574 Effects of fractures on reservoir deformation and flow modeling Mohammad Ali Bagheri and Antonin Settari Abstract: Fracture opening and closure caused by changes in the effective stress on fracture planes is well known. This phenomenon happens in aquifers and hydrocarbon reservoirs where production or injection events change the applied effective stress. Modeling this phenomenon requires a rigorous coupled geomechanics and flow simulator as well as detailed information regarding fracture properties, spacing, and orientation. A method has been developed to account for the effects of fractures on deformation in coupled reservoir and geomechanics simulation. The equivalent continuum approach was used where the constitutive matrix of individual blocks was calculated by means of an analytical formula that considers both rock and fracture deformation properties. The rock material is assumed to be isotropic and the normal stiffness of fractures varies with the applied effective stress according to a law that is typical for joints. The shear deformation of the fractures can be incorporated easily in the future. The general pseudo-continuum model was verified by comparing it with an approach using explicit modeling of the fractures. The fractured rock geomechanics model is then coupled to a single porosity, multiphase flow model. The deformations produce changes in the permeability tensor in both magnitude and orientation, which in turn influences compaction behavior. Key words: fracture deformation, equivalent moduli, coupling, reservoir compaction, reservoir simulation. Résumé : L’ouverture et la fermeture des fractures dues au changement de la contrainte effective sur le plan de fracture sont bien connues. Ce phénomène se produit spécialement dans les aquifères et les réservoirs d’hydrocarbure où la production ou l’injection change les contraintes effectives appliquées. La modélisation de ce phénomène requiert un simulateur couplé de géomécanique et d’écoulement, de même qu’une information détaillée sur les propriétés des fractures, leur espacement et leur orientation. On a développé une méthode pour prendre en compte les effets des fractures sur la déformation dans la simulation couplée de géomécanique et du réservoir. L’approche du continuum équivalent a été utilisée là où la matrice constitutive des blocs individuels a été calculée au moyen d’une formule analytique qui considère les propriétés de déformation de la roche de même que des fractures. On suppose que le matériau rocheux est isotrope et que la rigidité normale des fractures varient avec la contrainte effective appliquée selon un loi typique pour les joints. La déformation en cisaillement des fractures pourra être incorporée facilement dans le futur. Le modèle général du pseudo-continuum a été vérifié en le comparant avec une approche qui utilise la modélisation explicite des fractures. Le modèle géomécanique de la roche fracturée est ensuite couplé à un modèle d’écoulement multiphase à porosité simple. La déformation produit des changements dans le tenseur de perméabilité tant en amplitude qu’en orientation, qui à leur tour influencent le comportement du compactage. Mots clés : déformation par fractures, modules équivalents, couplage, compactage de réservoir, simulation de réservoir. [Traduit par la Rédaction] Bagheri and Settari Introduction As fluids are being produced reservoirs experience a pressure drop, which causes fractures to close. The reverse phenomenon is observed in injection wells used to pressurize the reservoirs. This closure and opening can be called “fracture breathing” and has been demonstrated by field measureReceived 18 February 2005. Accepted 15 December 2005. Published on the NRC Research Press Web site at http://cgj.nrc.ca on 12 May 2006. M. Bagheri1 and A. Settari.2 Department of Chemical and Petroleum Engineering, University of Calgary, 2500 University Dr., Calgary, AB T2N 1N4, Canada. 1 Present address: Sproule International Ltd., North Tower, Suite 900, 140 4 Ave. SW, Calgary, AB T2P 3N3, Canada. 2 Corresponding author (e-mail: asettari@ucalgary.ca). Can. Geotech. J. 43: 574–586 (2006) 586 ments. For instance, large variations in resistivity logs after changes in drilling mud density seen in a Gulf Coast well can be attributed to this phenomenon (Timko 1966). Fracture breathing can contribute to compaction of the reservoir and in turn to subsidence problems. From the point of view of reservoir engineering (fluid flow in porous media), this phenomenon is also of importance because of its effects on fracture porosity (fracture volume occupied by fluid) and fracture network permeability. Usually most of the reservoir fluid is stored in the rock matrix (as opposed to the fractures) so variation in fracture porosity is less important for pressure changes in the reservoir. In contrast, the changes in fracture permeability can have major influences on reservoir performance. Over the past decades, experimental, theoretical, and numerical studies have been carried out to account for the fracture deformation caused by change in applied stresses. These doi:10.1139/T06-024 © 2006 NRC Canada Bagheri and Settari 575 Fig. 1. Effect of fracture on normal deformation (modified from Bandis et al. 1983). studies cover different aspects of fracture deformation and provide some insight into its significance for fluid flow and overall deformation of fractured rocks. The stress–strain behavior of fractures is a nonlinear relationship depending crucially on the mechanical properties of the fractures. This type of nonlinear behavior has been shown in the experimental work of several investigators and has also been modeled mathematically. Goodman (1976) proposed a