Abstract
Porous polyimide (PPI) is a key material for the oil supply parts in the bearing units of the momentum wheels and reaction wheels. The permeability of PPI determines the oil supply rate in the bearing contact area, thus affecting the performance of the bearing units and wheels. However, few studies have focused on the effect of alternating stress on the permeability of PPI in running bearing units. Here, a study on the stress-dependent permeability of PPI under stress loading and unloading is conducted. After experimental and theoretical analyses, permeability is reduced at most by 37% when the effective stress σeff is between 0.75 and 26.45 MPa. The elastic-structural deformation of capillary cross-section and tortuosity sensitivity lead to a negative correlation between σeff and permeability, and permeability hysteresis. When σeff is greater than 15 MPa and the pressure gradient is greater than 40 MPa/m, the permeability decreases and then increases during stress unloading because of the blocking and unblocking of detached particles. A fitting formula for the tortuosity variation caused by detached particles is proposed to describe this permeability behavior. This study contributes to the application and improvement of PPI in bearing units and similar working conditions.


Introduction
A growing number of spacecraft need to maintain high control accuracy over a long service life. Momentum wheel or reaction wheel is often used as an attitude control actuator for long-lived spacecraft because of its high control accuracy, long duration and no need for propellant. Since the bearing unit is the main source of failure for the momentum wheel and reaction wheel, the long-term reliable operation of the bearing unit is critical for the spacecraft to perform its mission. To prevent lubrication failure of the bearing unit, the lubrication system needs to supply a small amount of oil to the bearings over a long period of time. This ensures that the bearing unit maintains a low and stable friction torque. Porous polyimide (PPI) is an ideal material for the lubrication system owing to its advantages, such as excellent mechanical properties, low density and thermal stability. , PPI is manufactured into the oil reservoir and the bearing retainer in the lubrication system. They are saturated with oil before the bearing unit is assembled, and they supply oil to the bearings by centrifugal force and heat when the wheels are in operation. This design can satisfy most of the lubrication needs of the bearing unit. Hence, PPI is widely used in lubrication system.
To improve the performance of PPI, many researchers have studied the tribological properties, − oil supply behavior, , oil exchange behavior and thermal stability of PPI. Current research suggests that the flow of oil in PPI is a seepage process, so Darcy’s law can be used to analysis the oil supply behavior. According to Darcy’s law, permeability is a key indicator to measure the oil supply capacity of PPI. Under actual working conditions, the oil reservoir and retainer are subjected to alternating stress due to the alternating temperature and the repeated impact. Furthermore, numerous studies have shown that the permeability of porous materials changes with stress, a phenomenon known as stress-dependent permeability. This indicates that the permeability of PPI may change during bearing running, which in turn affects the oil supply rate in the bearing contact area and the performance of the bearing units. However, fewer studies focus on the effect of stress on the permeability of PPI. An adequate study of the stress-dependent permeability of PPI is an important means of clarifying the change in the oil supply capacity of the oil reservoir and retainer under actual working conditions. It is of great significance for the evaluation, application and optimization of the oil supply capacity of PPI, and for ensuring the long-term reliable operation of the bearing units. It is also instructive for the application of PPI under working conditions similar to those of oil reservoir and retainer.
In the field of porous polymers, fewer scholars have studied the stress-dependent permeability of porous polymers under alternating stress. However, the stress-dependent permeability of rocks has been extensively studied. Since both rock and PPI are porous media, the methods and theories used to study the stress-dependent permeability of rocks are of important reference value for studying the stress-dependent permeability of PPI. Due to the differences in working conditions and material properties between PPI and rock, existing mechanisms and theories may not be fully applicable to PPI. Therefore, it is necessary to study the stress-dependent permeability of PPI.
Research into the stress-dependent permeability of rocks can be divided into experimental and theoretical studies. Experimental studies mainly observe the effect of stress on permeability by building a physical simulation system. Stress is determined by confining pressure, pore pressure and temperature. The study results show that the stress-dependent permeability of rocks directly affects oil/gas production, , geothermal production, the capacity of underground gas storage, etc. Effective stress is a key indicator for analyzing stress-dependent permeability. The mechanical behavior of the porous media microstructure is an important basis for revealing stress sensitivity. , Theoretical studies mainly include fitting models and physical models. Among these, although fitting models , can quantify stress-dependent permeability, they are lacking in the physical characteristics of porous media. Compared to fitting models, physical models mainly consider the elastic-structural deformation − and elasto-plastic deformation of the capillary cross-section, thus better reflecting the physical characteristics of porous media. As the research progressed, the physical models can describe not only stress-dependent permeability under alternative stress but also porosity stress sensitivity and permeability–porosity relationships. This makes the model reflect the physical characteristics of porous media more comprehensively and accurately.
Here, the stress-dependent permeability of PPI is studied experimentally and theoretically. First, physical quantities of PPI, such as permeability, porosity and pore size distribution, are measured to clarify the basic properties of the experimental samples. The representativeness of the PPI samples is confirmed by comparing their basic properties with those in the published literature. Second, the stress-dependent permeability experiment under a single stress cycle is conducted by a physical simulation system under the working conditions of the bearing units. The mechanism of stress-dependent permeability is revealed by analyzing the experimental results and the PPI microstructure. Third, the experimental results of stress-dependent permeability and stress-dependent porosity are analyzed using multiple theoretical models to further identify the influencing factors of stress sensitivity. A fitting formula for tortuosity variation is proposed and the Xu’s model is improved to describe the effect of detached particles on permeability. This theoretical analysis and improvement are of great significance for quantifying, comparing and improving the stress-dependent permeability of PPI.
Experimental Section
Materials
PPI samples are cylinders with a diameter of 2.5 ± 0.02 cm and a length of 5.0 ± 0.02 cm. There are 18 samples in total, numbered #1–#18. The samples are purchased from Lanzhou Zhongke Kailu Lubrication and Protection Technology Co., Ltd. The samples are made of polyimide (PI) particles with a diameter of 30 to 40 μm by cold pressing and then sintering technique. , The solid PI samples made from the same PI particles as the PPI are used for comparison. There are two PI samples in total, numbered #19 and #20. Test details for porosity ϕ and permeability k for all PPI samples, pore size distribution of PPI sample, mechanical properties of PPI and PI samples and structural characterization are provided in Appendix A.
The poly alpha-olefin (PAO) lubricant 4129, which is one of the most commonly used lubricants in the bearing units, is used in experiments on stress-dependent permeability and stress-dependent porosity. PAO 4129 is purchased from the Sinopec Chongqing Yiping Advanced Lubricating Oil Company and the viscosity equation for PAO 4129 is
| 1 |
where μoT is viscosity in Pa·s and T is temperature between 20 and 80 °C.
Stress-Dependent Permeability
The equipment for testing stress-dependent permeability is self-developed. The equipment consists of three parts: the fluid injection system, the fluid displacement system, and the fluid measurement system. A schematic diagram of the equipment is shown in Figure . The fluid injection system consists of a twin-cylinder pump and a piston container. This system pumps the oil into the sample under a constant inlet pressure. The inlet pressure ranges from 0.5 to 65 MPa. The fluid displacement system includes a core holder, a constant pressure pump, a pressure gauge and a thermotank, which are used to hold the PPI sample, provide a confining pressure, measure the inlet pressure, and maintain temperature, respectively. The confining pressure range is between 0 and 70 MPa, and the temperature range is between 25 and 150 °C. The fluid measurement system uses a 10 mL measuring cylinder and a timer to measure the outlet flow rate. The uncertainty of the equipment for measuring the permeability is 0.83%.
1.
Schematic diagram of the equipment for testing stress-dependent permeability.
