Immiscible Multiphase Flow Behaviours of Water-Oil-CO2 Ternary System Flooding Using X-ray CT

Carbon dioxide (CO2) injection into oil reservoirs has been widely accepted as an effective technique for enhanced oil recovery (EOR) after waterflooding. More recently, the development of carbon capture and storage (CCS) to reduce CO2 emissions has made CO2-based EOR increasingly attractive. Waterflooding is widely employed in China and, even after several years of water injection, considerable oil deposits remain in the thick, positive rhythm reservoirs in eastern China. The majority of these reservoirs may not be suitable for miscible CO2 flooding. The present work investigated immiscible CO2 flooding after waterflooding in such sites by laboratory trials. Series of large artificially-consolidated sandstone models with different levels of heterogeneity were used to simulate thick, positive rhythm oil reservoirs. Gaseous CO2 was continuously injected into these models at a constant injection pressure and an X-ray CT scanner (with a resolution of 0.7 × 0.7 mm) was used to monitor and record changes in the fluid saturation and migration. Based on the experimental results, it indicates that immiscible CO2 flooding following waterflooding is an efficient means of enhancing oil recovery. It is evident that both the reservoir heterogeneity and injection pressure differential affect oil recovery and CO2 distribution. The heterogeneity has a remarkable impact on oil recovery when the permeability differential between layers is lower than 2 millidarcy (md) or the permeability variation coefficient is less than 0.2. Fitting of the experimental results also demonstrates that there is an optimum pressure differential between inlet and outlet that maximizes oil recovery under specific inhomogeneous conditions.


INTRODUCTION
Enhanced oil recovery (EOR) technologies have become increasingly important in the petroleum industry.Generally, the primary recovery stage is relatively short, because the natural energy of reservoirs used in this stage is rapidly attenuated.A typical primary recovery ranges from 5% to 20% of the original oil in place (OOIP) (Blunt et al., 1993).For this reason, the reservoir pressure is often increased to a level at which the reservoir fluids can be obtained at the desired rate during the secondary recovery stage.The most widely used secondary recovery method is waterflooding, in which water is injected into the reservoir both to maintain the pressure and to drive the movement of the reservoir fluids (Bikkina et al., 2016).The secondary recovery is approximately 10% to 20% of the OOIP (Lake et al., 1992;Shaw and Bachu, 2002), meaning that approximately 50%-60% of the OOIP remains in the reservoir following the secondary recovery stage (Gao and Gu, 2013).Following more than 20 years of widely applied waterflooding, the majority of reservoirs presently generate a product stream with a water content of 80%, although this value can exceed 90% in mature oilfields in eastern China (Wang, 2009).However, this region also contains numerous thick, positive rhythm reservoirs with fluvial facies, in which the permeability of the lower reservoir is much higher than that of upper part (Wang et al., 2017).As a result of the combined effect of heterogeneity and gravitational differentiation (Yang et al., 2012(Yang et al., , 2014;;Jen et al., 2017), water breakthrough can occur in the lower part of such reservoirs, leading to low sweep efficiencies and high residual oil saturation in the upper reservoir during waterflooding (Yaghoobi et al., 1996;Brouwer et al., 2001;Zhou, 2008;Zhao et al., 2013;Wang et al., 2017).It has indicated that the reservoir permeability increases after waterflooding when the permeability is higher than 1500 millidarcy (md), denoting the unit for permeability (Ren et al., 2015).Long-term water injection promotes the development of high capacity channels (Jen et al., 2017) in the lower high permeability layers, thus exacerbating interlayer interference (Wang et al., 2017).
Gas injection has been widely recognized as a potential tertiary oil recovery technique for use after waterflooding (Bachu et al., 2004;Soong et al., 2014;Li et al., 2016).Having been in commercial use for 40 years now, the utilization of CO 2 removed from industrial sources for EOR has become increasingly attractive and significant to the petroleum industry in recent years (Godec et al., 2013;Huang and Tan, 2014;Duan et al., 2016;Sun et al., 2016).This is because CO 2 emissions from industrial processes, such as coal-fired power generation (Uemura et al., 2011) are currently a major contributor to the accelerating rise in the average global temperature and the consequent effects of climate change (Huang and Tan, 2014;Li and Li, 2014).CO 2 flooding technique (Shi et al., 2011a;Soong et al., 2014;Jean et al., 2015) is considered miscible, immiscible or near-miscible, based on the operating pressure, type of oil and reservoir characteristics.Miscible displacement is only possible at pressures above the minimum miscibility pressure (MMP) (Li et al., 2016;Jen et al., 2017).However, in most oilfields in eastern China, miscible flooding is difficult to achieve due to the limitations imposed by the reservoir conditions and oil properties (Liu et al., 2011).Therefore, it has become essential to study the feasibility and efficiency of immiscible CO 2 flooding after waterflooding, especially in thick, positive rhythm oil reservoirs.Unfortunately, because miscible CO 2 flooding (Jen et al., 2017) is considered to be more efficient, immiscible flooding (Yang et al., 2011) has not received sufficient attention relative to the importance of this technique in the mature oilfields of China.
To date, various laboratory experiments have investigated several aspects of CO 2 flooding, such as the gas/liquid saturation distribution (Shi et al., 2011b;Saeedi et al., 2011;Song et al., 2014;Soong et al., 2014), fluid migration characteristics (Ott et al., 2012;Wang et al., 2013;Behzadfar et al., 2014;Huang and Tan, 2014;Jean et al., 2015), changes in reservoir and fluid properties during flooding (Perrin and Benson, 2010;Alemu et al., 2011;Han et al., 2016;Liu et al., 2016), and factors affecting EOR recovery (Farajzadeh et al., 2010;Zhou, 2010;Iglauer et al., 2013).Imaging technologies, such as X-ray CT, are widely used in laboratories to allow the visual and intuitive examination of experimental data (Elkamel, 1998;Takashi et al., 2001;Kvamme et al., 2007;Chen et al., 2010;Soong et al., 2014).The CT scanning is typically used in conjunction with core flooding experiments on the laboratory scale.However, the two-phase fluid flooding experiments of laboratory scale cores make it impossible to visualize and analyse three-phase (e.g., water-oil-CO 2 system) flow behaviours in a heterogeneous thick oil reservoir.In addition, the difficulty in performing gas saturation calculations based on CT data results from a lack of three-phase fluid (CO 2 , water and oil) migration experiments.In this paper, the three-phase fluid (water-oil-CO 2 ) flooding behaviours in artificial cores with properties of in-situ oil reservoirs are visualized with X-Ray CT scanning.The multiphase fluid migration and distribution are characterized, and some dominant effects on CO 2 flooding efficiencies are also investigated.To our best knowledge, this is the first time to analyse three-phase fluid flooding behaviours of the (water-oil-CO 2 ) ternary system.The experimental results can be significant to improve the understanding of CO 2 enhanced oil recovery and CO 2 geological storage in depleted oil reservoirs.

