant rate until the coal–rock composite structure lost its bearing capacity. The loading scheme is illustrated in Fig. 4. In the specimen coordinate system, the Z direction corresponds to the vertical direction of the coal-rock layered specimen and is defined as the maximum principal stress direction during the final loading stage; the Y direction is the horizontal in-plane direction and is maintained as the intermediate principal stress direction; and the X direction is the horizontal unloading direction, corresponding to the minimum principal stress after roadway excavation. During the initial true-triaxial loading stage, σ1, σ2, and σ3 are applied along the Z, Y, and X directions, respectively. During excavation simulation, σ3 in the X direction is unloaded to simulate lateral stress relief, σ2 in the Y direction is held constant, and σ1 in the Z direction is subsequently increased until failure. Thus, the stress labels denote the principal stress magnitudes at each stage, while the X/Y/Z axes define the fixed spatial loading directions. ①Loading σ1, σ2, σ3 to 6.75 MPa at a rate of 0.1 MPa/s; ②Keep σ3 constant, load σ1、σ2 to 7.5 MPa at a rate of 0.1 MPa/s; ③Keep σ2 and σ3 constant, and load σ1 to 12 MPa at a rate of 0.1 MPa/s; ④Keep σ1, σ2 and σ3 constant for 3 min; ⑤Keep σ1 and σ2 constant, and unload σ3 to 0 at a rate of 0.1 MPa/s; ⑥Keep σ1, σ2 and σ3 constant for 2 min; ⑦Keep σ2 and σ3 constant, loading σ1 at a rate of 0.1 MPa/s until the coal-rock composite specimens failure. Fig. 4True triaxial loading-unloading path X, Y, and Z denote the unloading, constant intermediate-stress, and vertical loading directions, respectively. True triaxial loading-unloading path X, Y, and Z denote the unloading, constant intermediate-stress, and vertical loading directions, respectively. Experimental results Failure characteristics of specimens Figure 5 shows the failure characteristics of all group specimens after the true triaxial test. Fig. 5Intact and failed states of specimens with different coal-rock ratios. Intact and failed states of specimens with different coal-rock ratios. Both Group A and Group B specimens exhibited splitting or shear failure under true triaxial loading conditions. In Group A, a large open fracture developed on the front surface of the specimen, accompanied by a small number of locally distributed, non-penetrating fine fractures. The upper and lower surfaces remained intact. Two symmetric open fractures appeared on the left and right lateral surfaces, extending obliquely upward and downward at approximately 30° respectively, and penetrated the horizontal fractures at the upper and lower ends of the rear surface. The specimen failed along the transverse central axis, with intermittent internal sounds during loading. In Group B, two fractures similar to those in Group A developed within the coal mass and extended to the front surface of the rock mass. The fractures in the coal mass propagated parallel to the coal–rock interface, penetrating the coal mass and dividing it into two parts. In addition, fractures in the rock mass intersected with lateral fractures to form a conjugate shear plane, indicating stronger energy accumulation and release. A key fracture formed in the middle of the upper surface, extending vertically downward to the rock mass and horizontally along the Y direction of the loading system. The rear surface of the coal mass suffered severe damage, with about one quarter of the coal block peeling off. The specimen failed by splitting and bursting along the central axes of coal and rock masses respectively, with obvious dynamic manifestation and crisp cracking sounds. In Group C, fractures appeared at the top of the specimen due to vertical extrusion, and a major fracture also developed on the front surface, running parallel to the coal–rock interface. This fracture was connected with symmetrically distributed fractures on the lateral surfaces, forming a systematic X-shaped conjugate shear failure pattern. The main front fracture was exactly located at the coal-rock interface, connecting with symmetric fractures on both lateral surfaces. The lateral fractures extended at 30° obliquely upward and downward respectively, forming a typical semi-X-shaped conjugate shear feature. A distinct fracture formed in the middle of the upper surface, aligned with the X direction of the loading system. The specimen failed by shear along the coal-rock interface, with continuous stable deformation and crisp sounds during loading. When the specimen in Group D failed, splitting and bursting occurred from the interior to the exterior at both ends of the specimen, originating from the central axes of the coal mass and the rock mass, respectively. Two penetrating fractures formed in the coal mass, creating a semi-X-shaped conjugate shear failure zone along its entire length. Fractures in the rock mass were similar in morphology; however, the rock mass as a whole exhibited pronounced X-shaped conjugate shear failure characteristics, resulting in detachment of the lower surface. The lateral fractures in the coal mass extended at 18° obliquely upward and downward, forming a semi-X-shaped failure zone. The fractures in the rock mass penetrated all four surfaces, forming a complete X-shaped conjugate shear zone, which caused the entire lower surface of the rock mass to fall off. The dynamic manifestation was more significant in the rock mass, with crisp cracking sounds emitted during loading. The Group E specimens failed and lost overall structural continuity. During failure, splitting or shear failure occurred on both sides of the specimen, with the transverse central axis acting as the failure boundary. A fracture located at the center of the front surface generally inclined toward the upper left and penetrated the entire specimen, resulting in failure characteristics similar to those of plate-like fractures. The central front fracture inclined to the upper left and penetrated the whole specimen. The lateral fractures formed a semi-X-shaped pattern with mirror symmetry, extending at 27° obliquely upward and downward respectively. The tensile fractures on the rear surface connected