Study on rockburst mechanism and risk identification for isolated working face in extra-thick coal seam fold zone Authors: Yongqin Xie, Anye Cao, Geng Li, Weiwei Zhao, Cheng Zhang, Zhongjun Feng Published: 2026 ABSTRACT AbstractFaced with the challenges of high in-situ stress and elevated rockburst risk in isolated working faces within fold zones, this study takes an isolated working face in a specific coal mine as the engineering background. By integrating theoretical analysis, numerical simulation, and field practice, we investigated the evolutionary characteristics of overlying strata structure, clarified the variation law of mining-induced stress, revealed the rockburst mechanism of isolated working faces in fold zones of extra-thick coal seams, and established a dynamic risk identification model for the working face.The results show that when the minimum width of the coal pillar (95 m) during mining in the fold zone is significantly larger than the critical width (36.5 m), the overlying strata of the working face form a unique long-arm “F” structure. The cantilever effect of this structure causes the peak stress of the coal-rock mass in the working face advance area to increase to 42.1 MPa, with a stress concentration factor of 2.34. The superposition of the high static load in the advance area and the intense dynamic load released by the fracturing of the “F” structure leads to a significant decrease in the minimum principal stress and a corresponding increase in the maximum principal stress. This pushes the coal-rock mass beyond its strength limit, ultimately triggering a rockburst.To address the limitation of low early-warning accuracy (81.4%) associated with traditional single-index methods, a multi-index collaborative identification model based on microseismic data was constructed. Utilizing a Back-Propagation (BP) neural network algorithm, the model adaptively optimizes the coupling weights of six key indicators: b-value (0.04), P(b)-value (0.009), ∆F-value (0.185), AC-value (0.218), EEM-value (0.256), and S-value (0.292). This realizes a paradigm shift from “single-index alarm” to “multi-index collaborative discrimination”. Field validation demonstrates that the model improves the early-warning accuracy to 94.9%, significantly enhancing the capability for dynamic risk identification of rockbursts in the working face. The research conclusions provide a valuable reference for the safe mining of isolated working faces under similar geological and mining conditions.Keywords: Composite key strata, Composite effect, Rockburst, High energy microseismic events, Static and dynamic loadSubject terms: Engineering, Natural hazards, Solid Earth sciences Introduction With the continuous increase in coal mining depth, the geological conditions for coal seam extraction have become increasingly complex, and rockburst disasters have shown a marked rising trend. Numerous studies have confirmed that isolated working faces formed due to unreasonable mining layouts exhibit an extremely high propensity for rockbursts1–3. Owing to their special spatial occurrence characteristics, such working faces typically bear abnormally high abutment pressure, which is prone to causing severe fracturing of roadway surrounding rock and greatly increasing the difficulty of roadway maintenance4,5. When an isolated working face is superimposed on a synclinal structural zone, the asymmetric caving of overlying strata exacerbates the inhomogeneous distribution of in-situ stress, further pushing the rockburst risk of the working face to a critical level. Therefore, clarifying the rockburst-induced mechanism of isolated working faces under the coupling effect of synclinal axis structure and irregular coal pillar confinement is of great theoretical significance and engineering application value for ensuring the safe and efficient production of deep coal mines. At present, scholars have carried out systematic and in-depth research on the evolution of overlying strata spatial structure and stress distribution characteristics of conventional isolated working faces. For example, Dou Linming6, Jiang Fuxing7 et al. analyzed the evolution modes of overlying strata structures of different types of isolated working faces and revealed the degree and dynamic evolution law of stress concentration in surrounding rock. For isolated working faces adjacent to faults, Cao Anye8, Zhu Guang’an9et al. explored the characteristics of their overlying strata structures and the disaster-inducing correlation with fault slip instability. Gao Rui et al10. clarified that the superposition of the high static stress environment of isolated coal mass and the strong dynamic load disturbance induced by the fracturing of thick and hard overlying strata is the core controlling factor for the occurrence of rockbursts. Jia Chao et al11. elucidated the deformation and failure mechanism of roadways in isolated working faces under the condition of hard roof, while Liu Guolei et al12. attributed the high rockburst risk of isolated working faces to the static-dynamic load coupling effect generated by the overlying strata structure of adjacent goafs. Yang Sen et al13. studied the stress transfer mechanism of the lateral roof structure after the fracturing of overlying strata in isolated working faces, and Ji Ming et al14. proposed that the loading effect of the abutment pressure from large-scale pressure arches formed by overlying strata in extensive goafs on section coal pillars is the main inducement for the frequent occurrence of high energy mining tremors during the extraction of isolated working faces. Meanwhile, the academic community has also made considerable progress in the research on the stress field characteristics of fold structures in coal mines. Xia Yucheng et al15. pointed out through the analysis of stress distribution in fold structures that buckle folds are mainly formed by horizontal compressive tectonic stress. Zhou Ye et al16. found in their research that the maximum principal stress is always compressive stress during the deformation of single-layer folds, and its amplitude increases continuously with the increase of horizontal strain of folds. Wang Shengben et al17. discovered an obvious stress stratification phenomenon in the synclinal core of single-layer folds, where tensile stress is concentrated in the lower strata and compressive stress dominates in the upper strata. Kang Hongpu et al18. emphasized that the overall stress level in the synclinal axis zone is significantly higher, and the increase amplitude of horizontal stress is much larger than that of vertical stress. For multi-layer folds, Wang Yugang et al19. confirmed that the maximum horizontal stress is concentrated in the upper and lower hard rock layers, while the stress level in the middle soft rock layer is relatively low. Tang Long et al20., Jia Housheng et al21. further explored the evolution law of mining-induced stress in fold zones and optimized roadway support parameters based on the plastic failure characteristics of surrounding rock. Despite the fruitful achievements in the above two research fields, existing studies still have obvious limitations in the face of the composite geological condition of “synclinal axis + isolated working face”. Firstly, most studies on isolated working faces focus on regular coal pillar boundaries under simple geological backgrounds, and lack in-depth consideration of the superposition effect of synclinal tectonic stress and mining-induced stress. Secondly, research on synclinal structures is mostly aimed at roadway engineering or single working faces, and few studies involve the dynamic evolution of overlying strata structure, the static-dynamic load coupling disaster-inducing mechanism of isolated working faces under the confinement of irregular coal pillars, as well as the quantitative identification of rockburst risk under the coupling of multiple stress fields, and other core scientific issues have not been effectively solved. To address the aforementioned research gaps, this study takes an isolated working face in the