Method and Materials
In the present work, X-ray CT technology was employed to obtain accurate data from flooding tests involving three fluids (gaseous CO 2 , water and oil).Herein, we describe series of laboratory experiments carried out to improve our understanding of the multiphase flow processes and efficiency of immiscible CO 2 flooding following waterflooding in heterogeneous thick reservoirs.This work also examined the factors affecting oil recovery.A series of large-scale, artificially-consolidated sandstone models was used to simulate positive rhythm, heterogeneous, thick oil reservoirs.A schematic diagram of the apparatus, which was designed to allow both water and gas flooding trials, is shown in Fig. 1.An automatic displacement pump was used to inject water and oil from two different cylinders into the model reservoir.The resulting fluid flow from the model was subsequently separated by a liquid-gas separator.The flow rate and pressure were monitored via several different sensors located at the inlet and outlet of the model.Two gas dryers were used to dry the injected and separated CO 2 gas (that is, before and after injection) to protect the flow rate sensors.The data were recorded using a specially designed measurement system.A number of artificiallyconsolidated sandstone models were used to simulate thick reservoirs.These models were initially filled with quartz sand (80-200 mesh) followed by high pressure compaction for 48 h.A series of such models was made for different tests, each 500 mm in length and 250 mm in height but varying in thickness from 40 to 50 mm.The upper and lower part of each model (both of which were half the overall model height) functioned as the low and high permeability layers, respectively, in order to simulate thick, positive rhythm reservoirs (Fig. 2(a)).Each model was encapsulated within a 20 mm thick coating of a special epoxy resin for structural reinforcement and to ensure that the surrounding region was impermeable (Fig. 2(b)).
The appearance and physical properties of each of these sandstone models were quite similar to the natural sandstone found in the oilfields of eastern China (Table 1).Microscopy images (Fig. 3) show that the pore structural characteristics of the artificially-consolidated sandstone were also similar to that of the reservoir sandstone, and had the same proximate porosity and permeability as the actual sandstone (32.9% porosity and 1740 md permeability in the case of material from the Juggar Basin, as an example).The capillary curve of the artificial sandstone also indicated that the pore distribution of this material approximated that of natural sandstone.
Two simulated vertical wells with lengths of 250 mm and diameters of 8 mm were set at both ends of each model.A 6 mm diameter copper tube perforated at 10 mm intervals was placed in each well to simulate liner completion.The simulated oil used in the experiments was colourless, transparent, liquid paraffin with a molecular weight of 218.7 g mole -1 , a density of 0.8356 g cm -3 and a viscosity of 11.1 mPa•s at 25°C.Distilled water containing 5 wt% NaI (to enhance the contrast in the CT images) was used in all trials.The density and viscosity of the water were 1.040 g cm -3 and 0.894 mPa•s at 25°C, respectively.

