Multiscale Fracture Characteristics of Coal and Their Influence on Fracture Propagation

Abstract

The complex natural fracture system in coalbed methane reservoirs plays a critical role in hydraulic fracture propagation. This study investigates coal specimens collected from the subsurface of the Ordos Basin (North China). By integrating CT scanning, X-ray diffraction, nanoindentation, and compression tests, the authors systematically characterized the natural fracture development, mineral composition, and mechanical properties of the coal. Based on these data, a numerical model of coal formation was established using the discrete fracture network (DFN) method to evaluate the influence of natural fracture characteristics on hydraulic fracture propagation. Finally, based on a field fracturing case, recommendations for fracturing treatment design are proposed.

1. Introduction

Coalbed methane (CBM) development has undergone a notable evolution,, extending from shallow (<1000 m) to deep (>2000 m) formations, as exemplified by successful commercial production in regions such as Ordos Basin, China [1,2,3]. Current development strategies primarily rely on horizontal wells combined with large-scale hydraulic fracturing to enhance single-well productivity. Therefore, a thorough understanding of fracture propagation mechanisms in coal reservoirs is essential for improving efficiency and reducing development costs [4].

As an unconventional reservoir, coal is characterized by a complex natural fracture system, where the geometry and distribution of cleats significantly influence the initiation and propagation of hydraulic fractures [5,6,7,8]. Natural fractures in coal are typically distributed in a network consisting of high-angle face cleats and butt cleats, forming an orthogonal system [9,10]. These pre-existing fractures may either serve as pathways for hydraulic fracture extension or lead to substantial fluid loss, thereby affecting overall stimulation efficiency [11,12,13,14,15]. Consequently, investigating the multiscale fracture characteristics of coal and their influence on hydraulic fracture behavior is of both theoretical and practical significance for optimizing CBM development and maximizing stimulated reservoir volume (SRV).

Currently, scholars have conducted extensive research on the fracture propagation characteristics of coal. Tan, Huang, et al. investigated the effects of horizontal stress difference, elastic modulus, and fracturing fluid viscosity on hydraulic fracture propagation through physical simulation experiments [16,17,18,19]. Li, Teufel, et al. found that whether a fracture can cross coal bedding planes is influenced by the shear strength and frictional properties of these planes. Meanwhile, bedding planes may induce fracture branching or arrest propagation, thereby promoting the formation of complex fracture networks [20,21]. Wu et al. studied the impact of cluster spacing on fracture propagation and suggested that excessively close spacing leads to inter-cluster interference [22]. Liu et al. observed that by varying the injection rate during fracturing, multiple points can be simultaneously initiated, facilitating the formation of fracture networks [23]. Yang et al. conducted pulse fracturing tests on coal and found that cyclic loading, achieved through rapid changes in treatment pressure and frequency, resulted in more complex fractures [24]. CT scanning has become a widely used method for characterizing natural fractures in rock. Wang K and Xu, et al. [25,26] employed CT scanning to reconstruct digital cores of coal specimens and conducted numerical simulations of uniaxial compression tests to analyze the influence of cleats on coal strength. Similarly, Ju et al. [27] performed CT scanning on post-failure coal specimens to investigate fracture propagation during loading.

Numerical simulation serves as a key method for investigating hydraulic fracturing mechanisms. Tian, Xu, et al. performed numerical simulations of hydraulic fracturing in coal reservoirs using the finite element method, exploring the effects of injection rate, treatment scale, and fluid viscosity on fracture propagation in terms of height and length [25,26,27]. Wang, Xia, et al. discovered that the stress shadow generated during hydraulic fracturing significantly influences fracture propagation, potentially causing fracture deflection or growth arrest [28,29]. Conversely, excessive fracture spacing may lead to unstimulated zones, reducing stimulation efficiency [30,31]. Li et al. simulated fracture propagation during vertical well fracturing in coal reservoirs and concluded that increasing fluid viscosity and injection rate promotes uniform propagation of multi-layer fractures [20,32,33].

However, existing research on fracture propagation in coal lacks systematic investigation into the intrinsic properties of cleats, such as density, orientation, and size, and their influence on fracture propagation. There is a notable absence of multi-scale analyses [34,35,36,37,38] that integrate observations from the micro to macro scale and combine experimental with simulation approaches. Consequently, numerical models often fail to realistically represent the morphology of cleat systems. To address this, the present study employs FracMan 7.8 software to develop a reservoir model based on a 3D stochastic fracture modeling approach, which more accurately captures the characteristics of coal.