The experimental procedure is as follows. (1) Load the sample into the core holder. Set the confining pressure, inlet pressure, outlet pressure and temperature as the experimental conditions. The outlet pressure is 1 atm. (2) After oil appears at the outlet, wait 30 min for the seepage to stabilize. Measure the export volume during the last 5 min and then calculate the permeability according to Darcy’s law
| 2 |
where k σ is permeability considering effective stress; Q is the outlet flow rate; A σ and L σ are the cross-sectional area and length of a sample considering effective stress, respectively; p in and p out are the inlet pressure and outlet pressure, respectively. A σ and L σ are calculated according to Hooke’s law
| 3 |
where A 0 and L 0 are the cross-sectional area and length of a sample in a stress-free state, respectively; E p and νp are Young’s modulus and Poisson’s ratio of PPI, respectively; σeff is the effective stress. σeff is calculated by Terzaghi equation
| 4 |
where σ is the confining pressure; p is the mean relative pore pressure; αp is the thermal stress coefficient of PPI; T o is reference temperature, which is equal to 25 °C. The equation for αp is
| 5 |
(3) Increase the confining pressure step by step. Repeat step 2 at each confining pressure value. (4) After the confining pressure reaches the maximum value set in the experiment, decrease the confining pressure step by step. Repeat step 2 at each confining pressure value. The above is the whole process. After the experiment, the stress-dependent permeability can be analyzed by the relationship between k σ and σeff.
The experimental conditions are set based on the working conditions of the PPI retainer and cover the working conditions of the oil reservoir. The working conditions include contact pressure and temperature. First, the maximum contact pressure is around 5 MPa. This is calculated using the bearing explicit dynamic analysis model, taking the B7004C angular contact ball bearing bearing as an example. B7004C is a bearing commonly used in momentum wheels and reaction wheels. The working conditions for B7004C are as follows. The radial load is zero and the inner ring is fixed. The axial load ranges from 50 to 200 N. For the reaction wheel, the outer ring speed ranges from −2000 to 2000 rpm, while the speed of the momentum wheel ranges from 0 to 8000 rpm. Second, the allowable temperature of the wheel bearing is between −20 and 80 °C. Therefore, the experimental conditions of stress-dependent permeability are shown in Table . In Table , the range of the confining pressure covers the range of contact pressure of the retainer. The range of temperature is in the range of allowable temperature. The main focus is on the effect of temperature on permeability in the range from room temperature to high temperature. The inlet pressures are 0.5 MPa, 2 MPa, 4 and 6 MPa when the minimum confining pressures are 1 MPa, 3 MPa, 5 and 7 MPa, respectively. This is because the inlet pressure must be less than the confining pressure when measuring the stress-dependent permeability. According to eq , the maximum σeff for the experiment on stress-dependent permeability is 26.45 MPa. The experimental results of mechanical properties show that the PPI does not reach the yield point even at 60 MPa. This demonstrates that the experiment on stress-dependent permeability is carried out in the elastic deformation range of PPI.
1. Experimental Conditions of Stress-Dependent Permeability.
| experiment number | sample number | confining pressure (MPa) | inlet pressure (MPa) | temperature (°C) |
|---|---|---|---|---|
| SP1 | #4 | 25 | ||
| SP2 | #5 | 1, 2, 3, 4, 5, 6, 7, 8 | 0.5 | 50 |
| SP3 | #6 | 80 | ||
| SP4 | #7 | 25 | ||
| SP5 | #8 | 3, 4.3, 5.5, 6.8, 8 | 2 | 50 |
| SP6 | #9 | 80 | ||
| SP7 | #10 | 25 | ||
| SP8 | #11 | 5, 5.8, 6.5, 7.3, 8 | 4 | 50 |
| SP9 | #12 | 80 | ||
| SP10 | #13 | 25 | ||
| SP11 | #14 | 7, 7.3, 7.5, 7.8, 8 | 6 | 50 |
| SP12 | #15 | 80 |
Stress-Dependent Porosity
First, the porosity ϕσ considering the effective stress is
| 6 |
where V pσ is the pore volume considering the effective stress. V pσ can be written as the sum of two parts
| 7 |
where V po is the pore volume at the initial effective stress and ΔV pσ is the change in pore volume with effective stress. ΔV pσ is 0 at the initial effective stress. Therefore, to obtain the ϕσ, it is necessary to obtain V po and ΔV pσ. Compared to Figure , the equipment for testing ΔV pσ has an additional valve at the outlet to control whether the fluid flows out. The twin-cylinder pump is replaced by a metering pump for measuring ΔV pσ. Compared to the porosity measurement in Appendix A, the stress-dependent porosity experiment uses PAO 4129 instead of gas to more closer to the actual working conditions. The experimental procedure is as follows. (1) Load the standard metal core into the core holder. Set the confining pressure, pore pressure and temperature as the experimental conditions. (2) Wait 30 min for the pressure and chamber volume of the metering pump to reach a steady state, and then record the chamber volume. (3) Increase the confining pressure step by step. Repeat step 2 at each confining pressure value. (4) Calculate the volume change based on the chamber volume in steps 1–3. (5) Replace the standard metal core with a sample and repeat steps 1–3. (6) Calculate the volume change based on the chamber volume in step 5. (7) ΔV pσ is the volume change in step 5 minus the volume change in step 4. The above is the whole process. The experimental conditions are as follows. Samples #16, #17 and #18 are tested at 25 °C, 50 and 80 °C. At each temperature level, the pore pressure remains constant at 0.5 MPa. The confining pressure increases from 1 to 8 MPa (in increments of 1 MPa).
The settings of V po are as follows. At 25 °C, the V po is equal to the pore volume of a sample in a stress-free state, because the initial effective stress is only 0.5 MPa. The V po at 50 °C is set based on the experimental results at 25 °C and the procedure is as follows. (1) Calculate the ϕσ/ϕ based on the experimental results at 25 °C and eqs and . (2) Calculate the ϕσ/ϕ at the initial effective stress at 50 °C by linear fitting of ϕσ/ϕ at 25 °C. (3) Calculate the V po at 50 °C based on the calculated ϕσ/ϕ at the initial effective stress at 50 °C and eqs and . The above is the whole process. Similarly, the V po at 80 °C is set based on the experimental results at 50 °C.
Results and Discussion
Discussion of Basic Properties
Figure shows the porosity ϕ and permeability k of all PPI samples. From Figure , the maximum, minimum, mean and standard deviation of porosity are 26.65%, 20.81%, 23.71% and 1.5%, respectively. After analyzing the five porosities from Marchetti et al., Zhang et al. and Qiu et al. in Figure , the maximum, minimum, mean and standard deviation of porosity are 27%, 19%, 23.4% and 3.21%, respectively. It is seen that the porosities of PPI samples are similar to those in the published literature. In Figure , the maximum, minimum, mean and standard deviation of permeability are 7.52 mD, 4.58 mD, 6.43 mD and 0.85 mD, respectively. Since few relevant studies give the permeability of PPI, only the permeability of 2.1 mD in Marchetti et al. is used for comparison. It is seen that although the permeabilities of PPI samples are all greater than the permeability in Marchetti et al., they are still of the same order of magnitude.
2.

Porosity and permeability of PPI samples.
The pore size distribution of PPI sample is shown in Figure a. λp is the pore diameter. From Figure a, the maximum, minimum and average pore diameters are 33.2 μm, 6.6 × 10–3 and 5.54 μm, respectively. In particular, pores with a diameter of 10.33 μm have the highest distribution frequency. Compared with existing studies, the average pore diameter λavep in Marchetti et al. is smaller than the λavep of the PPI sample. The pore diameters with the highest distribution frequency λhdfp in Zhang et al. and Qiu et al. are also smaller than the λhdfp of the PPI sample.
3.
Pore size distribution of PPI obtained by mercury injection method (a) and fitted with fractal theory (b).