Experimental Set up and Procedure
Using six artificial consolidated sandstone models with different physical properties, two series (Ex1 and Ex2), and nine groups (Ex1-1 to Ex1-4 and Ex2-1 to Ex2-5, Ex1-4 shared data with Ex2-1) of experiments were conducted, varying the injection pressure and permeability ratio.The details of the models and the experimental trials are listed in Table 2.Each experimental trial can be divided into three main stages, including preparation, water flooding and gas flooding.To map out the fluid transport and distribution in the model during each stage, high resolution (0.7 × 0.7 mm) CT images of the model cross-sections from the   inlet to the outlet were taken at a regular interval of about 7.5 mm along the model length (thus giving a total of 60 images for every scan).The details of the experimental procedures during each stage are as follows: (1) Preparation.
The artificial consolidated sandstone model was cleaned and dried after simulated well drilling and liner assembly, then sealed in epoxy resin.CT scanning was conducted after holding the model under vacuum for 24 h.The simulated oil was then absorbed into the structure, followed by a further CT scan of the oil-saturated model.The absorbed oil volume was taken as equal to the pore volume so as to calculate the porosity of the model.(2) Waterflooding test.
After the preparation was complete, water was pumped into the model at a designated flow rate of 5 mL min -1 until the water content in the outgoing flow reached 98%.
During water injection, CT images of the model showing the distribution of oil and water in the model were recorded at regular time intervals.(3) Gas flooding test.After water flooding, gaseous CO 2 was injected into the model at a designated injection pressure (10-100 kPa) until no obvious changes in the production could be observed.The injection pressure was monitored and precisely adjusted to avoid any significant fluctuations.During these trials, CT scanning was conducted frequently, initially at intervals of 5 min and then at half hour intervals after gas breakthrough.The dynamic injection pressure (Yang et al., 2015), inlet and outlet gas flow rates, and the production of oil and water were also monitored and measured continuously.

Waterflooding
Following the initiation of waterflooding, the model went through a water-free production period, during which no water was produced and the oil recovery increased rapidly.Following the injection of approximately 0.2 pore volume (PV) of water, the water-free production period ended and a water-oil production period began (Fig. 4(d)).The resulting increase in oil recovery eventually began to plateau, while the water content in the outgoing flow increased dramatically.Both the oil and water production rates began to change much more gradually over time at water levels above 70%.The associated CT images demonstrate that the lower, high permeability layer was swept much earlier than the upper (Fig. 4(a)), low permeability layer.Even within the high permeability layer, the sweeping efficiency in the lower part was much greater than in the upper part (Fig. 4(b)).At the end of the trial, the high permeability layer was almost completely occupied by water, whereas only half the low permeability layer contained water (Fig. 4(c)).The oil saturation was also greatly decreased in the zones that had been swept with the water (Fig. 5).Viscous fingering (Yang et al., 2011) could be observed in both layers, but was more prominent in the high permeability layer.