hyperbolic function relating fracture closure to the applied normal stress. Youshinaka and Yamabe (1986) modeled the normal closure of fracture by an exponential function, and Barton and Choubey (1977) modeled the shear strength of fractures under shear stress. Bandis et al. (1983) have experimentally studied and mathematically formulated fracture deformation in detail. Their results have been utilized in this study. The concept of equivalent continuum is invoked in this work to treat fractured rock as a single media having the deformation properties of both intact rock and fractures. The resulting elastic equation is no longer linear because of the effect of nonlinear fracture deformation. Different approaches have been reported in the literature to define the equivalent properties of the continuum. From these, the method of Huang et al. (1995), which appears to be the most general and applicable, has been used to construct the elastic moduli of the fractured rock in our study. The main attraction of this method is that there are no constraints regarding the orientations of the fracture sets, which is important considering the general diversity of fractures in reservoirs. Knowing the deformation of the equivalent continuum allows precise calculation of the stress field, which in turn results in the calculation of deformation of fractures individually in space and time. This paper focuses on the rock deformation and its effect on the volumetric (porosity) behavior of reservoirs. Fracture deformations also cause variation of fracture tensor permeability. The permeability aspect of the coupling has been studied with results published elsewhere (Bagheri and Settari 2005). where σn is the normal stress, ∆ν is the change in fracture aperture, and “a” and “b” are constants that have physical meaning. For a value of σn → ∞, the closure reaches its maximum Single joint deformation [2] Here we follow the joint mechanics models developed by Bandis and co-workers (Bandis et al. 1983). In their laboratory investigations, interlocked block samples with natural unfilled joints were subjected to a series of loading–unloading cycles. During each test, the total deformation was recorded as a function of normal stress. In all cases, a nonlinear behavior was observed. Figure 1 shows a typical behavior of a single joint experiencing two cycles of a series of loading– unloading cycles in comparison with the behavior of an intact rock. The difference in total deformation of these two samples demonstrates the large influence of joints on normal deformation of rocks (the ratio of total deformation to the rock compression is around 3 at the end of compression). The stiffness of the jointed rock and that of the intact rock at high normal stresses becomes the same (i.e., the two curves are parallel), which indicates a complete closure of the joint. On decompression, a hysteretic behavior accompanied by a permanent set is observed. Large hysteresis and large permanent set are always seen in the first load cycle, while diminishing on subsequent cycles. In situ fractures likely react to Joint normal stiffness, which is the ratio of change of normal stress to normal closure, is defined as stress changes in a fashion similar to the third or fourth cycles of lab experiments. In other words, the hysteresis effect is not dominant in highly stressed in situ rocks. For our purposes, we neglected the hysteresis effect and used a single formula for the deformation of a joint in loading and unloading, which is assumed to best represent the deformation path of the joints. Normal closure – normal stress curve for both artificial and natural joints has a hyperbolic shape, given by the following equation: [1] [3] σn = ∆ν a − b∆ν ∆ν max ≅ kn = a b dσn d ∆ν The physical meaning of “a” is found from the derivative of eq. [1]. [4] kn = ∂σ n = ∂ ∆ν 1 b ⎞ ⎛ a ⎜1 − ∆ν⎟ ⎝ a ⎠ 2 At zero normal stress, normal stiffness is called initial normal stiffness and is shown to be [5] kni = ∂σ n 1 = ∂∆ν σ n = 0 a Substituting eqs. [2] and [5] into eq. [4] will result in © 2006 NRC Canada 576 [6] Can. Geotech. J. Vol. 43, 2006 kn = kni ⎛ ∆ν ⎞ ⎜1 − ⎟ ∆ ν ⎝ max ⎠ 2 Shear deformation of a rock sample containing a single fracture is also well described in the literature. In this case, the shear stiffness (which has a similar definition to the normal stiffness) is involved. However, including shear deformation of fractures in a full field study is somewhat difficult. First of all, many joint model parameters are involved, which requires extensive lab work. Secondly, in laboratory shear tests, the walls of the joints can move freely relative to each other and large shear deformations can be reached, which is not the case in hydrocarbon reservoirs. In this study, to verify the coded equivalent moduli of the jointed rock mass against analytical solutions, finite shear stiffness can be assigned to the joints. However, for reservoir application of this method, a very high stiffness has been chosen to minimize the effect of shear deformation. Equivalent moduli of a jointed rock mass The concept of an equivalent medium of a jointed rock mass (composed of two materials) is used to introduce a third material that behaves the same way as the jointed rock mass, under normal and shear stresses. The normal and shear moduli of this third material are functions of the joints and intact rock moduli. Normal and shear stiffness of joints along with their orientations and spacings are key parameters for constructing the equivalent material’s constitutive equation. In this section the constitutive relation of single fracture will be presented, and a general method for calculating the equivalent properties of the jointed