Since the pore size distribution of porous media usually has fractal characteristics, fractal theory is often used to describe the pore size distribution of porous media. However, few studies have mentioned the fractal characteristics of PPI. Therefore, the fractal scaling model is used to fit the pore size distribution of the PPI to confirm whether the PPI has fractal characteristics. In the fractal scaling model, the cumulative size distribution of pores should follow the following scaling law
| 8 |
where l is a certain pore diameter; N p is the number of pores whose diameters are equal to or larger than λp; λp,max is the maximum pore diameter; C is the number of pores with a diameter of λp,max; D f is the fractal dimension in two-dimension and 0 < D f < 2. The fitting result of the pore distribution of PPI using eq is shown in Figure b. From Figure b, the goodness of fit R 2 is 0.96 which indicates a good fit. D f and C of PPI are 1.8 and 104.03, respectively. In addition, D f can also be calculated by
| 9 |
where λp,min is the minimum pore diameter. D f is 1.82 by substituting ϕ, λp,min and λp,max of sample #1 into eq . The D f calculated using eq is similar to the D f obtained by fitting the pore size distribution using eq . The above analysis shows that the pore size distribution of PPI has fractal characteristics. The pore size distribution of PPI can be described using the fractal scale model, and the stress-dependent permeability of PPI can be analyzed using the model based on fractal theory. Second, fractal characteristics are also important for material performance.
The mechanical properties of PI and PPI are as follows. Young’s modulus E s, Poisson’s ratio νs and thermal expansion coefficient αTs of PI are 3.05 GPa, 0.42 and 5.47 × 10–5 °C–1, respectively. Young’s modulus E p, Poisson’s ratio νp and thermal expansion coefficient αTp of PPI are 1.69 GPa, 0.38 and 4.88 × 10–5 °C–1, respectively. Compared to the mechanical properties of PI, the Young’s modulus, Poisson’s ratio and thermal expansion coefficient of PPI are reduced. In Wypych, E s is between 1.2 and 3.8 GPa, νs is between 0.15 and 0.42, and αTs is between 2 × 10–5 and 5.4 × 10–5 °C–1. The mechanical properties of PI samples are basically the same as those of PI in Wypych. Since few relevant studies give the Poisson’s ratio of PPI, only Young’s modulus and thermal expansion coefficient are used for comparison. In Zhang et al., Young’s modulus of PPI PIB-1 and PIB-2 are 2.72 and 3.71 GPa, respectively. They are similar to Young’s modulus of sample #2. In Marchetti et al., the thermal expansion coefficient of PPI is 4.2 × 10–5 °C–1, which is similar to that of sample #3.
The cross-sectional morphology of the PPI is shown in Figure . From Figure , the particles are firmly bonded to each other, and the voids in PPI are pores rather than fractures. Some of the particles have a diameter similar to that of the raw material particles (30 μm–40 μm), while the remaining particles have a smaller diameter. In combination with the preparation process of PPI, it is seen that some particles are not deformed after processing (undeformed PI particles), while the remaining particles are deformed into a combination of small particles (deformed PI particles). A further magnification of the cross-section shows that the pores of the PPI are surrounded by particles. This pore structure is the same as that in Zhang et al. , and Wang et al. After connecting the centers of the particles surrounding the pores, it is found that the pore structures have triangular, quadrilateral and pentagonal structures. The pore diameters in Figure are similar to the pore diameters in Figure a. This demonstrates the accuracy of the pore size distribution obtained by mercury intrusion.
4.
SEM micrograph of the cross-section of PPI.
After comparison and discussion, it can be seen that the basic properties of the PPI samples are similar to most of the basic properties of PPI in the published literature. This indicates that the stress-dependent permeability of the PPI samples can, to some extent, represent the stress-dependent permeability of existing PPI. Research based on the PPI samples is important for revealing the mechanism of stress-dependent permeability and evaluating stress-dependent permeability of existing PPI.
Experimental Analysis of Stress-Dependent Permeability
Experimental results of the stress-dependent permeability of PPI are shown in Figure . After analyzing the extreme values in Figure , it is found that normalized permeability is 0.99 at the minimum effective stress 0.75 MPa. It demonstrates that the lubricant permeability of PPI is close to the permeability of PPI, and the interaction between PI and PAO 4129 has a minor effect on permeability. The normalized permeability reaches the minimum value of 0.63 when the effective stress σeff is between 21 and 25 MPa. Some experimental results are more scattered in Figure , such as the results of experiment SP1. A similar phenomenon has been observed in rocks, but the mechanism of the phenomenon is incompletely studied. The possible reason for the PPI sample is as follows. The PI particles in the PPI samples are both deformed and undeformed, as shown in Figure . The heterogeneity in particle shape may lead to heterogeneity in pore structure. This reason causes the local pore deformations in the PPI samples and the interactions between the local pore deformations to have a random effect on the permeability under σeff, leading to a scatter in the normalized permeability.
5.
Experimental results of stress-dependent permeability.
According to the stress-dependent permeability behavior, Figure can be divided into two regions. Region I is less than 15 MPa and Region II is greater than 15 MPa. In Region I, permeability basicaly shows a decreasing trend during stress loading and an increasing trend during stress unloading (Type 1). In each set of experiments, permeability under stress loading is greater than permeability under stress unloading when σeff is the same. This means that PPI has permeability hysteresis under alternating σeff. The above results show that the stress-dependent permeability behavior of PPI in Region I is typical. The difference between Regions I and II is the changes in permeability under stress unloading. The stress-dependent permeability behavior in Region II under stress unloading can be divided into three types. Type 1: The permeability increases in experiments SP2 and SP3. Type 2: The permeability decreases in experiments SP9 and SP12. Type 3: The permeability first decreases and then increases in experiments SP5 and SP6. This indicates that the permeability changes in Region II under stress unloading vary from experiment to experiment.
To discuss the influence of σeff on permeability stress sensitivity and permeability hysteresis, the normalized permeability range (NPR) and the hysteresis loss of normalized permeability (HLNP) are used to quantify the permeability stress sensitivity and permeability hysteresis, respectively. Among these, the NPR is the difference between the maximum and minimum normalized permeabilities under stress loading or unloading. The NPR is selected because it can reflect the permeability stress sensitivity even if the stress-dependent permeability behavior are different. The HLNP is the difference between normalized permeabilities at the initial effective stress under stress loading and unloading. The NPR and the HLNP need to be further processed to eliminate the influence of the σeff range and retain the influence of the σeff magnitude. The processing method is to obtain the mean value of the NPR or the HLNP for experiments having the same temperature. This is because experiments with the same temperature have similar mean σeff, and experiments with the same inlet pressure have similar σeff ranges. After processing, the mean values of the NPR and the HLNP under different mean σeff are shown in Figure .
6.

Influence of mean effective stress on the normalized permeability range (NPR) and the hysteresis loss of normalized permeability (HLNP).
In Figure , the mean σeff of experiments under 25 °C, 50 and 80 °C are 4.4 MPa, 12.9 and 23.1 MPa, respectively. The larger the mean value of the HLNP, the greater the permeability hysteresis. The larger the mean value of the NPR, the greater the permeability stress sensitivity. According to Figure , permeability stress sensitivity under stress loading and permeability hysteresis decrease with increasing mean σeff. Permeability stress sensitivity under stress unloading first decreases and then increases with increasing σeff. This is because the stress-dependent permeability behavior in Region II under stress unloading has changed compared to that in Region I. The experimental results of stress-dependent permeability show that the permeability of the oil reservoir and retainer will decrease under alternating stress during the operation of the bearing units. This will cause a reduction in the oil supply rate in the bearing contact area and affect the performance of the bearing units. Therefore, the stress-dependent permeability of PPI should be considered in the design of bearing units to ensure the normal operation of the bearing units.