folded zone of an ultra-thick coal seam in a rockburst-prone coal mine in Gansu Province as the engineering background, and adopts an integrated research approach combining theoretical analysis, mechanical modeling, numerical simulation and field practice. It systematically reveals the dynamic evolution law of the spatial structure of overlying strata and the spatiotemporal distribution characteristics of surrounding rock stress in the folded zone of ultra-thick coal seams, and clarifies the formation mechanism of the unique long-arm F-shaped overlying strata structure and its stress concentration effect. Combined with the energy accumulation and release characteristics of key stratum fracturing, it quantitatively analyzes the static-dynamic load coupling disaster-causing mechanism of rockbursts in such isolated working faces. Aiming at the limitation of low early-warning accuracy of traditional methods, microseismic precursor indicators are selected from the temporal, spatial and intensity dimensions to construct a multi-index collaborative rockburst risk identification model based on a Back-Propagation (BP) neural network, realizing the adaptive optimization of indicator coupling weights and the quantitative classification of hazard levels. Finally, the engineering applicability of the research results is verified through field practice. The findings of this study can fill the theoretical gaps in the rockburst mechanism and risk identification under the composite geological condition of “folded zone + ultra-thick coal seam + isolated working face”, and provide important theoretical support and practical reference for the safe mining of deep ultra-thick coal seams under similar complex geological and mining conditions. Engineering background The present study was conducted in the 1504 mining district of a coal mine in Gansu Province, China. The district’s primary extraction is the No. 5 coal seam, which possesses an average thickness of 41.78 m and demonstrates a notable propensity for rockburst. The present study focuses on Panel 150,404, which is located in the south-western part of the district. The dimensions of the panel are such that its mining height is 12.1 m, with surface elevations ranging from + 1520 m to + 1665 m and corresponding seam floor elevations between + 1054 m and + 1163 m. This results in an average mining depth of approximately 480 m. The coal seam dips at an average angle of 10°. The mining advance rate is controlled at 3.2 m per day. The operation employs fully-mechanized top-coal caving mining technology, with support provided by ZFT14000/25/40D hydraulic supports in conjunction with advanced support measures. The ma-terial and haulage roadways for the panel are developed at an average elevation of 1090 m. The panel has an average strike length of 1448.9 m and a dip width of 160 m. During the preliminary mining stage (within 0 ~ 340 m from the setup room), the panel is located within the hinge zone of a syncline.The direction of the panel’s advance forms an angle of 13° with the synclinal axis. The specific morphological parameters of the syncline are listed in Table 1. Table 1Parameters related to synclinal morphology.NameHub length/mWavelength/mAmplitude/mAspect ratioInter-wing angle/°Slope angle/°Syncline66231800703.66137.46.6 Parameters related to synclinal morphology. As the final upper-seam panel to be extracted in the 1504 district, Panel 150,404 is bordered on the north by the goaf of the mined-out Panel 150,403, separated by a 15-m-wide barrier pillar. To the south, it is adjacent to the goafs of several panels (250,202 Upper, 250,203 Upper, 250,204 Upper, 250,205 Upper) from the neighbouring 2502 district, separated by irregularly-shaped, sawtooth-barrier pillars ranging from 20 to 170 m in width. The western boundary is delineated by the goaf of Panel 250,203 Lower. The eastern side is delineated by the primary roadway network of the 1504 dis-trict, encompassing the return airway, track rise, and belt rise. Consequently, Panel 150,404 exists in an asymmetric, three-sided goaf-surrounded “isolated” configuration. The general layout is illustrated in Fig. 1a. Working face overview. (a) 150,404 working face layout; (b)Key layer information. In accordance with the fundamental principles of key stratum theory22,23, the analysis of the overburden structure identifies three principal layers. The first of these is a silty sandstone, situated at a depth of 17.6 m above the coal seam with a thickness of approximately 12.55 m, which is classified as the lower sub key stratum. The second is a fine sandstone, located at a depth of 58.77 m above the seam, with a thickness of about 20.28 m, which serves as the upper sub-key stratum. The third is a silty sand-stone, situated at a depth of 166.45 m above the seam, with a thickness of roughly 68.64 m, which is identified as the main key stratum. The general layout is illustrated in Fig. 1b. Structural characteristics of overlying strata and stress evolution behavior of isolated working faces Evolutionary characteristics of overlying rock space structures Prior to the mining of Panel 150,404, the overburden load was borne by the panel coal mass, the adjacent irregular barrier pillars, and the section barrier pillars. The southwestern side of the working face is adjacent to the 2502 goaf, which has a width of approximately 2000 m. This is significantly greater than the average mining depth of approximately 480 m. This finding suggests that the roof in this area has undergone complete collapse and is now subject to significant mining activity. Conversely, the mean width of the 150,403 goaf to the northeast is a mere 160 m, which is significantly less than its mining depth. Surface subsidence data collected on 1 March 2025 indicate settlement values ranging from 29 to 1308 mm for monitoring points QX37 to QX53 above this area. The measured subsidence profile is displayed in Fig. 2. Subsequent analysis was conducted using the probability integral empirical method24. The results indicate that: Ground subsidence curve. In the formula: Zmax is the maximum surface subsidence; w is the subsidence factor (0.4 ~ 0.95); m is the mining height of the panel, taken as 12.1 m; β is the dip angle of the panel, taken as 12°. In accordance with the established parameters, the calculated empirical range for the maximum surface subsidence Zmax above the 150,403 goaf is determined to be between 4704 mm and 11,172 mm. However, the field-measured surface subsidence values were substantially lower than this theoretical range. This discrepancy indicates that the fracture zone in the overlying strata above the 150,403 goaf had a limited vertical extent, which was insufficient to induce large-scale fracturing and movement in the upper key stratum. The capacity for a working face to establish an interactive overburden spatial structure through coordinated movement with the strata overlying adjacent goafs is constrained by the widths of the intervening barrier pillars25. It is imperative to un-dertake a comprehensive evaluation of the distribution of roof fracture lines and the inherent stability of the pillars in order to ascertain the critical pillar width required for the formation of a stable overburden spatial structure. Its expression is given as follows26–28: In the formula: Ld denotes the distance from the key stratum fracture line to the coal wall, which is determined from field measurements as Ld = 10 m; ρ is the width of the elastic zone within the coal pillar, taken as 0.9 m based on the hardness of the coal; R represents the width of the plastic zone in the coal pillar; ad is the maximum width for dynamic instability of the coal pillar. In the formula: m is the height of the coal pillar; λ is the lateral pressure coefficient, λ = μ/(1-μ); φ0、c0 are the internal friction angle and cohesion, respectively, at the interface between the coal and the roof/floor strata; k is the stress concentration coefficient; μ is Poisson’s ratio; γ is the average unit weight of the overlying strata; H is the burial depth; Px is the support