4.1. DEM Simulation Model and Procedure
To comprehensively analyze the effects of damage evolution, failure processes, and unloading rates on the mechanical properties of shale specimens under confining pressure unloading conditions, the DEM numerical model [
43,
44] of the shale specimen was established by Particle Flow Code 2D (PFC
2D). The PFC model simulates the movement and interaction of many finite-sized particles. The particles are rigid bodies with finite mass that move independently of one another and can both translate and rotate. Particles interact at pair-wise contacts through an internal force and moment. Contact mechanics are embodied in particle-interaction laws that update the internal forces and moments. The time evolution of this system was computed via the DEM, which provides an explicit dynamic solution to Newton’s laws of motion [
45].
Figure 8a shows the shale specimen at a bedding inclination of 60°, and the bedding planes can be observed on the surface.
Figure 8b presents the established numerical model of a shale specimen in PFC
2D. In the model, the parallel bond (PB) model and smooth joint (SJ) model were employed to establish the particle contact in the shale matrix and bedding plane, respectively. In the PFC model shown in
Figure 8b, a parallel bond can be envisioned as a set of elastic springs with constant normal and shear stiffnesses, uniformly distributed over a cross-section lying on the contact plane and centered at the contact point, and the parallel bonds can transmit both force and moment between the pieces. As depicted in
Figure 8c, the smooth-joint model simulates the behavior of a planar interface with dilation regardless of the local particle contact orientations along the interface. The behavior of a frictional or bonded joint can be modeled by assigning smooth-joint models to all contacts between particles that lie on opposite sides of the joint. By this DEM model of a shale specimen in PFC
2D, the mesomechanical characteristics and damage evolution during the confining pressure unloading process can be analyzed and revealed in-depth.
The micro-parameter calibration of the shale model in PFC
2D was according to the results of the conventional triaxial compression test in Yang’s research [
8]. The “trial and error” method was applied to calibrate the micro-parameters. In the calibration process, the simulated results of every calibration test were compared with the experimental results and the ones that could mostly reflect the experimental results of the set of mesoscopic parameters, namely for the model specimen calibration parameters in the end. By using the “trial and error” method for micro parameters calibration, a group of micro parameters for the shale numerical model was determined, and the main micro parameters are listed in
Table 2. To illustrate the good agreement of the calibration results with the experimental results,
Figure 9 presents the calibrated failure mode of the shale model under conventional triaxial compression conditions by PFC
2D. In the simulation results of
Figure 9, the micro-crack band forms a macro-fracture. Compared to the experimental failure mode shown in
Figure 7a, it can be seen that the macro-fractures in the specimens of 0°, 60°, and 90° are similar to the tensile and shear fractures in
Figure 7a. The simulated failure modes of the shale specimens under different bedding inclinations and confining pressures show good agreement with the experimental results. The determined micro-parameters in
Table 3 reflect the mechanical properties of the shale specimens in laboratory test results under different conditions.
After the micro-parameter calibration of the shale model, the confining pressure unloading simulation can be conducted.
Table 4 lists the simulation scheme of the confining pressure unloading test. From
Table 4, it can be seen that the simulation designs test the conditions more than the laboratory test. Compared with the laboratory test, the simulation model under six bedding inclinations, four initial unloading stresses, and two unloading rates was added to the test. On the other hand, once the PFC
2D model had been established, the numerical specimen had almost no discreteness, and the computational efficiency was high compared to the experimental study; therefore, more variable factors of the mechanical properties under the confining pressure unloading condition can be studied.