Figure shows the stress-dependent permeability mechanism of PPI based on the basic properties of PPI, experimental results of stress-dependent permeability and related theories. − In Figure , the pores in PPI are taken as capillary bundles. The capillary size distribution is the pore size distribution in eq . The factors leading to stress-dependent permeability can be divided into the factors of the capillary cross-section and the factors of the capillary length. The factors of the capillary cross-section include elastic deformation and structural deformation. The elastic deformation of capillary cross sections is caused by the elastic deformation of PI particles because the pores in a PPI are surrounded by PI particles. All the pore structures in a PPI have elastic deformation. The structural deformation of capillary cross sections is caused by the structural instability of pore structures such as the quadrilateral and pentagonal structures in the PPI. In Figure , the quadrangular structure is selected to represent the pore structures with structural instability. The triangular structure only has elastic deformation because triangles have good structural stability. Under stress loading, elastic and structural deformation together cause the pores to narrow, thus causing a decrease in permeability as the σeff increases. Under stress unloading, elastic deformation recovers but structural deformation does not recover. This results in the pores opening as the σeff decreases, thus causing an increase in permeability. However, the permeability cannot return to its original value, i.e. there is a permeability hysteresis under alternating σeff.
7.
Stress-dependent permeability mechanism of PPI.
The factors of the capillary length include tortuosity sensitivity and detached particles. Tortuosity sensitivity is the change in tortuosity under σeff. This is because, in a PPI, capillaries consist of pores that are connected by throats in multiple directions. Fluid meanders through capillaries in the PPI and tortuosity is used to describe the tortuosity of capillaries. Under σeff, connectivity changes due to the deformation of capillary cross sections, which in turn alters the flow path of fluid and induces changes in tortuosity. The effect of detached particles is analyzed as follows. The elastic-structural deformation and tortuosity sensitivity can explain the stress-dependent permeability behavior of PPI, except for the permeability trend under stress unloading in Region II. This indicates that there are still other factors that affect the stress-dependent permeability. Particle blockage is most likely to affect the stress-dependent permeability under stress unloading in Region II. This is because the lubricant color changed from colorless to yellow after the experiments SP5 and SP6. It suggests that the outlet lubricant contains PI since only PPI (yellow) and PAO 4129 (colorless) are used in the experiment. Therefore, particle blockage is likely to exist in PPI. To clarify the presence and mechanism of particle blockage, PI in the outlet lubricant is extracted and observed. The observation results are shown in Figure .
8.
Micrograph (a) and size distribution (b) of the detached particles.
First, PI in the outlet lubricant is in particle form, as shown in Figure a. Second, the statistical results for 100 particle diameters are shown in Figure b. In Figure b, the particle diameter ranges from 2 to 11 μm, with a larger proportion of particles having a diameter of 3–7 μm. The size of particles in the outlet lubricant is closer to the size of deformed particles in the PPI. Besides, no particles are found adsorbed on the pores in the PPI because the particles are firmly bonded to each other in Figure . Therefore, the particles in the outlet lubricant are detached particles from the deformed particles in the PPI. Compared to Figure a, the minimum diameter of detached particles is larger than the minimum diameter of pores. This indicates that some pores can be blocked by detached particles. Based on the above analysis, the source and size of the detached particles suggest that there is a particle blockage in the PPI.
The influence mechanism of detached particles on stress-dependent permeability of PPI is as follows. When the σeff is greater than or equal to the critical effective stress, some particles will detach from the solid skeleton under stress loading. However, the detached particles are not migrated by the lubricant because of the squeezing of the solid skeleton. Under stress unloading, the detached particles are released from the solid skeleton by the lubricant propulsion and the recovery of solid skeleton deformation, resulting in particle migration and particle blockage. The blockage causes a reduction in the flowable channels and a change in tortuosity. Permeability first decreases under stress unloading because the effect of blockage on permeability is greater than the effect of pores opening. As the σeff is further reduced, more detached particles are released, which in turn leads to a further reduction in permeability. When the pore diameters return to being larger than the detached particle diameters, the detached particles are migrated away from the blocked pores and the blocked pores are unblocked. Thus, permeability begins to increase with decreasing σeff. The above mechanism causes the permeability of SP5 and SP6 to first decrease and then increase during stress unloading. For SP9 and SP12, permeability only decreases during stress unloading because the σeff ranges of SP9 and SP12 are smaller than those of SP5 and SP6. The pore diameters do not return to larger than the detached particle diameters and unblocking does not occur. For SP2 and SP3, permeability increases under stress unloading because the pressure gradients of SP2 and SP3 (10 MPa/m) are smaller than the pressure gradients of SP5 and SP6 (40 MPa/m). The pressure gradient is the change in fluid pressure per unit length in the direction of fluid flow. The pressure gradient is calculated by (p in – p out)/L 0. Under a smaller pressure gradient, the force of the lubricant pushing against the detached particles cannot overcome the force between the detached particles and the solid skeleton. As a result, the detached particles cannot be released from the solid skeleton and blocking does not occur.
It is worth noting that the influence mechanism of particle blockage in PPI under alternating σeff is different from the mechanism in most existing studies. In Kozhevnikov et al. and Zhang et al., particle blockage occurs during stress loading. Therefore, particle blockage does not reduce the permeability under stress unloading. If the detached particles in PPI are also released during stress loading, the permeability under stress unloading will not decrease. However, the result of this hypothesis does not match the experimental results of PPI. This further indicates that the detached particles in PPI are released during stress unloading. One reason for the unique particle blockage mechanism of PPI is the blocking particles in PPI are different from those in other porous media. In Kozhevnikov et al., the blocking particles are the absorbed particles. When the pressure gradient exceeds a critical value, the absorbed particles are moved by the fluid, and particle blockage occurs during stress loading rather than stress unloading. In Zhang et al., the blocking particles are the crushed coals. Particle blockage occurs when the coals are crushed under stress loading. For PPI, the blocking particles are the particles detached from the solid skeleton and the detachment occurs because of internal squeezing of the solid skeleton. Therefore, even if the particles detach from the solid skeleton, they are constrained by the solid skeleton due to the increased internal squeezing under stress loading. Under stress unloading, internal squeezing decreases with decreasing σeff, which in turn leads to the release, migration and blockage of detached particles.
Theoretical Analysis of Stress-Dependent Permeability
To further clarify and quantify the stress-dependent permeability mechanism of PPI, Xu’s model, Tan’s model, Cao’s model and Huang’s model are used for theoretical analysis. Xu’s model considers the elastic-structural deformation of the capillary cross-section and tortuosity sensitivity. Tan’s model considers the elastic deformation of the capillary cross-section. Huang’s model considers the elasto-plastic deformation of the capillary cross-section. Cao’s model considers the elastic-structural deformation of the capillary cross-section. Different models calculate the same factors in distinct ways. Since the fractal scaling law in this paper has an additional coefficient C compared to the fractal scaling law in Xu’s model, the fractal scaling law in Xu’s model needs to be replaced with eq , and Xu’s model also needs to be rederived. The permeability k σ and porosity ϕσ in the rederived Xu’s model are
| 10 |
| 11 |
where g pσ is the elastic deformation term for capillary diameter; τσ is the tortuosity considering effective stress; h qσ is the structural deformation term of quadrilateral structures; γ is the proportion index of the quadrilateral structures; g sσ is the elastic deformation term for the diameter of solid particles; λsoA is the solid equivalent diameter in the two-dimensional cross-section at the initial effective stress; λp,max,o and λp,min,o are the maximum and minimum capillary diameters at the initial effective stress, respectively; and R λ = λp,min,o/λp,max,o. Detailed formulas for the parameters in eqs and are given in Xu et al.
In the theoretical analysis, it is not comprehensive to only use the theoretical models to fit the experimental results of stress-dependent permeability. This is because influencing factors of stress sensitivity can not only cause stress-dependent permeability but also stress-dependent porosity. A theoretical model can be considered applicable to PPI when both the stress sensitivities of permeability and porosity are well described by the theoretical model. Therefore, experimental data on stress-dependent porosity under stress loading is also used in the theoretical analysis. The reason for only using the stress-dependent porosity under stress loading is that the porosity is only reduced by a maximum of 11% under stress loading. Since the stress-dependent porosity is also affected by the elastic-structural deformation, the porosity stress sensitivity under stress unloading is less than the porosity stress sensitivity under stress loading. The theoretical model not only fits the experimental data on the stress-dependent permeability but also the experimental data on the stress-dependent porosity. The model also predicts the stress-dependent porosity from the fitting results of the stress-dependent permeability and vice versa. To conduct a more comprehensive and accurate theoretical analysis, the differences between the solved and the real physical properties, the consistency of the solved parameters, and the differences between the predicted and the real permeability–porosity relationships are also further discussed. The better the consistency of the parameters, the better the theoretical applicability. The porosity sensitivity exponent α is used to describe the permeability–porosity relationships and is obtained by linear fitting the stress sensitivities of permeability and porosity as follows
| 12 |
where k o and ϕo are the permeability and porosity at the initial effective stress, respectively.
