- Email: [email protected]
Friction between pearlitic steel surfaces
Wear 432-433 (2019) 102910
Contents lists available at ScienceDirect
Wear journal homepage: www.elsevier.com/locate/wear
Friction between pearlitic steel surfaces
T
∗
Casey Jessop , Johan Ahlström Department of Industrial and Materials Science, Chalmers University of Technology, Gothenburg, Sweden
A R T I C LE I N FO
A B S T R A C T
Keywords: Rail steel Rail-wheel tribology Rolling friction Sliding wear Third-body layer
Experiments rubbing two pearlitic rail steel surfaces against each other were done using an axial-torsion test rig. After the experiments were completed, the surfaces were observed under stereomicroscope, and the wear debris were examined. The aim of the experiments was to evaluate the variation of friction characteristics between two surfaces of pearlitic rail steel, mimicking the friction which occurs between the faces of a crack during rolling contact fatigue (RCF) loading. For both dry and wet conditions, the sliding velocity, the angles of rotation, and the average normal force were varied. It was found that the most significant effect of changing angles of rotation is on the formation of ridges on the contact surface with large angles, leading to a higher friction coefficient. The presence of water reduces the friction coefficient and leads to less deformation and wear in the contact surface.
1. Introduction
well as the presence of fluids and wear debris. These phenomena remain less studied than the crack initiation associated to squats. However, crack face friction has a major influence on the crack growth since the friction as well as the surface asperities which are present on the crack faces have an influence on the stress intensity factors (SIF) in shear sliding (Mode II) and tearing (Mode III) loading, as well as on crack closure (Mode I loading). Research done on shear mode crack propagation has also shown that the main driving forces for branching of Mode II cracks are crack face friction and wedging, caused by surface asperities [10], particularly if the contacting crack faces are closed [11]. The present study evaluates the friction between two pearlitic R260 rail steel surfaces in different combinations of compression and torsional loading under dry and wet conditions in order to determine the influence of these parameters on the friction coefficient.
In the railway industry, tribology is an important topic since the contact in between rails and wheels is subjected to varying temperature and climatic conditions in operation, which affect both friction and wear. Another important concern is the effect of environment on crack face tribology, that is, the contact between crack faces during crack propagation due to rolling contact fatigue (RCF) loading. The effect of crack face friction on the propagation of subsurface cracks has been studied in the past [1,2], however, since these cracks have no opening to the surface, the penetration of water and other fluids was not considered in the models. Nevertheless, the presence of water or another fluid is thought to be necessary for the propagation of squat and squattype defects in rails which are surface breaking cracks driven by rolling contact fatigue loading [3,4]. It is thought that the friction has an important effect on crack propagation and branching [3,5–7]. Furthermore, the presence of third-body materials such as oxides or sheared bulk material, is also an important consideration [5,6]. Several studies have been conducted using, for example, pin-on-disk and test rig experiments, and the analysis of third-body materials present in field as a consequence of wheel-rail contact in different stick and slip conditions, has been compared to laboratory experiments [4,8,9]. Squat and squat-type defects are of particular interest in the field of RCF in rails. The darkened surface associated to these defects is thought to originate from a depression of the surface, due to the loss of material and wear occurring between the crack faces, which causes the surface of the rail to be worn at a different rate than the rest of the rail. The growth of squats is significantly influenced by the crack face friction as
∗
2. Experimental The aim of the experiments is to evaluate the friction coefficient between two surfaces of pearlitic rail steel, mimicking the friction which occurs between the faces of a crack in a rail during repeated wheel-rail contact. Two halves of a test bar (shown schematically in Fig. 1), representing the crack faces, are subjected to a normal force and rotated back and forth against each other in different conditions. The test bars were designed with a steering pin going into a roller bearing to avoid any lateral translations between the halves, yielding a well-controlled sliding path without disturbing the torque measurement. The product of the normal stress and the friction coefficient is equivalent to
Corresponding author. E-mail address: [email protected] (C. Jessop).
https://doi.org/10.1016/j.wear.2019.05.025 Received 9 January 2019; Received in revised form 16 May 2019; Accepted 23 May 2019 Available online 28 May 2019 0043-1648/ © 2019 Elsevier B.V. All rights reserved.
Wear 432-433 (2019) 102910
C. Jessop and J. Ahlström
Fig. 1. (a) Top view of R260 rail steel surfaces in machined condition, and (b) schematic of test bar used in friction experiments with dimensions in mm.
machine offset. The exact position of the spots in true contact are not known, but the reasonably thin (2 mm), circular contact band enables an approximate quantification according to (Equation (1)):
Table 1 Experiment parameters.