resistance provided by the rockbolts/standing support to the rib of the entry. The following parameters were applied, as derived from field-measured data: m = 12.1 m, φ0 = 25°, H = 480 m, and c0 = 3.5 MPa. It is evident from the in-situ tests conducted at the working face that the following parameters are observed: a relaxation time of 0.8, a mean relaxation time of 0.4, and a constant of 2.7. The influence of support resistance was not considered in this analysis. Substituting these values into the formula yields a critical pillar width W1 = 36.5 m. It can thus be concluded that the evolution of the overburden spatial structure during panel advance can be characterised as follows. In instances where the pillar width W1 is constrained to a maximum of 36.5 m, the overburden stratum situated above the working face is capable of undergoing coordinated movement with the overlying stratum on both adjacent sides, thereby establishing an interactive spatial structure. This configuration manifests as an asymmetrical “T”-shaped structure, as illustrated in Fig. 3a. Conversely, when W1 > 36.5 m, the irregular pillar acts as a barrier, thereby preventing the fracture zone in the overburden from connecting with the goaf of the 2502 district. In this particular instance, structural interaction occurs exclusively with the overburden situated above the 150,403 goaf, thereby giving rise to a long-arm “F”-shaped structure, as illustrated in Fig. 3b. Overlying rock structures of the working face. (a) asymmetric “T”-shaped structure; (b) long-arm “F” configuration. Field measurements have been conducted which indicate that the minimum width of the irregular pillar on the haulage roadway side is 95 m during mining within the synclinal hinge zone. This dimension is conducive to the formation of a long-arm “F”-shaped overburden structure. As the unsupported span of the overbur-den increases, tensile failure initially occurs in the lower key stratum. Subsequent to this process, the resulting rock blocks undergo rotation and compression. This process is accompanied by the successive rotational instability of the lower key stratum in both the long and short arms of the “F”-shaped structure. Consequently, the upper key stra-tum also undergoes tensile failure. The rotational instability of its blocks further in-duces passive destabilization in the underlying strata, ultimately leading to the rup-ture of the main key stratum. As the sequence of overburden fracturing progresses, a cyclical pattern becomes evident. Figure 4 presents a schematic representation of the local fracturing morphology of the overburden. Schematic diagram of overlying rock failure process. Mechanical analysis of stress distribution in isolated working face surrounding rock Figure 5 presents the computational model used to analyze the stress distribution in the surrounding rock of the working face. To simplify the analysis and elucidate the underlying mechanics, the problem is simplified to a plane-strain condition, and a sin-gle synclinal structural unit is selected as the focus of study. Mechanical stress model of isolated working face surrounding rock. (a) Coal seam occurrence state; (b) Simplified mechanical model of coal seams. Let the stress function be defined as29: In the formula: A is an undetermined coefficient; C is the vertical in-situ stress within the rock strata; P is the horizontal confining stress; L is the syncline wavelength. The aforementioned stress function satisfies the biharmonic equation: Substituting Eq. (5) into (7) yields: The relationship between the undetermined coefficient A and the syncline wave-length L in the above equation can be determined through both geometric and physical equations: In the formula: E is the elastic modulus of the rock layer; μ is the Poisson’s ratio. Substituting Eq. (7) into (8) and integrating yields: While y = 0 and x = 4nL (n = 0, ± 1, ± 2…) in Eq. (9), ω = u. Setting x = 0 yields the relationship: In the formula: G = E/2(1 + μ) represents the stiffness modulus of the rock layer. Substituting Eq. (10) into (7) yields: The model’s undulating geometry has been analysed, with the result that the stress components, σx and σy, have been found to exhibit a roughly cosine-periodic distribu-tion along the x-axis. In addition, it has been determined that τxy follows an approxi-mately sine-periodic pattern. In locations where x = 4nL (n = 0, ± 1, ± 2…), correspond-ing to the synclinal hinge, it can be demonstrated that both σx and σy attain their maximum values. Conversely, in the vicinity of x = 4nL + L, which corresponds to the limb region of the syncline, τxy attains its maximum magnitude. Along the y-axis, at y = 0, both σx and σy are zero, and the absolute value of τxy is minimised, though it remains greater than zero. At y = h/2, the absolute magnitudes of σx, σy, and τxy are maximised, with σx being compressive and σy tensile. In a similar manner, at y = -h/2, the absolute values of all stress components also reach their maximum, but in this case, σx is tensile and σy is compressive. From the relevant data in Table 1, it can be calculated that the maximum values of σx and σy are 44.3 MPa and 43.4 MPa, respectively, and τxy reaches a maximum value of 41.2 MPa in the limb zone. Isolated working face numerical simulation analysis Establishment of numerical models The mining process for Panel 150,404 was simulated using the FLAC3D numerical code. A numerical model was developed using integrated borehole geological data and the actual geological conditions of the mining district. The dimensions of the model are 2400 m × 1200 m × 420 m (Strike × Dip × Vertical). In the model coordinate system, the Y-axis is aligned with the strike of the coal seam, the X-axis is aligned with the dip direction and the Z-axis is vertical. The boundary conditions were set as follows: horizontal displacements were fixed at both ends. Based on in-situ stress measurements, a confining stress with a gradient of 0.025 MPa/m and an initial value of 15.75 MPa was applied to the base of the model and its X-direction sides. For the Y-direction sides, a confining stress with a gradient of 0.035 MPa/m and an initial value of 22.05 MPa was applied. The top of the model was set as a free surface and a gravitational stress of 2.5 MPa was applied, corresponding to the actual overburden depth. Figure 6 illustrates the overall configura-tion of the model. Numerical simulation layout of the working face. (a) Overall model condition; (b) Working face 150,404 condition. The mechanical parameters of the intact coal-rock mass, as determined by labora-tory testing of samples collected from Panel 150,404’s mining area, are consistent with field measurements. These parameters are detailed in Table 2. Due to variations in the degree of caving, the load-bearing behaviour and deformation modulus differ between the goafs on the two sides of the panel. In a fully caved goaf, the lower bulking factor and tighter contact between rock fragments result in a higher equivalent elastic mod-ulus. This enables the goaf to carry a portion of the overburden load. Conversely, an inadequately caved goaf has a significantly lower equivalent elastic modulus, meaning the load is diverted to the surrounding pillars and the working face. Following relevant research and empirical formulas30, the equivalent deformation modulus for a fully caved goaf is taken as half that of intact rock, while for an inadequately caved goaf it is taken as one-fifth. Table 2Mechanical parameters of coal and rock.LithologyDensity/(kg·m3)Bulk modulus/GPaShear modulus/GPaCohesion/MPaInternal friction angle/(°)Mudstone22002.01.62.526Sandy mudstone24204.01.92.737Siltstone27003.63.65.037Argillaceous siltstone24803.63.52.037Fine sandstone27003.73.45.037Coal14001.10.83.525 Mechanical parameters of coal and rock. Stress evolution patterns at the working face As shown in Fig. 7, the cloud diagrams illustrate the variation laws of the maximum and minimum principal stresses in the strike direction of the working face after excavation. Both theoretical derivation and numerical simulation