The theoretical models are solved using the genetic algorithm (GA) based on experimental data. The detailed solution method, parameter settings and the solved parameters for all models are shown in Appendix B. A theoretical analysis of Xu’s model without tortuosity sensitivity is also conducted to illustrate the effect of tortuosity sensitivity on the stress sensitivity of PPI. To demonstrate the results concisely and clearly, some of the abbreviations and symbols used in this section are explained as follows. (1) f_pe: The permeability fitted from experimental data of permeability. (2) f_po: The porosity fitted from experimental data of porosity. (3) p_pe: The permeability predicted from experimental data of porosity. (4) p_po: The porosity predicted from experimental data of permeability. (5) α_pe: The porosity sensitivity exponent α predicted from experimental data of permeability. (6) α_po: The porosity sensitivity exponent α predicted from experimental data of porosity. (7) MRe_*: The mean relative error of*. (8) SdR_*: The standard deviation of the relative errors of*. (9) NTS: Without the tortuosity sensitivity.
Table lists the means and standard deviations of relative errors for the fitting and predicting results with respect to experimental data of PPI. From Table , expect the fitting results of Xu’s (NTS), Tan’s and Huang’s models for permeability, the fitting results of all models for permeability as well as porosity are in good agreement with the experimental data. In the prediction task, compared to other models, Xu’s model obtains the lowest MRe_p_pe and SdR_p_pe. MRe_p_po and SdR_p_po obtained by Xu’s model are closer to the minimum MRe_p_po obtained by Tan’s model and the minimum SdR_p_po obtained by Huang’s model, respectively. This suggests that Xu’s model has a good performance in predicting stress sensitivities. In Table , the smaller average values of MRe and SdR obtained by Xu’s model are highlighted in bold type. This points to the fact that Xu’s model has highest average accuracy in fitting and predicting the stress sensitivities of PPI compared to other models. In each model, MRe_p_po and SdR_p_po are less than MRe_p_pe and SdR_p_pe, respectively. This indicates that the accuracy in predicting the stress-dependent porosity is higher than the accuracy in predicting the stress-dependent permeability for each model.
2. MRe and SdR for Fitted and Predicted Results of Different Models on PPI.
| model | MRe_f_pe ± SdR_f_pe (%) | MRe_p_pe ± SdR_p_pe (%) | MRe_f_po ± SdR_f_po (%) | MRe_p_po ± SdR_p_po (%) | average value (%) |
|---|---|---|---|---|---|
| Xu’s | 4.06 ± 4.24 | 9.19 ± 4.51 | 3.74 ± 2.02 | 6.86 ± 2.31 | 5.96 ± 3.27 |
| Xu’s (NTS) | 8.65 ± 4.38 | 32.71 ± 8.58 | 3.74 ± 2.02 | 21.75 ± 5.51 | 16.71 ± 5.12 |
| Tan’s | 9.56 ± 6.28 | 63.88 ± 25.2 | 0.23 ± 0.12 | 5.36 ± 3.63 | 19.76 ± 8.81 |
| Huang’s | 17.27 ± 8.11 | 28.63 ± 8.4 | 1.52 ± 0.48 | 5.76 ± 3.02 | 13.3 ± 5 |
| Cao’s | 4.95 ± 4.9 | 25.44 ± 9.76 | 2.92 ± 2.02 | 12.39 ± 4.07 | 11.43 ± 5.19 |
Figure shows the comparison of Young’s modulus and Poisson’s ratio solved by different models with their experimentally measured values. E s and νs are Young’s modulus and Poisson’s ratio of the solid skeleton, respectively. In Figure a, the E s solved by Huang’s model are closest to the experimentally measured value. For other models, compared to the experimentally measured values, the E p solved by Xu’s and Xu’s (NTS) models are higher, while the E s solved by Tan’s and Cao’s models are lower. Even for Huang’s model, the minimum relative error between the solved E s and the experimentally measured value is 76%. A possible reason for the large relative error is that the existing models do not comprehensively consider the mechanical properties of PI when calculating the elastic deformation of capillary cross sections, ultimately leading to a large relative error between the solved E s and the experimentally measured value. In Figure b, the νp solved by Xu’s model are closest to the experimentally measured value. Compared to the experimentally measured values, the νp solved by Xu’s (NTS) model from porosity data is higher, while the Poisson’s ratio solved by other models are lower. This shows that all models have large errors for solving Young’s modulus of PPI or PI, while Xu’s model is more accurate for solving Poisson’s ratio of PPI.
9.
Young’s modulus (a) and Poisson’s ratio (b) solved by different models.
To discuss the consistency of parameters solved by different models, Figure shows the MRe and SdR of all the parameters solved by different models. MRe and SdR of each model in Figure are the average values of MRe and SdR from all the solved parameters. For each parameter, the relative error is calculated by . par_f_pe is a parameter solved by the model based on permeability. par_f_po is the corresponding parameter solved by the model based on porosity. From Figure , it is apparent that both MRe and SdR of Xu’s model are the minimum among all models. It demonstrates that Xu’s model has the best consistency in the representation of physical parameters of PPI.
10.

MRe and SdR of all the parameters solved by different models.
Figure shows the experimental porosity sensitivity exponent α, α_pe and α_po under conditions of stress loading, respectively. In Figure , the experimental α is 4.07. α_pe and α_po of Xu’s model are closest to the experimental α compared to the other models. It is means that Xu’s model can better predict the permeability–porosity relationships of PPI. Except for Tan’s model, α_pe is similar to α_po in each model. α_pe and α_po of Xu’s, Xu’s (NTS), Huang’s and Cao’s model are smaller than the experimental data, while α_pe and α_po of Tan’s model are larger than the experimental data. Figure a shows the comparison of fitting and prediction results of Xu’s model for stress-dependent permeability with respect to the experimental results, while Figure b is for porosity. From Figure , the predicted permeability stress sensitivity is smaller than the experimental permeability stress sensitivity while porosity is the opposite. Therefore, permeability changes slower with the porosity under stress loading compared to the experimental data. It explains why α_pe and α_po of Xu’s model are smaller than the experimental α. The same reason applies to other models.
11.

Comparison of porosity sensitivity exponent α under stress loading between all models’ predicted results and experimental data.
12.
Fitting and prediction results of Xu’s model for stress-dependent permeability (a) and stress-dependent porosity (b).
After model comparison, Xu’s model is the most accurate in describing the stress sensitivities of PPI. Furthermore, it can be seen from the comparison of Xu’s model and Xu’s (NTS) model that the influence of tortuosity sensitivity on the stress sensitivities of PPI is essential. In Xu’s model, the influencing factors include the elastic-structural deformation of the capillary cross-section and tortuosity sensitivity, which are consistent with the influencing factors in Figure . This further indicates that exactly the factors in Figure influence the stress-dependent permeability of PPI. In addition, the fitting results of Xu’s model for stress-dependent permeability in Figure suggest that the structural deformation disappears after 2.27 MPa. The E p, νp, deformation parameter β, the proportion index of the quadrilateral structures γ and the tortuosity sensitivity exponent ϵ of PPI from Xu’s model based on permeability data are 9.94 GPa, 0.49, 0.22, 0.37 and 2.16, respectively. For the rocks in Xu et al., the maximum values of E p, νp, β, γ and ϵ are 95.7 GPa, 0.38, 0.045, 0.093 and 2.12, while the minimum values are 34 GPa, 0.2, 0.02, 0.033 and 0.01, respectively. Compared to rocks, PPI has a lower E p and higher νp, β and γ. This indicates that the elastic-structural deformation of PPI is more sensitive to stress, and fewer capillaries in PPI have structural deformation. According to the ϵ empirical equation for rocks in Xu et al., the ϵ calculation result for sample #1 is 0.06, which is lower than the 2.16 from permeability data. This suggests that the ϵ empirical equation for rocks proposed by Xu et al. does not apply to PPI, and the ϵ value of PPI is higher than that of rocks with the same permeability and porosity.