with lateral fractures, causing the upper and lower surfaces to detach, and the lower part of the specimen produced a large number of fine fragments. Fig. 6Full stress-strain curves of coal-rock composite specimens under the true triaxial loading-unloading path. Full stress-strain curves of coal-rock composite specimens under the true triaxial loading-unloading path. Figure 6 presents the complete stress-strain curves of the specimens in Groups A–E. ε₁, ε₂, ε₃ and εv denote the maximum principal strain, intermediate principal strain, minimum principal strain, and volumetric strain of the specimens, respectively. Volumetric strain is defined as the algebraic sum of the three principal strains, as expressed by the equation: εv = ε₁ + ε₂ + ε₃. As shown in Fig. 6, all specimens exhibit four typical deformation stages: compaction, elastic deformation, yield and failure. The deformation characteristics of each strain component are analyzed as follows: Maximum principal strain (ε₁): It is the dominant factor controlling specimen volumetric expansion, with the largest positive strain value and most sensitive response to loading. Groups A, C and E developed a clear yield plateau after reaching the critical failure threshold, indicating sufficient plastic deformation and fully mobilized bearing capacity, thus their impact tendency is weakened. In contrast, Groups B and D exhibited distinct inflection points on their ε₁ curves at the critical failure threshold, suggesting underdeveloped plastic deformation along the maximum principal stress direction, thus their impact tendency is enhanced.Intermediate principal strain (ε₂): It remains nearly constant during compaction and elastic stages with minimal deformation, showing low sensitivity to loading-unloading and negligible contribution to volumetric strain.Minimum principal strain (ε₃): All specimens show obvious strain rebound at the unloading point, reflecting elastic deformation characteristics. After entering the plastic stage, the X-direction deformation transforms from compression to dilation, with strain values changing from positive to negative.Volumetric strain (εv): It generally undergoes stable growth, maintenance and reduction stages. Group D shows earlier transition from positive to negative volumetric strain, indicating enhanced dilatancy with increasing coal ratio. Group E’s volumetric strain is highly consistent with ε₃ evolution, and exhibits a prominent plastic plateau at failure. Maximum principal strain (ε₁): It is the dominant factor controlling specimen volumetric expansion, with the largest positive strain value and most sensitive response to loading. Groups A, C and E developed a clear yield plateau after reaching the critical failure threshold, indicating sufficient plastic deformation and fully mobilized bearing capacity, thus their impact tendency is weakened. In contrast, Groups B and D exhibited distinct inflection points on their ε₁ curves at the critical failure threshold, suggesting underdeveloped plastic deformation along the maximum principal stress direction, thus their impact tendency is enhanced. Maximum principal strain (ε₁): It is the dominant factor controlling specimen volumetric expansion, with the largest positive strain value and most sensitive response to loading. Groups A, C and E developed a clear yield plateau after reaching the critical failure threshold, indicating sufficient plastic deformation and fully mobilized bearing capacity, thus their impact tendency is weakened. In contrast, Groups B and D exhibited distinct inflection points on their ε₁ curves at the critical failure threshold, suggesting underdeveloped plastic deformation along the maximum principal stress direction, thus their impact tendency is enhanced. Intermediate principal strain (ε₂): It remains nearly constant during compaction and elastic stages with minimal deformation, showing low sensitivity to loading-unloading and negligible contribution to volumetric strain. Intermediate principal strain (ε₂): It remains nearly constant during compaction and elastic stages with minimal deformation, showing low sensitivity to loading-unloading and negligible contribution to volumetric strain. Minimum principal strain (ε₃): All specimens show obvious strain rebound at the unloading point, reflecting elastic deformation characteristics. After entering the plastic stage, the X-direction deformation transforms from compression to dilation, with strain values changing from positive to negative. Minimum principal strain (ε₃): All specimens show obvious strain rebound at the unloading point, reflecting elastic deformation characteristics. After entering the plastic stage, the X-direction deformation transforms from compression to dilation, with strain values changing from positive to negative. Volumetric strain (εv): It generally undergoes stable growth, maintenance and reduction stages. Group D shows earlier transition from positive to negative volumetric strain, indicating enhanced dilatancy with increasing coal ratio. Group E’s volumetric strain is highly consistent with ε₃ evolution, and exhibits a prominent plastic plateau at failure. Volumetric strain (εv): It generally undergoes stable growth, maintenance and reduction stages. Group D shows earlier transition from positive to negative volumetric strain, indicating enhanced dilatancy with increasing coal ratio. Group E’s volumetric strain is highly consistent with ε₃ evolution, and exhibits a prominent plastic plateau at failure. Results of acoustic emission ringing and energy counting Ringing count is one of the most commonly used AE evaluation parameters. When an AE event reaches the sensor, it induces oscillatory signals within the sensor; each oscillation waveform that exceeds the preset electrical threshold is recorded as one ringing count. Figure 7 shows the relationship between the maximum principal strain and the AE ringing count for Groups A–E as a function of time. Fig. 7Relationships among strain, time, and AE ringing counts for specimens with different coal-rock ratios. Relationships among strain, time, and AE ringing counts for specimens with different coal-rock ratios. As