reveal that the horizontal stresses σx and σy reach their maximum values at the synclinal axis, while the shear stress τxy peaks at the limbs of the syncline, showing consistent laws. Field in-situ stress measurements indicate that the maximum principal stress in the syncline area of the working face ranges from 23.5 MPa to 25 MPa. Quantitatively, the simulation shows that the peak maximum principal stress at the axis (42.1 MPa) deviates from the theoretical value (44.3 MPa) by approximately 5.2%, verifying the applicability of the theoretical model. In terms of shear stress, the simulation demonstrates that the shear strain energy density at the limbs is 23% ~ 31% higher than that at the hinge area, which confirms the theoretical judgment of the τxy peak. However, since the theoretical model does not consider the effects of rock mass joints and mining-induced stress superposition, the location of the extreme point shifts 15 ~ 20 m toward the hinge compared with the theoretical prediction. Stress evolution nephogram of the coal seam at the working face. (a) Maximum principal stress; (b) Minimum principal stress. With the advancement of the working face, the concentration zones of the maximum and minimum principal stresses are mainly distributed on both sides of the synclinal axis. The unique long-arm “F” structure formed in the overlying strata of the working face leads to a significant increase in the stress peak of the coal and rock mass ahead of the working face. The stress concentration degree between the synclinal axis and the material roadway is noticeably higher than that on the haulage roadway side, which is attributed to the different widths of the barrier pillars on the two sides. When the working face advances 250 m from the setup room, the distance between the synclinal axis and the haulage roadway is only 10 m, resulting in a remarkable rise in the surrounding rock stress. Subsequently, the high stress zone spreads over the entire working face, with the peak values of the maximum and minimum principal stresses reaching 50 MPa and 25 MPa, respectively. When advancing to 350 m, the increase in the stress peak becomes relatively gentle, but the scope of the stress concentration zone between the synclinal axis and the roadway expands significantly. The prominent difference in horizontal elevations between the two gateways, combined with the unequal pillar widths, jointly results in a distinctly non-uniform stress distribution across the working face. This enlarges the difference between the maximum and minimum principal stresses near the end of the material gateway, raising the risk of rockburst in working face. As shown in Fig. 8, the variation pattern of the maximum principal stress with the advancing distance of the working face is clearly observable. As shown in Fig. 8a and b, the areas of variation for the maximum and minimum principal stresses in the advance zone of the working face are largely consistent. As the working face advances, the maximum principal stress increases from 32.6 MPa to 42.1 MPa, an average increase of 29.1%, while the minimum stress increases only 19.6%. The maximum influence ranges of the two are 200 m and 150 m, respectively. As illustrated in Fig. 8c, when the working face advanced from 50 to 150 m away from the setup room, the average maximum principal stress in the area 10 m ahead of the working face increased from 27.7 MPa to 40.0 MPa. However, after the advancing distance exceeded 150 m, the rate of stress increase slowed down significantly. As also shown in Fig. 8d, during the working face advance from 50 to 150 m from the setup room, the average minimum principal stress in the area 10 m ahead of the working face increased from 11.6 MPa to 15.9 MPa. After the advancing distance exceeded 150 m, the rate of stress increase accelerated significantly. With continued advancement, the stress peaks on both sides of the roadway increased. Notably, the stress increase on the inner rib (the solid coal side of the working face) was significantly greater than that on the outer rib (the panel pillar side). Furthermore, the location of the stress peak gradually migrated towards the solid coal zone of Panel 150,404. Stress distribution of the working face. (a) Maximum principal stress variation curve along the strike; (b) Maximum principal stress variation curve along the strike; (c) Maximum principal stress variation curve along the dip; (d) Minimum principal stress variation curve along the dip. Mechanism of rockburst in isolated working face Key fracture energy accumulation characteristics The preceding analysis demonstrates that the stress state of the surrounding rock changes as the synclinal axis moves closer to the haulage roadway. Despite the relatively modest increase in maximum principal stress on the haulage roadway side, the overall stress distribution in the panel shifts towards the haulage roadway side. The stress levels in the coal mass on the side of the material roadway are found to be sig-nificantly higher(by more than 50%)than in the central area of the working face. As the area of the Goaf expands, the load borne by the key stratum above it con-comitantly increases. Prior to the fracturing of the key stratum, an augmentation in load transfer through the canti-levered strata results in an escalation of the static load level ahead of the face. This phenomenon is characterised by the presence of elevated, extensive stress concentration zones at both extremities of the working face. The elastic energy that is released when the key stratum structure fractures originates primarily from the bending strain energy of the roof. In accordance with the cantilevered roof energy storage model30,31, the calculation of this energy is as follows: In the formula: η is the seismic efficiency, defined as the proportion of elastic en-ergy released during strata fracture that is radiated as seismic waves, Its value ranges from 0.36% to 3.6%32; E is the elastic modulus of the cantilevered section of the key stratum; q is the uniformly distributed load acting on the cantilevered key stratum; l, b, and h₂ are the fracture span, width, and thickness of the key stratum. The quantity of energy discharged upon the fracturing of the key stratum can be cal-culated by means of Eq. (12). This energy release has been shown to generate a significant amount of dynamic load disturbance, which is a contributing factor to the initiation of rockbursts at the working face. The dynamic energy emanating from a fracture in the key stratum propagates through the coal-rock mass in the form of seis-mic waves. This has the effect of disturbing the in-situ stress field, thereby promoting the continued propagation of internal fractures. Concurrently, the vibration induced in the coal-rock mass engenders cumulative damage at the particle level. This build-up of damage brings the mass closer to its critical state of dynamic instability, thereby increasing the risk of a rockburst occurring at the working face. Dynamic-static load superposition-induced surge mechanism A thorough evaluation of the stability of the surrounding rock at the working face must take into account the combined effect of maximum and minimum principal stresses. The present study employs the Mohr–Coulomb failure criterion to evaluate coal-rock failure33. As demonstrated in the extant literature, the following relationship exists between the maximum principal stress σ₁, the minimum principal stress σ₃ and the stress components in a plane coordinate system σx, σy, τxy34: In the formula: σ₁ is the maximum principal stress; σ₃ is the minimum principal stress. When the dynamic energy released by key stratum fractures propagates through the coal-rock mass as seismic waves, the wave energy undergoes varying degrees of attenuation influenced by the fracture density of the rock mass and the propagation distance. The relationship between the stress disturbance induced within the medium by the seismic waves and their propagation distance can be expressed as follows35: In the formula: σpm is the normal stress generated at