In
Table 4, the stress loading mode was adopted for confining pressure unloading. The displacement loading mode was employed for axial stress loading, and the loading rate was 0.05 m/s. The loading rate in PFC was completely different from those in the actual physical world. In PFC modeling, since the calculation logic in PFC is fundamentally based on the dynamic mode governed by Newton’s second law, the time step (Δ
t) in each calculation cycle was chosen to be an infinitely small value, especially for static analysis. For instance, the loading rate of 0.05 m/s used for the confining pressure unloading simulation in this paper could be translated to 2.1 × 10
−6 mm/step, which implies it requires more than 200,000 steps to move a loading plate 1 mm. Hence, while physically, a 0.1 m/s loading rate is unreasonably high, this rate was slow enough in the PFC simulation [
46]. To compare the two loading modes, the axial stress loading rate value should be converted to the stress loading mode. In the PFC simulation, the loading rate of 0.05 m/s could be translated to 2.1 × 10
−6 mm/step. According to the result of the conventional triaxial compression test [
8], the average elasticity modulus value of the shale specimen under confining pressure 20 MPa was 31.5 GPa; thus, the axial loading rate of 2.1 × 10
−6 mm/step could be converted to 0.06615 MPa/100 steps (calculated by linear stress-strain relationship). Therefore, the ratios of the confining pressure unloading rate to the axial stress loading rate were 0.38, 0.76, 1.13, 1.51, 3.78, and 7.56, respectively, and it can be seen that the ratios of the confining pressure unloading rate to the axial stress loading rate were in the range of 0.3~8, covering a variety of test conditions from low to high. Hereby, this simulation scheme is reasonable.
4.2. Results Comparison of the DEM Simulation with Laboratory Tests
In laboratory tests, the strength and deformation parameters and failure modes of shale specimens at 0°, 60°, and 90° bedding inclinations under different confining pressure unloading rates were obtained. In this PFC2D simulation research, a variety of tests under different confining pressure unloading rates were conducted; however, the logic of the loading rate in the PFC procedure was different from that in the physical world, and the two could not be connected directly. Therefore, the reliability of simulation results cannot be verified by comparing the laboratory test results from the specific conditions of one-to-one correspondence analysis. Hereby, the reliability of the simulation test results was verified by comparing the strength envelope of the shale specimens under confining pressure unloading. However, the failure mode could not be compared with the laboratory test due to the differences in the unloading rate and stress state at the failure point.
Figure 10 presents the failure strength envelope comparison of shale specimens at inclination inclinations of 0°, 60°, and 90° under confining pressure unloading and conventional triaxial compression tests in a simulation test and a laboratory test. The simulation results and laboratory test results in
Figure 10 include conventional triaxial compression strength (TCS) and confining pressure unloading strength (CUS). It can be seen from
Figure 10a, b that the simulation results of the specimens at 0° and 60° bedding inclinations are in good agreement with the laboratory test results, and the confining pressure unloading strength (CUS) points obtained from the simulation test are scattered near the envelope of conventional triaxial compression strength (TCS) in the laboratory test. In addition, all the TCS points in the simulation test also fall within the dispersion range of the TCS points obtained in the experiment. It can be concluded that the simulation test of confining pressure unloading in this research reflects well the laboratory test results. At the same time, it also can be seen that the simulated CUS points are more centralized, have less difference from the simulated TCS envelope, and the TCS and CUS values in laboratory tests show large discreteness. Therefore, it can be seen that the numerical simulation results show less discreteness and can better reflect the effect of single variable factors on the test results. In addition, as can be seen from
Figure 10c, for the specimen at a bedding inclination of 90°, although there is a great difference between the simulated strength and the laboratory test strength, the linear fitting slope of the simulated strength is close to the laboratory test result, and the variation of the two is highly consistent.
On the other hand, by observing the simulated CUS values, it can be seen that the CUS values under different axial stress unloading points (
σU) have different distribution regions. When the
σU increased from 50%
σP to 95%
σP, the CUS value changed from a high confining pressure region to a low confining pressure region. As seen in
Figure 10, the failure confining pressure corresponding to the CUS under
σU distribution ranges of 50%
σP, 70%
σP, 85%
σP, and 95%
σP are about 20~15 MPa, 15~10 MPa, 10~5 MPa, and 5~0 MPa, respectively. This is because the greater
σU is, the greater the failure strength is, and the corresponding confining pressure on the strength envelope is larger. In addition, it also can be seen from
Figure 10 that the CUS value of the same
σU value decreases with an increasing confining pressure unloading rate,
VU, as shown by the arrows of different colors in
Figure 10. The arrows indicate the variation paths of CUS values with the
VU increase. This is because the greater the confining pressure unloading rate, the greater the reduction of confining pressure in the same period of time, while the increase of axial pressure in this period of time is relatively small; thus, the failure strength will be lower.
