Beruflich Dokumente
Kultur Dokumente
pore microstructure to hydrodynamic properties of porous media such as granular packs and sandstones. In the GPT model, viscous dissipation is assumed to be localized only in pore throats while grain size determines the spacing between pore throats, i.e. the porelgrain number density. By using this model, homogeneous porous media are described by three geometrical parameters dh, lh and It characterizing respectively the hydrodynamic diameter of pore throat, 2 its length and the spacing between pore throats. The permeability k is predicted to vary as dh or 4 d2 for unconsolidated grain packs and as dh for sandstones cemented by uniform grain g overgrowth. If the porosity of the overgrowth is constant and independent of its thickness, the - Q 1 ) ] 4 where c is a shape constant model prediction is k = c dg,2[(0 - Q1) I (0, , the initial porosity characterizing initial detrital grains of diameter dg,, 0 the porosity, a before cementation, and @ I , the residual porosity reached at the end of mineral deposit occuring when pore throat and permeability has droped to zero, i.e. at the "hydrodynamic" percolation threshold. Since all model parameters (dh, lh, It, dgo, c and @I) have a clear physical meaning and were determinated in the reported experiments, the proposed model has been experimentally validated. The experimental data concern both unconsolidated porous media such as sphere and sharp-edged grain packs as well as purely quartzitic sandstones. Polymer depletion and latex deposition methods were used to determine dh, while conventional ~ . mercury injection tests were used to get another estimation of pure throat diameter d ~The values of d~~ and dh were found to be proportional as expected theoretically, but for porous media having a sharp edged pore geometry, d~~ is significantly larger than dh ( 1.7 < d ~ ~ l < d 1.9). h As a consequence, the filtration or exclusion pore diameter dmin, which is relevant for predicting permeability damage by straining and pore bridging is strongly overestimated if assimilated to d ~ ~ .
INTRODUCTION
The transport properties of porous media are determined by their microstrucure, i.e. geometry and topology of interconnected pore space. Transport properties include thermal and electrical conductivity as well as permeability to various types of fluids from Newtonian fluids, such as water, oil and gas to complex "colloidal" fluids containing polymers, micelles, microernulsions, emulsions or fine solid particules, possibly carrying contaminants. Due to the importance of such transport properties for water supply, environmental issues as well as for
chemical and petroleum engineering, extensive investigations using sometimes sophisticated techniques were conducted since some decades. However, these techniques which can provide useful information are often costly and time consuming, so that their use is often limited to research studies or specific applications. In this study, very simple analytical equations containing few parameters, having a clear physical meaning and often easy to determine by experiments, are derived by using the "grain and pore throat" model, which describes the porous media as a bundle of unit cells, each consisting of a grain and a pore throat. The predictions of this model are in good agreement with the reported experimental data obtained by using relatively model porous media such as unconsolidated sphere and sharp-edged grain random packs and sandstones cemented by silicalquartz overgrowth. Extension of this new model to predict transport properties in sandstones cemented by more complex overgrowth is in progress.
THEORETICAL
The basic law of non-inertial viscous fluid hydrodynamics states that the flux inside a capillary (Poiseuille law or around successive grains (Stokes law), i.e. the flux per unit cell, i.e. per cross section d,1 or d 2 for a given pressure drop per unit length is proportional to the power four of characteristicf ength (capillary or grain diameter):
This scaling law (Eq. 1) can be derived very simply from viscosity definition or from NavierStokes equations if inertia term is neglected. Thus, the equation 1 is valid as long Reynolds number remains low (Re < 1 for flow through porous media or around isolated spheres). Only the proportionality constant (omitted in Equation I ) depends on geometry, i.e. shape and arrangement of capillaries or grains. For systems having compIex geometries, such as porous media, it is very convenient to define the equivalent hydrodynamic diameter &, which can be calculated by:
rL
where s is the curvilinear abcissa, L the total length of a stream line and dh is the local hydrodynamic diameter, i.e. the diameter of a cylindrical capillary having the same length ds, and giving the same pressure drop for the same flow rate. Interestingly, the equation 2 shows that for geometries such as that of porous media where the constricted sections, commonly called "pore throats", are much smaller than "pore bodies", the viscous dissipation can be considered as being localized only in pore throats. This approximation is quite acceptable, since the pore body-to-pore throat radius is quite large, varying from values around 3 for random
monosized sphere packs, to around 5 for random packs of sharp-edged grains with a narrow size distribution, and increasing up to much higher values for cemented sandstones. As a consequence, modelling granular porous media as successive pore throats of diameter &, and length ih where viscous dissipation is localized seems reasonnable to predict permeability and its variations with grain size and cementation. In this approach, grains acts as "spacers" determining the number of pore throats per unit length, section or volume. For other purposes, for instance to model colloid deposition and subsequent permeability damagel, grains and pore throats are also collectors. Grains collect most of colloids with a negligible effect on permeability, while the colloids collected in pore throats by surface deposition or pore bridging2 determine permeability damage. This "grain and pore throat" model can be used without any difficulty to derive simple equations relating permeability to microstructure characteristics of sand packs and sandstones:
For unconsolidated porous media, where grains are in contact without cementation, the spacer between pore throats It is slightly smaller than grain diameter dg (where dg is the diameter of the sphere having the same volume), and related to porosity by:
Since porosity is practically constant (0.38 c 4r c 0.42 for random packs of monosized spheres or grains) and independent of grain size, the number of pores per unit cross section scales like di2 or, :d so that the equation 1 gives:
It must be noted that any attempt to link permeability and porosity is theoretically incorrect as long as the exact process leading to a change in porosity is not known. In this respect, the commonly used Kazeny-Carman relation cannot be considered as a model, but only as an empirical correlation giving an acceptable fit of experimentd data in some cases.