rock mass, developed by Huang et al. (1995), is critically reviewed. Constitutive equation of a single joint The general constitutive equation for a joint has the following form: [7] ∆␦ I = DIJ ∆ J where ␦ I is the displacement vector (1 by 3), J is the traction vector (1 by 3), and DIJ is the constitutive matrix (3 by 3). The traction vector bears the direction and magnitude of the stress acting on the joint surface, and can be expressed by Cauchy’s formula [8] ∆ j = ∆ijn i where “n” is the vector normal to the joint surface. The local coordinate system of every individual joint is composed of three vectors in the direction of the normal, strike, and dip angles. The components of the displacement and traction vectors in joint local coordinate systems are in these directions. Joint structure restricts having more than one normal deformation and two shear deformations (dimensions of the displacement vector). Generally, each of these deformations can be produced by all three components of traction. For instance, a normal opening or closure can be caused by shear stresses as well as normal stresses. Such normal deformation of the fracture (increase of aperture) caused by shear stress is known as “dilation” and results in an increase in fracture volume. It is also possible that a normal stress acts like a shearing factor on the joint surface and creates shear deformation. The rate of this deformation, as well as the rate of dilatancy, is connected to the joint surface properties. Often it is assumed that the deformation characteristic of the joint in the dip direction is the same as in the strike direction. Using this assumption, and neglecting the effect of dilatancy behavior and shear strain due to normal stress, eq. [7] can be expanded in extremely simple form as [9] ⎧∆δ N ⎫ ⎡ k1n ⎪ ⎪ ⎢ ⎨ ∆δ S ⎬ = ⎢ 0 ⎪ ∆δ T ⎪ ⎢ 0 ⎩ ⎭ ⎢⎣ 0 1 ks 0 0 ⎤ ⎧∆τ N ⎫ ⎥⎪ ⎪ 0 ⎥ ⎨ ∆τ S ⎬ 1 ⎥ ⎪ ∆τ ⎪ T⎭ ks ⎥ ⎦⎩ This equation will be used to find the constitutive relation of jointed rocks next. Constitutive equation of jointed rock masses A constitutive equation to calculate the overall deformation of rocks containing joints is desired because it helps to simplify the overall expanded form of the equation and also reduces the number of equations without neglecting the effect of joints. Huang et al. (1995) proposed a stress–strain model for a fractured rock mass containing several fracture sets. In their method, the strains and stresses have been considered as tensors and so the final equation is more general compared to some other methods. The following paragraphs briefly describe this method. Changes in total strain due to stress changes consist of two components: one from the intact rock and the second from the fractures [10] ∆ij = ∆ Iij + ∆ Jij Equation [10] can be rewritten for both media to show the relation of each component in response to the total stress change. [11] ∆ Iij = C Iijkl ∆kl [12] ∆ Jij = C Jijkl ∆kl Note that lower case subscripts indicate the global coordinate system. Each joint set has its own local coordinate system, which is indicated by upper case subscripts. Based on the principle of energy conservation, strain energy stored in fractures due to stress is equal to the total work done by the traction at each joint that caused joint movement. Thus M [13] ij∆ Jij = 1 jf ∆␦ jf A f ∑ V f =1 where “M” is the total number of fracture sets, “ f ” indicates individual fracture sets and “Af ” is the total area of the f th fracture set, which is equal to the volume of the representative block divided by the average spacing between the frac© 2006 NRC Canada Bagheri and Settari 577 tures in the set. Substituting Cauchy’s formula into eq. [13], the strain of fractured rock due to movement of the fractures is given by M [14] ∆Jij = ∑ n if ∆␦ jf f =1 Fig. 2. Dimensions of the block used for linear verification of the model: (a) intact rock, (b) fractured rock. 1 Sf where “S” is the spacing between the fractures. The traction and displacement vectors can be transformed between the global coordinate system and the fracture set local coordinate system. For instance [15] ∆␦ jf = L jJf ∆␦ Jf [16] ∆ If = L Iif ∆ if where “L” contains the direction cosines between the local and global coordinate systems. Substituting eqs. [7], [9], [15], and [16] into eq. [14], the general constitutive equation for joints is obtained M [17] f f C Jijkl = ∑ n if L jJf D JL L Ll n kf f =1 1 Sf The expanded form of this equation for a rock containing three sets of fractures is presented in Appendix A, and differences found between the present work and that of Huang et al. (1995) are discussed. Model verification In this section, the aim is to verify the 3D finite element model developed using this approach in linear and nonlinear conditions. Here, the term “linear” means that the normal and shear stiffness values of fractures are assumed constant, in contrast to the “nonlinear” condition where the normal stiffness of fractures changes according to eq. [4]. Linear verification Elastic deformation Consider a quarter of symmetry of a cubic block of intact rock as shown in Fig. 2a. The dimensions of the simulated element are 0.5 m in the “x” and “y” directions and 1 m in the “z” direction. Boundary condition of fixed surfaces is applied on three faces of the cube, which have a point in common including the bottom of the block. A typical Young’s modulus of 100 GPa is assigned to the intact rock material. A σn of 1 MPa creates a deformation along the z direction equal to 0.01 mm. Now consider a horizontal fracture with a constant (corresponding to the given stress level) normal stiffness of kn intersecting the intact rock according to Fig. 2b. Such a fracture increases rock mass deformation