4.3. DEM Simulation Results of Unloading Confining Pressure Tests on Shale
Figure 11 shows the stress-strain curves under different unloading rates in the confining pressure unloading simulation tests. Due to a large number of simulation test conditions and limited space, only shale specimens at bedding inclinations of 0°, 60°, and 90° and under unloading stresses
σU of 50%
σp and 95%
σP are shown in
Figure 11. As can be seen from
Figure 11, the failure strength and failure confining pressure of the specimen at an unloading stress
σU of 50%
σp are lower than those at 95%
σP, which corresponds to the results shown in
Figure 10. On the other hand, it can be seen from
Figure 11 that the confining pressure unloading rate has a significant effect on the mechanical properties: the greater the confining pressure unloading rate of the shale specimen, the greater the loading rate of deviatoric stress. Specifically, it can be seen from the stress-strain curves that the slope of the deviatoric stress curve increases after unloading, and the higher the unloading rate, the greater the slope of the deviatoric stress curve. The increase in the confining pressure unloading rate accelerates the increasing rate of deviating stress and also accelerates the decreasing rate of confining pressure. Therefore, the confining pressure drops to a low level rapidly in a short time, resulting in a decrease in the failure strength of the shale specimen.
In addition, it can be concluded that the low unloading stress level leads to a large effect of the confining pressure unloading rate on the mechanical properties; when the unloading stress level is close to the strength under monotonic loading conditions, the effect of the confining pressure unloading rate on the mechanical properties become weak. This is because the high unloading stress level decreases the time that the specimen takes to reach the failure strength. Meanwhile, the confining pressure unloading makes the specimen fail more easily; however, the effect of the unloading rate cannot be shown in such a short time.
In
Figure 10, the comparison of the confining pressure unloading strength envelope between the simulation test and the laboratory test of specimens at bedding inclinations of 0°, 60°, and 90° is presented. To comprehensively analyze the confining pressure unloading strength characteristics of specimens at different bedding inclinations in the simulated test,
Figure 12 shows the confining pressure unloading strength envelope of the other four groups of shale specimens at bedding inclinations of 15°, 30°, 45°, and 75° obtained by simulation. From
Figure 12, as a whole, the confining pressure unloading strength (CUS) points under different unloading stress levels and unloading rates are scattered on both sides of the envelope of conventional triaxial compression strength (TCS). However, it can be found from the observation of the CUS distribution at different loading rates that with the confining pressure unloading rate increases, CUS points will be distributed more above the TCS envelope (as shown in the blue oval dotted box in
Figure 12), which means the strength of the shale specimen increases as the confining pressure unloading rate increases. By observing the distribution of all the CUS points, it can be found that the CUS values are higher than the TCS values on the whole, which indicates that the loading path of the unloading confining pressure and increasing axial pressure enhances the failure strength of shale specimens to a certain extent. This phenomenon can be analyzed from another point of view. The loading path of confining pressure unloading and increased axial pressure substantially increase the loading rate of the shale specimens. Numerous current studies have confirmed that the higher the loading rate is, the stronger the specimen is.