For sandstones formed by cementation of detrital sand beds by diagenetic grain growth, porosity decreases during diagenesis, while the number of grain (or pore throats) remains constant. In addition the spacer remains constant and can be estimated from the equation 3 where dg is equal to dg,, the size of the "initial" detrital sand grain. Under these conditions, the permeability is expected to vary as: k Ma; (5)
From these considerations it becomes easy to calculate k from overgrowth characteristics. For simple cases where the overgrowth has a "statistically" constant thickness ed, uniformly
distributed on the grain, i.e. equally in pore bodies and in pore throats, the equation 5 gives:
where is the hydrodynamic diameter of the initial detrital sand bed and edh = cl ed with cl is a constant depending on the overgrowth characteristics. For an amorphous deposit of constant thickness around the initial grain cl = 1, while for a crystalline deposit with crystallisation planes having a random orientation, we would expect a value of cl around \% Before cementation, ed = 0, and k = ko leading to:
ah(
with c a constant characterizing initial detrital grain shape and packing. The cementation of a sandstone is due to a surface deposit from a water oversaturated in ions (for instance silicates). As a consequence, we can derive a simple equation between @ and ed:
where f, is a surface shape factor (the ratio between grain surface and that of a sphere having the same volume) and Q0 the porosity of initial sand bed before cementation. The cementation reaches a maximum when water flow drops to zero, i.e. when the fraction of pore throats closed by the overgrowth becomes equal to the non-percolation threshold (around 75% in three dimensions). At this final stage of cementation, pore bodies are not entirely fulfilled by the overgrowth so that there is an irreductible residual porosity al, such as:
It should be emphasized that Ql is only the residual pore volume remaining at the end of the overgrowth, and results from the relation between dhand O (Equation 10). Consequently, 0 1 cannot be assimilated to a "dead-end pore" volume since this last concept is related to flow
properties. Indeed in our GPT model, viscous dissipation occurs only in pore throats, not in pore bodies. Interestingly, the equation 11 is very similar to the scaling relation derived from numerical simulation of Stokes flow through random packing of various types of overlapping spheres3. This suggests that the universal curve (Fig. 7 in Ref. 3) is obtained because viscous dissipation occurs quasi-exclusively in pore throats as assumed in our model and the effective specific surface per unit volume remains approximately constant during deposit overgrowth. The equations 9,10 and 11 have been derived by assuming the specific mass of overgrowth is constant and equal to that of detrital grains. This hypothesis is quite valid for Fontainebleau sandstones where the detrital grains are almost purely quarzitic as well as the overgrowth415. Very recently6, cathodoluminescence has been used to identify and quantify the overgrowth in sandstones, showing that in Fontainebleau sandstones the overgrowth may reach up to 35%, i.e. a nearly complete occluding of the initial porosity, in agreement with our analysis (see below experimental section). Pore size is currently evaluated by mercury injection, or more generally by intrusion of a nonwetting phase in the porous medium738.9. In this type of experiment, the cumulative volume of non-wetting phase introduced in the porous medium is recorded as a function of pressure, and the results are analyzed in terms of pore throat radius distribution using the Laplace equation (Ro = 2 y cos 8/Pc) where y is the interfacial tension, 0 the wetting angle and PC the capillary pressure. Independently of the refinement of the analysis process, mercury intrusion remains a static non-directional process. More precisely, mercury invades pore bodies by the largest pore throats while the directional character of fluid flow implies the preferential use of pore throats located upstream and downstream. Thus, the equality between pore dimensions determined by mercury injection and flow experiments would be obtained only for unrealistic porous media where all pore throats would have the same size. Thus for real porous media, dh must be smaller than d ~ . In addition &, as defined by Equation 2 gives a mean value of pore throat hydrodynamic diameters, which are of course larger than the minimum pore throat diameter dmin relevant for characterizing the straining of particles by steric exclusion. Then, dmin should be smaller than dh, SO that: (12) drnin < dh < d~~ For tronconical pores, we can derive a simple equation between dmin and dh:
where dm, is the maximum diameter of pore bodies which is nearly equal to dg, so that dmin should be approximately two times smaller than dh, in sharp-edged geometries.