shown in Fig. 7, the AE activity of all specimens exhibited a staged evolution during the loading process. In situ rock stress loading stage: AE events were frequent, and the AE rate increased. As internal fractures closed, the integrity of the specimens improved. After entering the elastic stage, a continuous and complete stress-transfer framework was formed. Consequently, the number of AE events decreased, and the AE characteristics of coal–rock composite specimens with different coal-rock ratios became similar.Stress maintenance stage: AE activity remained relatively stable, and specimens with different coal-rock ratios exhibited similar AE ringing-counting characteristics.X-direction unloading stage: AE events increased slightly due to the development of internal fractures; however, the overall trend remained consistent, with only local differences. Group A specimens exhibited significantly fewer AE events. With increasing coal-rock ratio, both the frequency and amplitude of AE events gradually increased, indicating that the coal mass was less compact than the rock mass.Z-direction loading to failure stage: AE activity was significantly intensified. In specimens with coal-rock ratios of 0 and 0.5, an AE “quiet period” was observed prior to failure. In contrast, specimens with coal-rock ratios of 0.33, 0.67, and 1 exhibited continuously increasing AE activity until failure. In situ rock stress loading stage: AE events were frequent, and the AE rate increased. As internal fractures closed, the integrity of the specimens improved. After entering the elastic stage, a continuous and complete stress-transfer framework was formed. Consequently, the number of AE events decreased, and the AE characteristics of coal–rock composite specimens with different coal-rock ratios became similar. In situ rock stress loading stage: AE events were frequent, and the AE rate increased. As internal fractures closed, the integrity of the specimens improved. After entering the elastic stage, a continuous and complete stress-transfer framework was formed. Consequently, the number of AE events decreased, and the AE characteristics of coal–rock composite specimens with different coal-rock ratios became similar. Stress maintenance stage: AE activity remained relatively stable, and specimens with different coal-rock ratios exhibited similar AE ringing-counting characteristics. Stress maintenance stage: AE activity remained relatively stable, and specimens with different coal-rock ratios exhibited similar AE ringing-counting characteristics. X-direction unloading stage: AE events increased slightly due to the development of internal fractures; however, the overall trend remained consistent, with only local differences. Group A specimens exhibited significantly fewer AE events. With increasing coal-rock ratio, both the frequency and amplitude of AE events gradually increased, indicating that the coal mass was less compact than the rock mass. X-direction unloading stage: AE events increased slightly due to the development of internal fractures; however, the overall trend remained consistent, with only local differences. Group A specimens exhibited significantly fewer AE events. With increasing coal-rock ratio, both the frequency and amplitude of AE events gradually increased, indicating that the coal mass was less compact than the rock mass. Z-direction loading to failure stage: AE activity was significantly intensified. In specimens with coal-rock ratios of 0 and 0.5, an AE “quiet period” was observed prior to failure. In contrast, specimens with coal-rock ratios of 0.33, 0.67, and 1 exhibited continuously increasing AE activity until failure. Z-direction loading to failure stage: AE activity was significantly intensified. In specimens with coal-rock ratios of 0 and 0.5, an AE “quiet period” was observed prior to failure. In contrast, specimens with coal-rock ratios of 0.33, 0.67, and 1 exhibited continuously increasing AE activity until failure. The energy of an acoustic emission signal represents a signal-derived relative parameter computed from the amplitude and duration of detected elastic waveforms, rather than constituting an absolute measurement of physical energy release. Crucially, AE energy should not be directly equated with the strain energy associated with microcrack propagation, as its quantification is substantially influenced by material-induced wave attenuation, sensor-specimen coupling efficiency, and acquisition system threshold parameters. Nevertheless, AE energy demonstrates significant positive correlation with microcrack initiation and propagation intensities, thereby serving as a robust indirect metric for characterizing relative damage progression and energy dissipation rates across loading phases. This study employs AE energy measurements to comparatively analyze damage evolution dynamics among varying coal-rock ratio groups, explicitly acknowledging its function as a relative damage assessment tool rather than an absolute mechanical energy quantification method. Figure 8(a)–(e) show the strain–AE energy curves of the coal–rock composite structures for Groups A–E, respectively. Fig. 8Relationships between strain and AE energy for specimens with different coal-rock ratios. Relationships between strain and AE energy for specimens with different coal-rock ratios. As shown in Figs. 7 and 8, the relationships among maximum principal strain–time–AE energy for coal–rock composite specimens with different coal-rock ratios are generally consistent with the previously described relationships among maximum principal stress–time–AE ringing count. When the coal-rock ratio is 0 and 0.5, the AE energy count during the failure process of the specimens is positively correlated with the internal stress difference. Before the instability, the AE energy curve exhibited a distinct anomalous stage. Subsequently, after a short period of relative stability, AE activity increased again and reached its peak at specimen failure. This behavior indicates that part of the accumulated elastic impact energy was dissipated before instability, thereby weakening the impact tendency of the specimens. When the coal-rock ratio