a given location by the propagation of the P-wave; σsm is the shear stress generated at the corresponding location by the S-wave; ρ is the density of the coal-rock medium; Cp and Cs are the propagation velocities of the P-wave and S-wave, respectively, within the coal-rock medium; L₀ is the central distance between the dynamic load source and the impact location; λ is the attenuation coefficient of the dynamic load. In accordance with the Mohr–Coulomb criterion, the superposition of static and dynamic loads has the potential to induce coal-rock failure through three primary mechanisms, namely: The first scenario is characterised by an increase in σ₁, whilst maintaining σ₃ constant. The second scenario is marked by a decrease in σ₁, while maintaining σ₃ constant, with the possibility of σ₃ becoming negative. The third sce-nario is typified by a simultaneous increase in σ₁ and decrease in σ₃, as illustrated in Fig. 9. When the dynamic load released by a key stratum fracture is superimposed on the static load of the coal-rock mass at the working face (see Fig. 10), the stress disturbance induced by the dynamic load can cause the minimum principal stress, designated here as σ₃, to decrease significantly or even transition into tensile stress. It is important to note that the σ1 required for failure decreases as σ3 diminishes. This means that failure of the coal-rock mass may occur under a lower σ1 than would be required under pure static loading conditions. In the context of these conditions, the relationship between the maximum principal stress σ₁ and the minimum principal stress σ₃ at failure under dynamic load disturbance can be expressed as follows: Dynamic failure criterion of coal-rock mass based on Mohr–Coulomb theory. Mechanism of dynamic and static load impact instability of the working face. Substituting Eqs. (11) and (12) into Eqs. (13) and (15) allows for the determination of whether the coal-rock mass enters a failure state upon the initial fracture of the key stratum. During the initial fracturing of the key stratum, the working face is located in the high static load zone of the synclinal axis. The ultimate fracture of the key stratum releases a substantial amount of energy, generating strong dynamic loads in the form of P-waves and S-waves. The P-waves first create tensile cracks, followed by S-waves that induce shear slip. The superposition of static and dynamic loads causes a significant reduction in the minimum principal stress, or even a shift to tensile stress, with a corresponding increase in the maximum principal stress. According to the Mohr–Coulomb criterion, the coal-rock mass thus reaches its strength limit and fails under a lower maximum principal stress, triggering a rockburst. Therefore, the core mechanism of rockburst incubation is the coupling effect of “high tectonic static load + strong dynamic load from fracturing + abrupt stress state transition”. Research on rockburst risk identification in isolated working areas Strong mine shock event confirmed Rockburst is fundamentally a result of coupled static and dynamic loading. Isolated working face located within a synclinal axis are subject to elevated levels of static stress within the surrounding rock mass. In this critical context, strong mining-induced seismic events, which represent a pronounced increase over the regional microseismic background, release dynamic energy that further destabilises the stress equilibrium. This process has been shown to substantially elevate the risk of rockburst. Consequently, the systematic analysis of microseismic monitoring data to identify temporal evolution patterns and precursor characteristics of major seismic events has become a critical methodology for assessing rockburst hazard. The preliminary identification of major mining-induced seismic events was facilitated by the analysis of microseismic monitoring data from Panel 150,404 (January to February 2025, Fig. 11). The data demonstrate a clear correlation between key microseismic parameters, with event frequency and released energy exhibiting an approximately linear relationship. This observed correlation is consistent with the energy-frequency distribution described by the Gutenberg-Richter law36, indicating statistical self-similarity in the seismic activity of this region. Consequently, this pattern provides a potential basis for identifying precursors to larger seismic events. Distribution characteristics of microseismic events at the working face. In the formula: E represents microseismic energy; N denotes the total number of microseismic events with energy not less than E; both the a value and b value are constants characterising seismic activity levels, with the a value being positively correlated with regional hazard, a lower b value indicates higher impact hazard in the area. As demonstrated in Fig. 12, a discernible negative correlation exists between the magnitude and frequency of microseismic events. The plot reveals a marked shift in energy levels. Subsequent to this transition, there was a marked decline in the number of events that exceeded the background seismic level, accompanied by an increase in the dispersion of the data. It is evident that, on the basis of this characteristic change, the energy value corresponding to the shift point, 103.93 J (converted to 8511.38 J), was established as the critical threshold for the classification of strong mining-induced seismic events. The application of this threshold to the data from Panel 150,404 indicates that 13 such strong seismic events occurred during the mining period, as detailed in Fig. 13. Energy-frequency curve of microseismic events at the working face. High energy event dates and energy. Characteristics of microseismic precursor information In the context of a syncline structure, this study identifies six microseismic indicators across temporal, spatial, and intensity dimensions: the b-value, P(b)-value, microseismic activity scaling ΔF, microseismic activity degree S-value, algorithmic complexity AC-value, and equivalent energy magnitude EEM-value. The b-value and P(b)-value are of primary importance in the characterisation of the distribution of microseismic events, thereby unveiling the dynamic process of energy accumulation within the syncline structure area. The S-value, AC-value, and EEM-value have been shown to reflect the intensity of microseismic activity, thus aiding in the analysis of the evolution of stress state in the rock mass. The microseismic activity scaling ΔF is further utilised to evaluate the stability and potential risk level of microseismic activity in this area. The precise definitions of these indicators are outlined in the following section: (1) b-value The calculation of the b value is shown in Eq. (17)37, When this value continuously decreases, or briefly rises and then accelerates its decline, it indicates the occurrence of a strong mining tremor event: In the formula, m denotes the number of magnitude categories; Ni represents the total number of microseismic events within the i-th magnitude category. (2) P(b)-value The P(b) value is a common means of characterising the distribution of low-magnitude microseismic events, Reflecting the stability of stress accumulation. Research has demonstrated37 that prior to high energy events, the frequency of low energy microseismic activity frequently exhibits an increasing trend, which can be captured through variations in the P(b) value. The precise definition of P(b) is provided below. In the equation, Mi denotes the energy level of the microseismic event. (3) △F-value The microseismic activity scale, designated as ∆F, is a monitoring index that has been established on the basis of seismological theory. When its value demonstrates a sustained increase or remains elevated over an extended period, it typically indicates a significantly increased probability of high energy mine seismic events38. In the formula, T represents the number of days; M denotes energy. (4) S-value The microseismic activity degree S-value is a comprehensive index utilised for the evaluation