You are on page 1of 10

4th INTERNATIONAL SYMPOSIUM MAINTENANCE AND REHABILITATION OF PAVEMETNS AND TECHNOLOGICAL CONTROL Belfast, 2005

Fatigue Damage Accumulation measured using Different Laboratory Techniques


Y.K. Choi, N.H. Thom, A.C. Collop
Nottingham Centre for Pavement Engineering, University of Nottingham, Nottingham, United Kingdom.

ABSTRACT: This paper reports on an extensive series of fatigue tests carried out on a particular dense graded asphalt mixture. Three different tests were conducted. Two were relatively conventional, namely uniaxial and trapezoidal; the third consisted of trafficking, by means of a small solid-tyred wheel, a beam of asphalt mixture continuously supported by an elastic foundation. Tests were carried out at three different temperatures. In each case detailed data has been recovered and the accumulation of damage during the test is represented by a loss of (average) stiffness. The principal objective of the paper is to present the data set, in particular the comparison between the three different tests and the different test temperatures, since this will be of general research interest. However, discussion is also presented regarding the likely reasons for the different numerical outcomes of the different tests, as well as the reasons for the general underlying pattern of behaviour found. Finally, comment is made on the application of such test data to crack development in a real pavement under a full-scale wheel load. KEY WORDS: Asphalt, Fatigue Cracking, Laboratory Testing 1 INTRODUCTION

The issue of fatigue cracking of asphalt and the factors which influence it has become the subject of intense research over the last few years. In practical pavement design, the Perpetual Pavement concept as introduced in the United States or, more conservatively, Long Life Pavements as defined by the UK Highways Agency, are based on the assumption that an asphalt base has the capability to survive undamaged indefinitely. Yet it is certain that tensile stresses and strains are induced in such pavement bases under traffic loads. The question therefore remains: do such strains actually cause permanent fatigue damage or is healing sufficient to restore the material as quickly as it is damaged? Molenaar (2004) presents clear evidence that permanent damage does indeed take place in practice. Turning to the laboratory assessment of fatigue, the key questions surround the interpretation of the loss of material stiffness modulus which is always found to take place during a test. Secondary questions concern the effect of loading rate and the sometimes remarkable degree of healing which some researchers have recorded. However, the purpose of this paper is not to address any of these issues directly, but rather to present a set of fatigue data, derived from three types of laboratory test, including testing under a moving wheel. Comments will certainly be made regarding the reasons or suspected reasons for the behaviour found, but the prime purpose is to make public a most interesting and complete test sequence.

TEST EQUIPMENT USED

2.1 Two-Point Trapezoidal Test The test used in this work was one developed at Nottingham during the 1990s (Rowe, 1996). It differs from the standard European test in that a larger specimen was used, but conceptually the test is the same. Figure 1 shows the arrangement diagrammatically.
40 mm

1 2
350 mm

Height

max =

3PL 4bd 2

40 mm 120 mm

Bending Stress

Figure1: Dimensions and stress distribution of Two-Point Trapezoidal test The specimens were manufactured by cutting the required shape from a slab, the slab having been made in a roller compactor. Initial slab size was 404 280 60mm. Slabs were first divided longitudinally in two and then trimmed to 40mm thickness by removing the relatively rough and voided upper portion. The trapezoidal shape was sawn from each halfslab using a specially designed sawing jig. Finally, steel end plates were attached to each specimen using an epoxy glue. The test itself was carried out in a temperature-controlled chamber. The load was applied to the top end plate by means of an electro-dynamic actuator and movement of the plate was measured by means of a Linear Variable Differential Transformer (LVDT). Tests were carried out under sinusoidal loading (about the zero load state) at a frequency of 10Hz. As shown in Figure 1, the point of maximum stress occurs at mid-height for this particular geometry. 2.2 The Uniaxial Test The trapezoidal fatigue test is representative of a class of tests which induce bending into a specimen, the other commonly used arrangement being 4-point beam bending. In contrast, it was decided that a pure tension-compression test should also be conducted. This is a much more difficult test to carry out, first with regard to specimen design and then with regard to manufacture. The aim is to generate a stress field which is almost perfectly uniaxial, but one for which the location of fracture is broadly known. This inevitably means that a necked specimen shape has to be used, leading to lower stresses at the ends, particularly at the interface between the specimen and the loading equipment. After significant initial experimentation, the specimen shape shown in Figure 2 was selected.