in the z direction according to [18] ⎛ 1 1⎞ δz = σn⎜ + ⎟H ⎝ Skn E ⎠ where “S” is the spacing between the identical fractures in the sample and “H” is the height of the sample. For instance, if normal stiffness of the fracture is 20 GPa/m, the total computed deformation with one fracture will be 0.06 mm. It is apparent from eq. [18] that the higher the number of fractures perpendicular to the z direction, the higher will be the deformation. However, for this special case where only one normal stress is applied, any configuration of vertical fractures has no effect on the deformation in the vertical direction. This was verified for different types of vertical fractures. Further modeling verifications were also done by adding 180° to the strike of the fractures and by reproducing the results of the test problems with a multiblock numerical model. Now consider a block intersected by three orthogonal fracture sets. These fracture sets (presented as single lines) are arranged according to Fig. 3. The objective is to find the overall strain components of the rock mass under a uniaxial normal stress of 1 MPa on the top of the specimen. Amadei and Goodman (1981) proposed an analytical solution to this problem using transformation of the stress–strain relationship. The strain components of the resulting deformation are ⎡ sin 2 2α ⎛ 1 1 1 1 ⎞ [19a] ε x = ⎢ + − − ⎜ ⎟ ⎢⎣ 4 ⎝ kn1S1 kn2 S2 ks1S1 ks2 S2 ⎠ − v⎤ σn E ⎥⎦ ⎛ v⎞ [19b] ε y = ⎜ − ⎟ σ n ⎝ E⎠ ⎡ 1 cos 4 α sin 4 α [19c] ε z = ⎢ + + kn1S1 kn 2 S2 ⎣E + 1 ⎞⎤ sin 2 2α ⎛ 1 + ⎜ ⎟ ⎥ σn 4 ⎝ ks1S1 ks2 S2 ⎠ ⎥⎦ ⎡ ⎛ cos 2 α sin 2 α ⎞ [19d] γ xz = ⎢sin 2α ⎜ − ⎟ kn 2 S2 ⎠ ⎝ kn1S1 ⎢⎣ − sin 4α ⎛ 1 1 ⎞⎤ + ⎜ ⎟ ⎥ σn 4 ⎝ ks1S1 ks2 S2 ⎠ ⎥⎦ To calculate the strain components with the numerical stress model, different values of joint properties were assigned to the fracture sets, which are shown in Table 1, along with the other relevant data. Boundary conditions of fixed surfaces were applied on the right (y = Ly) and bottom (z = 0) faces to restrict them from moving in the “y” and “z” directions, respectively. In addition, the front edge (x = Lx © 2006 NRC Canada 578 Can. Geotech. J. Vol. 43, 2006 Fig. 3. Three orthogonal joint sets intersecting an intact rock (test model of Amadei and Goodman 1981). Table 2. Comparison of numerical and analytical results in terms of strain components. Uniaxial test εx εy εz γ xz Triaxial test Numerical 0.006 859 Analytical 0.006 859 1 –0.000 490 0.059 965 0.051 140 –0.000 490 0 0.059 965 4 0.051 139 8 Numerical Analytical 0.036 205 0.000 131 0.066 335 0.092 728 0.036 205 4 0.000 131 1 0.066 334 6 0.092 728 0 ⎡⎛ 1 − 2 v ⎞ cos 4 α sin 4 α [20c] ε z = ⎢⎜ + ⎟+ kn1S1 kn2 S2 ⎣⎝ E ⎠ + Table 1. Rock sample properties (modified from Amadei and Goodman 1981). Lx = Ly = Lz (m) E (kPa) ν α (°) kn1 (kPa/m) kn2 (kPa/m) kn3 (kPa/m) ks1 (kPa/m) ks2 (kPa/m) ks3 (kPa/m) S1 (m) S2 (m) S3 (m) 1.0 1.0×106 0.49 36 1.0×107 2.0×108 3.0×109 1.5×107 1.5×108 1.5×109 0.001 0.002 0.003 and z = 0) of the bottom face of the block was also restricted from moving in the “x” direction. The resulting strain components from the numerical model and those from the analytical solution are identical, as shown in Table 2. Again, further tests were run using the same configurations in different coordinate systems obtained by cyclic substitution of the axes. The previous results were also reproduced with a large number of computational elements. In a triaxial compression test where three identical stresses of 1 MPa are applied in the “x,” “y,” and “z” directions, the strain components found by Amadei and Goodman (1981) are ⎡⎛ 1 − 2 v ⎞ sin 4 α cos 4 α [20a] ε x = ⎢⎜ + ⎟+ kn1S1 kn2 S2 ⎣⎝ E ⎠ + 1 ⎞⎤ sin 2 2α ⎛ 1 + ⎜ ⎟ ⎥ σn 4 ⎝ kn1S1 kn 2 S2 ⎠ ⎥⎦ ⎡⎛ 1 − 2 v ⎞ 1 ⎤ [20b] ε y = ⎢⎜ ⎟+ ⎥ σn ⎝ ⎠ E k n 3S3 ⎦ ⎣ 1 ⎞⎤ sin 2 2α ⎛ 1 + ⎜ ⎟ ⎥ σn 4 ⎝ kn1S1 kn 2 S2 ⎠ ⎥⎦ ⎡⎛ 1 ⎤ 1 ⎞ [20d] γ xz = ⎢⎜ − ⎟ sin 2α ⎥ σ n ⎢⎣⎝ kn1S1 kn2 S2 ⎠ ⎥⎦ In this case the computed strain components are again identical to those obtained from analytical relations (eq. [20]) as shown in Table 2. Again, further tests were run using the cyclic substitution of axes and a large number of elements. Poroelastic deformation Another method of linear verification is the poroelasticity approach, which governs the elastic behavior of porous media saturated with a fluid. The general approach is the theory of poroelasticity as first presented by Biot (1941). When the porous medium is not an isotropic material, the classic isotropic poroelasticity theory is no longer valid and the more general anisotropic poroelasticity theory must be invoked. Consider an isotropic saturated porous rock having the same dimensions as the intact rock of Fig. 2a, intersected by three mutually perpendicular fracture sets whose normal vectors are parallel to the coordinate axes. This configuration is shown in Fig. 4. The equivalent continuum resulting from the presence of fractures intersecting an isotropic rock is an orthotropic medium. In this case, we apply the general form of poroelasticity equations [21] = C – (C′Bp/3) where is the strain tensor, is the stress tensor, p is the pore pressure, C is the compliance matrix for orthotropic material (fractures plus isotropic rock), C ′ is a function of the sum of all terms in the compliance matrix, porosity, fluid compressibility etc. (Cheng 1997), and B is the general representation of Skempton’s pore pressure parameter and has different values in different directions as follows: [22a] Bx = 3 (C11 + C12 + C13) − C s C′ [22b] By = 3 (C 21 + C 22 + C 23) − C s C′ © 2006 NRC Canada Bagheri and Settari 579 Fig. 4. Orthotropic porous rock for verification of the model using poroelasticity theory. Table 3. Results of poroelasticity verification using