Rate of rotation Angles of rotation Normal force
Min
Max
0.1°/sec ± 0.5° −0.5 kN
5°/sec ± 10° −8 kN
μ = T/F·r
Where μ is the friction coefficient, T is torque [kN·mm], F is the normal force [kN], and r is the average radius in the contact [mm]. Tests were run in dry and wet conditions at room temperature. For the wet set up, a funnel filled with deionized water was attached to the test bar, as shown in Fig. 2. The surfaces are initially dry; the water is only added to the funnel after the faces of the two halves are pressed together with the full axial load, and the contact surface is thus sealed. That is, the water only surrounds the test bar, and any penetration which occurs is due to the intrusion of water in between the steel surfaces during twisting. The rotation sensor of the machine was found to give a sufficient measure of the relative motion/rotation of the steel halves for larger angles of rotation. This can be shown from the relation between the rotation measured by the machine sensor and that measured by the extensometer. The correlation is found using the following relation (Equation (2)):
Table 2 Detailed parameters of tests included in results. Some tests are repeated, but referred to by different names for clarify of presentation. (Tests W3 and W4 were run in wet conditions.) Test
Rate of rotation (°/sec)
E 0.1 Direction D1 1 D2 1 D3 5 D4 5 Rotation R1 0.5 R2 0.5 R3 0.5 Torque response T1 5 T2 5 T3 5 T4 0.1 Normal force F1 5 F2 5 F3 5 Dry vs wet W1 1 W2 0.5 W3 1 W4 1
Angles of rotation (°)
Normal force (kN)
0.5
−4
5 5 5 5
−4 −4 −4 −4
0.5 10 2
−4 −4 −4
5 1 5 1
−0.5 −2 −4 −4
5 5 5
−0.5 −2 −4
5 10 10 5
−4 −4 −4 −8
(1)
tan(φ) = r·θ/L
(2)
Where φ is the torsional strain [rad] measured by the extensometer on the test specimen, r·θ is the torsional rotation measured by the machine rotation sensor, and L is the distance between the extensometer pins [mm]. Fig. 3 (Test E) shows the direct relation between the rotation measured from the machine and that from the extensometer. The first cycle (shown as the dashed line) is different from the remaining cycles in the experiment. The points lie on two lines with slope equal to 1 (blue line is a reference). The reason for the two paths, and the
the friction stress which is present between opposing crack faces under ideal contact conditions [10]. All tests were run in a MTS 809 axialtorsion servo-hydraulic testing machine with closed loop control. The tests were run in axial force, and torsional rotation control, and in some tests an extensometer was used to measure torsional rotation locally with high precision. The test bars were in machined condition. The effect of the different conditions is measured using the friction coefficient as a basis for comparison. The parameters which are varied are the sliding velocity (rate of rotation), the angles of rotation (stroke/length of sliding movement), and the average normal force (see Table 1). A more detailed description of the test parameters for given tests is included in Table 2. After the experiments were completed, the surfaces were observed using stereomicroscopy and the wear products evaluated. The friction coefficient is determined by measuring the torque during the experiments. That is, the friction is determined by the torque required to rotate the sample under a given normal force. The torque values were centred after the experiments to compensate for the
Fig. 2. Experimental setup for tests run in wet conditions. The test bar is mounted into the machine and not until after full axial loading, the funnel is filled with deionized water. 2
Wear 432-433 (2019) 102910
C. Jessop and J. Ahlström
Fig. 3. Correlation between the machine rotation (r·θ/L) and the relative rotation of the test specimen (tan(φ)) during testing. The “points” lie on two lines with slope 1:1, showing the direct relation between the rotation measured by the machine and the extensometer. (The first cycle is shown with dashed line, and a 1:1 slope is plotted for reference.)
“hysteresis loop” is the rotation direction, that is, the top path is the clockwise rotation, and the bottom path is the counter-clockwise direction. This graph makes it evident that movement is identical during the slide (slope 1:1), but there is a delay before true sliding starts after each turning point. As the machine is controlled by the feedback from the rotation sensor it keeps to the given angle range (i.e. +0.5° to −0.5°), while the extensometer shows that actual sliding angle is somewhat lower (ca +0.4° to −0.4°). The deviation of some 0.1° leading to a transition between the two straight lines is due to elastic and plastic strains in the machine, the gripping, and the test bars on the change in direction of rotation of the machine. However, on sliding under constant torque, the relation is constant and this provides evidence for the built-in rotation sensor being a sufficient measure of relative movement in the contact surface during the sliding motion.
Furthermore, the effect of the angles of rotation on the directional dependence can be seen in Fig. 6. The experiments R1 and R2 were repeated for the same overall distance of contact. Test R1 was run for 1000 cycles with range ± 0.5°, and Test R2 was run for 50 cycles with range ± 10°. It can be seen that the variation during each rotation is decreased with when the angles of rotation are increased. It was also found that the friction coefficient stabilizes already within the first few cycles in Test R2, while it takes nearly 350 cycles to stabilize in Test R1 (Fig. 6d). It should be noted that these cycles are equivalent to approximately 25 mm and 65 mm of accumulated travel, as measured by the machine rotation sensor, respectively. An additional test, R3, was run to verify the stabilization time, and it was found that for ± 2°, the friction coefficient stabilized after approximately 50 mm of travel. 3.2. Normal force
3. Results Fig. 7 shows the torque versus angle of rotation for several experiments, T1-T4 in Table 2. It can be seen that the torque increases with increasing compressive normal force. This is expected since the torque is directly related to the applied normal force, as previously mentioned. However, there seems to be a more stable behaviour between the torque and rotation with larger angles of rotation that is, less spread in the measured values of the torque, which in turn influences the friction coefficient calculations. Variations of the friction coefficient measurements with respect to the direction of rotation are observed, however the effect is diminished with larger compressive forces. For this reason, most experiments were conducted with higher compressive forces of 2 kN or 4 kN, yielding nominal contact pressures of 70 MPa and 140 MPa, respectively. Fig. 8 shows the effect of varying the normal force on the friction coefficient. It can be seen from Table 4 that the average values are similar, however the spread between the data points as well as the difference between the maximum and minimum values is significant for small normal force, and decreases with increasing normal force, as seen in Fig. 8.
The results from the experiments evaluating the effect of varying the rate of rotation (0.1°/sec to 5°/sec), the angles of rotation ( ± 0.5°–10°), as well as the average normal contact force (−0.5 kN to −8 kN) on the friction coefficient are presented, and the effect of each parameter is treated in its own section. 3.1. Rotation The friction coefficient seems to be somewhat dependent on the direction of the rotation, that is, the friction coefficient calculated for the clockwise and counter-clockwise rotation vary. This can clearly be seen in Fig. 4, where experiments were repeated with first cycle starting in clockwise (+) and counter-clockwise (−) directions. Higher values are measured for one direction, that is, one half of each cycle. However, the maximum and minimum values of the friction coefficient, as well as the average, remain approximately the same in all tests irrespective of the initial rotation direction, as summarized in Table 3 and Fig. 4. Furthermore, no significant difference is observed in the wear and deformation present on the specimen surfaces (Fig. 5). It can also be seen from Fig. 4 that the rate of rotation does not have a significant effect on the measured friction coefficient. All parameters except the rate of rotation were kept the same for both series. Tests D1 and D2 were run at 1°/s and Tests D3 and D4 were run at 5°/s. The minimum and maximum values of the friction coefficient are approximately the same, for both sets of tests with the same rate, but the average friction coefficient is slightly increased with an increased rate of rotation.