Figure 13 shows the failure modes of shale specimens at bedding inclinations of 0°, 60°, and 90° under different confining pressure unloading rates when the unloading axial stress
σU is 50%
σP. It can be seen from
Figure 11 that the overall failure modes of shale specimens at the same bedding inclination are similar under different confining pressure unloading rates. The specimens at a bedding inclination of 0° in
Figure 13a mainly failed by V-shaped shear fractures, while the specimens at a bedding inclination of 60° in
Figure 13b mainly failed by shear fractures along the bedding plane, and the specimens at a bedding inclination of 90° in
Figure 13c are dominated by an oblique single shear fracture plane. In particular, from
Figure 13, it can be further observed that the failure mode of simulated specimens under the lowest confining pressure unloading rate (
VU = 0.025 MPa/100 steps) is different from those at the high confining pressure unloading rate (
VU = 0.500 MPa/100 steps). In
Figure 13a, the shale specimen at a bedding inclination of 0° has two narrow “V-shaped” crack bands, and the distribution of microcracks is concentrated when
VU = 0.025 MPa/100 steps, however, when
VU increases to 0.500 MPa/100 steps the two “V-shaped” crack bands become wide. Moreover, the distribution of microcracks is scattered, and a small number of fine axial crack bands are also distributed. In
Figure 13b, the shale specimen at a bedding inclination of 60° has two main crack bands along the bedding plane (as the yellow line indicates in the first image of
Figure 13b) at
VU = 0.025 MPa/100 steps and other small local crack clusters are also dispersed. However, when the
VU increases to 0.500 MPa/100 steps, the micro-cracks are mainly distributed beside the two bedding planes and accompanied by several fine crack zones propagated along the axial stress. In addition, as shown in
Figure 13c, the shale specimen at a bedding inclination of 90° has a wide oblique crack zone and relatively dispersed crack distribution at
VU = 0.025 MPa/100 steps, while the shale specimen has a narrow oblique crack zone and concentrated crack distribution at
VU = 0.500 MPa/100 steps.
Figure 14 shows the failure modes of shale specimens at bedding inclinations of 0°, 60°, and 90°under different confining pressure unloading rates when the unloading axial stress
σU is 95%
σP. It can also be seen from
Figure 14 that the overall failure modes of shale specimens with the same bedding inclination are similar under different confining pressure unloading rates. The shale specimens at a 0° bedding inclination in
Figure 14a mainly present V-shaped shear fracture planes, while the specimens at a 60° bedding inclination in
Figure 14b are mainly shear fracture planes along the bedding plane, and the specimens at a 90° bedding inclination in
Figure 14c are dominated by an oblique single shear fracture plane. From
Figure 14, it can be further observed that the failure modes of the numerical specimens at the lowest unloading rate (
VU = 0.025 MPa/100 steps) are different from those at the highest unloading rate (
VU = 0.500 MPa/100 steps). In
Figure 14a, the shale specimen at a 0° bedding inclination has only one wide oblique crack zone (indicated by the yellow wireframe in the first image) when
VU = 0.025 MPa/100 steps, while there are two fine “V” shaped crack zones (indicated by the yellow lines in the last image) when
VU = 0.500 MPa/100 steps. The shale specimen at the 60° bedding inclination in
Figure 14b has two fine crack zones at
VU = 0.025 MPa/100 steps, one along the bedding plane and the other through the bedding plane. However, when at
VU = 0.500 MPa/100 steps, the cracks are distributed near multiple bedding planes, and no single crack zone is concentrated or significant. As shown in
Figure 14c, the crack distribution of the shale specimen at a 90° bedding inclination is concentrated, and the oblique crack zone is significant at
VU = 0.025 MPa/100 steps, while the crack distribution is dispersed at
VU = 0.500 MPa/100 steps.
By comparing the failure modes at the low unloading stress level shown in
Figure 13 and the high unloading stress level shown in
Figure 14, it can be found that when compared with the low unloading stress level (
Figure 13,
σU = 50%
σP), the failure shale specimens under the high unloading stress level (
Figure 14,
σU = 95%
σP) have wider crack bands and more microcracks, which are more widely distributed. This is because the shale specimen under a high unloading stress level has a higher stress level when it fails, and the specimen has more damage accumulation and more fully developed cracks, resulting in its crack zone and the crack distribution range being wider. On the other hand, by comparing the unloading confining pressure failure mode shown in
Figure 13 and
Figure 14 to the conventional triaxial compression failure mode shown in
Figure 9, it can be found that the micro-cracks in a failure specimen under conventional triaxial compression are more concentrated and more distinct, and the shale specimen fails due to oblique shear fracture under high confining pressure. When under confining pressure unloading test, the micro-crack in the failure specimen is dispersive, and the crack band is not distinct; in particular, there are several thin and narrow tensile crack bands extending along the axial stress direction, which indicates that the specimen will be a tensile failure under the confining pressure unloading.
4.4. Discussion on the Failure Mechanism of Unloading Confining Pressure Tests on Shale
(1) Analysis of micro-crack evolution during the confining pressure unloading process.