either Cheng’s approach (1997) or the current study. ∆σx ∆σy ∆σz [22c] Bz = [23] −v E 1 + S 1k E 2 n2 −v E 0 εx εy εz 0.0 –0.000 893 0 –0.000 797 9 Note that “Cs” is the grain (solid) modulus. These equations are based on micromechanical analysis (e.g., Cheng 1997). Following the discussion presented in the section entitled “Constitutive equation of jointed rock masses”, the orthotropic compliance matrix for the specimen to be substituted in eq. [21] is derived as 3 (C 31 + C 32 + C 33) − C s C′ ⎡1 + 1 ⎢ E −vS1 kn1 ⎢ E ⎢ −v E C=⎢ 0 ⎢ ⎢ 0 ⎢ ⎢ 0 ⎣ 636 kPa 0.0 0.0 −v E −v E 1 + S 1k E 3 n3 0 0 0 0 0 0 0 2(1 + v ) + S 1k + S 1k E 1 s1 2 s2 0 0 0 0 2(1 + v ) + S 1k + S 1k E 2 s2 3 s3 0 0 0 0 Young’s modulus of 1 GPa, Poisson’s ratio of 0.2, and a very high grain (solid) modulus of 3.0 × 1013 kPa are assigned to the porous rock. Each fracture set has a normal stiffness of 2 × 108 kPa/m. Fracture spacings of 0.05 m, 0.03 m, and 0.07 m are assigned to the fracture sets whose normals are parallel to either the “x,” “y,” or “z” axis, respectively. Moreover, a uniform initial compressive stress of 6000 kPa is imposed on the system. The specimen experiences an increase in pore pressure while its front (x = Lx), back (x = 0), right (y = Ly) and bottom (z = 0) faces are restricted from moving in the “x,” “x,” “y,” and “z” directions, respectively. The analysis is isothermal. A single step increase in pore pressure from its initial value of 1 kPa to 1000 kPa creates changes in stresses and strains, as shown in Table 3. The model and the analytical method of Cheng (1997) are in perfect agreement. Nonlinear verification In the previous section, joints had constant parameters and there was no limitation on the amount of deformation a joint could undergo. As a result of constant joint parameters, the overall compliance matrix was also constant, and a single step load was sufficient to find the overall deformation. While convenient for testing, this can result in unrealistic behavior. In this section, a nonlinear joint model (discussed in the section entitled “Single joint deformation”) is assigned to the joints and as a result, normal stiffness is no longer a constant. High shear stiffness is assumed to eliminate its effect ⎤ ⎥ 0 ⎥ ⎥ 0 ⎥ 0 ⎥ ⎥ 0 ⎥ 2(1 + v ) 1 1 ⎥ + + E S 1 ks1 S 3 ks3 ⎦ 0 on deformation. A linear variation of normal stress with time is applied on the top of a rock sample containing a single horizontal fracture. The Young’s modulus of the intact rock and the size of the block were picked from a range of data given in Bandis et al. (1983) for their experiments. The initial stress was zero. Figure 5 shows the effect of fracture on the overall deformation of the rock mass and qualitatively verifies results of the model. Normal deformation of intact rock was assumed to be linear in the model, while the experiment has a small deviation from a straight line. When the system is at low stress, a small increase in normal load will create a relatively high total deformation, which demonstrates high deformability of fracture at low stress. At high stresses the slope of the rock mass deformation curve is almost equal to that of intact rock, indicative of reaching the maximum closure of the fracture. This can be seen from the fracture deformation, which is also plotted in Fig. 5. The vertical trend of this curve indicates asymptotically decreasing closure across the fracture (mathematically an infinite normal load is required for a complete closure). Accuracy of the nonlinear solution The dynamic problem must be solved using load steps. Depending on the initial effective stress level, choosing proper load increments may be very important for the accuracy of the computations, owing to the nonlinearity of the stiffness. In our implementation, the fracture normal stiffness is updated one load step behind the load increment. In © 2006 NRC Canada 580 Can. Geotech. J. Vol. 43, 2006 Fig. 5. Verification of the model using Bandis experiment (1983). Fig. 6. Effect of load steps on the overall deformation of rock mass (Bandis experiment 1983). other words, equivalent properties at each load increment are computed from the rock and fracture stiffnesses at the end of the preceding increment. In a loading cycle, a higher deformation than actual will result, owing to increasing stiffness of fractures with time. The reverse behavior is expected in the case of load decrement where smaller expansion would be computed. To ensure an accurate deformation solution, a sensitivity analysis has been performed. Figures 6 and 7 show the effect of load step size on the accuracy of the deformation of jointed rock mass and of the joint itself. Larger load steps obviously result in larger errors, and in some cases a negative fracture aperture was calculated (e.g., for a load increment of 1 MPa the final fracture closure is about 0.1 mm, which is larger than the maximum possible closure of 0.08 mm). A load step of 10 kPa is sufficiently accurate as seen in comparison with the 5 kPa step. Obviously, the optimum step depends on the joint properties used and the range of the load. It would be desirable to develop a dimensionless criterion for automatic selection of the load steps. This has potentially large implications for the computing efficiency of the coupled solutions of fluid flow and © 2006 NRC Canada Bagheri and Settari 581 Fig. 7. Effect of load steps on single fracture deformation (Bandis experiment 1983). geomechanics in fractured rock masses. If the stress path includes the nonlinear region of joint deformation, an accurate solution will require either a small load (i.e., time) increments or iterative updating of the stiffness matrix during the load increment. This situation