115 mm

163 mm

81 mm

200 mm

Curved (stress-yy)
0.10

Center
0.09

Middle Outside

0.08

0.07

Height of specimen (m)

0.06

0.05

0.04

0.03

0.02

0.01

0.00 100000

150000

200000

250000

300000

350000

400000

450000

500000

550000

600000

650000

700000

Stress (pa)

Figure 2: Uniaxial specimen used plus FE analysis of stress state Because of the inevitable non-uniformity of stress state in such a specimen, a Finite Element (FE) analysis was carried out and the results for stress in the axial direction are shown in Figure 2. This revealed a relatively moderate difference between the stress within the body of the specimen and that at the edge, which meant that a minor correction had to be applied to the average stress in order to determine the maximum value. Having established that the shape shown in Figure 2 was appropriate for uniaxial testing, it was then necessary to develop a suitable method of specimen manufacture. This was achieved by means of a split mould (Figure 3) and a precisely defined compaction protocol. The method which was found to produce suitably uniform specimens of appropriate density and void content was first to compact using a vibrating hammer with a 35mm diameter head in two lifts. A top platen (suitably greased) was added and the specimen transferred into a static press. A predetermined compressive load was applied for 45 seconds to the top platen, and the specimen was then turned over and a further 45 seconds of pressure applied to the bottom platen. The specimen was extracted from the mould the following day and upper and lower

test plates glued to the ends> The two plates were set exactly parallel to each other by means of a purpose-designed jig.
115 mm

163 mm 81 mm 200 mm

Figure 3: Uniaxial specimen, mould and static press used for compaction The testing itself was carried out in a servo-hydraulic apparatus and loading was applied at a frequency of 10Hz. Deformation over the central part of the specimen was monitored during the test by means of a pair of LVDTs attached to purpose-designed glued-on studs. 2.3 Beam on Elastic Foundation under Moving Wheel Load The final test carried out in this series was intended to form a step toward the realistic conditions experienced by an asphalt layer in a road and its principal distinctive feature was that the load was applied by means of a moving wheel. This therefore induced reversing shear stresses in a similar way to that experienced in a pavement; it also induced an approximately realistic combination of compressive and tensile stress as the wheel approached and left each section of the beam. Figure 4 shows a schematic of the test.
Loading length = 224mm 50 mm Beam 60 mm

150 mm

Rubber Foundation 404 mm

60 mm

Figure 4: Dimensions used in Beam on Elastic Foundation tests The beam and its rubber support were housed in a mould and the whole arrangement placed in a facility normally used for wheel track rutting tests. The wheel had a solid rubber tyre; its width was just under 50mm. The rate of testing was about 40 passes to the minute, giving a pulse duration at the bottom of the beam of a little under a second. For practical reasons, restraints were placed around the beam. Side restraints added negligible stress but ensured that the beam remained centrally placed within the mould, directly under the path of the wheel. Additionally, the top surface was restrained at each end.

This simulated the effect of a continuous asphalt pavement and ensured that there was no permanent bending induced in the beam. A key issue in this test was the means by which specimen damage was monitored. As in the other types of test, the specimens were painted white to facilitate observation of cracking. However, it was considered essential to obtain a measure of the effective stiffness of the beam during the test and this was achieved by means of pairs of strain gauges glued centrally to the underside of each beam. A series of FE computations was carried out with different beam stiffness moduli and with the wheel in different positions in order to determine the maximum expected compression and tension at the location of the strain gauges. Figure 5 shows the resulting stresses with the wheel placed centrally.
1.50E+06

Surface Middle
1.00E+06

Bottom

Stress (Pa)

5.00E+05

0.00E+00 0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

0.40

0.45

-5.00E+05

-1.00E+06

Beam length (m)

Figure 5: Beam stresses computed with the wheel positioned centrally By analysing the FE data, it was possible to determine a relationship between strain as measured by the gauges (difference between maximum tension and maximum compression) and effective beam stiffness modulus. 3 TEST DATA

The data presented in the following sections all relate to a single asphalt mixture. It was a 14mm maximum aggregate size dense graded wearing course with a carboniferous limestone aggregate and 4.9% of a straight run bitumen with measured penetration 38dmm. 3.1 Two-Point Trapezoidal Test Specimens were tested at three temperatures, 10, 20 and 30C, and at three load levels at each temperature. Tests were conducted under load control, equal loads being applied in the two directions. In addition to cyclic deflection, mean position of the top of the beam was monitored and found to deviate only very slightly from the initial state.