Figure 13 and
Figure 14 are unlike conventional triaxial compression failure, which only produces oblique shear crack bands; the confining pressure unloading leads to many fine axial tensile crack bands in the specimen, which indicates that the failure mechanism of shale specimens under confining pressure unloading conditions is different from that under conventional triaxial compression. Therefore,
Figure 15 shows the micro-crack evolution of the shale specimen during the confining pressure unloading simulation test. Similarly, due to space limitations, only shale specimens at bedding inclinations of 0°, 60°, and 90°, under unloading stresses (
σU) of 50%
σp and 95%
σp, and unloading rates (
VU) of 0.025, 0.100, and 0.500 MPa/100 steps are presented in
Figure 15. In addition, it should be noted that for the shale specimens under the same bedding inclination and unloading stress, the stress-strain curves before the start of confining pressure unloading are consistent; therefore, to clearly show the details of stress and micro-crack variation after confining pressure unloading start, the curves before unloading starts are deleted, and only the curves after the confining pressure unloading are depicted in
Figure 15.
As shown in
Figure 15, there is a difference in the number of microcracks between the specimens under the two unloading stress levels because the higher the unloading stress level is, the more damage accumulation and more microcracks are generated. The number of microcracks begins to increase rapidly after confining pressure unloading starts, which indicates that confining pressure unloading will strengthen the damage accumulation of shale specimens. For the shale specimens under different bedding inclinations, PB tensile cracks are mainly generated in specimens under a 0° bedding inclination, SJ shear micro-cracks are mainly generated in specimens under a 60° bedding inclination before failure, and PB tensile cracks and SJ tensile cracks increase rapidly after failure. The micro-crack evolution of specimens under a 90° bedding inclination is similar to those under a 60° bedding inclination.
To further observe the micro-crack evolution curves before and after confining pressure unloading in
Figure 15, it can be found that after the start of confining pressure unloading, the number of PB tensile micro-crack and SJ tensile micro-crack increases rapidly, and the increasing rate of the SJ shear micro-crack is stable, which indicates that the confining pressure unloading mainly produces tensile damage. At the same time, seen from the micro-crack evolution curves under different confining pressure unloading rates, it can be further concluded that the confining pressure unloading leads to more tensile damage, as the red and blue curves depict in
Figure 15, with the increase in the confining pressure unloading rate, the rate of PB tensile cracks and SJ tensile crack increases, which reveals that the greater the confining pressure unloading rate, the more serious the tensile failure is. This is in line with the micro-crack destruction patterns shown in
Figure 13 and
Figure 14.
(2) Analysis of strain energy evolution during the confining pressure unloading process.
The failure process of the rock material is essentially a process of transformation and release of various energies; in particular, the evolution of strain energy can reflect the internal mechanisms of rock failure to a certain extent. When the rock specimen is under conventional triaxial compression (
σ1 >
σ2 =
σ3), the total strain energy,
U, of the rock element is contributed to via the axial stress, which does positive work,
U1, and the confining pressure, which does negative work,
U3; thus, the total strain energy
U of the rock element in the whole process of triaxial compression can be expressed as:
where
U1 and
U3 are the axial strain energy and radial strain energy, respectively.
U1 and
U3 can be obtained by integration of the stress-strain curve and are normally calculated by the area summation of tiny trapezoids, according to the definition of integral calculation. The calculation equations are as follows:
where
ε1 and
ε3 are the axial and radial strain, respectively,
n is the total number of trapezoids of the stress-strain curve, and
i is the segmentation points.
In addition, the total strain energy
U of the rock element can be divided into two parts, the elastic strain energy
Ue, which is stored in the specimen and can be released by unloading, and the dissipated energy
Ud, which leads to plastic deformation and crack propagation in the rock specimen [
47].
The releasable strain energy stored in the element
Ue is related to the unloading elastic modulus,
E, and Poisson’s ratio,
μ, after the rock element damage, and the elastic strain energy,
Ue, can be calculated by using the following equation [
47].