is similar to other nonlinear problems of computational rock mechanics. Fig. 8. Hypothetical reservoir coupled with the stress model. Application in reservoir engineering The aforementioned stress model has been coupled with a single porosity flow model to study the flow behavior of hydrocarbon reservoirs. The resulting simulator solves both flow equations and stress–strain equations iteratively. The coupling term in this simulator is the porosity term. The details of the coupling method for the simulator used in this work are found in Settari and Mourits (1994). As the effective stress acting on the fracture plane changes, fracture will deform according to eq. [1]. This results in a variation in fracture normal stiffness over time, which in turn makes the compliance matrix of the fractured rock mass a nonlinear function of stress. In the coupled flow and geomechanics simulator, this matrix is recalculated in the stress module of the system iteratively within a time step. The ability of the model to handle fracture deformations will produce more accurate elemental volumetric strain as well as a more realistic stress field. The former leads to a more realistic determination of the porosity variation and total deformation of the reservoir and overburden rocks. The latter provides the necessary input to calculate variation in fracture aperture and permeability by some typical law. The main interest in modeling the variation of the aperture of fractures in time is to allow for the calculation of the fracture full permeability tensor, which is likely to have changing orientation and will evolve with time. Inclusion of fracture deformation in a full field coupled simulator also al- lows the contribution of fractures to compaction and in turn to reservoir subsidence to be considered. The presence of deforming fractures alters the calculation of volumetric strain and in turn the porosity in the flow model. The volumetric strain of each fractured element in the stress model is updated every time step based on new values of normal stiffness. Volumetric strain εv corresponding to a given confining stress state σkl is [24] εv = εx + εy + εz 3 ⎡ ⎛ 1 − 2v ⎞ ⎤ = ⎢3 ⎜ ⎟ σ avg + ∑ C ijkl δ ij δ kl σ kl ⎥ m =1 ⎢⎣ ⎝ E ⎠ ⎥⎦ where m = i, j, k, l. For a hydrostatic triaxial confining stress, the equivalent bulk modulus of fractured rock is [25] 3 ⎡ ⎛ 1 − 2v ⎞ ⎤ k * = ⎢3 ⎜ + ⎟ ∑ C ijklδ ij δ kl ⎥ ⎢⎣ ⎝ E ⎠ m =1 ⎥⎦ −1 © 2006 NRC Canada 582 Can. Geotech. J. Vol. 43, 2006 Fig. 9. Effect of fractures on vertical deformation of deformable matrix containing incompressible oil. Fig. 10. Effect of fractures on vertical deformation of nondeformable matrix containing incompressible oil. In this work, we restrict ourselves to coupled problems where the flow equations are solved as a single porosity media. In this case, the equivalent bulk modulus, k*, is sufficient to calculate the porosity changes and compressibility of the equivalent media in the flow model. This is in contrast to the dual porosity flow equations where two sets of equations are solved, for fluid in the fractures and in the matrix. In such a system, one needs to calculate the bulk moduli and compressibilities of the fractures and the matrix separately from the stress solution. An extension of this work to the dual porosity flow systems is described in Bagheri and Settari (2005). The development of the pseudo-continuum presented here in terms of total stresses is also valid in terms of effective stresses, provided that the fluid pressure difference between the matrix and the fractures is small (see Bagheri (2006) for details). Treatment of nonlinearity in the equivalent properties of jointed rocks in the coupled simulator was implemented differently from what was described for a single joint in the section entitled “Accuracy of the nonlinear solution.” The normal stiffness used in the equiva© 2006 NRC Canada Bagheri and Settari 583 Fig. 11. Effect of fractures on reservoir pressure and production of incompressible oil. Fig. 12. Effect of fractures on vertical deformation of deformable matrix containing compressible oil. lent moduli of fractured rock mass is the average of the old and the new (current) normal stiffness. This parameter is updated at every iteration loop during a time step. Averaging the normal stiffness of fractures results in better accuracy and will alleviate the problem of controlling the load (time) steps. However, monitoring of stress (or pressure) changes over the time step is necessary to ensure accuracy. It is also noted that fractures in reservoirs usually do not start deforming from zero effective stress (except in highly overpressured reservoirs). In most cases fractures start deforming from the less sensitive part of the fracture deformation path. The effect of production on deformation of fractures will now be studied using test examples. The effect of several parameters, including intact rock stiffness and fluid compressibility, was studied. Figure 8 shows a hypothetical single phase oil reservoir, which contains dead oil. The grid is 4 by 4 by 1 with ∆x = ∆y = ∆z = 10 m. The initial in-situ porosity is 35%, and the reservoir permeabilities are 1000 D in all directions to cause uniform pressure changes throughout the © 2006 NRC Canada 584 Can. Geotech. J. Vol. 43, 2006 Fig. 13. Effect of fractures on reservoir pressure and production of compressible oil. entire reservoir. Oil has a constant formation volume factor of 1.0 (zero oil compressibility), viscosity of 1.0 cP (1 P = 0.1 Pa·s), and a density of 800 kg/m3. The initial fluid pressure is 14 450.8 kPa at a depth of 805 m. The rock properties were based on data from Bandis et al. (1983). The elastic modulus is 51 