Gas Flooding
Because the injection pressure applied during these trials was much lower than the MMP, the CO 2 was maintained in a gaseous state throughout each test.In addition, under the experimental pressure and temperature conditions, the CO 2 solubility (Soong et al., 2014;Jean et al., 2015) in water was negligible.When the breakthrough of water occurred, the gas flooding was carried out.After gas injection was initiated, the production of both oil and water increased rapidly.The amount of product obtained during this initial period accounted for more than 60% of the entire production from gas flooding.In contrast to the results obtained from waterflooding, the water content in the outgoing flow continually declined until no water could be observed in the effluent.The model subsequently entered a water-free production period during which the oil recovery accumulated very slowly (Fig. 6   .When CO 2 began to enter the high permeability layer, approximately two thirds of the low permeability layer had already been swept, and the majority of the residual oil and water remaining after waterflooding had been displaced (Fig. 6(c)).This stage corresponded to the water-oil production period of gas flooding.Following this period, the CT values of the images obtained from the model tended to be relatively low, demonstrating that the pores were largely empty (Fig. 6(d)).
The CT images continued to exhibit a bright blue block in the front part of the model within both the low and high permeability layers, indicating a much drier zone with very low oil and water saturation (Fig. 6(e)).The anterior border of this dry zone in the high permeability layer was slightly ahead of that in the low permeability layer, although both were irregular.This stage corresponded to the water-free production period of gas flooding.At the completion of each trial, approximately 2000 PV of CO 2 (under the experimental pressure and temperature conditions) had been injected, but only one third of the model had transitioned into a dry zone.Considering the cost of CO 2 and the time required to inject the gas, this technique therefore appears to exhibit minimal efficiency with regard to oil recovery.Although the pressure difference between inlet and outlet during each test was relatively stable, the inlet gas flow increased gradually up to a constant rate of 1000 to 1200 mL min -1 , after the injection volume reached 500 PV (Fig. 7(a)).The ratio of the inlet to outlet gas flows fluctuated significantly at first and then decreased to close to unity.The ratio of the cumulative inlet to outlet gas flows was in the range of 1.02 to 1.01 (Fig. 7(b)), and from 1% to 2% of the initial CO 2 remained in the model.It is observed that the applied pressure difference is proportional to the cumulative injected CO 2 mass, which is consistent with in-situ numerical results (Jen et al., 2017;Wang et al., 2017).

Oil Recovery
The oil recoveries obtained from waterflooding trials ranged from 40% to 65%, while the majority of gas flooding tests gave recoveries in the range of 15%-25%, with the lowest and highest values being 9.35% and 30.24% (Fig. 8), respectively.These data demonstrate that CO 2 flooding after waterflooding can improve oil recovery from thick oil layers with positive rhythms.

Effect of Heterogeneity on Oil Recovery
The oil recovery from waterflooding decreased as the permeability differential or permeability variation coefficient increased.This result is consistent with experimental results  previously published (Chun et al., 1995;Yang and Gu, 2005;Yang et al., 2005).Under the present laboratory conditions, when the permeability differential was less than 2md or the permeability variation coefficient was below 0.2, the oil recovery decreased rapidly as the model became more heterogeneous.After these respective turning points of 2 or 0.2, the effect of heterogeneity was not as obvious (Fig. 9).In contrast to the data obtained with waterflooding, oil recovery during gas flooding was found to increase with increasing heterogeneity.Distinct changes in the data trends were evident at a permeability differential of 2 md and a permeability variation coefficient of 0.2 (Fig. 10).Following these transition points, the data plots were found to become flat.During the water-free production period in the gas flooding trials, peak oil recovery also appeared at the same turning points.

Effects of the Pressure Differential on Oil Recovery
The oil recovery pressure differential plots were found to be parabolic (Fig. 10).With increasing injection pressure, the recovery initially increased gradually and then decreased (Fig. 11).The phenomena could be interpreted by physical backgrounds.As injection pressure increased, the pressure gradient was elevated.Thus, the flow velocity increased, which accelerated the migration of the fluid-fluid interface.Therefore, the oil recovery initially increased.When the injection pressure further increased, the flow speed was also increased, the breakthrough of CO 2 gas firstly occurred.Then CO 2 mass occupied some sections of the outlet flow areas (Yang et al., 2012).This fact had strong effects on oil recovery characterized by a decoupled phase movement of the oil and CO 2 .Based on fitting of the experimental results, there is an optimum pressure differential that maximizes oil recovery under specific inhomogeneous conditions.An artificial neural network (Dickson et al., 2006) indicated that a given reservoir will exhibit maximum recovery under a specific set of conditions during immiscible flooding, which agrees with the current experimental results.Throughout the oil-water production stage of gas flooding, the relationship between recovery and pressure differential was similar to that of the total recovery and pressure differential.In contrast, during the water-free production period, the recovery was found to simply increase along with the pressure differential (Fig. 11).