For Types 2 and 3 under stress unloading in Region II, the effect of detached particles on tortuosity is not considered in Xu’s model. Therefore, this paper proposes the formula for τσ under stress unloading as follows
| 13 |
where τo is the tortuosity at the initial effective stress; τnϕ and τnp are the normalized tortuosity considering the effects of elastic-structural deformation and detached particles, respectively; a τ and b τ are the coefficients in the fitting formula of τnp; σeffmax is the maximum σeff under stress unloading. Equation works when σeff is greater than the critical effective stress. τnϕ is from Xu’s model, while the fitting formula of τnp is proposed in this paper. τnp ≥ 1 and is only active under stress unloading. The coefficients a τ and b τ are obtained as follows. First, make the normalized permeability of an experiment at the σeffmax the same as the normalized permeability of Xu’s model by changing the intercept of the fitting result of Xu’s model under stress unloading. This is because the fitting results of Xu’s model in Figure are hard to equate to the normalized permeability at the σeffmax in each experiment condition. Second, obtain τnp by dividing the fitted normalized permeability under stress unloading by the experimental normalized permeability. Finally, obtain the a τ and b τ by fitting τnp using the fitting formula in eq . Figure shows τnp in SP5, SP6, SP9 and SP12 and the fitting results. In Figure , the fitting results can describe the trend of τnp, but there is still a difference between the fitting results and the experimental data. This is because the number of detached particles, the size of detached particles, the size of blocked pores, and the connectivity between blocked pores and other pores vary among different PPI samples. The above variation causes differences in the effect of detached particles on τnp in different PPI samples, ultimately leading to differences between the fitting results and the experimental data. After substituting τnp into Xu’s model, the fitting results of normalized permeability in SP5, SP6, SP9 and SP12 under stress unloading are shown in Figure a. The MRe and SdR of fitting results are 4.22% and 1.27%, respectively. This indicates that, although there are differences between the fitting results and the experimental data in Figure , the fitting formula proposed in eq can well describe the stress-dependent permeability caused by detached particles.
13.

Fitting results of normalized tortuosity τnp considering the effect of detached particles.
Conclusion
In this paper, the stress-dependent permeability of PPI under stress loading and unloading has been studied experimentally and theoretically. Based on the working conditions of the bearing units, the experimental study measured the lubricant permeability of PPI at different temperatures, inlet pressures and confining pressures using a physical simulation system. The experimental results illustrate that the permeability is reduced by up to 37% when the effective stress σeff is between 0.75 and 26.45 MPa. Both the permeability hysteresis and the negative correlation between permeability and σeff are caused by the elastic-structural deformation of capillary cross-section and tortuosity sensitivity. When σeff is greater than 15 MPa and the pressure gradient is greater than 40 MPa/m, the detachment, release, blocking, unblocking and migration of solid skeleton particles cause the permeability to first decrease and then increase under stress unloading.
The theoretical study uses different theoretical models to analyze the experimental results. For a comprehensive analysis, the experimental results of stress-dependent porosity under stress loading are also used. The theoretical results show that the model, which has the same influencing factors as PPI, has the best performance in fitting stress sensitivities, predicting stress sensitivities, predicting the permeability–porosity relationships and solving physical parameters. This further indicates the influencing factors of stress-dependent permeability in PPI. The tortuosity variation and the permeability behavior influenced by detached particles are described by proposing a fitting formula and improving a theoretical model, respectively. This study has a good potential for the design, evaluation and improvement of the oil supply capacity of PPI.
Acknowledgments
This work was supported by the Key Program of National Natural Science Foundation of China (U1637206) and project under contract number 110070162400040.
Appendix A
The PPI samples are dried at 95 °C for 24 h before conducting experiments related to porosity and permeability. For experiments on stress-dependent permeability and stress-dependent porosity, the PPI samples also need to be saturated with lubricant after drying. The saturation method is as follows. (1) Place the sample in the saturation chamber and vacuum it; (2) inject lubricant into the saturation chamber and increase the chamber pressure to 10 MPa; (3) maintain this pressure for 8 h to fully saturate the sample with the lubricant. The above is the whole process.
Porosity ϕ and permeability k for all PPI samples are measured using helium porosimeter and gas permeameter, respectively. Both instruments are from Jiangsu Tuochuang Scientific Instrument Co., Ltd. In the porosity measurement, the reference chamber pressure is 0.4 MPa before the helium expands into the sample chamber. In the permeability measurement, k is obtained by measuring the gas permeability at different absolute mean pressures and then eliminating the gas slippage effect using the Klinkenberg equation. The gas permeability is measured according to Darcy’s law. The confining pressure for the gas permeability measurement is 1 MPa. Absolute mean pressures include 0.175 MPa, 0.2 and 0.225 MPa.
Pore size distribution is measured using half of sample #1 and the mercury injection method. The other half of sample #1 is used for structural characterization. Mercury injection apparatus is the AutoPore9505 from Micromeritics Instrument Corp. U.S.A. The intrusion pressure ranges from 0 to 206.8 MPa. The pore diameter λp is calculated by
| 14 |
where σHg is the surface tension of mercury, which is equal to 0.48 N/m; θHg is the contact angle of mercury on PI, which is equal to 130°; p c is the intrusion pressure.
For the relevant calculations and analyses, Young’s modulus E p, Poisson’s ratio νp and thermal expansion coefficient αTp of PPI are measured. E p and νp are measured using sample #2 and the RTR-1500 rock triaxial testing system from GCTS U.S.A. The axial stress ranges from 0 to 60 MPa, and its growth rate is 0.16 MPa/s. αTp is measured using sample #3 and the 402 Expedis Supreme from NETZSCH Group. The test temperature ranges from 25 to 120 °C, and the experimental atmosphere is helium. The mechanical properties of PI are also measured for comparison. Sample #19 is used for Young’s modulus E s and Poisson’s ratio νs of PI. Sample #20 is used for thermal expansion coefficient αTs of PI.
Images of PPI and detached particles are taken with the JSM-7500F field emission scanning electron microscopy from Japan Electronics Co., Ltd. The PPI is chiseled and the fractured surface is used for observation. The detached particles are extracted from the lubricant flowing out of the equipment outlet in Figure . The extraction process is as follows. (1) Mix lubricant (10 mL) with ethyl alcohol (10 mL) for 5 min. (2) After the mixed solution has stratified, take the upper layer as the dispersion liquid of detached particles. (3) Concentrate the dispersion liquid to 5 mL at 80 °C. The above is the whole process. The detached particles can be observed after the dispersion liquid has been dropped onto a substrate and the alcohol has completely evaporated. Both PPI and detached particles require spray-gold treatment before the SEM observation. Besides, the IX73 biological microscope from OLYMPUS is used to measure the size of detached particles.