is 0.33, 0.67 or 1.0, the AE ringing count was positively correlated with the internal stress difference during specimen failure. This result indicates that the elastic impact energy accumulated prior to instability was not effectively dissipated, leading to a more pronounced impact tendency. Numerical modeling Model description The numerical simulation of coal-rock composite structure is carried out by PFC3D, and the discrete element simulation should be consistent with the indoor test. The model size, coal seam and rock layer physical and mechanical parameters are based on the test results of the samples made of No. 4 coal and its floor rock layer in Mengcun Coal Mine. The numerical model with the same size is constructed, and the particles are bonded with different microscopic parameters. The bonding method is parallel bonding, which can truly simulate the internal cementation structure of the material by adding bonding keys between particles. Parallel bonding can transmit bending moment and torque, and more accurately reflect the bending strength and anisotropic mechanical behavior. The bonding key can be gradually broken with the load, and the whole process of material from micro-crack propagation to macro-fracture is visually presented. The strain energy will be released in the form of kinetic energy when the bonding is broken, that is, the bonding fracture can be regarded as the acoustic emission process under the discrete element simulation. In the numerical model, the interface between coal and mudstone was represented by parallel bonds with calibrated contact stiffness and bond strength. This treatment is an idealized description of the bonded laboratory interface; it is suitable for identifying the relative roles of stiffness mismatch, interface bond breakage, and energy conversion, but it does not claim to reproduce all mesoscopic heterogeneities of a natural coal-rock contact (Table 3). Before the parameters are applied to the true triaxial PFC numerical simulation of the coal-rock composite structure, uniaxial compression, Brazilian splitting, triaxial compression and other tests must be carried out to calibrate the mesoscopic parameters. According to the parameter calibration principle described by Potyondy D31, the stiffness calibration is first carried out, and the macroscopic elastic modulus of the rock is matched by adjusting the particle contact modulus; then the tensile strength and bond strength calibration are carried out, and the strength parameters of the cementation are adjusted to match the macroscopic strength characteristics of the rock (such as uniaxial compressive strength and Brazilian splitting strength) under the premise of keeping the stiffness parameters unchanged. The microscopic parameters of coal seam and rock stratum are finally determined (Table 4), and the macroscopic mechanical properties are highly matched with the indoor test mechanical properties. Considering that cross-transitional zones commonly exist between coal seams and rock formations in practical scenarios, where the overall rock strength exceeds that of coal but presents challenges in boundary delineation, the parallel bond parameters of rock should be selected for the coal-rock contact interface to better reflect actual geological conditions. The size of the true triaxial PFC model of the coal-rock composite structure is a cube with a side length of 100 mm. The particle size of the model is 1.8 mm~3 mm, the maximum/minimum particle size ratio is 1.66, the local damping coefficient is 0.7, time step 1 × 10⁻7s. The stability of the simulation has been better realized under the condition of ensuring the gradation effect. The total number of particles is about 11,500 (Fig. 9); for quasi-static problems such as compression tests, the particle density can be set to be similar to the real rock formation. Fig. 9PFC3D model of the coal-rock composite specimen. PFC3D model of the coal-rock composite specimen. Table 3Calibrated mesoscopic parameters used in the PFC3D model.TypesMicroscopic parametersMudstoneCoalBall partBall density2500 kg/m31295 kg/m3Normal-to-shear stiffness ratio3.02.8Friction coefficient10.577Bond partParallel bond tensile strength5.53 MPa3.70 MPaParallel bond cohesion10.02 MPa8.41 MPaParallel bond effective modulus4.80GPa1.49GPa Table 3Calibrated mesoscopic parameters used in the PFC3D model.TypesMicroscopic parametersMudstoneCoalBall partBall density2500 kg/m31295 kg/m3Normal-to-shear stiffness ratio3.02.8Friction coefficient10.577Bond partParallel bond tensile strength5.53 MPa3.70 MPaParallel bond cohesion10.02 MPa8.41 MPaParallel bond effective modulus4.80GPa1.49GPa Calibrated mesoscopic parameters used in the PFC3D model. Table 4Comparison of macroscopic properties obtained from laboratory tests and numerical calibration.ParametersMudstoneMudstoneCoalCoalPFCTestPFCTestModulus of elasticity8.58 GPa8.51 GPa2.70 GPa2.69 GPaUniaxial tensile strength1.59 MPa1.68 MPa1.19 MPa1.23 MPaUniaxial compressive strength29.57 MPa28.75 MPa19.9 MPa19.37 MPa Table 4Comparison of macroscopic properties obtained from laboratory tests and numerical calibration.ParametersMudstoneMudstoneCoalCoalPFCTestPFCTestModulus of elasticity8.58 GPa8.51 GPa2.70 GPa2.69 GPaUniaxial tensile strength1.59 MPa1.68 MPa1.19 MPa1.23 MPaUniaxial compressive strength29.57 MPa28.75 MPa19.9 MPa19.37 MPa Comparison of macroscopic properties obtained from laboratory tests and numerical calibration. Results Based on the complex true triaxial stress path set by the experiment, the evolution of acoustic emission signal can be clearly divided into three main logical stages: the initial multi-axial loading stage (0 ~ 120s), the middle unloading and voltage stabilizing stage (120 ~ 487.5s) and the final fracturing loading stage (487.5s to failure). The acoustic emission count rate represents the frequency of micro-crack generation, while the energy rate reflects the magnitude of elastic energy released by crack propagation. By comparing the AE response maps of five groups of specimens with different coal-rock height ratios, the non-uniformity of coal-rock combination structure and the significant influence of component