It should be noted that although the form of Equation (5) is the same as the strain energy calculation formula in linear elasticity mechanics, it is aimed at the linear unloading process in the nonlinear process of a rock element. With the damage aggravation of the rock element under external action, the strength gradually decays. When the releasable elastic strain energy of a rock element reaches the surface energy required for failure, the element will fail and be released in the form of elastic surface energy.
To further study the failure mechanism of the shale specimen in the process of confining pressure unloading, the strain energy evolution curves of the shale specimen under the confining pressure unloading simulation test are depicted in
Figure 16; in particular, the evolution curves of lateral strain energy and dissipation energy are shown in
Figure 17.
Due to the space limitation,
Figure 16 only presents the results of specimens at bedding inclinations of 0°, 60°, and 90°, under unloading stresses
σU of 50%
σp and 95%
σp, and a confining pressure unloading rate of 0.100 MPa/100 steps. As can be seen from
Figure 16, before the confining pressure unloading starts, the lateral strain energy,
U3, increases very slowly and remains at a very low level. In this period, the confining pressure is constant, and the low-level energy dissipation is mainly caused by the small lateral deformation of the specimen and almost all the work performed by the axial stress is converted into strain energy inside the specimen. Therefore, the
U1 and
U curves in
Figure 16 almost coincide before the unloading point. However, after the start of confining pressure unloading, the
U3 curve changes significantly and increases rapidly, indicating that the released lateral energy of the specimen begins to increase at this time. This is due to a decrease in the confining pressure, which leads to the increase of lateral deformation and results in the increase of lateral energy release. At the same time, the growth rate of the total dissipation energy,
Ud, also increases, indicating that the strain energy stored inside the specimen begins to accelerate its release. Therefore, the
U1 and
U curves in
Figure 16 begin to separate significantly after the unloading point, and the gap between them becomes larger and larger. By observing the strain energy evolution curves under the different conditions in
Figure 16, it can be concluded that although the specimens are under different bedding inclinations and unloading stress levels, the strain energy evolution law after the confining pressure unloading is similar, and the increase of dissipation energy is caused by the increase of lateral strain energy release, corresponding to the intensification of the internal crack damage in
Figure 15.
To further study the evolution of lateral strain energy
U3 and the total dissipated energy
Ud of the shale specimen during the confining pressure unloading process, the curves of
U3 and
Ud under different confining pressure unloading rates are depicted in
Figure 17. Due to the space limitation, only the results of specimens at bedding inclinations of 0°, 60°, and 90° and under unloading stresses (
σU) of 50%
σp and 95%
σp are presented in
Figure 17.
As can be seen from the overall variation of
U3 and
Ud in
Figure 17, the energy release curves under different unloading rates are concentrated under the high unloading stress point (95%
σp) and dispersed under the low unloading stress point (50%
σp). This is because the accumulated strain energy in the specimen is very large under a high unloading stress level. At this time, the specimen approaches the critical point of energy release and failure, leading to the energy dissipation rate being similar under different confining pressure unloading rates. By comparing the specimens under different bedding inclinations, it can be seen that the energy dissipation of specimens at a 0° bedding inclination has a slow rate and low magnitude before the confining pressure unloading starts, while the energy dissipation of specimens at a 90° bedding inclination is distinct before the confining pressure unloading starts. In addition, by comparing the evolution curves of
U3 and
Ud under different confining pressure unloading rates, it can be found that the higher the confining pressure unloading rate is, the higher the increase rate of
U3 and
Ud, and the faster the energy dissipation. The results further reveal that the confining pressure unloading results in the increase of lateral strain energy release and the acceleration of total dissipated energy release, which is the internal mechanism driving the specimen to failure during the confining pressure unloading process.
Based on the above analysis of micro-cracks and strain energy evolution of the shale specimens during confining pressure unloading, it can be concluded that during the process of confining pressure unloading, the continuous decrease of confining pressure weakens the lateral constraint of the specimen and changes in the internal stress state, leading to the increase of tensile micro-cracks and damage, and resulting in a large number of thin tensile crack bands extend along the axial direction in the specimen. Moreover, the increase in tensile micro-crack damage causes the energy dissipation to increase rapidly, which leads the specimen to the final failure.