3.3. Dry vs wet After the friction experiments, the surfaces of the test bars were investigated. Fig. 9 shows the large deformation on the used test bars which occurred for testing in dry, room-temperature condition, as well as the significant amount of third body particles (wear debris) which can be observed when the halves are separated. The surface topography resembles that found in squat defects, which have been examined in previous work [12]. The ridges which are present on the crack faces, 3
Wear 432-433 (2019) 102910
C. Jessop and J. Ahlström
Fig. 4. Friction coefficient as a function of time, showing the effect of the direction of rotation.
approximately the same regardless of the direction and rate of rotation. However, the difference between the maximum and minimum values decreases with increasing angles of rotation, as shown in Fig. 6. A possible reason for the directional dependence is the starting topography of the test bars due to machining. That is, the as-machined surface forms a spiral on the steel halves which has a preferred direction of motion in the contact. The directional dependence of the tests can be associated to real-life applications, in terms of the wheel/rail contact, as well as the contact between the crack faces. During maintenance, rail grinding, which is common practice, might have an impact of the friction coefficient in the contact between the wheel and rail. Furthermore, the crack morphology could lead to preferential motion between the crack faces due to the shape and direction of the ridges. In the experiment, the direction of sliding is restricted by the design of the test pieces, and the two halves are prevented from any non-circular relative movement. In a squat crack, depending on the traction in the wheel-rail contact surface relative to the crack shape and position will allow for more free relative sliding within the crack. This will affect the local condition for mixed mode crack propagation. A smaller range of angles of rotation would be less affected by the starting surface roughness. Furthermore, the effect of the initial roughness on the formation of ridges is more significant at larger range of angles. The surfaces after testing, shown in Fig. 13 show little deformation and wear in the test with small range of angles of rotation, and the periphery remains untouched (shown in detail in Fig. 14). Meanwhile, the larger angles of rotation lead to a more uniform deformation along the entire contact surface (Fig. 13). Similarly, it has been found that an increased stabilized friction coefficient results due to larger rotation angles due to the larger tangential force induced by a more severe surface [13]. It can also be seen that the layer of debris on
Table 3 Friction coefficient values for direction of rotation. Test number
Direction
Rate of rotation (°/sec)
Average friction coefficient
D1 D2 D3 D4
+ – + –
1 1 5 5
0.617 0.625 0.653 0.632
shown in Fig. 9c are thought to originate from the relative sliding between the crack faces during the passage of wheels on the rails. Fig. 10 shows the surfaces of the test bars after testing in dry versus wet conditions. It is clearly seen that the deformation and wear debris are present to a greater extent on the dry test bar. Furthermore, it was observed that any wear which occurred on the surfaces of the wet test bars was near the inner radius of contact (Figs. 10 and 11). That is, the exterior remains almost the same as the initial state. It can also be seen in Fig. 12 that the friction coefficient is decreased when the tests are run in wet conditions. The average friction coefficient calculated for dry and wet conditions are 0.63 and 0.38, respectively.
4. Discussion 4.1. Rotation With respect to rotation, the parameters which were varied were the range of angles and the rate of rotation. It was found that the friction coefficient is affected by the direction of the rotation (clockwise versus counter-clockwise). The maximum and minimum values remain
Fig. 5. Surface of test bars after testing with first cycle starting in (a) clockwise and (b) counter-clockwise rotation direction. No significant difference is observed in the wear and deformation present on the specimen surfaces. 4
Wear 432-433 (2019) 102910
C. Jessop and J. Ahlström
Fig. 6. Friction coefficient as a function of time for increasing angles of rotation in Tests (a) R1, (b) R3, and (c) R2. The time to stabilization in the three tests is shown in (d).
clockwise versus the counter-clockwise rotation direction is less significant for higher loads, as shown in Fig. 8. This can also be related to the torque, since the friction coefficient is simply a factor between the normal force applied and the measured torque.
the small angles sample is uniform and quite dense. The one on the larger angles is more “loose” and accumulates in certain areas. Some areas remain untouched on the surface as well, and this seems to coincide with where the wear debris accumulate. The presence of these debris is thought to have an effect of the stabilization of the friction coefficient, through the “dynamic balance between debris escape and formation at contacting surfaces” [14].
4.3. Dry vs wet The test bar surfaces were observed after testing in dry and wet conditions. It was observed that there is large deformation occurring for the dry tests, especially at large angles of rotation, accompanied by a large amount of wear debris. However, the test bars from wet condition remain almost in the initial state, with some deformation and debris accumulation in the centre of the contact surface. This is in agreement with the friction coefficient values which have been calculated. That is, the larger friction coefficients calculated for dry testing conditions suggest more plastic deformation occurring, since friction is due in part to the deformation of the materials which are in contact. It seems the formation of ridges on the surface would increase the friction occurring between the steel halves, and this effect is more present with larger angles of rotation, as previously discussed. In wet test condition, the deformation and debris accumulation occurs only in the smallest radius of the contact surfaces. This suggests
4.2. Normal force The results presented show that an increase in the normal force causes an increase in the torque measured. As mentioned, this is to be expected, since the torque is directly related to the applied normal force for a given friction coefficient. However it was found that the behaviour is more stable with increased normal loads. This could be related to the initial surface texture present on the steel test bars, for which a higher compressive load would keep the entire surface active as the contact between the ridges remain the same throughout the rotation, causing less fluctuation. Furthermore, the directional dependence of the friction coefficient remains for tests in the range of normal forces examined. However, the spread between the friction coefficient values calculated in the
Fig. 7. Effect of varying normal force and angles of rotation on torque response. 5
Wear 432-433 (2019) 102910
C. Jessop and J. Ahlström
Fig. 8. Friction coefficient with respect to time for different normal forces equivalent to: (a) 17.5 MPa, (b) 70 MPa, and (c) 140 MPa (Tests F1, F2, and F3, respectively). The difference between the maximum and minimum values decreases with increasing normal force.
5. Conclusions
Table 4 Friction coefficient values for varying normal force. Test number
Normal force (kN)
Average contact pressure (MPa)
Average friction coefficient
F1 F2 F3
0.5 2 4
17.5 70 140
0.679 0.590 0.653
As a result of the present study, it can be concluded that: (1) The machined finish of the steel surfaces leads to an initially low coefficient of friction. After some run-in time, the coefficient of friction increases until it stabilizes. The friction coefficient then decreases slightly with more use. In terms of this effect on crack growth would be that the products which gather between the crack faces due to wear, will lead to a decreased coefficient of friction, since this third body layer acts as a lubrication. It is likely that such lubrication within a crack, facilitates crack growth by increasing the net stress intensity at the crack tip. (2) The repeated twisting of the two steel faces leads to the formation of asperities on the faces. This feature similar to the “ridges” which have been observed on the crack faces of so-called squats in railway rails. The compressive forces added to the twisting motion, imposes preferential deformation of the materials. It is thought that the
that the intrusion of water during twisting prevents wear of the outermost radius, which may be due to lower shear stress between the steel surfaces. The observation that wear debris in the inner part of the contact patch seem to prevent wear damage in the outer contact, supports the hypothesis of a wedge effect caused by the wear debris. If accumulation of wear debris in between the two halves of the test bar leads to separation, the corresponding mechanism in a squat crack will be crack opening by local wedge effects. This increases the crack growth rate by adding a Mode I crack opening component. 6
Wear 432-433 (2019) 102910
C. Jessop and J. Ahlström
Fig. 9. (a) Wear particles between surfaces after testing. (b) Surface of test bar after friction experiments resembling (c) detail of ridges on squat crack face.