proportion on damage evolution mechanism are revealed. The obtained AE count rate results are roughly similar to the experimental AE ring count. Fig. 10Simulated crack distributions and temporal evolutions of AE count and energy rate. Simulated crack distributions and temporal evolutions of AE count and energy rate. As shown in Fig. 10(a), the acoustic emission signal of group A samples presents the typical characteristics of “brittle-ductile transformation”. In the middle unloading stage, that is, in the process of unloading σ3 from 6.75 MPa to 0, the counting rate of the pure mudstone group remained at a very low level, almost in the “AE silence period”, and the counting rate remained at a very low level and the energy rate had no obvious fluctuation. This shows that the dense mudstone structure has a strong resistance to the removal of lateral confining pressure, and no obvious unloading induced damage occurs. After entering the final loading stage, the sample showed a very long elastic energy storage stage, and the AE count began to climb slowly until the time reached about 830s; at about 840s, the abrupt increase in counting rate in a medium amplitude, and the energy rate rose synchronously, reaching the peak value in an instant (counting rate > 10 × 103, energy rate > 2.0 × 104), and the sample suffered a sudden macroscopic damage. The first damage changed the internal structure of the rock, and the particle skeleton was rearranged again, thus increasing the ductility of the material and making the sample undergo “brittle-ductile transformation”. Therefore, when the internal cracks are compacted again for secondary fracturing, although the abrupt increase in counting rates greatly, the energy rate is not improved compared. With the combination of rock and coal, the AE response characteristics of the samples changed fundamentally, and the most significant difference appeared in the middle unloading and voltage stabilization section (300s ~ 487.5s interval). Compared with pure mudstone, the B, C, and D group samples (Fig. 10b-d) showed an obvious AE count isolated peak at about 370s~400s, and produced different degrees of AE energy release. This shows that when the lateral constraint is reduced, due to the difference in the elastic modulus and Poisson’s ratio of coal and mudstone, uncoordinated deformation occurs at the junction, resulting in the germination of early microcracks. Although some energy dissipation was generated, it did not affect the isolated peak shape of the AE count at the time of final failure for the B and D group samples, and only advanced the final failure time to about 600s~700s, and the impact tendency of the samples was still obvious. For group C, which has the most active and complex dynamic response, unloading is no longer just a change in stress and its boundary conditions, but directly becomes the driving force for inducing damage. The final failure stage presents a “double peak” or “multi-peak” structure, indicating that the failure process has changed from a single main shear failure to a complex mode in which coal and rock matrix rupture and interface slip coexist. The AE precursor peak in the unloading stage has extremely high intensity, and its energy rate is even close to the initial compaction stage. After entering the final loading section, from 550s, the AE signal continued to oscillate at high frequency, showing obvious characteristics of “progressive failure”. The group of samples experienced a long process of damage accumulation before macroscopic failure, and the impact elastic energy accumulated inside was partially dissipated, and the impact tendency of the samples was weakened. As shown in Fig. 10(e), although the pure coal group also had a weak response during the σ3 unloading stage, it was far less intense than the composite. This indirectly proves that the strong unloading AE signal in the composite mainly comes from the interaction and stiffness mismatch of the “coal-rock interface”. The final failure process of the pure coal group was extremely rapid. The counting rate began to increase sharply from about 650s, and the energy rate suddenly increased to more than 3.0 × 104 at 690s. This feature shows that in the absence of mudstone base constraint, the pure coal sample shows extremely high brittleness under the true triaxial unloading path, the crack penetration speed is extremely fast, the energy release is concentrated, and the impact tendency is relatively obvious. To further elucidate the relationship between acoustic emission energy and mechanical energy, we performed quantitative energy partitioning analysis using the PFC model. The total mechanical energy input to the specimen system comprises three primary components: elastic strain energy stored in intact particles and bonds, kinetic energy associated with particle motion and specimen instability, and dissipated energy encompassing bond breakage energy, frictional energy, and plastic deformation energy. The simulated AE energy represents a fraction of the dissipated energy, specifically corresponding to elastic energy release during bond fracture. Our simulation results demonstrate that cumulative simulated AE energy constitutes approximately 18–26% of total dissipated energy across all experimental groups, with its evolutionary trend exhibiting strong consistency with the specimen’s cumulative absorbed mechanical energy. These findings validate AE energy as an effective indicator of overall energy release processes, while emphasizing its distinction from total strain energy release or kinetic energy generation during terminal failure events. Discussion The failure of coal-rock composite structures under deep mining conditions is a multi-mechanism coupled process controlled by stress path, material heterogeneity and energy evolution. Conventional uniaxial/pseudo-triaxial studies have difficulty reproducing the actual excavation-induced stress adjustment process, leading to deviations in failure mechanism interpretation. Systematically interprets the experimental and numerical results from three core perspectives: stress-path effect, coal-rock stiffness mismatch, and energy