EXPERIMENTAL
The experimental results concern the following relationships I ) permeability vs grain size for granular packs at constant porosity, 2) permeability vs porosity for sandstones, cemented by grain overgrowth and 3) hydrodynamic diameters, and mean pore throat dimaters derived from mercury injection vs permeability for both granular packs and sandstones.
1) The polymer depletion methodlo>l3,that consists of injecting a well-known non-adsorbing polymer in a porous medium while measuring its apparent viscosity in the Newtonian regime. Then, the hydrodynamic mean diameter is derived from the apparent-to-actual viscosity ratio, using the "depletion model"14.
2) The latex deposition method12 that consists of depositing a monolayer of well-known latexes in a porous media and then measuring the subsequent decrease in permeability. This method was found to give results very similar to the polymer depletion methodl2.
Pore throat diameters derived from mercury injection tests d~~ were calculated from the pressure required to invade 50% of pore volume.
For unconsolidated granular packs, dh and d~~ are plotted vs permeability for glass bead and silicon carbide packs (Fig. 4). Both dh and d~~ varies as the square root of permeability as predicted by Equation 4. An interesting result is that d~~ = 1.7 dh for SIC packs. For glass beads, simulation results5 show that 50% of pore filling is reached through pore throats having a size equal to 0.39 dg while the value of dh versus dg can be derived from the best fit of dh versus k in Fig. 4 (db 1.15 [2 ( 8 k / ~ ) ~ . 5 and ] ) k versus dg in Fig. 1 (k 13.6 104 d,2) so that, we obtain: d~~ = 1.2 dh for bead packs.
For sandstones consolidated by deposit growth, both d~~ and dh vary as the power 0.25 of permeability (Fig. 5) as predicted by Equation 5 and d~~ 1.9 dh. Since dmin is expected to be around two times smaller than dh (Eq. 13), we can expect that mercury injection overestimates by a factor 3 to 4 the mean value of minimum restrictions between pore bodies located on the flow path of fine particles. Thus, for homogeneous sandstones and sand packs, we expect a strong overestimation of dminwhen estimated by d ~ ~ : d~~ " 3 to 4 dmin
1996 SCA Conference Paper Number 9607 Using these values, the deposit thickness may be calculated from the equation 9:
where 0.41 is the initial porosity, a value consistent with data on Sic packs and with extrapolation for Fontainebleau sandstones towards maximum permeabilities. For a 10% porosity sandstone core, fs ed derived from the equation 14 is 13.14 pm while edh = d h d 2 - dh/2 = 12.37 pm, SO that W e obtain: Since c l values are expected to vary between 1 (amorphous deposit) and fi (cristalline we deposit), we obtain 1.06 < fs < 1.5 from Eq. 15. Since deposit is mainly quartzitic4~5~6, obtain f, 1.5, a quite reasonnable value for Fontainebleau sandstone grains.
CONCLUSIONS
A new structural model of granular porous media, referred as the "grain and pore throat model" was used to derive theoretically simple analytical equations relating transport properties to microstructure of grain packs and sandstones. All model predictions are in agreement with experiments. Particularly, the equation relating permeability to porosity in sandstones is as simple as Kozeny-Cman empirical equation but is based on a correct physics and gives a very good fit of experimental data in the low permeability range. In addition, this study shows that pore throat sizes are substantially smaller than that derived from conventional analysis of mercury injection tests.
ACKNOWLEDGEMENTS
The authors thanks Institut Fran~ais du Petrole for permission to publish these results. They also wish to acknowledge 0. Vizika and C. Roque for kindly providing mercury injection data.