Appendix B
The reason for using the genetic algorithm (GA) to solve the theoretical models is that the parameter solving problems in the models are all multiparameter optimization problems and GA has been widely used in multiparameter optimization problems. Compared with other optimization algorithms such as the ant colony algorithm and simulated annealing algorithm, probabilistic exploration and parallel computation enable GA to search for the optimal solution faster and more accurately under a given solution space. In addition, the solution capability of GA is not limited by the application domain. Therefore, this paper uses GA to solve the parameters in the models. The solution process of GA contains initial population generation, selection and crossover. Real coding is used for chromosomes. Each gene locus corresponds to a parameter to be solved and the coded value satisfies its range of values. The roulette selection and the uniform crossover are applied to individuals to generate the next generation. The mean relative error between theoretical value and experimental value is set as the objective function for GA. According to Xu et al., the population size, number of generations and crossover fraction are set as 200, 20, and 0.6 for GA, respectively. The solution procedure of theoretical models is as follows. (1) Set the solution range for the model parameters. (2) Solve the model parameters 10 times using GA. (3) Calculate the mean of the results of each parameter given by 10 times of GA. (4) Predict the stress-dependent porosity when the experimental data is permeability and vice versa. (5) Calculate the porosity sensitivity exponent α. The above is the whole process. For the stress loading process, the experimental data of stress-dependent permeability and stress-dependent porosity for model fitting includes the results for all experimental conditions. For the stress unloading process, the model needs to fit the stress-dependent permeability for each experimental condition.
The parameter settings for different models are as follows. First, all models include two input parameters for elastic properties: Young’s modulus and Poisson’s ratio. Xu’s model uses Young’s modulus E p and Poisson’s ratio νp of the porous media, while others use Young’s modulus E s and Poisson’s ratio νs of the solid skeleton. Although the elastic parameters are defined differently in different models, the elastic parameters of PPI are closer to those of the PI, so the range of the elastic parameters is the same in different models. The upper limits of Young’s modulus and Poisson’s ratio are 10 GPa and 0.5, while the lower limits are 0.01 GPa and 0.1, respectively. Some models require an initial effective stress, where both the normalized permeability and normalized porosity are 1, and the initial effective stress cannot be 0. Experimental results show that the normalized permeability is 0.99 at the minimum effective stress 0.75 MPa and the minimum effective stress in the stress-dependent porosity experiment is 0.5 MPa. Therefore, the initial effective stress is set as 0.5 MPa. Permeability, porosity and pore size distribution of sample #1 are set as the relevant parameters at the initial effective stress. In Xu’s model, tortuosity at the initial effective stress is set as 151.62. The upper limits of the deformation parameter, the proportion index of the quadrilateral structures and the tortuosity sensitivity exponent are 0.5, 0.5 and 6, while the lower limits are 0.2, 0.2 and 2, respectively. The solution range of parameters for Xu’s model is also used for Xu’s model without tortuosity sensitivity (NTS). Since the tortuosity sensitivity exponent cannot be obtained from the porosity data, the tortuosity sensitivity exponent obtained from the permeability data is directly used to predict the stress-dependent permeability. In Huang’s model, the upper limits of the outer to inner radius ratio, the boundary to inner radius ratio and the elastic limit pressure are 8, 8, and 26.45 MPa, while the lower limits are 1, 1, and 0.5 MPa, respectively. The pressure inside the capillary is set as 1.56 MPa, which is the average of the pore pressures in Table . This is because Huang’s model needs to fit all the data under stress loading, but the experimental conditions have different inlet pressures. In Cao’s model, the upper limits of the deformation rate and the deformation degree are 3 and 3, while the lower limits are 0 and 0, respectively. According to Figure , the particle radius is set as 10 μm. In Tan’s model, according to Figure , the ratio of the maximum diameter of solid cluster to the minimum diameter of solid cluster is set as 1 × 10–2. The characteristic radius of particles is set as 10 μm.
The solved parameters for all models are as follows. For Xu’s model, Young’s modulus E p, Poisson’s ratio νp, deformation parameter β, the proportion index of the quadrilateral structures γ and the tortuosity sensitivity exponent ϵ based on permeability data are 9.94 GPa, 0.49, 0.22, 0.37 and 2.16, while E p, νp, β and γ based on porosity data are 9.99 GPa, 0.50, 0.21 and 0.50, respectively. For Xu’s (NTS) model, E p, νp, β and γ based on permeability data are 4.83 GPa, 0.26, 0.22 and 0.20, while the values based on porosity data are the same as those of Xu’s model based on porosity data. For Tan’s model, Young’s modulus E s and Poisson’s ratio νs based on permeability data are 0.566 GPa and 0.10, while the values based on porosity data are 0.057 GPa and 0.10, respectively. For Huang’s model, E s, νs, the outer to inner radius ratio, the boundary to inner radius ratio and the elastic limit pressure based on permeability data are 0.72 GPa, 0.16, 4.76, 1.68, and 14.67 MPa, while the values based on porosity data are 5.38 GPa, 0.28, 5.27, 4.03, and 5.42 MPa, respectively. For Cao’s model, E s, νs, the deformation rate and the deformation degree based on permeability data are 0.084 GPa, 0.17, 0.84 and 0.27, while the values based on porosity data are 0.042 GPa, 0.16, 0.39 and 1.83, respectively.
The authors declare no competing financial interest.
References
- Tafazoli M.. A study of on-orbit spacecraft failures. Acta Astronaut. 2009;64:195–205. doi: 10.1016/j.actaastro.2008.07.019. [DOI] [Google Scholar]
- Sathyan K., Hsu H. Y., Lee S. H., Gopinath K.. Long-term lubrication of momentum wheels used in spacecraftsAn overview. Tribol. Int. 2010;43:259–267. doi: 10.1016/j.triboint.2009.05.033. [DOI] [Google Scholar]
- Hall, J. Advances in Spacecraft Technologies, 1st ed.; InTech: Rijeka, 2011; pp 75–98. [Google Scholar]
- Zhang P., Zhang K., Chen X., Dou S., Zhao J., Li Y.. Influence of coagulation bath temperature on the structure and dielectric properties of porous polyimide films in different solvent systems. ACS Omega. 2020;5:29889–29895. doi: 10.1021/acsomega.0c04103. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Lv P., Dong Z., Dai X., Qiu X.. Flexible polydimethylsiloxane-based porous polyimide films with an ultralow dielectric constant and remarkable water resistance. ACS Appl. Polym. Mater. 2019;1:2597–2605. doi: 10.1021/acsapm.9b00484. [DOI] [Google Scholar]
- Xu Z., Li J., Li K., Shi C., Zhou N., Hua L., Yang Q., Zhao J., Qing T.. Strengthening wear resistance of porous polyimide with organic silicon coating. J. Appl. Polym. Sci. 2025;142:e56597. doi: 10.1002/app.56597. [DOI] [Google Scholar]
- Wang J., Zhao H., Huang W., Wang X.. Investigation of porous polyimide lubricant retainers to improve the performance of rolling bearings under conditions of starved lubrication. Wear. 2017;380–381:52–58. doi: 10.1016/j.wear.2017.03.008. [DOI] [Google Scholar]