of the intensity of microseismic activity. The quantification of microseismic intensity is achieved by integrating multiple parameters, including event frequency, energy release magnitude, spatial clustering of seismic sources, and temporal correlation of intensity. A sustained increase or prolonged high level of the S-value is generally regarded as a primary indicator of heightened risk of rockburst40. The calculation formula for this index is presented as Eq. (20). In the formula, M denotes energy. (5) AC-value The algorithmic complexity AC-value is a quantitative metric derived from seismological theory. The objective of the present study is to characterise the evolution of mine seismic activity sequences during mining operations and to reflect the temporal distribution characteristics of microseismic events39. When this value abruptly shifts from a chaotic state to an ordered state, accompanied by enhanced periodicity in the time-series signal, it often indicates the occurrence of a high energy event. The precise definition of this metric is given by Eq. (21). In the formula, n is the number of energy level changes within a specified time interval; M is calculated as the difference between Mmax and Mmin with an additional constant of 1. (6) EEM-value The equivalent energy magnitude EEM parameter serves as an indicator for quantifying the average energy release level of microseismic events within a specific region. Research has demonstrated that anomalous fluctuations in this parameter are closely associated with substantial changes in the stress state of the rock mass. When this value exhibits a stepwise jump, it often indicates the occurrence of a high energy event. The calculation formula is provided below. In the formula, m* represents the normalized energy level of the microseismic event. Within the statistical window of the studied isolated working face at the synclinal axis (with a 5 day time window and 1 day sampling interval), all the aforementioned indicators exhibited clear precursory characteristics prior to the occurrence of 13 high energy events. As demonstrated in Fig. 14a-d, the mean b-value was found to be 0.01. Of the 13 high energy events, eight exhibited b-values below the mean value. Despite the b-values of the remaining five events exceeding the mean, the majority were at or near their minima, or exhibited a rapidly decreasing trend. This finding indicates that b-values typically decrease rapidly 1 2 days prior to a high energy event, attaining or approaching their nadir at the time of the event, and subsequently exhibiting a gradual recovery. A similar pattern was observed for the P(b)-value: although only four events had P(b)-values below average, in the remaining nine cases the value was either in a low range or exhibited a rapid decline before the event. The behaviours exhibited by the microseismic activity scale ΔF and the microseismic activity degree S-value prior to high energy events were found to be essentially consistent with those of the P(b)-value. Seismic precursor indicator time series curve. (a) b-value; (b) P(b)-value; (c) ΔF-value; (d) S-value; (e) AC-value; (f) EEM-value. As illustrated in Fig. 14e and f, the mean algorithmic complexity AC-value was determined to be 0.19. A total of ten events exhibited AC-values that were below the average value, while the remaining three events were at or near their minima. The average equivalent energy magnitude EEM parameter was found to be 0.001, indicating relatively quiet microseismic activity in this isolated working face area, with low-energy events predominating. In summary, before high energy events, each warning indicator displayed pronounced anomalous variations towards either high or low values, demonstrating the potential efficacy of these indicators in predicting major mine seismic events. Impact risk identification model construction (1) Data standardization In consideration of the varying dimensions among the indicators, standardisation is imperative to enhance data consistency and comparability, thereby ensuring the elimination of dimensional differences between positive and negative indicators. The specific calculation formula is provided below. Positive indicators: Negative indicators: In the formula, Ri denotes the i-th value of a given early warning indicator; Rmax represents the maximum value within that early warning indicator; Rmin denotes the minimum value within that early warning indicator. (2) Hazard level classification As illustrated in Fig. 12, three distinct inflection points are observed on the energy-frequency curve, with coordinates of (3.12, 3.05), (3.49, 2.72), and (3.93, 1.79), respectively. In the vicinity of each inflection point, the frequency of microseismic events exhibits a significant decrease, accompanied by an increase in data dispersion. This shift in distribution characteristics indicates a phased evolution in the stress state and fracture mechanism of the surrounding rock system of the working face, corresponding to a progressive escalation of rockburst risk. Therefore, the aforementioned inflection points can serve as quantitative critical criteria for classifying different levels of rockburst hazard at the working face. Based on the inflection characteristics of the energy-frequency curve, the quantitative energy thresholds for rockburst hazard classification at the working face are determined as follows: the critical energy value for weak rockburst hazard is Er = 103.12 = 1318.26 J, for moderate rockburst hazard, Ez = 103.49 = 3090.30 J, and for strong rockburst hazard, Eq = 103.93 = 8511.38 J. The maximum energy value of microseismic events occurring each day will be used to determine the corresponding rockburst hazard level for that day, Table 3. Table 3Correspondence between rockburst risk levels and labels.Single day maximummicroseismic energyCorresponding rockburst risk levelCorresponding labelLess than 1318.26 Jno rockburst risk11318.26 J ~ 3090.30 Jweak rockburst risk23090.30 J ~ 8511.38 Jmoderate rockburst risk3Greater than 8511.38 Jstrong rockburst risk4 Correspondence between rockburst risk levels and labels. (3) Construction of the rockburst risk identification model The integration of multiple precursor indicators derived from microseismic monitoring is achieved through the utilisation of a back-propagation neural network (BPNN), thereby facilitating the conversion of qualitative assessments into quantitative identifications. The configuration of the resulting multi-indicator microseismic risk identification model is illustrated in Fig. 15. Multi-indicator risk identification model for microseismic monitoring. (4) Training of the risk identification model Subsequent to the standardisation of the data, the dataset was divided into three subsets: In the context of the neural network model, the term “training set” is employed to denote the set of data that is used to train the network. This set is specifically utilised for training the neural network model. The second point to consider is the validation set. This is employed for the purpose of evaluating the generalisation ability of the model, as well as optimising the network architecture. The third point pertains to the test set, which is applied for the purpose of final performance evaluation after the completion of model training. The sample proportions for the training, validation, and test sets were set to 7:2:1. First, the candidate range for the number of hidden layer nodes (4 to 13) was determined based on empirical formulas. Subsequently, using a trial-and-error method, microseismic events from the 150,404 working face (January to February 2025) were taken as samples. With consistent training parameters (learning rate of 0.01, target error of 0.001), networks with different node counts were trained repeatedly 10 times, and the accuracy of the validation set was used as the evaluation metric. The trial calculation results showed that when the number of nodes increased from 4 to 8, the validation set accuracy continuously improved from 84.8% to 94.9%. However, when the number of