accumulation-release mechanism, to reveal the intrinsic dynamic instability mechanism of coal-rock composites under realistic roadway excavation stress paths. Stress-path effect on failure mode transition The true triaxial loading-unloading path adopted in this study accurately reproduces the stress evolution process of roadway floor surrounding rock: initial three-dimensional in-situ stress, minimum principal stress (σ3) unloading to form a free surface, maximum principal stress (σ1) continuous loading until failure. This specific stress path fundamentally changes the failure characteristics of coal-rock composites compared with monotonic loading conditions. The experimental results show that all specimens experienced significant strain rebound during the σ₃ unloading stage, and the minimum principal strain transformed from compression to dilation after entering the plastic stage. This indicates that lateral unloading not only changes the boundary stress conditions but also induces initial damage inside the specimens, providing pre-existing cracks for subsequent fracture propagation. For rock-dominated specimens (Group A, coal ratio 0%), the dense mudstone structure has strong resistance to lateral confining pressure removal, resulting in minimal unloading-induced damage and a relatively stable splitting-shear failure mode with a main fracture dip angle of 30°. In contrast, coal-dominated specimens (Group E, coal ratio 100%) exhibit high brittleness under the unloading path, with extremely fast crack penetration (main fracture dip angle 27°) and concentrated energy release, producing a large number of fine fragments during failure. For composite specimens with intermediate coal ratios, the unloading stage directly becomes the driving force for damage initiation: Group B (coal ratio 33%) shows one-quarter coal block spalling caused by unloading-induced interface cracks, while Group C (coal ratio 50%) forms a complete X-shaped conjugate shear failure plane originating from the interface. This stress-path effect indicates that excavation-induced unloading is the key trigger of interface damage. In engineering, timely active support should be installed immediately after roadway excavation to suppress unloading-induced microcrack initiation, especially in composite strata with intermediate coal-rock ratios. Coal-rock stiffness mismatch and interface control mechanism The most significant feature of coal-rock composite structures is the large difference in mechanical properties between the two media. As shown in Table 1, the elastic modulus of mudstone (8.51GPa) is 3.16 times that of coal (2.69GPa), while the Poisson’s ratio of coal (0.34) is 30.8% higher than that of mudstone (0.26). This stiffness mismatch leads to uncoordinated deformation at the coal-rock interface under loading, resulting in stress concentration and interface bond breakage, which is the core mechanism controlling the failure of composite specimens. Both experimental AE monitoring and PFC numerical simulations confirm the existence of interface-induced damage. Composite specimens (Groups B, C, D) show distinct AE precursor peaks during the unloading and stress readjustment stages (370s~400s), which are absent in pure coal or pure rock specimens. Numerical simulations further reveal that these precursor peaks correspond to the breakage of parallel bonds at the coal-rock interface. When the lateral constraint is reduced, the coal mass with lower stiffness and higher Poisson’s ratio tends to expand more than the rock mass, leading to tensile stress concentration at the interface. Once the interface bond strength is exceeded, microcracks initiate at the interface and propagate into both coal and rock masses. For specimens with a coal ratio of 50% (Group C), the interface effect is most pronounced. The main fracture is exactly located at the coal-rock interface, connecting with symmetric fractures on both lateral surfaces to form a typical X-shaped conjugate shear failure pattern. This indicates that when the thicknesses of coal and rock are comparable, the interface becomes the weakest link of the composite structure, and the failure process is completely controlled by interface slip and debonding. For specimens with coal ratios of 33% and 67%, the interface still plays an important role in controlling fracture propagation, but the failure mode gradually transitions to rock-dominated or coal-dominated instability as one component becomes dominant. This interface control mechanism suggests that the stability of composite strata is not determined by the strongest rock but by the bonding strength at the coal-rock interface. Therefore, improving the interface bonding strength through grouting reinforcement and reducing stress concentration at the interface are key to preventing dynamic disasters in composite strata. For roof/floor coal retention design, coal-rock ratios around 67% should be avoided, and interface grouting reinforcement is recommended to improve bonding strength and prevent interface-controlled failure. Energy accumulation-release mechanism and rock burst proneness The dynamic instability of coal-rock composites is essentially an energy-driven process, in which the accumulated elastic strain energy is suddenly released and converted into kinetic energy. To quantitatively characterize this process, we established an energy balance framework based on the first law of thermodynamics, neglecting the internal damping energy dissipation of the testing machine. The total external work input by the testing machine (W0) is divided into the energy absorbed by the specimen (Us) and the elastic energy stored in the testing machine (Uj):2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$W_{0} = U_{{\mathrm{s}}} + U_{{\mathrm{j}}} = \sum {W_{i} } ,i = 1,2,3$$\end{document} Where Us is the energy absorbed by the specimen, Uj is the elastic energy stored in the testing machine, Wi is the work done in all directions of the loading system, and W₁, W₂, and W₃ are the work done by the testing machine in the X, Y, and Z directions, respectively. Wi can be calculated by integrating the loading force fi along the corresponding