The Stages of Gas Flooding
The CT images and production data suggest that two entirely different stages occurred during gas flooding: oilwater and water-free production.The displacement processes and mechanisms associated with both stages are discussed in this section, using the Ex2-5 trial as a typical example.

(a) Water-oil Production
During this stage, the residual water and oil remaining after waterflooding were rapidly displaced on a large scale.Because the CO 2 was not miscible with oil during the gas injection process, three phases (oil, water and gaseous CO 2 ) were always present in the model.Therefore, buoyancy, capillary and driving forces were acting on the injected CO 2 .According to previously published results, the interfacial tension between CO 2 and water is typically in the range of 20-70 N m -1 (Chang and Chen, 2005;Sutjiadi-Sia et al., 2008;Georgiadis et al., 2010a, b;Wang et al., 2011;Yang et al., 2011;Wang et al., 2017), and the contact angle of CO 2 on a quartz surface in a CO 2 -brine-quartz system is approximately 65-80° (Sarmadivaleh et al., 2015;Al-Yaseri et al., 2016).Mercury penetration experiments were conducted to obtain the capillary force of mercury in the pore throats of the experimental model, in order to determine the CO 2 capillary force (Plug et al., 2008) where σ Hg = 480 mN m -1 is the mercury interfacial tension, θ Hg = 140° is the mercury contact angle, P Hg is the mercury capillary force, σ CO2 is the interfacial tension of CO 2 at the experimental temperature and pressure, and θ CO2 is the contact angle of CO 2 at the experimental temperature and pressure.The ratio between the capillary force, buoyancy and driving force for CO 2 was determined to be approximately (1-17): 36: (35-357), indicating that both buoyancy and driving force were important (Yang et al., 2012;Jen et al., 2017;Wang et al., 2017).The driving force continuously added to the total energy in the model to force CO 2 forward.
In addition, buoyancy prompted CO 2 to gradually migrate upward and to pool when not immediately discharged, such as when blocked by liquid (Yang et al., 2011;Jen et al., 2017).When this pooled CO 2 reached a specific amount, it flowed downward, and this expansion of the sweep area both horizontally and vertically increased the contact between CO 2 , oil and water until almost all the pore fluid was eventually displaced (Fig. 12).Under the experimental pressure and temperature used in the present work, the injected CO 2 was compressible.The pore framework and internal fluid served to obstruct and compress the CO 2 to a certain extent, thus increasing the internal gas pressure (Soong et al., 2014;Yang et al., 2014;Wang et al., 2017).When this pressure exceeded the resistance, the CO 2 began to expand and to move around the fluid until the resistance again blocked the flow (Jen et al., 2017).Repetitions of this cycle resulted in flow rate and pressure pulses (Yang et al., 2015(Yang et al., , 2016)).The effect of confinement and the presence of a shear gradient in the flow geometry, the flow-induced deformation of fluid interfaces, deformation-flow interaction play a key role in the ternary system flooding (Yang et al., 2012).Both shear and extensional effects (Yang et al., 2011) take place in the flow behavior of CO 2 flooding.Mechanisms of buoyancy and surface tension variation with concentration also contribute to the flow behaviors during CO 2 flooding (Yang et al., 2014).
(b)Water-Free production A considerable amount of additional oil was obtained during this stage.The CT images showed that areas with high gas saturation appeared near the inlet and slowly expanded.Prior work has demonstrated that strong interactions between oil molecules and mineral surfaces can lead to the adsorption of oil on the rock surfaces in the absence of water.To simplify the experimental preparation, simulated irreducible water was omitted during the current study.Therefore, a dry pore framework was initially contacted by the oil.The application of a vacuum also assisted in filling the small pores and throats (which are difficult to sweep with water) with oil, and increased the degree of contact and attachment between the oil and the model structure.The oil in these regions accounted for a portion of the residual oil after waterflooding.During water-flooding, this oil also prevented the injected water from contacting and adhering to the pore surfaces.Therefore, the water was easily displaced by the CO 2 .At the point at which the effluent contained almost no water, it can be assumed that, without initial plugging of the pores by water, the oil adhering to the pore surfaces was exposed to the CO 2 , stripped from the pore surfaces, and carried out by the continuous CO 2 flow.In addition, the highly mobile gaseous CO 2 was also  able to more easily sweep the small pores and throats (Soong et al., 2014;Jean et al., 2015;Wang et al., 2017), which also contributed to the water-free recovery of oil.However, because of the relatively small amount of residual oil, the oil recovery during the water-free production stage was much lower than that during the water-oil production stage, representing only about 5% extra oil recovery.This process is therefore not overly efficient when considering the amount of time and energy required to inject the CO 2 (Huang and Tan, 2014;Li and Li, 2014).It should be noted that the proposed displacement process and mechanism during this stage are preliminary, and that the adsorption and desorption (Jean et al., 2015) of liquids on solid surfaces is a complex surface chemistry process.It pointed that mineral dissolution and mineral precipitation can occur in the host deposit altering its characteristics for CO 2 storage over time (Soong et al., 2014;Jen et al., 2017).In the presence of water, the adsorption characteristics of the pore surfaces and mineral dissolution may become more complex, further studies will be necessary to fully understand the relevant process.