GPa, and the Poisson’s ratio is 0.2. The initial horizontal stresses are 16 000 kPa and the initial vertical stress is 18 000 kPa over the entire reservoir depth. A set of equal roughness horizontal fractures is assumed to intersect all grid blocks of the model. These fractures have a normal stiffness of 19.35 GPa/m at zero stress, a shear stiffness of 5 × 1099 kPa/m (a static parameter as discussed earlier), and an aperture of 0.149 655 mm. It is assumed that the fracture can close completely if enough stress is applied (zero roughness fracture). Several configurations of fracture intensity and rock properties have been considered. Corresponding initial apertures of the fractures and initial normal stiffnesses were calculated using the relations from Bandis et al. (1983). It is worth noting that these correlations may produce unrealistic data for low strength materials (especially for initial aperture and maximum joint closure). A vertical well was located in the cell (1,1,1), producing 1 m3 of oil at surface conditions until the reservoir pressure was depleted to 500 kPa. Depending on the case, the model was run for 50 or 500 days. Figure 9 depicts the deformation of the structure for various fracture intensities. In these figures (Figs. 9–13), a conventional reservoir with no fractures is indicated as “Conventional Res.”; a fractured reservoir with 1 m fracture spacing is labeled “Low Intensity”; the case with fracture spacing of 10 cm is labeled “High Intensity”; and the case with 1 cm spacing is denoted as “Very High Intensity.” As fracture intensity increases, the overall deformation also increases. The amount of contribution of the fractures shows the importance of considering fractures in bulk deformation. At the end of the reservoir life, the cal- culated fracture contribution to the deformation of the high intensity fractured reservoir lies exactly on the top of a curve, which is obtained by subtraction of the overall deformation of the conventional reservoir from that of the high intensity case. These two curves are not equal at the beginning of production simply because of the fracture contribution to reservoir compaction, which causes a difference in pressure decline of these two reservoirs. The gap between these two curves is due to the rate of matrix deformation. For a nonfractured reservoir, the rate of pressure change is higher than that in a fractured reservoir. As a result, the matrix will deform faster in the conventional reservoir. However, because of the contribution of fracture deformation, the overall rate of the deformation in the high intensity case is higher. This makes the difference in the overall deformations of these two cases lower than the calculated fracture contribution in deformation. This difference disappears when very high Young’s moduli of rock matrix are used. Figure 10 shows such a case and demonstrates that for nondeformable rocks, all of the deformation in the model is caused by fractures. The effect of fractures on compaction can also be observed from their impact on fluid flow. Higher intensity of fractures increases the compaction of the reservoir, which in turn results in a slower pressure decline and maintains longer production. This fact is demonstrated in Fig. 11 for an incompressible oil reservoir. In a more realistic reservoir where fluid is compressible, the effect of fractures on solid deformation remains the same; but their impact on fluid flow parameters tends to diminish. Figure 12 shows the impact of fracture intensity on the same reservoir containing oil with compressibility of 10–5 kPa–1. Slower pressure decline in a compressible system masks the influence of fractures on pressure maintenance. As a result, the gap observed in an incompressible system is not seen here. Comparison of this figure and Fig. 9 reveals © 2006 NRC Canada Bagheri and Settari that the overall deformations in both systems are exactly the same. When a fluid system is compressible, fractures do not play a significant role in pressure decline until their density becomes large. Figure 13 shows the effect of fracture compaction on fluid flow. In this figure, the solid line curve, which has the lowest slope, represents pressure drop in a fractured reservoir with spacing equal to 1 cm (very high intensity fractured reservoir, which is too dense for most actual reservoirs). Fracture spacing must be very small to result in an appreciable influence on reservoir pressure. This fact emphasizes that the main effect of fractures on fluid flow coupling is not the effect of fracture deformation on compressibility. The main coupling to fluid flow is through the effect of fracture deformation on reservoir permeability as shown in Bagheri and Settari (2005). However, fracture deformation, as shown by the examples herein, plays an important role in the overall deformation of the system, regardless of system compressibility. Conclusions A reliable methodology for the determination of equivalent moduli of jointed rock mass was presented and implemented for modeling fracture contribution to deformation and flow solution in a coupled geomechanical and reservoir simulator. The methodology is flexible enough to accommodate any number of fracture sets of any fracture orientation and properties. The interesting feature is that the deformation characteristics of fractures are included without treating them explicitly. In addition to modeling the role of fractures in overall deformation, fracture aperture can be calculated as a parameter needed for the calculation of the dynamic permeability tensor on an element basis. Deformation of fractures plays an important role in the overall deformation of fractured rocks, but has only