Fig. 10. Surfaces of test bar after testing in (left) Test W2 (dry) and (right) Test W3 (wet) conditions. The more pronounced deformation and wear in the dry conditions is apparent and a reflection of the larger coefficients of friction calculated. Fig. 11. Surfaces of test bar after testing in (left) Test W2 (dry) and (right) Test W3 (wet) conditions. The more pronounced deformation and wear in the dry conditions is apparent and a reflection of the larger coefficients of frictions calculated. The periphery of the contact surface in the wet tests remains in original condition, while the inner radius is worn and corroded.
peak values) decreases with increasing normal force. (5) The average coefficient of friction decreases when water is present between the steel surfaces. This decrease leads to less wear, and a smoother, more uniform deformation of the surfaces. The deformation is concentrated in the inner radius of contact. It appears that the presence of water between the surfaces would lead to no wear/deformation as far as the water penetrates.
surface asperities which appear on the crack faces of rolling contact fatigue cracks both contribute to crack closure, and steer the relative displacement between the crack faces and thus affect the SIFs related to Mode II and III. (3) Increasing the angles of rotation increases the formation of ridges on the two surfaces, and leads to higher maximum coefficient of friction. Larger angles of rotation (e.g. ± 10°) also lead to more significant directional dependence as compared to very small (e.g. ± 0.5°) range. This effect is probably due to the initial machined surface finish. (4) The average coefficient of friction is similar for most experiments, however the variation during the tests (i.e. the difference between
The effect of crack face friction on crack propagation is an important issue which requires further investigation. The purpose of the present study is to determine the changes in coefficient of friction from different parameters. The results are expected to provide suitable data 7
Wear 432-433 (2019) 102910
C. Jessop and J. Ahlström
Fig. 12. Calculated friction coefficient for tests in dry and wet conditions.
Fig. 13. Surface deformation after same “sliding distance”. The rate of rotation and normal force are the same for both tests, the only varied parameter is the range of angles: (a) ± 0.5°, and (b) ± 10°.
Joint Technology Initiative Shift2Rail through contract No. 730841, as well as partly in the project In2Track2 under grant agreement No 826255. Appendix A. Supplementary data Supplementary data to this article can be found online at https:// doi.org/10.1016/j.wear.2019.05.025. References [1] A.D. Hearle, K.L. Johnson, Mode II stress intensity factors for a crack parallel to the surface of an elastic half-space subjected to a moving point load, J. Mech. Phys. Solids 33 (1) (1985) 61–81. [2] S. Sheppard, J.R. Barber, M. Comninou, Short subsurface cracks under conditions of slip and stick caused by a moving compressive load, J. Appl. Mech. 52 (1985) 811–817. [3] A.F. Bower, The influence of crack face friction and trapped fluid on surface initiated rolling contact fatigue cracks, J. Tribol. 110 (1988) 704–711. [4] D.I. Fletcher, P. Hyde, A. Kapoor, Investigating fluid penetration of rolling contact fatigue cracks in rails using a newly developed full-scale test facility, J. Rail Rapid Transit 221 (2007) 35–44. [5] K. Cvetkovski, J. Ahlström, M. Norell, C. Persson, Analysis of wear debris in rolling contact fatigue cracks of pearlitic railway wheels, Wear 314 (2014) 51–56. [6] U. Olofsson, Y. Zhu, S. Abbasi, R. Lewis, S. Lewis, “Tribology of the wheel-rail contact – aspects of wear, particle emission and adhesion, Int. J. Veh. Mech. Mobil. 51 (7) (2013) 1091–1120. [7] S. Simon, A. Saulot, C. Dayot, X. Quost, Y. Berthier, Tribological characterization of rail squat defects, Wear 297 (2013) 926–942. [8] U. Olofsson Sundvall, Influence of leaf, humidity and applied lubrication on friction in the wheel-rail contact: Pin-on-disc experiments, Proc. Inst. Mech. Eng. 218 (3) (2004) 235–242. [9] Y. Berthier, S. Descartes, M. Busquet, E. Niccolini, C. Desrayaud, L. Baillet, The role and effects of the third body in the wheel – rail interaction, Fatigue Fract. Eng. Mater. Struct. 27 (2004) 423–436.
Fig. 14. Surface after testing. The deformation and accumulation is concentrated in the centre of the contact surface, while the periphery essentially remains as initial condition, including machining groves.
for crack face friction in upcoming experimental and numerical studies of RCF crack propagation. Acknowledgments This work has been a part of research activities within the Centre of Excellence CHARMEC (CHAlmers Railway MEChanics, www.charmec. chalmers.se). It is partly financed within the European Horizon 2020 8
Wear 432-433 (2019) 102910
C. Jessop and J. Ahlström
[13] X. Wang, D. Wang, D. Zhang, S. Ge, J. Alexander, Effect of torsion angle on tensiontorsion multiaxial fretting fatigue behaviors of steel wires, Int. J. Fatigue 106 (2018) 159–164. [14] D. Wang, D. Zhang, S. Ge, “Fretting – fatigue behavior of steel wires in low cycle fatigue, Mater. Des. 32 (10) (2011) 4986–4993.
[10] N. Gates, A. Fatemi, Friction and roughness induced closure effects on shear-mode crack growth and branching mechanisms, Int. J. Fatigue 92 (2016) 442–458. [11] M. Kaneta, M. Suetsugu, Y. Murakami, Mechanism of surface crack growth in lubricated rolling/sliding spherical contact, J. Appl. Mech. 53 (1986) 354–360. [12] C. Jessop, J. Ahlström, L. Hammar, S. Fæster, H.K. Danielsen, 3D characterization of rolling contact fatigue crack networks, Wear 366–367 (2016) 392–400.