CONCLUSIONS
The immiscible CO 2 flooding after water flooding was examined, employing consolidated sandstone models that simulated heterogeneous, positive rhythm, thick oil reservoirs.The behaviors of various flowing substances (gaseous CO 2 , water and oil) after CO 2 injection were determined based on X-ray CT.The experimental results indicate that CO 2 can efficiently displace the oil remaining in the upper part of the model after water flooding.Immiscible CO 2 flooding is therefore an effective method for enhancing oil recovery after waterflooding, with an additional oil recovery of more than 10%.The experimental data also suggest that both the reservoir heterogeneity and the pressure differential affect oil recovery and CO 2 migration.The heterogeneity had a remarkable impact on oil recovery when the permeability differential between layers was lower than 2md or the permeability variation coefficient between layers was less than 0.2.The decoupled phase movement of the oil and CO 2 had strong effects on oil recovery.The experimental investigation may significantly advance understanding in multiphase flow physics associated with CO 2 geological storage and utilization.The large models used in this study made it difficult to suitably contain high pressure levels.
Future experiments under high-pressure conditions are needed to obtain additional data.

Fig. 3 .
Fig. 3. Microscopy images of the artificial consolidated sandstone and actual reservoir sandstone from the Juggar Basin: (a) artificial consolidated sandstone, in which the purple areas are pores and the grey and black areas are the solid frame, and (b) material from an in-situ reservoir sandstone, in which the blue areas shows pores and tan and black areas show the solid frame.
Fig. 4. The results of water flooding trials: after injecting water at (a) 0.29 PV, (b) 1 PV and (c) 4 PV, and (d) plots showing the oil recovery and water content during water flooding.

Fig. 5 .
Fig. 5.The oil saturation distribution in the test model during water flooding as determined from CT images of: (a) the lower high permeability layer and (b) the upper low permeability layer.

Fig. 6 .
Fig. 6.The data obtained from gas flooding after injecting CO 2 for (a) 10 min, (b) 40 min, (c) 100 min, (d) 1080 min and (e) 3142 min, and (f) a plot of the oil recovery and water content during gas flooding.

Fig. 7 .
Fig. 7. Inlet and outlet monitoring data during gas flooding: (a) changes in the instantaneous inlet flow rate and inlet-outlet pressure difference during gas flooding and (b) variations in the ratio of the inlet and outlet flow rates during gas flooding.

Fig. 8 .
Fig. 8.The oil recovery values from various trials.

Fig. 9 .
Fig. 9.The relationships between reservoir heterogeneity and oil recovery during water flooding: (a) oil recovery as a function of the permeability ratio between layers and (b) oil recovery as a function of the permeability variation coefficient between layers.

Fig. 10 .
Fig. 10.The relationships between reservoir heterogeneity and oil recovery during gas flooding: (a) oil recovery as a function of the permeability ratio between layers and (b) oil recovery as a function of the permeability variation coefficient between layers.

Fig. 11 .
Fig. 11.The relationship between the pressure differential and oil recovery during gas flooding.

Table 1 .
Properties of the artificially-consolidated sandstone models.

Table 2 .
Experimental conditions and corresponding model properties.