minor impact on the flow behavior of fractured reservoirs, if its effect on permeability is not considered. Neglecting the contribution of fractures on overall deformation could lead to unrealistic predictions of deformation, especially in highly fractured rocks. The effect of fracture deformation on reservoir permeability also plays a significant role in fluid flow of fractures reservoir. This coupling will have appreciable influence especially in a low compressibility system. References Amadei, B., and Goodman, R.E. 1981. A 3-D constitutive relation for fractured rock masses. In Proceedings of the International Symposium on the Mechanical Behavior of Structured Media, Ottawa. pp. 249–268. Bagheri, M. 2006. Modeling geomechanical effects on the flow properties of fractured reservoirs. Ph.D thesis, University of Calgary, Calgary, Alta. Bagheri, M., and Settari, A. 2005. Modeling of geomechanics in naturally fractured reservoirs. In Proceedings of the Reservoir Simulation Symposium, The Woodlands, Tex., 31 January – 2 February 2005. Society of Petroleum Engineers, Richardson, Tex. Bandis, S.C., Lumsden, A.C., and Barton, N.R. 1983. Fundamentals of rock joint deformation. International Journal of Rock Me- 585 chanics and Mining Science and Geomechanics Abstracts, 20: 249–268. Barton, N.R., and Choubey, V. 1977. The shear strength of rock joints in theory and practice. Rock Mechanics, 10: 1–54. Biot, M.A. 1941. General theory of three dimensional consolidation. Journal of Applied Physics, 12: 155–164. Cheng, A.H.-D. 1997. Material coefficient of anisotropic poroelasticity. International Journal of Rock Mechanics and Mining Sciences, 34: 199–205. Goodman, R.E. 1976. Methods of geological engineering in discontinuous rocks. West Publication Company. Huang, T.H., Chan, C.H., and Yang, Z.Y. 1995. Elastic moduli for fractured rock mass. Rock Mechanics and Rock Engineering, 28(3): 135–144. Settari, A., and Mourits, F.M. 1994. Coupling of geomechanics and reservoir simulation models. In Proceedings of Computer Methods and Advances in Geomechanics Conference, Morgantown, W. Va., 22–28 May 1994. Edited by H.J. Siriwardane and M.M. Zaman. A.A. Balkema, Rotterdam, The Netherlands. Vol 3, pp. 2151–2158. Timko, D.T. 1966. A case against oil muds. The Log Analyst, 1966: 4. Youshinaka, R., and Yamabe, T. 1986. Joint stiffness and the deformation behavior of discontinuous rock. International Journal of Rock Mechanics and Mining Science and Geomechanics Abstracts, 23(1): 19–28. List of symbols a, b constants A total area of a fracture set in a block B matrix of Skempton’s pore pressure parameters Cs grain (solid) modulus C compliance matrix D constitutive matrix E Young’s modulus G shear modulus H height of sample k joint stiffness k* bulk modulus L dimensions of the sample L transformation matrix n joint normal vector p pore pressure S joint spacing v Poisson’s ratio V volume α dip angle δ Kronocker delta ␦ joint deformation vector ε normal strain strain tensor γ shear modulus ν normal deformation σ stress stress tensor traction vector Superscripts f individual fracture set I intact rock J joint © 2006 NRC Canada 586 Can. Geotech. J. Vol. 43, 2006 M number of joint sets Fig. A1. Test example used to expand eq. [A1]. Subscripts avg mean (average) i initial i,j,k,l tensor indices I,J,L tensor indices j joint max maximum n normal N,S,T joint coordinate system (normal, strike, dip) s shear t total (joint + rock) v volumetric x,y,z directions Appendix A Figure A1 shows an element of intact rock, which is intersected by three sets of fractures. The first two sets of fractures are orthogonal to the third set, and have an angle of θ = 2α between them. These two sets are assumed to have the same normal and shear stiffness, kn, ks, and spacing, S, which are different from that of the third joint set parameters, kn3, ks3, S3. This example is presented in Huang et al. (1995), where eq. [A1] has been used to derive the equivalent moduli, neglecting the dilatancy effect. M [A1] f f C Jijkl = ∑ n if L jJf D JL L Ll n kf f =1 1 Sf The resulting moduli according to Huang et al. (1995) are as follows: [A8] 1 1 2 sin(2α) cos 2 α = + G xy S3ks3 Sks [A9] 2 sin 2 (2α)(ks + kn ) 1 = G yz Skn ks [A10] 1 1 2 sin(2α) sin 2 α = + G zx S3ks3 Sks where G is the shear modulus of the rock mass. It seems however that eqs. [A8]–[A10] are incorrect. This can be demonstrated by setting α = 0 and neglecting the third fracture set for clarity without losing the generality of the discussion. In this case all shear terms are equal to zero, while we have two horizontal sets of fractures intersecting the “z” axis with an angle of 90°. The shear modulus in “zx” and “zy” should be affected by the fracture shear deformation term, 1/Sks, where in this case it has to be multiplied by a factor of 2 owing to the presence of two sets of fractures with the same properties. Equations [A8]–[A10] have been re-derived in this work using the same method, as follows: [A2] 1 1 = Ex kn3S3 [A3] 1 2 sin 2 α(kn cos 2 α + ks sin 2 α) = Ey ks kn S [A4] 1 2 cos 2 α(kn sin 2 α + ks cos 2 α) = Ez ks kn S [A11] 1 1 2 sin 2 α = + G xy S3ks3 Sks [A5] − kn − ks sin 2 (2α) 2 kn ks S [A12] 2 [sin 2 (2α)ks + cos 2 (2α)kn ] 1 = G yz Skn ks [A6] − [A13] 1 1 2 cos 2 α = + G zx S3ks3 Sks [A7] − vyz Ez vxy Ex =− =− vzy Ey vyx Ey =− =0 vzx v = − xz = 0 Ez Ex These new formulae give the expected shear terms for the case of α = 0 and in other simple cases where eqs. [A8]– [A10] are unable to do so. © 2006 NRC Canada
0
You can add this document to your study collection(s)
Sign in Available only to authorized usersYou can add this document to your saved list
Sign in Available only to authorized users(For complaints, use another form )