9
Contents lists available at ScienceDirect
Wear journal homepage: www.elsevier.com/locate/wear
Friction between pearlitic steel surfaces
T
∗
Casey Jessop , Johan Ahlström Department of Industrial and Materials Science, Chalmers University of Technology, Gothenburg, Sweden
A R T I C LE I N FO
A B S T R A C T
Keywords: Rail steel Rail-wheel tribology Rolling friction Sliding wear Third-body layer
Experiments rubbing two pearlitic rail steel surfaces against each other were done using an axial-torsion test rig. After the experiments were completed, the surfaces were observed under stereomicroscope, and the wear debris were examined. The aim of the experiments was to evaluate the variation of friction characteristics between two surfaces of pearlitic rail steel, mimicking the friction which occurs between the faces of a crack during rolling contact fatigue (RCF) loading. For both dry and wet conditions, the sliding velocity, the angles of rotation, and the average normal force were varied. It was found that the most significant effect of changing angles of rotation is on the formation of ridges on the contact surface with large angles, leading to a higher friction coefficient. The presence of water reduces the friction coefficient and leads to less deformation and wear in the contact surface.
1. Introduction
well as the presence of fluids and wear debris. These phenomena remain less studied than the crack initiation associated to squats. However, crack face friction has a major influence on the crack growth since the friction as well as the surface asperities which are present on the crack faces have an influence on the stress intensity factors (SIF) in shear sliding (Mode II) and tearing (Mode III) loading, as well as on crack closure (Mode I loading). Research done on shear mode crack propagation has also shown that the main driving forces for branching of Mode II cracks are crack face friction and wedging, caused by surface asperities [10], particularly if the contacting crack faces are closed [11]. The present study evaluates the friction between two pearlitic R260 rail steel surfaces in different combinations of compression and torsional loading under dry and wet conditions in order to determine the influence of these parameters on the friction coefficient.
In the railway industry, tribology is an important topic since the contact in between rails and wheels is subjected to varying temperature and climatic conditions in operation, which affect both friction and wear. Another important concern is the effect of environment on crack face tribology, that is, the contact between crack faces during crack propagation due to rolling contact fatigue (RCF) loading. The effect of crack face friction on the propagation of subsurface cracks has been studied in the past [1,2], however, since these cracks have no opening to the surface, the penetration of water and other fluids was not considered in the models. Nevertheless, the presence of water or another fluid is thought to be necessary for the propagation of squat and squattype defects in rails which are surface breaking cracks driven by rolling contact fatigue loading [3,4]. It is thought that the friction has an important effect on crack propagation and branching [3,5–7]. Furthermore, the presence of third-body materials such as oxides or sheared bulk material, is also an important consideration [5,6]. Several studies have been conducted using, for example, pin-on-disk and test rig experiments, and the analysis of third-body materials present in field as a consequence of wheel-rail contact in different stick and slip conditions, has been compared to laboratory experiments [4,8,9]. Squat and squat-type defects are of particular interest in the field of RCF in rails. The darkened surface associated to these defects is thought to originate from a depression of the surface, due to the loss of material and wear occurring between the crack faces, which causes the surface of the rail to be worn at a different rate than the rest of the rail. The growth of squats is significantly influenced by the crack face friction as
∗
2. Experimental The aim of the experiments is to evaluate the friction coefficient between two surfaces of pearlitic rail steel, mimicking the friction which occurs between the faces of a crack in a rail during repeated wheel-rail contact. Two halves of a test bar (shown schematically in Fig. 1), representing the crack faces, are subjected to a normal force and rotated back and forth against each other in different conditions. The test bars were designed with a steering pin going into a roller bearing to avoid any lateral translations between the halves, yielding a well-controlled sliding path without disturbing the torque measurement. The product of the normal stress and the friction coefficient is equivalent to
Corresponding author. E-mail address: [email protected] (C. Jessop).
https://doi.org/10.1016/j.wear.2019.05.025 Received 9 January 2019; Received in revised form 16 May 2019; Accepted 23 May 2019 Available online 28 May 2019 0043-1648/ © 2019 Elsevier B.V. All rights reserved.
Wear 432-433 (2019) 102910
C. Jessop and J. Ahlström
Fig. 1. (a) Top view of R260 rail steel surfaces in machined condition, and (b) schematic of test bar used in friction experiments with dimensions in mm.
machine offset. The exact position of the spots in true contact are not known, but the reasonably thin (2 mm), circular contact band enables an approximate quantification according to (Equation (1)):
Table 1 Experiment parameters.