direction with respect to the displacement Si (measured by the extensometer along the same direction) over the fi–Si curve:3\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$W_{i} = \int_{0}^{{s_{i} }} {f_{i} } ds_{i} ,i = 1,2,3$$\end{document} Where fi is the loading force along the corresponding direction, which exhibits a linear relationship with the deformation li of the testing machine:4\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$f_{i} = k_{i} l_{i} ,i = 1,2,3$$\end{document} The total displacement measured by the extensometers includes both the testing machine deformation and the specimen deformation. The machine deformation li was calculated using the above linear relationship, and the actual specimen deformation was obtained by subtracting li from the total measured displacement. All stiffness values were calibrated before the tests through 5 repeated no-load cyclic loading experiments: the vertical (Z-direction, σ1) loading frame stiffness was 1000kN/mm, and the horizontal (X-direction σ3, Y-direction σ2) loading frame stiffness was 400kN/mm. The elastic strain energy Uj stored in the testing machine was calculated by linear integration of fi with respect to li and subtracted from the total external work to obtain the net energy absorbed by the specimen:5\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$U_{j} = \sum\limits_{{i = 1}}^{3} {\int_{0}^{{l_{i} }} {f_{i} } } \cdot dl_{i} = \sum\limits_{{i = 1}}^{3} {\frac{{f_{i}^{2} }}{{2k_{i} }}} ,\;i = 1,2,3$$\end{document} The main sources of energy calculation uncertainty include the ± 0.5% measurement error of the force sensor and the ± 0.1% error of the extensometer, resulting in a total relative uncertainty of less than 2% for the cumulative absorbed energy values. The net energy absorbed by the specimen is further partitioned into elastic strain energy (Ue) stored in the specimen and dissipated plastic strain energy (Ud):6\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$U_{s} = W_{0} - U_{j} = U_{e} + U_{d}$$\end{document} Based on the above energy calculation framework, we obtained the cumulative absorbed energy and stage-wise energy evolution for all coal-rock ratio groups. As shown in Fig. 11, the peak strength of specimens decreases monotonically with increasing coal-rock ratio, while the cumulative absorbed energy first increases and then decreases, reaching a maximum at a coal ratio of 67%. This non-correspondence between strength and energy indicates that energy evolution is a more reliable indicator for evaluating rock burst proneness than strength alone. For the eight repeated tests in each group, the quantitative results in Fig. 11 were expressed as mean values with error ranges. The relative scatter of peak strength was limited, showing a monotonic decrease with increasing coal ratio, whereas the cumulative absorbed energy showed larger ratio-dependent variation and reached its maximum in Group D (coal-rock ratio 2:1, coal ratio 67%). The standard deviation ranges of the repeated tests indicate that the first-increase-then-decrease trend of cumulative absorbed energy is larger than the within-group scatter, supporting the conclusion that cumulative absorbed energy is more sensitive than peak strength for evaluating rock burst proneness under this loading-unloading path. Fig. 11Relationship among cumulative absorbed energy, peak strength, and coal-rock ratio. Relationship among cumulative absorbed energy, peak strength, and coal-rock ratio. At the same time, these results demonstrate a strong positive correlation between the cumulative absorbed energy of coal-rock specimens and the severity of their dynamic instability failure: specimens with higher cumulative absorbed energy exhibit more pronounced impact characteristics and more violent failure processes. This indicates that the magnitude of cumulative absorbed energy is a key factor controlling the occurrence and intensity of impact failure under the specific experimental conditions of this study. The stage-wise energy evolution data in Fig. 12 further reveal the internal mechanism of this trend. Most of the cumulative absorbed energy is concentrated in the initial loading stage and the post-peak failure stage, and the energy absorption in the failure stage is 3–10 times that in the initial loading stage. Fig. 12Stage-wise evolution of absorbed energy for different coal-rock ratios. Stage-wise evolution of absorbed energy for different coal-rock ratios. The reason for the first-increase-then-decrease trend of cumulative absorbed energy can be explained by the combined effect of strength and plasticity. When the coal ratio is low (0%-33%), the composite structure is dominated by rock with high strength but low plasticity. Although the strength is high, the plastic deformation capacity is limited, resulting in relatively low energy absorption. When the coal ratio increases to 67%, the composite structure has both sufficient strength provided by the rock component and certain plastic deformation capacity provided by the coal component. This combination allows the specimen to absorb and store a large amount of elastic strain energy before failure, leading to the most significant impact instability characteristics (the energy increment in the failure stage is 10 times that in the initial loading stage). When the coal ratio further increases to 100%, the strength of the composite structure decreases significantly, and the plastic deformation of coal dissipates part of the energy, resulting in a decrease in cumulative absorbed energy. PFC numerical simulations further reveal the energy conversion process during final instability. When the interface bonds break, the stored strain energy is rapidly released, and a large portion of it is converted into kinetic energy of particles, which drives the dynamic ejection of coal and rock fragments. This confirms that interface bond breakage is not only a sign of damage accumulation but also a key factor promoting the conversion of strain energy into kinetic energy. With respect to the generalizability of the core finding of this study, the conclusion that cumulative absorbed energy is more suitable than