- Qiu Y., Wang Q., Wang C., Wang T.. Oil-containing and tribological properties of porous polyimide containing lubricant oil. Tribol. 2012;32:538–543. [Google Scholar]
- Zhang D., Wang T., Wang Q., Wang C.. Selectively enhanced oil retention of porous polyimide bearing materials by direct chemical modification. J. Appl. Polym. Sci. 2017;134:45106. doi: 10.1002/app.45106. [DOI] [Google Scholar]
- Marchetti M., Meurisse M. H., Vergne P., Sicre J., Durand M.. Analysis of oil supply phenomena by sintered porous reservoirs. Tribol. Lett. 2001;10:163–170. doi: 10.1023/A:1009026123907. [DOI] [Google Scholar]
- Bertrand P. A., Carré D. J.. Oil exchange between ball bearings and porous polyimide ball bearing retainers. Tribol. Trans. 1997;40:294–302. doi: 10.1080/10402009708983658. [DOI] [Google Scholar]
- Zhang D., Wang C., Wang Q., Wang T.. High thermal stability and wear resistance of porous thermosetting heterocyclic polyimide impregnated with silicone oil. Tribol. Int. 2019;140:105728. doi: 10.1016/j.triboint.2019.04.012. [DOI] [Google Scholar]
- Mi K., Xu Y., Lei Y., Wang J., Shen P., Zhang X.. Numerical modeling of multiphysics constitutive relations governing water migration into variably saturated sorptive shale. ACS Omega. 2025;10:6948–6964. doi: 10.1021/acsomega.4c09681. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Mukherjee M., Vishal V.. Gas transport in shale: A critical review of experimental studies on shale permeability at a mesoscopic scale. Earth-Sci. Rev. 2023;244:104522. doi: 10.1016/j.earscirev.2023.104522. [DOI] [Google Scholar]
- Teklu T. W., Li X., Zhou Z., Abass H.. Experimental investigation on permeability and porosity hysteresis of tight formations. SPE J. 2018;23:672–690. doi: 10.2118/180226-PA. [DOI] [Google Scholar]
- Liang Z., Jiang Z., Wu W., Guo J., Wang M., Nie Z., Li Z., Xu D., Xue Z., Chen R., Han Y.. Study and classification of porosity stress sensitivity in shale gas reservoirs based on experiments and optimized support vector machine algorithm for the silurian longmaxi shale in the Southern Sichuan Basin, China. ACS Omega. 2022;7:33167–33185. doi: 10.1021/acsomega.2c03393. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Bohnsack D., Potten M., Freitag S., Einsiedl F., Zosseder K.. Stress sensitivity of porosity and permeability under varying hydrostatic stress conditions for different carbonate rock types of the geothermal Malm reservoir in Southern Germany. Geotherm. Energy. 2021;9:15. doi: 10.1186/s40517-021-00197-w. [DOI] [Google Scholar]
- Li J., Zhao G., Qi Z., Yin B., Xu X., Fang F., Yang S., Qi G.. Stress sensitivity of formation during multi-cycle gas injection and production in an underground gas storage rebuilt from gas reservoirs. Pet. Explor. Dev. 2021;48:968–977. doi: 10.1016/S1876-3804(21)60081-2. [DOI] [Google Scholar]
- Zhong X., Zhu Y., Liu L., Yang H., Li Y., Xie Y., Liu L.. The characteristics and influencing factors of permeability stress sensitivity of tight sandstone reservoirs. J. Pet. Sci. Eng. 2020;191:107221. doi: 10.1016/j.petrol.2020.107221. [DOI] [Google Scholar]
- Guo X., Zou G., Wang Y., Wang Y., Gao T.. Investigation of the temperature effect on rock permeability sensitivity. J. Pet. Sci. Eng. 2017;156:616–622. doi: 10.1016/j.petrol.2017.06.045. [DOI] [Google Scholar]
- Kong Q., Wang H., Xu W.. Experimental study on permeability and porosity evolution of sandstone under cyclic loading and unloading. Chin. J. Geotech. Eng. 2015;37:1893–1900. doi: 10.11779/CJGE201510018. [DOI] [Google Scholar]
- Liu Y.. Experimental analysis of coal permeability evolution under cyclic loading. J. Chin. Coal Soc. 2019;44:2579–2588. doi: 10.13225/j.cnki.jccs.2019.0353. [DOI] [Google Scholar]
- Tan X. H., Li X. P., Liu J. Y., Zhang L. H., Fan Z.. Study of the effects of stress sensitivity on the permeability and porosity of fractal porous media. Phys. Lett. A. 2015;379:2458–2465. doi: 10.1016/j.physleta.2015.06.025. [DOI] [Google Scholar]
- Cao N., Lei G.. Stress sensitivity of tight reservoirs during pressure loading and unloading process. Pet. Explor. Dev. 2019;46:138–144. doi: 10.1016/S1876-3804(19)30013-8. [DOI] [Google Scholar]
- Liu K., Yin D., Sun Y.. The mathematical model of stress sensitivities on tight reservoirs of different sedimentary rocks and its application. J. Pet. Sci. Eng. 2020;193:107372. doi: 10.1016/j.petrol.2020.107372. [DOI] [Google Scholar]
- Xu R., Wang L., Zhao X., Chi J.. A permeability hysteresis model for fractal porous media based on elastic-structural deformation of capillary cross section. Fractals. 2023;31:2350019. doi: 10.1142/s0218348x23500196. [DOI] [Google Scholar]
- Xu R., Wang L., Zhao X., Mao Y.. Characterizing permeability-porosity relationships of porous rocks using a stress sensitivity model in consideration of elastic-structural deformation and tortuosity sensitivity. J. Rock Mech. Geotech. Eng. 2024;16:3437–3451. doi: 10.1016/j.jrmge.2024.01.020. [DOI] [Google Scholar]
- Huang Z., Su X., Li Y., Zhang K., Yao J.. Stress sensitivity analysis of fractal porous media based on the elasto-plastic thick-walled cylinder model. Fractals. 2021;29:2150162. doi: 10.1142/S0218348X21501620. [DOI] [Google Scholar]
- Bao M., Wang L., Wu J., Zheng D., Zhang C., Zhao X.. The influence on dynamic characteristics of ball bearings with consideration of cage pocket shapes within concurrently rotating of inner and outer rings. Tribol. Int. 2025;202:110263. doi: 10.1016/j.triboint.2024.110263. [DOI] [Google Scholar]
- Yu B., Cai J., Zou M.. On the physical properties of apparent two-phase fractal porous media. Vadose Zone J. 2009;8:177–186. doi: 10.2136/vzj2008.0015. [DOI] [Google Scholar]
- Yu B., Li J.. Some fractal characters of porous media. Fractals. 2001;09:365–372. doi: 10.1142/s0218348x01000804. [DOI] [Google Scholar]
- Wang J. Z., Ma J., Ao Q. B., Zhi H., Tang H. P.. Review on fractal analysis of porous metal materials. J. Chem. 2015;2015:427297. doi: 10.1155/2015/427297. [DOI] [Google Scholar]
- Wypych, G. Handbook of Polymers, 2nd ed.; ChemTec Publishing: Toronto, 2016; pp 431–436. [Google Scholar]
- Wang B., Qin Y., Shen J., Wang G.. Study on stress sensitivity of lignite reservoir under salinity and pH composite system. Energy Explor. Exploit. 2018;36:464–487. doi: 10.1177/0144598717740957. [DOI] [Google Scholar]
- Chen L., Zhang D., Zhang W., Guo J., Yao N., Fan G., Zhang S., Wang X., Wu P.. Experimental investigation on post-peak permeability evolution law of saturated sandstone under various cyclic loading–unloading and confining pressure. Water. 2022;14:1773. doi: 10.3390/w14111773. [DOI] [Google Scholar]
- Ebner M., Wood V.. Tool for tortuosity estimation in lithium ion battery porous electrodes. J. Electrochem. Soc. 2015;162:A3064–A3070. doi: 10.1149/2.0111502jes. [DOI] [Google Scholar]
- Kozhevnikov E. V., Turbakov M. S., Riabokon E. P., Gladkikh E. A.. Apparent permeability evolution due to colloid migration under cyclic confining pressure: On the example of porous limestone. Transp. Porous Media. 2024;151:263–286. doi: 10.1007/s11242-023-01979-5. [DOI] [Google Scholar]
- Zhang C., Zhang L., Zhao Y., Wang W.. Experimental study of stress–permeability behavior of single persistent fractured coal samples in the fractured zone. J. Geophys. Eng. 2018;15:2159–2170. doi: 10.1088/1742-2140/aac12e. [DOI] [Google Scholar]
- David C., Wong T., Zhu W., Zhang J.. Laboratory measurement of compaction-induced permeability change in porous rocks: Implications for the generation and maintenance of pore pressure excess in the crust. Pure Appl. Geophys. 1994;143:425–455. doi: 10.1007/BF00874337. [DOI] [Google Scholar]
- Goldberg, D. E. Genetic Algorithms in Search, Optimization, and Machine Learning, 1st ed.; Addison-Wesley Publishing Company: Canada, 1989; pp 59–88. [Google Scholar]