Rate of rotation Angles of rotation Normal force
Min
Max
0.1°/sec ± 0.5° −0.5 kN
5°/sec ± 10° −8 kN
μ = T/F·r
Where μ is the friction coefficient, T is torque [kN·mm], F is the normal force [kN], and r is the average radius in the contact [mm]. Tests were run in dry and wet conditions at room temperature. For the wet set up, a funnel filled with deionized water was attached to the test bar, as shown in Fig. 2. The surfaces are initially dry; the water is only added to the funnel after the faces of the two halves are pressed together with the full axial load, and the contact surface is thus sealed. That is, the water only surrounds the test bar, and any penetration which occurs is due to the intrusion of water in between the steel surfaces during twisting. The rotation sensor of the machine was found to give a sufficient measure of the relative motion/rotation of the steel halves for larger angles of rotation. This can be shown from the relation between the rotation measured by the machine sensor and that measured by the extensometer. The correlation is found using the following relation (Equation (2)):
Table 2 Detailed parameters of tests included in results. Some tests are repeated, but referred to by different names for clarify of presentation. (Tests W3 and W4 were run in wet conditions.) Test
Rate of rotation (°/sec)
E 0.1 Direction D1 1 D2 1 D3 5 D4 5 Rotation R1 0.5 R2 0.5 R3 0.5 Torque response T1 5 T2 5 T3 5 T4 0.1 Normal force F1 5 F2 5 F3 5 Dry vs wet W1 1 W2 0.5 W3 1 W4 1
Angles of rotation (°)
Normal force (kN)
0.5
−4
5 5 5 5
−4 −4 −4 −4
0.5 10 2
−4 −4 −4
5 1 5 1
−0.5 −2 −4 −4
5 5 5
−0.5 −2 −4
5 10 10 5
−4 −4 −4 −8
(1)
tan(φ) = r·θ/L
(2)
Where φ is the torsional strain [rad] measured by the extensometer on the test specimen, r·θ is the torsional rotation measured by the machine rotation sensor, and L is the distance between the extensometer pins [mm]. Fig. 3 (Test E) shows the direct relation between the rotation measured from the machine and that from the extensometer. The first cycle (shown as the dashed line) is different from the remaining cycles in the experiment. The points lie on two lines with slope equal to 1 (blue line is a reference). The reason for the two paths, and the
the friction stress which is present between opposing crack faces under ideal contact conditions [10]. All tests were run in a MTS 809 axialtorsion servo-hydraulic testing machine with closed loop control. The tests were run in axial force, and torsional rotation control, and in some tests an extensometer was used to measure torsional rotation locally with high precision. The test bars were in machined condition. The effect of the different conditions is measured using the friction coefficient as a basis for comparison. The parameters which are varied are the sliding velocity (rate of rotation), the angles of rotation (stroke/length of sliding movement), and the average normal force (see Table 1). A more detailed description of the test parameters for given tests is included in Table 2. After the experiments were completed, the surfaces were observed using stereomicroscopy and the wear products evaluated. The friction coefficient is determined by measuring the torque during the experiments. That is, the friction is determined by the torque required to rotate the sample under a given normal force. The torque values were centred after the experiments to compensate for the
Fig. 2. Experimental setup for tests run in wet conditions. The test bar is mounted into the machine and not until after full axial loading, the funnel is filled with deionized water. 2
Wear 432-433 (2019) 102910
C. Jessop and J. Ahlström
Fig. 3. Correlation between the machine rotation (r·θ/L) and the relative rotation of the test specimen (tan(φ)) during testing. The “points” lie on two lines with slope 1:1, showing the direct relation between the rotation measured by the machine and the extensometer. (The first cycle is shown with dashed line, and a 1:1 slope is plotted for reference.)
“hysteresis loop” is the rotation direction, that is, the top path is the clockwise rotation, and the bottom path is the counter-clockwise direction. This graph makes it evident that movement is identical during the slide (slope 1:1), but there is a delay before true sliding starts after each turning point. As the machine is controlled by the feedback from the rotation sensor it keeps to the given angle range (i.e. +0.5° to −0.5°), while the extensometer shows that actual sliding angle is somewhat lower (ca +0.4° to −0.4°). The deviation of some 0.1° leading to a transition between the two straight lines is due to elastic and plastic strains in the machine, the gripping, and the test bars on the change in direction of rotation of the machine. However, on sliding under constant torque, the relation is constant and this provides evidence for the built-in rotation sensor being a sufficient measure of relative movement in the contact surface during the sliding motion.
Furthermore, the effect of the angles of rotation on the directional dependence can be seen in Fig. 6. The experiments R1 and R2 were repeated for the same overall distance of contact. Test R1 was run for 1000 cycles with range ± 0.5°, and Test R2 was run for 50 cycles with range ± 10°. It can be seen that the variation during each rotation is decreased with when the angles of rotation are increased. It was also found that the friction coefficient stabilizes already within the first few cycles in Test R2, while it takes nearly 350 cycles to stabilize in Test R1 (Fig. 6d). It should be noted that these cycles are equivalent to approximately 25 mm and 65 mm of accumulated travel, as measured by the machine rotation sensor, respectively. An additional test, R3, was run to verify the stabilization time, and it was found that for ± 2°, the friction coefficient stabilized after approximately 50 mm of travel. 3.2. Normal force
3. Results Fig. 7 shows the torque versus angle of rotation for several experiments, T1-T4 in Table 2. It can be seen that the torque increases with increasing compressive normal force. This is expected since the torque is directly related to the applied normal force, as previously mentioned. However, there seems to be a more stable behaviour between the torque and rotation with larger angles of rotation that is, less spread in the measured values of the torque, which in turn influences the friction coefficient calculations. Variations of the friction coefficient measurements with respect to the direction of rotation are observed, however the effect is diminished with larger compressive forces. For this reason, most experiments were conducted with higher compressive forces of 2 kN or 4 kN, yielding nominal contact pressures of 70 MPa and 140 MPa, respectively. Fig. 8 shows the effect of varying the normal force on the friction coefficient. It can be seen from Table 4 that the average values are similar, however the spread between the data points as well as the difference between the maximum and minimum values is significant for small normal force, and decreases with increasing normal force, as seen in Fig. 8.
The results from the experiments evaluating the effect of varying the rate of rotation (0.1°/sec to 5°/sec), the angles of rotation ( ± 0.5°–10°), as well as the average normal contact force (−0.5 kN to −8 kN) on the friction coefficient are presented, and the effect of each parameter is treated in its own section. 3.1. Rotation The friction coefficient seems to be somewhat dependent on the direction of the rotation, that is, the friction coefficient calculated for the clockwise and counter-clockwise rotation vary. This can clearly be seen in Fig. 4, where experiments were repeated with first cycle starting in clockwise (+) and counter-clockwise (−) directions. Higher values are measured for one direction, that is, one half of each cycle. However, the maximum and minimum values of the friction coefficient, as well as the average, remain approximately the same in all tests irrespective of the initial rotation direction, as summarized in Table 3 and Fig. 4. Furthermore, no significant difference is observed in the wear and deformation present on the specimen surfaces (Fig. 5). It can also be seen from Fig. 4 that the rate of rotation does not have a significant effect on the measured friction coefficient. All parameters except the rate of rotation were kept the same for both series. Tests D1 and D2 were run at 1°/s and Tests D3 and D4 were run at 5°/s. The minimum and maximum values of the friction coefficient are approximately the same, for both sets of tests with the same rate, but the average friction coefficient is slightly increased with an increased rate of rotation.