This study integrates microscale experiments (micro-CT imaging and nanoscale mechanical testing), core-scale mechanical experiments, and field-scale numerical simulations of hydraulic fracturing to systematically elucidate the multiscale characteristics of natural fractures in coal and their governing mechanisms on hydraulic fracture propagation. The nanoscale mechanical differences between organic matter and quartz in coal were quantified and linked to macroscopic mechanical behavior. The nonlinear effects of natural fracture density, orientation, and size on hydraulic fracture length and stimulated volume were quantitatively evaluated, and key controlling parameters were identified. Based on the results, it is concluded that fracture propagation in coal is jointly governed by the properties of its microscopic components and the multiscale natural fracture network.

2. Materials and Methods

Deep coalbed methane reservoirs are typically buried at depths exceeding 1500 m. In this study, experimental investigations were conducted on core specimens obtained from a depth of 1500 m in the eastern margin of the Ordos Basin.

2.1. CT Scanning Test

To investigate the spatial distribution characteristics of cleats in coal, CT scanning was performed on two coal specimens in this study. During CT scanning, coal specimens were wrapped in thin foam to prevent artifacts caused by shaking. Each scan lasted 45 min, achieving a spatial resolution of 5.0 μm. A high-sensitivity detector was employed during CT scanning to obtain clear CT images. To clarify the fracture characteristics of the rock mass, a spherical coordinate system was introduced for the evaluation of rock fractures (Figure 1).

2.2. XRD Test

Mineral composition is a key factor influencing the mechanical properties of rocks. Specimens measuring 5 × 5 × 5 mm were cut from the surface of two coal specimens for XRD analysis. Each group of rock specimens was ground into a powder with particle diameters less than 0.08 mm. The powder was then placed in an X-ray diffractometer to obtain diffraction patterns. Based on the diffraction signal characteristics of different minerals, the types and contents of minerals in the coal were identified [39].

2.3. Nanoindentation Test

In this study, nanoindentation tests were conducted on coal specimens retrieved from a depth of 1500 m using the TI Premier nanoindentation system (Bruker Corporation, Billerica, MA, USA). The small specimen size required for this technique helps mitigate the challenges associated with acquiring downhole coal specimens and the limited number of available specimens. Tests were performed on one coal specimen. Owing to the uneven surface, loose texture, and the presence of numerous micro-fractures in the coal, multiple preparation steps were necessary to achieve the surface smoothness required for testing [40]. The specimen preparation procedure is illustrated in Figure 2: (1) wire cutting was used to obtain flat rock specimens, which were then divided into small fragments; (2) the fragments were embedded in epoxy resin and allowed to solidify before demolding; (3) the solidified specimens were roughly polished using a polishing machine; (4) fine sandpaper was employed to polish the specimen surface until a mirror finish was achieved.

Given the inherently low hardness of coal, the maximum load during indentation tests was consequently reduced. A load of 8 mN was set as the termination signal. The test procedure for each indentation point included three stages: loading, holding, and unloading, with durations of 5 s, 2 s, and 5 s, respectively. Indentation points were arranged in a 4 × 5 array with a spacing of 50 μm (Figure 3).

After the test, the elastic modulus of the tested material was calculated using the classical method proposed by Oliver and Pharr [41] for determining micromechanical parameters. The calculation formula is as follows:

$$E_r={\displaystyle \frac{\sqrt{\pi }}{2\beta }}{\displaystyle \frac{S}{\sqrt{A}}}$$

(1)

where $E_r$ is the reduced elastic modulus; $A$ is the projected contact area between the indenter and the material (${\mathsf{\mu }\mathrm{m}}^2$); $\beta$ is the indenter correction factor, which is 1.034 for the indenter used in this study; and $S$ is the slope of the upper part of the unloading curve.

2.4. Uniaxial/Triaxial Compression Test

Compression tests in this study utilized coal specimens taken from a depth of 1500 m in the Ordos Basin. Wire cutting technology was used during specimen preparation to prevent water-induced damage to the coal. Due to the abundant organic matter from lignocellulosic plant tissues during coalification, coal exhibits low strength, well-developed cleats, and a complex natural fracture network. Conventional standard core specimens (diameter 25 mm, height 50 mm) cannot accurately reflect the influence of cleats on coal strength. Therefore, the coal specimens used in these tests had dimensions of 50 mm diameter and 100 mm height (Figure 4a), exhibiting greater randomness in cleat spatial distribution.

Uniaxial compression tests and triaxial compression tests under a confining pressure of 2 MPa were conducted on two sets of coal specimens (

Table 1. Compression test parameter.