REFERENCES
I.
2.
3.
4.
5.
L. Nabzar, G. Chauveteau and C. Roque, "A New Model for Formation Damage by particle Retention", SPE 31 119, International Symposium on Formation Damage Control, Lafayette, Louisiana, USA, February, 14-15, 1996. C. Roque, G. Chauveteau, M. Renard, G. ThibauIt, M. Bouteca and J. Rochon: "Mechanism of Formation Damage by Retention of Particles Suspended in Injection Water", SPE 301 10, European Formation Damage Control Meeting, The Hague, May, 1516, 1995. N.S. Martyrs, S. Torquato and D.P. Bentz, "Universal Scaling of Fluid Permeability for Sphere Packing" Physical Review E, 50, 1, 403-408, 1994. C. Jacquin: "Correlation entre la permeabilite et les caracteristiques georn6triques du Grks de Fontainebleau", Rev. Inst. Fran~. du Petrole, XIX, 921-937, 1984. T. Bourbie, 0. Coussy et B. Zinszner: "Acoustique des milieux poreux", Ed. Technip, 1985.
B. Markchal, A. Chartier, I)h. Blanc, M. Thiry, N. Guilhaumou and .C Durand, "New Insight on Diagenetic Overgrowth by Cathodoluminescence", 1996, Int. Symp. of the Society of Core Analysis, Sept. 8-10, Montpellier, France. A.H. Thompson, A.D. Katz and R.A. Raschke: "Estimation of Absolute Permeability From Capillary Pressure Measurements", SPE 16794, 62nd Annual Technical Conference and Exhibition, Dallas, TX, Sept. 27-30, 1987. D. Zhou and E.H. Stenby: "Interpretation of Capillary Pressure Curves Using Invasion Percolation Theory", Transport in Porous Media (1 I), 17-31, 1993. D.M. Smith, D.P. Gallegos and D.L. Sterner: "Mercury Porosimetry in Random Sphere Packings: Breakthrough Pressure and Pore Structure Determination", Transport in Porous Media (53), 11-22, 1987. G. Chauveteau, "Rodlike Polymer Solution Flow Through Fine Pores: Influence of Pore Size on Rheological Behavior", J. Rheol., 26, 2, 111-142, 1982. S.L. Bryant, P.R. King and D.W. Mellor: "Network Model Evaluation of Permeability and Spatial Correlation in a Real Random Sphere Packing", Transport in Porous media ( l l ) , 53-70, 1993. S.Y. El Attar and G. Chauveteau, "Permeability Control by Colloid Deposition", submitted for publication. G. Chauveteau and A. Zaitoun, "Basic Rheological Behavior of Xanthan Polysaccharide Solutions in Porous Media: Effect of Pore Size and Polymer Concentration", Europ. Syrnp. on Enhanced Oil Recovery Proc., Elsevier, Sequoia, Lausanne, 197-212, 1981. G. Chauveteau, M. Tirrell and A. Omari, "Concentration Dependence of the Effective Viscosity of Polymer Solution in SmalI Pores with Repulsive or Attractive Walls", J. Colloid and Interface Sci., 100, 1, 41-54, 1984.
'
" "
Sharp
0.01
10
Fig 1: Permeability vs grain diameter for unconsolidated granular porous media ( 9 = 41%) : k, = 0.00056dgl-98 for S i c grain packs and k2 = 0.00136dg1*92 for glass bead packs; k,IkI = 2.43
Permeability (pm2)
Fig2: Correlation between porosity and perrneabiIity for three different series of Fontainebleau sandstones. Data are from : Ref. 4 ( series I ) , Ref. 2 ( ) . The dotted lines are the series 2) and Ref. 10 ( 0 series 3 best fits (slope 0.25) of experimental data for series 2 and 1 in the high permeability range
0.001
1
2)
10
0 - 9,
) and permeability
for 0 1 = 2 % showing the good agreement with the model (Eq 1 0 ) . The straight line is the best fit of experimental data : k = 23~4.(Q-Q,)~-~*
d ~ g Sharp-edged grains
0
0
dh
( S i c grains)
dh
0.1
1 Permeability (Fm2)
Fig 4: Characteristic pore throat diameters of unconsolidated granular packs.The dotted line corresponds to the capillary model relationship : d,=2(~k/$)O*~
0.01
0.1 Permeability ( p n 2 )
Fig 5: Characteristic pore throat diameters for FontainebIeau sandstones showing the 0.25 power law dependence between dh, dHg and permeability.