3.3. Dry vs wet After the friction experiments, the surfaces of the test bars were investigated. Fig. 9 shows the large deformation on the used test bars which occurred for testing in dry, room-temperature condition, as well as the significant amount of third body particles (wear debris) which can be observed when the halves are separated. The surface topography resembles that found in squat defects, which have been examined in previous work [12]. The ridges which are present on the crack faces, 3
Wear 432-433 (2019) 102910
C. Jessop and J. Ahlström
Fig. 4. Friction coefficient as a function of time, showing the effect of the direction of rotation.
approximately the same regardless of the direction and rate of rotation. However, the difference between the maximum and minimum values decreases with increasing angles of rotation, as shown in Fig. 6. A possible reason for the directional dependence is the starting topography of the test bars due to machining. That is, the as-machined surface forms a spiral on the steel halves which has a preferred direction of motion in the contact. The directional dependence of the tests can be associated to real-life applications, in terms of the wheel/rail contact, as well as the contact between the crack faces. During maintenance, rail grinding, which is common practice, might have an impact of the friction coefficient in the contact between the wheel and rail. Furthermore, the crack morphology could lead to preferential motion between the crack faces due to the shape and direction of the ridges. In the experiment, the direction of sliding is restricted by the design of the test pieces, and the two halves are prevented from any non-circular relative movement. In a squat crack, depending on the traction in the wheel-rail contact surface relative to the crack shape and position will allow for more free relative sliding within the crack. This will affect the local condition for mixed mode crack propagation. A smaller range of angles of rotation would be less affected by the starting surface roughness. Furthermore, the effect of the initial roughness on the formation of ridges is more significant at larger range of angles. The surfaces after testing, shown in Fig. 13 show little deformation and wear in the test with small range of angles of rotation, and the periphery remains untouched (shown in detail in Fig. 14). Meanwhile, the larger angles of rotation lead to a more uniform deformation along the entire contact surface (Fig. 13). Similarly, it has been found that an increased stabilized friction coefficient results due to larger rotation angles due to the larger tangential force induced by a more severe surface [13]. It can also be seen that the layer of debris on
Table 3 Friction coefficient values for direction of rotation. Test number
Direction
Rate of rotation (°/sec)
Average friction coefficient
D1 D2 D3 D4
+ – + –
1 1 5 5
0.617 0.625 0.653 0.632
shown in Fig. 9c are thought to originate from the relative sliding between the crack faces during the passage of wheels on the rails. Fig. 10 shows the surfaces of the test bars after testing in dry versus wet conditions. It is clearly seen that the deformation and wear debris are present to a greater extent on the dry test bar. Furthermore, it was observed that any wear which occurred on the surfaces of the wet test bars was near the inner radius of contact (Figs. 10 and 11). That is, the exterior remains almost the same as the initial state. It can also be seen in Fig. 12 that the friction coefficient is decreased when the tests are run in wet conditions. The average friction coefficient calculated for dry and wet conditions are 0.63 and 0.38, respectively.
4. Discussion 4.1. Rotation With respect to rotation, the parameters which were varied were the range of angles and the rate of rotation. It was found that the friction coefficient is affected by the direction of the rotation (clockwise versus counter-clockwise). The maximum and minimum values remain
Fig. 5. Surface of test bars after testing with first cycle starting in (a) clockwise and (b) counter-clockwise rotation direction. No significant difference is observed in the wear and deformation present on the specimen surfaces. 4
Wear 432-433 (2019) 102910
C. Jessop and J. Ahlström
Fig. 6. Friction coefficient as a function of time for increasing angles of rotation in Tests (a) R1, (b) R3, and (c) R2. The time to stabilization in the three tests is shown in (d).
clockwise versus the counter-clockwise rotation direction is less significant for higher loads, as shown in Fig. 8. This can also be related to the torque, since the friction coefficient is simply a factor between the normal force applied and the measured torque.
the small angles sample is uniform and quite dense. The one on the larger angles is more “loose” and accumulates in certain areas. Some areas remain untouched on the surface as well, and this seems to coincide with where the wear debris accumulate. The presence of these debris is thought to have an effect of the stabilization of the friction coefficient, through the “dynamic balance between debris escape and formation at contacting surfaces” [14].
4.3. Dry vs wet The test bar surfaces were observed after testing in dry and wet conditions. It was observed that there is large deformation occurring for the dry tests, especially at large angles of rotation, accompanied by a large amount of wear debris. However, the test bars from wet condition remain almost in the initial state, with some deformation and debris accumulation in the centre of the contact surface. This is in agreement with the friction coefficient values which have been calculated. That is, the larger friction coefficients calculated for dry testing conditions suggest more plastic deformation occurring, since friction is due in part to the deformation of the materials which are in contact. It seems the formation of ridges on the surface would increase the friction occurring between the steel halves, and this effect is more present with larger angles of rotation, as previously discussed. In wet test condition, the deformation and debris accumulation occurs only in the smallest radius of the contact surfaces. This suggests
4.2. Normal force The results presented show that an increase in the normal force causes an increase in the torque measured. As mentioned, this is to be expected, since the torque is directly related to the applied normal force for a given friction coefficient. However it was found that the behaviour is more stable with increased normal loads. This could be related to the initial surface texture present on the steel test bars, for which a higher compressive load would keep the entire surface active as the contact between the ridges remain the same throughout the rotation, causing less fluctuation. Furthermore, the directional dependence of the friction coefficient remains for tests in the range of normal forces examined. However, the spread between the friction coefficient values calculated in the
Fig. 7. Effect of varying normal force and angles of rotation on torque response. 5
Wear 432-433 (2019) 102910
C. Jessop and J. Ahlström
Fig. 8. Friction coefficient with respect to time for different normal forces equivalent to: (a) 17.5 MPa, (b) 70 MPa, and (c) 140 MPa (Tests F1, F2, and F3, respectively). The difference between the maximum and minimum values decreases with increasing normal force.
5. Conclusions
Table 4 Friction coefficient values for varying normal force. Test number
Normal force (kN)
Average contact pressure (MPa)
Average friction coefficient
F1 F2 F3
0.5 2 4
17.5 70 140
0.679 0.590 0.653
As a result of the present study, it can be concluded that: (1) The machined finish of the steel surfaces leads to an initially low coefficient of friction. After some run-in time, the coefficient of friction increases until it stabilizes. The friction coefficient then decreases slightly with more use. In terms of this effect on crack growth would be that the products which gather between the crack faces due to wear, will lead to a decreased coefficient of friction, since this third body layer acts as a lubrication. It is likely that such lubrication within a crack, facilitates crack growth by increasing the net stress intensity at the crack tip. (2) The repeated twisting of the two steel faces leads to the formation of asperities on the faces. This feature similar to the “ridges” which have been observed on the crack faces of so-called squats in railway rails. The compressive forces added to the twisting motion, imposes preferential deformation of the materials. It is thought that the
that the intrusion of water during twisting prevents wear of the outermost radius, which may be due to lower shear stress between the steel surfaces. The observation that wear debris in the inner part of the contact patch seem to prevent wear damage in the outer contact, supports the hypothesis of a wedge effect caused by the wear debris. If accumulation of wear debris in between the two halves of the test bar leads to separation, the corresponding mechanism in a squat crack will be crack opening by local wedge effects. This increases the crack growth rate by adding a Mode I crack opening component. 6
Wear 432-433 (2019) 102910
C. Jessop and J. Ahlström
Fig. 9. (a) Wear particles between surfaces after testing. (b) Surface of test bar after friction experiments resembling (c) detail of ridges on squat crack face.