No.	Specimen Size (mm)	Loading Method
Coal C-2	Φ50 × 100	2 MPa Confining Pressure
Coal C-4	Φ50 × 100	Uniaxial

). Although the in situ stress at the target reservoir depth is higher than 2 MPa, the purpose of this experiment is to compare the differences in the mechanical response of coal under conditions with and without confining pressure, and to obtain basic strength parameters (cohesion and internal friction angle) for numerical model calibration. The test results obtained under low confining pressure are sufficient to meet the requirements for establishing the Mohr–Coulomb strength envelope. Considering the well-developed fractures and loose structure of the specimens, a displacement-controlled loading mode was adopted on the MTS rock testing system (Figure 4b). The axial loading rates of two rock specimens were both 0.002 mm/s.

2.5. Hydraulic Fracturing Numerical Simulation

Small-scale fracture networks cannot fully replicate the actual conditions encountered during hydraulic fracturing. This study established a 3D numerical model of fractured coal reservoirs using FracMan software and DFN method. The model aims to investigate the influence of specific fracture attributes on propagation of hydraulic fracture.

The primary development target for deep coalbed methane in the current study area is at a depth of 2000 m, where the stress regime is characterized as normal faulting. The in situ stress values in the numerical model of this study are established based on this background. The maximum horizontal principal stress is oriented north–south with a magnitude of 42 MPa. The minimum horizontal principal stress is 36 MPa, and the vertical stress is 48 MPa.

In the hydraulic fracturing simulation, the cleat system in the coal formation was set as high-angle fractures, with butt cleats orthogonal to face cleats [1]. The elastic modulus, Poisson’s ratio, cohesion, and friction angle of the reservoir were obtained based on the experimental results of this study.

The simulation considered a horizontal well with three-stage fracturing, three clusters per stage, a stage length of 50 m, and a cluster spacing of 20 m. The treatment scale was set at 2500 m³ per stage with a pumping rate of 18 m³/min. These fracturing treatment parameters are representative of those currently used in deep coalbed methane development.

3. Results

3.1. Natural Fracture Development Characteristics

CT scanning revealed the natural fracture distribution in the rock specimens. Three-dimensional visualizations of the scanned coal specimens showed a distinct and complex cleat system (Figure 5). The CT scan results are shown in

Table 2. Cleat development characteristics in coal specimens.

Core No.	$\mathit{\phi }$			$\mathit{\theta }$
	Angle Range/°	Fracture Count/cm3	Avg Length/mm	Angle Range/°	Fracture Count/cm3
C-2	(0, 20)	1.67	1.29	(−180, 90)	3.34
	(20, 40)	3.23	1.18	(−90, 0)	4.41
	(40, 60)	3.74	1.21	(0, 90)	3.90
	(60, 80)	4.74	1.28	(90, 180)	5.05
	(80, 90)	3.31	1.37
C-4	(0, 20)	0.3	7.05	(−180, 90)	0.7
	(20, 40)	0.4	7.16	(−90, 0)	0.3
	(40, 60)	0.5	6.44	(0, 90)	1.3
	(60, 80)	0.9	6.63	(90, 180)	0.7
	(80, 90)	0.9	8.25

. Specimen C-1 has a fractal dimension of 2.14, a porosity of 2.96%, and an average length ranging from 1.18 to 1.37 mm. Specimen C-4 has a fractal dimension of 2.21, a porosity of 2.73%, and an average length ranging from 6.44 to 8.25 mm. As shown in Figure 5, two sets of cleats exist in the core, oriented parallel and perpendicular to the specimen axis.

3.2. Mineral Composition

X-ray diffraction analysis was used to determine the mineral composition and contents of the coal. According to the mineral content

Table 3. Mineral content percentage.

No.	Quartz	Plagioclase	Organic Matter	Dolomite	Calcite	Pyrite	Kaolinite	Illite
C-1	12.9%	0	87.1%	0	0	0	0	0
C-4	27.2%	0	72.8%	0	0	0	0	0

, the coal primarily consisted of quartz and organic matter, with organic matter constituting approximately 80% of its mass.

3.3. Micromechanical Properties

Nanoindentation tests were performed on smoothly polished coal specimens, involving loading, load holding, and unloading, yielding 20 sets of load-penetration curves (Figure 6). The curves indicated that the nanoindenter penetrated quartz twice and organic matter 18 times, suggesting a quartz content of approximately 10% in the tested area. Under the same load, the penetration depth in quartz was much smaller than in organic matter, and the indentation curve was steeper, reflecting the significantly higher hardness and strength of quartz compared to organic matter. The indentation curves for organic matter were more dispersed, with penetration depths ranging from 900 to 1600 nm, indicating the complex and highly discrete nature of the organic matter components in coal.