peak strength for evaluating rock burst proneness is not limited to the tested coal-mudstone system. This finding is derived from the fundamental energy-driven failure mechanism of heterogeneous composite structures: the dynamic instability of coal-rock composites is essentially a process involving the sudden release of accumulated strain energy, rather than a simple strength-based failure. Therefore, for most coal-rock combinations (e.g., coal-sandstone, coal-limestone) with significant differences in elastic modulus, strength and Poisson’s ratio, the lack of a positive correlation between peak strength and rock burst proneness will persist, and cumulative absorbed energy will remain a more robust descriptor. However, for coal-rock systems with negligible mechanical property differences or interface bonding strength significantly lower than the strengths of the coal and rock components, the failure mode may be dominated by interface debonding and slip rather than internal strain energy accumulation, and the applicability of this conclusion requires further experimental verification. This energy mechanism provides a theoretical basis for rock burst risk assessment. In engineering practice, we should not only focus on the strength of surrounding rock but also pay more attention to the energy accumulation state of coal-rock composite structures. For areas with high energy accumulation potential, targeted energy release measures such as hydraulic fracturing and borehole pressure relief should be taken to reduce the risk of rock burst. In dynamic disaster monitoring, AE precursor peaks during unloading can be used as early warning indicators; energy-based criteria are preferred over strength for rock burst risk assessment, especially in high-energy accumulation zones. Although this study has revealed the fracture characteristics and failure mechanisms of coal-rock composites under complex true triaxial stress paths, some limitations still exist. First, this study only considered the influence of coal-rock height ratio under a specific stress path, and the effects of different interface bonding strengths, stress loading rates and dynamic disturbance intensities were not investigated. Second, although 8 parallel specimens were prepared for each group to ensure the reliability of the results, the inherent heterogeneity of natural coal and rock materials may still lead to minor discreteness in individual test data. Subsequent studies will expand the sample size, consider more complex engineering conditions, and combine field microseismic monitoring data to further verify and improve the conclusions of this study. Conclusion The constituent ratio strongly controls the macroscopic failure mode of the coal-rock composite specimen. With increasing coal-rock ratio, failure evolves from relatively stable splitting-shear damage in rock-dominated specimens to abrupt coal-dominated instability, and interface-controlled conjugate shear failure becomes most prominent at intermediate ratios.AE ringing counts and AE energy exhibit clear staged evolution under the true triaxial loading-unloading path. Composite specimens show distinct precursor peaks during unloading and stress readjustment because of the mismatch in elastic modulus and Poisson’s ratio between coal and mudstone, demonstrating the diagnostic value of interface-sensitive AE monitoring.The cumulative absorbed energy first increases and then decreases with coal-rock ratio, reaching a maximum at a coal-rock ratio of 67%, whereas peak strength decreases monotonically. Energy evolution therefore provides a better descriptor of rock burst proneness and impact tendency than strength alone. PFC3D simulations confirm that interface bond breakage promotes the transition of stored strain energy into kinetic energy during final instability. The constituent ratio strongly controls the macroscopic failure mode of the coal-rock composite specimen. With increasing coal-rock ratio, failure evolves from relatively stable splitting-shear damage in rock-dominated specimens to abrupt coal-dominated instability, and interface-controlled conjugate shear failure becomes most prominent at intermediate ratios. The constituent ratio strongly controls the macroscopic failure mode of the coal-rock composite specimen. With increasing coal-rock ratio, failure evolves from relatively stable splitting-shear damage in rock-dominated specimens to abrupt coal-dominated instability, and interface-controlled conjugate shear failure becomes most prominent at intermediate ratios. AE ringing counts and AE energy exhibit clear staged evolution under the true triaxial loading-unloading path. Composite specimens show distinct precursor peaks during unloading and stress readjustment because of the mismatch in elastic modulus and Poisson’s ratio between coal and mudstone, demonstrating the diagnostic value of interface-sensitive AE monitoring. AE ringing counts and AE energy exhibit clear staged evolution under the true triaxial loading-unloading path. Composite specimens show distinct precursor peaks during unloading and stress readjustment because of the mismatch in elastic modulus and Poisson’s ratio between coal and mudstone, demonstrating the diagnostic value of interface-sensitive AE monitoring. The cumulative absorbed energy first increases and then decreases with coal-rock ratio, reaching a maximum at a coal-rock ratio of 67%, whereas peak strength decreases monotonically. Energy evolution therefore provides a better descriptor of rock burst proneness and impact tendency than strength alone. PFC3D simulations confirm that interface bond breakage promotes the transition of stored strain energy into kinetic energy during final instability. The cumulative absorbed energy first increases and then decreases with coal-rock ratio, reaching a maximum at a coal-rock ratio of 67%, whereas peak strength decreases monotonically. Energy evolution therefore provides a better descriptor of rock burst proneness and impact tendency than strength alone. PFC3D simulations confirm that interface bond breakage promotes the transition of stored strain energy into kinetic energy during final instability.