Fig. 10. Surfaces of test bar after testing in (left) Test W2 (dry) and (right) Test W3 (wet) conditions. The more pronounced deformation and wear in the dry conditions is apparent and a reflection of the larger coefficients of friction calculated. Fig. 11. Surfaces of test bar after testing in (left) Test W2 (dry) and (right) Test W3 (wet) conditions. The more pronounced deformation and wear in the dry conditions is apparent and a reflection of the larger coefficients of frictions calculated. The periphery of the contact surface in the wet tests remains in original condition, while the inner radius is worn and corroded.
peak values) decreases with increasing normal force. (5) The average coefficient of friction decreases when water is present between the steel surfaces. This decrease leads to less wear, and a smoother, more uniform deformation of the surfaces. The deformation is concentrated in the inner radius of contact. It appears that the presence of water between the surfaces would lead to no wear/deformation as far as the water penetrates.
surface asperities which appear on the crack faces of rolling contact fatigue cracks both contribute to crack closure, and steer the relative displacement between the crack faces and thus affect the SIFs related to Mode II and III. (3) Increasing the angles of rotation increases the formation of ridges on the two surfaces, and leads to higher maximum coefficient of friction. Larger angles of rotation (e.g. ± 10°) also lead to more significant directional dependence as compared to very small (e.g. ± 0.5°) range. This effect is probably due to the initial machined surface finish. (4) The average coefficient of friction is similar for most experiments, however the variation during the tests (i.e. the difference between
The effect of crack face friction on crack propagation is an important issue which requires further investigation. The purpose of the present study is to determine the changes in coefficient of friction from different parameters. The results are expected to provide suitable data 7
Wear 432-433 (2019) 102910
C. Jessop and J. Ahlström
Fig. 12. Calculated friction coefficient for tests in dry and wet conditions.
Fig. 13. Surface deformation after same “sliding distance”. The rate of rotation and normal force are the same for both tests, the only varied parameter is the range of angles: (a) ± 0.5°, and (b) ± 10°.
Joint Technology Initiative Shift2Rail through contract No. 730841, as well as partly in the project In2Track2 under grant agreement No 826255. Appendix A. Supplementary data Supplementary data to this article can be found online at https:// doi.org/10.1016/j.wear.2019.05.025. References [1] A.D. Hearle, K.L. Johnson, Mode II stress intensity factors for a crack parallel to the surface of an elastic half-space subjected to a moving point load, J. Mech. Phys. Solids 33 (1) (1985) 61–81. [2] S. Sheppard, J.R. Barber, M. Comninou, Short subsurface cracks under conditions of slip and stick caused by a moving compressive load, J. Appl. Mech. 52 (1985) 811–817. [3] A.F. Bower, The influence of crack face friction and trapped fluid on surface initiated rolling contact fatigue cracks, J. Tribol. 110 (1988) 704–711. [4] D.I. Fletcher, P. Hyde, A. Kapoor, Investigating fluid penetration of rolling contact fatigue cracks in rails using a newly developed full-scale test facility, J. Rail Rapid Transit 221 (2007) 35–44. [5] K. Cvetkovski, J. Ahlström, M. Norell, C. Persson, Analysis of wear debris in rolling contact fatigue cracks of pearlitic railway wheels, Wear 314 (2014) 51–56. [6] U. Olofsson, Y. Zhu, S. Abbasi, R. Lewis, S. Lewis, “Tribology of the wheel-rail contact – aspects of wear, particle emission and adhesion, Int. J. Veh. Mech. Mobil. 51 (7) (2013) 1091–1120. [7] S. Simon, A. Saulot, C. Dayot, X. Quost, Y. Berthier, Tribological characterization of rail squat defects, Wear 297 (2013) 926–942. [8] U. Olofsson Sundvall, Influence of leaf, humidity and applied lubrication on friction in the wheel-rail contact: Pin-on-disc experiments, Proc. Inst. Mech. Eng. 218 (3) (2004) 235–242. [9] Y. Berthier, S. Descartes, M. Busquet, E. Niccolini, C. Desrayaud, L. Baillet, The role and effects of the third body in the wheel – rail interaction, Fatigue Fract. Eng. Mater. Struct. 27 (2004) 423–436.
Fig. 14. Surface after testing. The deformation and accumulation is concentrated in the centre of the contact surface, while the periphery essentially remains as initial condition, including machining groves.
for crack face friction in upcoming experimental and numerical studies of RCF crack propagation. Acknowledgments This work has been a part of research activities within the Centre of Excellence CHARMEC (CHAlmers Railway MEChanics, www.charmec. chalmers.se). It is partly financed within the European Horizon 2020 8
Wear 432-433 (2019) 102910
C. Jessop and J. Ahlström
[13] X. Wang, D. Wang, D. Zhang, S. Ge, J. Alexander, Effect of torsion angle on tensiontorsion multiaxial fretting fatigue behaviors of steel wires, Int. J. Fatigue 106 (2018) 159–164. [14] D. Wang, D. Zhang, S. Ge, “Fretting – fatigue behavior of steel wires in low cycle fatigue, Mater. Des. 32 (10) (2011) 4986–4993.
[10] N. Gates, A. Fatemi, Friction and roughness induced closure effects on shear-mode crack growth and branching mechanisms, Int. J. Fatigue 92 (2016) 442–458. [11] M. Kaneta, M. Suetsugu, Y. Murakami, Mechanism of surface crack growth in lubricated rolling/sliding spherical contact, J. Appl. Mech. 53 (1986) 354–360. [12] C. Jessop, J. Ahlström, L. Hammar, S. Fæster, H.K. Danielsen, 3D characterization of rolling contact fatigue crack networks, Wear 366–367 (2016) 392–400.
9