Based on the load-indentation depth curves, the elastic modulus of organic matter in coal was calculated to range from 2.5 GPa to 5.67 GPa, with an average of 4.15 GPa. In contrast, quartz exhibited a much higher elastic modulus, varying between 95.60 GPa and 100.20 GPa and averaging 97.90 GPa.

The relationship between mineral elastic modulus and penetration depth showed a strong linear correlation. A greater penetration depth corresponded to a smaller elastic modulus (Figure 7).

3.4. Rock Strength and Elastic Parameters

Uniaxial compression and triaxial compression tests (confining pressure 2 MPa) were conducted on the coal specimens. Under uniaxial loading, the failure load of the coal specimen was 15.20 MPa, with an elastic modulus of 3.99 GPa and Poisson’s ratio of 0.33. Under triaxial loading with 2 MPa confining pressure, the failure load was 37.71 MPa, with an elastic modulus of 4.85 GPa and Poisson’s ratio of 0.31. Based on the stress–strain curves (Figure 8), under uniaxial compression, the curve still exhibited fluctuations after the stress reached its peak; in contrast, under confining pressure, the stress–strain curve dropped rapidly after reaching its peak. The failure characteristics of the coal rock transitioned from plastic failure to brittle failure.

Fitting the strength versus confining pressure data from uniaxial/triaxial tests and plotting Mohr’s circles (Figure 9) based on the Mohr-Coulomb criterion yielded a cohesion of 2.27 MPa and an internal friction angle of 56.80° for the coal.

3.5. Numerical Simulation Results

Through numerical simulation, this section investigated the influence of natural fracture density, their orientation relative to the maximum horizontal principal stress, and their individual dimensions on hydraulic fracture propagation. Based on rock mechanics test results, the coal reservoir elastic modulus was set to 4 GPa, Poisson’s ratio was set to 0.3; reservoir matrix cohesion was set to 3 MPa, and internal friction angle was set to 60°.

Due to limitations in computational efficiency, the natural fracture density in the formation model was defined as the number of fractures per meter along the wellbore direction. The equivalent face cleats were modeled as rectangles with an equivalent radius of 10 m and an aspect ratio of 2.5, while the butt cleats were modeled as rectangles with an equivalent radius of 5 m and the same aspect ratio of 2.5. When investigating the influence of fracture dimension on hydraulic fractures, the equivalent radius of the face cleats was varied, while the dimensions of the butt cleats were kept constant.

Table 4. Numerical simulation geological and engineering parameters.

No.	Parameter	Baseline Value	Values Investigated
1	Natural fracture density (1/m)	0.3	0.1, 0.3, 0.5, 0.7
2	Angle between natural fracture and ${\sigma }_H$ (°)	30°	0, 30, 45, 60
3	Natural fracture size (m)	10	5, 10, 15, 20

presents both the baseline parameters for the numerical model and the values adopted in the subsequent sensitivity analysis. Figure 10 depicts the naturally fractured formation model constructed using these baseline parameters.

3.5.1. Influence of Natural Fracture Density

As shown in Figure 11, hydraulic fracture length decreases with increasing fracture density. The SRV showed a fluctuating but generally decreasing trend. When fracture density increased from 0.1 to 0.7, hydraulic fracture length decreased by approximately 25.6%, and SRV decreased by approximately 14.8% (Figure 12).

3.5.2. Influence of Natural Fracture Orientation

As shown in Figure 13, varying the angle between natural fractures and the maximum principal stress (0° to 60°) showed that the farthest distance of hydraulic fractures from the wellbore gradually decreased, and the SRV also progressively declined. As shown in Figure 14, when the angle increased from 0° to 60°, hydraulic fracture length decreased by approximately 44.7%, and SRV decreased by approximately 38.8%.

3.5.3. Influence of Natural Fracture Dimension

As shown in Figure 15, simulation results indicated a sharp drop in hydraulic fracture length when the natural fracture dimension reached 10 m, followed by a gradual decline. As shown in Figure 16, the SRV showed a similar sharp drop at the same size threshold, after which no significant further change was observed.

4. Discussion

4.1. Discussion of Experimental Results

CT scan results of coal revealed a complex distribution of cleats within the specimens, with the presence of approximately orthogonal cleat sets. This naturally occurring networked fracture system provides a fundamental condition for efficient CBM extraction. Under high pressure, the rock mass is prone to fragmentation along cleats, potentially leading to sand production.