nodes exceeded 8, the training set accuracy continued to rise while the validation set accuracy began to decline (dropping to 93.1% at 10 nodes), indicating significant overfitting. Balancing accuracy and generalization capability, the optimal number of hidden layer nodes was determined to be 8. This result is highly consistent with the optimization outcome, validating the rationality of the network design. As illustrated in Fig. 16, the performance curves of the neural network are evident during the training, validation, and testing phases. As demonstrated in the accompanying figure, the minimum mean squared error (MSE) of 0.005 was attained at the 37th iteration. Performance curves for training, validation, and testing of neural networks. As demonstrated in Fig. 17, the error exhibited a direct correlation with the number of iterations during the training process of the neural network. As demonstrated in the accompanying figure, the error decreased fourfold following the 20th iteration and once more after the 32nd iteration. It should be noted that at this point, the training had not yet reached its conclusion. However, when the iteration reached the 42nd step, the error increased continuously for six consecutive occurrences, leading to the termination of the training process. Iterative process of sample error optimization. As illustrated in Fig. 18, the prediction-fitting curve of the neural network model for the 150,404 working face is presented. The findings suggest a high degree of consistency between the predicted values and the measured data, with the optimal fit observed in the extreme intervals close to 1 or 0, where the expected and predicted values nearly coincide. Consequently, the proposed model can be employed in the construction of a mine safety risk identification system. Comparison between predicted and actual values in the 150,404 working face. Application effect verification In order to evaluate the prediction performance of rockburst risk identification on the working face, the R‑scoring method41 was employed for quantitative assessment. The forecast performance score R can be calculated using Eq. (25). Subsequent to the training of the BP neural network, the mean values of the weight parameters were obtained. The final weights corresponding to each indicator are as follows: b‑value (0.065), P(b)‑value (0.015), ∆F value (−0.298), S‑value (0.471), AC‑value (−0.351), and EEM value (−0.412). In the formula, Ai represents the weights of each indicator after quantification; ai represents the weights of each indicator before quantification; n denotes the number of early warning indicators. The final weights of the indicators were determined as follows: b‑value(0.04), P(b)‑value(0.009), ∆F-value(0.185), S‑value(0.292), AC‑value(0.218), and EEM-value(0.256). The integrated microseismic early‑warning results calculated using these weights are shown in Fig. 19. The analysis shows that during the total 59 day monitoring period, the model achieved 56 accurate predictions and 3 misjudgments for microseismic events, with two overestimations and one underestimation of the risk level. The prediction accuracy of the model for each hazard level is summarized in Table 4. Impact risk identification results. Table 4Risk identification model performance evaluation.Hazard levelHazard level 1Hazard level 2Hazard level 3Hazard level 4AccuracyReports53081394.9%Occurrences632813Rccuracy rate83%95%100%100%R-value0.6530.87511 Risk identification model performance evaluation. The existing microseismic monitoring and identification system for this working face employed the following threshold criteria: a daily total energy of 4.57 × 104 J, a daily event frequency of 15, and a daily maximum energy of 9.82 × 103 J. The triggering of a rockburst hazard warning was initiated by the exceeding of any one of these thresholds, otherwise, the condition was classified as non-hazardous, Table 5. The corresponding risk identification results are displayed in Fig. 20. Table 5Impact risk identification results.Risk stateRisk presentRisk absentAccuracyReports84081.4%Occurrences1148Accuracy rate72.7%83.3%R-value0.6230.526 Impact risk identification results. Impact risk identification results of the original method. Discussion Engineering applicability of the long-arm “F” structure theory and stress evolution model The long-arm “F”-shaped overburden structure model proposed in this study, established based on the geological and mining characteristics of isolated working faces in synclinal hinge zones of extra-thick coal seams, exhibits clear engineering applicability for similar working conditions. The critical coal pillar width calculation method, derived from field measurements and theoretical analysis, can effectively determine the overburden structure morphology during mining and provide a quantitative basis for early-stage rockburst risk prediction. The model has been demonstrated to reveal stress distribution characteristics, including stress concentration at the synclinal core and asymmetric stress evolution in the two roadways. These characteristics can guide the zoning of rockburst prevention and control and the determination of key pressure relief locations. However, the application of the model is constrained by the matching degree of geological and mining parameters. Factors such as coal-rock mechanical parameters, syncline structural morphology, mining height, and advance speed all influence the simulation results. In practical engineering applications, parameter calibration should be conducted based on borehole geological data, in-situ stress measurements, and real-time monitoring data to avoid blind application. Furthermore, the current model primarily focuses on the static-dynamic load coupling effect during mining, while the impacts of dynamic disturbance factors such as blasting operations and hydraulic support movement on overburden stability require further investigation and refinement. Optimization space and application extension of the multi-index collaborative identification model The microseismic multi-index collaborative identification model based on BP neural network, which selects six indicators from the temporal, spatial, and intensity dimensions and achieves multi-source data fusion through adaptive weight optimization, has improved the early warning accuracy of rockburst risk in isolated working faces within folded zones to 94.9%, effectively addressing the low accuracy issue of traditional single indicator methods. The energy thresholds determined based on inflection points of the energy-frequency curve provide quantitative criteria for graded early warning and graded prevention and control of rockburst. The model still has optimization potential in the following aspects: the time window can be further refined according to mining advance speed to capture short-term precursors, future work should integrate multi-source monitoring data such as mine pressure monitoring (support resistance, roof subsidence), stress monitoring (coal mass stress, surrounding rock stress), and displacement monitoring to enhance the comprehensiveness and accuracy of risk identification, at the algorithmic level, the BP neural network can be optimized by incorporating particle swarm optimization and genetic algorithm to address issues of slow convergence and susceptibility to local optima. In terms of application extension, the model’s indicator system can be adjusted according to different rockburst-inducing mechanisms, enabling its popularization to other complex geological conditions such as isolated working faces near faults, working faces with thick and hard roofs, and deep high stress working faces. Furthermore, by integrating the model with mine safety monitoring and early warning platforms, real-time data input, automatic indicator calculation, and intelligent output of risk warning results can be achieved, forming a closed-loop technical system of “dynamic identification—graded early warning—targeted prevention and control” for rockburst risk, thereby providing more intelligent technical support for the safe mining of deep coal seams under complex geological conditions. Conclusions The present study systematically investigated the evolution of overburden struc-tures, the distribution of stress resulting from mining activities, and the mechanisms of rockburst in isolated working face within synclinal hinge zones. The primary conclusions that can be drawn from this analysis are as follows. When the minimum width of the irregular barrier pillar exceeds the critical value of 36.5 m (95 m in the field application), the overlying strata fail to form a connected fracture zone with the adjacent goafs, thereby developing a unique long-arm “F”-shaped overlying structure. Within this structural model, the stress distribution of the surrounding rock exhibits distinct zonal characteristics: the normal stress components reach their peak values in the synclinal core, where the shear stress is zero, conversely, τxy attains its maximum in the synclinal limbs. Vertically, the absolute values of σx and σy peak at y = ± h/2 and exhibit opposite tensile-compressive characteristics, which substantially elevates the risk of rockburst incidents at the working face. When the minimum width of the irregular barrier pillar exceeds the critical value of 36.5 m (95 m in the field application), the overlying strata fail to form a connected fracture zone with the adjacent goafs, thereby developing a unique long-arm “F”-shaped overlying structure. Within this structural model, the stress distribution of the surrounding rock exhibits distinct zonal characteristics: the normal stress components reach their peak values in the synclinal core, where the shear stress is zero, conversely, τxy attains its maximum in the synclinal limbs. Vertically, the absolute values of σx and σy peak at y = ± h/2 and exhibit opposite tensile-compressive characteristics, which substantially elevates the risk of rockburst incidents at the working face. Numerical simulation results indicate that the maximum principal stress in the abutment pressure zone ahead of the working face increases from an initial 32.6 MPa to a peak of 42.1 MPa, representing a cumulative increase of 29.1%, with the stress concentration factor rising to 2.34. The influence range of the high-stress zone in front of the working face expands from 20 to 40 m, accompanied by a pronounced asymmetric stress distribution. Specifically, the stress increment on the inner rib (solid coal side) of the haulage roadway is significantly higher than that on the outer rib (pillar side), and the stress peak gradually shifts toward the solid coal zone of the working face, forming a highly heterogeneous high-stress concentration area. Numerical simulation results indicate that the maximum principal stress in the abutment pressure zone ahead of the working face increases from an initial 32.6 MPa to a peak of 42.1 MPa, representing a cumulative increase of 29.1%, with the stress concentration factor rising to 2.34. The influence range of the high-stress zone in front of the working face expands from 20 to 40 m, accompanied by a pronounced asymmetric stress distribution. Specifically, the stress increment on the inner rib (solid coal side) of the haulage roadway is significantly higher than that on the outer rib (pillar side), and the stress peak gradually shifts toward the solid coal zone of the working face, forming a highly heterogeneous high-stress concentration area. The incubation of rockbursts in isolated working faces within fold zones is the result of the superposition of static stress concentration and dynamic load disturbance induced by key stratum fracturing. On the one hand, the long-arm “F”-shaped overlying structure establishes a high static stress environment in the surrounding rock of the working face. On the other hand, the fracturing of key overlying strata generates intense dynamic load disturbances. Under the combined effect of these factors, the minimum principal stress of the coal-rock mass decreases sharply, rendering the coal-rock mass highly susceptible to reaching the critical state of dynamic instability and ultimately triggering rockburst disasters. The incubation of rockbursts in isolated working faces within fold zones is the result of the superposition of static stress concentration and dynamic load disturbance induced by key stratum fracturing. On the one hand, the long-arm “F”-shaped overlying structure establishes a high static stress environment in the surrounding rock of the working face. On the other hand, the fracturing of key overlying strata generates intense dynamic load disturbances. Under the combined effect of these factors, the minimum principal stress of the coal-rock mass decreases sharply, rendering the coal-rock mass highly susceptible to reaching the critical state of dynamic instability and ultimately triggering rockburst disasters. A microseismic multi-index risk identification model based on a Back-Propagation (BP) neural network was established. Six precursor indicators, namely the b-value, P(b)-value, ΔF-value, S-value, AC-value, and EEM-value, were selected from the temporal, spatial, and intensity dimensions to effectively characterize the precursory features of high energy microseismic events. Through data standardization, hazard level classification, and model training optimization, the model realizes the quantitative fusion and collaborative discrimination of multi-source microseismic information. Field application results demonstrate that the model improves the rockburst early-warning accuracy of the working face from 81.4% to 94.9%, significantly enhancing the capability for rockburst risk identification and early warning in isolated working faces within fold zones of extra-thick coal seams. A microseismic multi-index risk identification model based on a Back-Propagation (BP) neural network was established. Six precursor indicators, namely the b-value, P(b)-value, ΔF-value, S-value, AC-value, and EEM-value, were selected from the temporal, spatial, and intensity dimensions to effectively characterize the precursory features of high energy microseismic events. Through data standardization, hazard level classification, and model training optimization, the model realizes the quantitative fusion and collaborative discrimination of multi-source microseismic information. Field application results demonstrate that the model improves the rockburst early-warning accuracy of the working face from 81.4% to 94.9%, significantly enhancing the capability for rockburst risk identification and early warning in isolated working faces within fold zones of extra-thick coal seams. Acknowledgements This research is funded by the projects: National Natural Science Foundation of China (52274098), National Key Research and Development Program (2022YFC3004603), Jiangsu Province International Collaboration Program- Key national industrial technology research and development cooperation projects (BZ2023050), These supports are gratefully acknowledged. Author contributions Yongqin Xie and Anye Cao were responsible for the drafting of the main body of the paper and the securing of funding. Geng Li and Weiwei Zhao constructed the theoretical model and performed numerical simulations. The theoretical and simulation results were analysed by Cheng. Zhang. and Zhongjun. Feng, who also collected and organised the data. The document was subjected to review by all authors. Funding This research was funded by National Natural Science Foundation of China, 52274098, Anye Cao, National Key Research and Development Program, 2022YFC3004603, Anye Cao, Jiangsu Province International Collaboration Program- Key national industrial technology research and development cooperation projects, BZ2023050, Anye Cao. Data availability The datasets used or analyzed during the current study are available from the corresponding author on reasonable request. Declarations Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Data Availability Statement The datasets used or analyzed during the current study are available from the corresponding author on reasonable request.