The organic matter content in coal is high (80%), and nanoindentation tests demonstrated its low elastic modulus (average 4.15 GPa). Uniaxial and triaxial tests showed the elastic modulus of coal specimens to be around 4 GPa, indicating that the abundant organic matter dictates the low elastic modulus characteristic of coal. The stress–strain curves of coal exhibited fluctuations after reaching peak compressive strength, suggesting plastic behavior. It is inferred that post-failure shear sliding is inhibited by cleat surfaces, causing stress to rise again after initial failure.

The low elastic modulus of coal indicates that during hydraulic fracturing, even low net pressure can induce significant deformation, resulting in increased fracture width, which in turn leads to greater fracturing fluid consumption.

4.2. Numerical Simulation and Discussion

When conducting hydraulic fracturing in coal reservoirs with developed natural fractures, the influence of these fractures must be carefully considered. The complex cleat system exerts a complicated influence on hydraulic fracture propagation. Increased natural fracture density leads to significant fluid consumption near the wellbore, inhibiting the extension of hydraulic fractures to the far field. Unlike coal fracturing, in shale gas development, fracturing operations typically target areas with higher natural fracture density, or alternatively, aim to increase artificial fracture density to enhance the SRV [42].

As the angle between natural fractures and the minimum principal stress decreased, the farthest extent of the main hydraulic fractures from the wellbore shortened. However, fracture propagation length in the direction of the minimum principal stress gradually increased according to the simulation images, minimizing unstimulated areas between fractures and achieving more comprehensive stimulation. As noted by Zhao et al. [43], the cleat system in coal exerts a guiding effect on hydraulic fractures. During fracturing operations, the angle between hydraulic fractures and the predominant orientation of natural fractures can be adjusted to achieve optimal stimulation performance.

Smaller natural fractures are more easily activated and opened by fracturing fluid. As fracture size increased, the resulting fracture network became simpler. Smaller natural fractures (like butt cleats) were less likely to be opened, leaving larger unstimulated areas between primary fractures.

While this study establishes a link between the microscopic properties of coal and hydraulic fracturing, certain limitations remain. Although the natural fracture model captures the three-dimensional orthogonal cleat system, it still assumes planar fracture surfaces. Due to equipment limitations, the confining pressure applied in the experiments was lower than the in situ stress. Furthermore, the difficulty of acquiring in situ coal specimens restricted the specimen quantities, which precludes more systematic conclusions.

4.3. Case Study Discussion

Geological and engineering conditions in the field are often more complex than those in laboratory experiments and numerical simulations. This study selected a hydraulic fracturing case from a coal reservoir at approximately 2000 m depth on the eastern margin of the Ordos Basin. Reservoir pressure was 29 MPa, with a horizontal stress difference of 5 MPa. Fracturing fluid volume was 4092 m³ injected at 18 m³/min. The fracturing stage was 60 m long with two clusters. A pump-off/flowback diversion technique was employed during fracturing. As shown in Figure 17, the treatment curve showed a significant pressure increase after a 90 min shut-in compared to the previous stage, indicating the opening of new fractures. The micro-seismic monitoring results show that after the re-pumping, the number of distant fractures significantly increased. Furthermore, isolated micro-earthquake (Figure 18) events were observed, indicating that hydraulic pressure had altered the original stress field, and that natural fractures were activated without the direct effect of fluid pressure.

To address the challenges of limited far-field fracture propagation and small SRV during hydraulic fracturing in deep coalbed methane reservoirs, incorporating a pump shutdown during the treatment can promote fracture extension into unstimulated regions by leveraging changes in the stress field.

5. Conclusions

This study analyzed the characteristics of natural fractures in coal and their influence on hydraulic fracture propagation in coal formation through experimental testing and numerical simulation. It was found that dense natural fractures and low elastic modulus of coal reservoirs hinder the extension of hydraulic fractures to the far field in the near-wellbore region. In hydraulic fracturing design, cluster spacing can be optimized based on natural fracture density, and the angle between hydraulic fractures and the natural fracture system can be optimized by adjusting perforation angles, thereby enhancing the far-field propagation capability of fractures.

Due to the difficulty of acquiring in situ coal core specimens and the limitations of laboratory scale, the number of specimens in this study is relatively small, preventing a systematic understanding of the mechanical properties of coal. The numerical simulations still simplify fractures and do not consider the influence of fluid–solid interaction on fracture propagation, which can be possibly modeled by multiphysics numerical methods [44,45,46]. Future research should further investigate based on fluid–solid coupling to improve the realism of simulations.