In the majority of tests, failure occurred approximately as expected, around mid height of the specimen, or slightly lower. Since the distribution of stress (and therefore damage) is not uniform around the mid height (see Figure 1), a failure rather lower than mid height is logical. However, in some cases, at 10C, failure occurred next to the glue at the base of the specimen, and these results have therefore been discarded. Figure 6 shows a typical plot of stiffness reduction during a test.
9000

Effective Stiffness Modulus (MPa)

8000 7000 6000 5000 4000 3000 2000 1000 0 0 2000 4000 6000 8000 10000 12000 14000 16000

Number of Load Applications

Figure 6: Trapezoidal fatigue test, 20C, 1750kPa maximum stress, 10Hz The overall form of Figure 6 is very familiar from numerous publications (e.g. Bodin et al, 2003, Baaj et al, 2003, Thom et al, 2002), illustrating the three phases: 1) a sharp initial stiffness loss; 2) an approximately linear portion; 3) final failure. Figure 7 illustrates the effect of temperature and load level on the shape of this stiffness reduction curve by normalising both axes and by averaging data from several specimens in each case.
1 0.9 0.8 Relative Stiffness 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0 0.2 0.4 0.6 0.8 1 Proportion of Life Used Low strain Medium strain High strain

10C

20C 30C

Figure 7: Stiffness reduction data from all trapezoidal fatigue tests

Figure 7 shows very clearly the quite different form of behaviour applying at different temperatures. The initial stiffness loss is much less at 10C than at 20 or 30C, for example, which would appear to be evidence against the commonly stated view that this stiffness loss is due to localised bitumen heating. Heating certainly occurs (researchers typically measure a degree or two e.g. Bodin et al, 2003), but it would not appear to be adequate to explain the 30C data. Nor would it explain why the shape of the stiffness reduction curve is almost unaffected by the level of stress or strain applied. With regard to the actual number of load applications sustained, Figure 8 is a conventional strain-life plot, where the strain plotted is that applying over the first few cycles, before significant material damage occurs.
1000
Trapezoidal - 10C Trapezoidal - 20C

Microstrain

Trapezoidal - 30C

100 1000 10000 100000 1000000

Number of Load Applications at Failure

Figure 8: Initial strain-life plot for all trapezoidal tests With the exception of one of the 10C tests, all the data clearly support the use of a linear relationship between initial applied strain and life to failure. Equally clearly, in this case, with this particular binder, the 30C data lies well above the 20 and 10C data. It is particularly interesting to note that, although approximately the same strain-life characteristic appears to apply at both 10 and 20C, the loss of stiffness was much more rapid at 20C, which means that the strain during the majority of the test would have been much higher in the 20C specimens than those at 10C. 3.2 The Uniaxial Test Uniaxial tests were also carried out at 10, 20 and 30C and also over a range of load levels. However, problems with the glue between specimen and platen at 10C meant that only two valid tests were conducted at that temperature. The pattern of stiffness reduction was visually similar to that from the trapezoidal test and the effect of strain level on that pattern was similarly slight. An average stiffness reduction curve was therefore obtained at each temperature, using normalised parameters and utilising data from several strain levels. In Figure 9, this is compared to a similar averaged and normalised plot for the trapezoidal tests, and it is immediately apparent that the pattern from the two tests is identical. Whilst it is certain that stress distribution is quite different in the two tests, and one might therefore expect damage to accrue at different rates, the way in which that damage affects the overall stiffness of the specimen is surprisingly similar.

1 0.9 0.8 Relative Stiffness 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0 0.2 0.4 0.6 0.8 1 Proportion of Life Used

10C

20C 30C
Trapezoidal Uniaxial

Figure 9: A comparison of stiffness reduction patterns in trapezoidal and uniaxial tests Next, Figure 10 shows the failure data from uniaxial fatigue tests on a strain-life plot.
1000 Trapezoidal - 10C Trapezoidal - 20C Trapezoidal - 30C Uniaxial - 10C Uniaxial - 20C Uniaxial - 30C

Microstrain
100 1000

10000

100000

1000000

Number of Load Applications at Failure

Figure 10: Initial strain-life plot including uniaxial fatigue data The degree of difference between the two test methods in terms of the number of load applications to failure at a given strain level is surprisingly large. However, the same trend was noted by Di Benedetto et al (2003) from tests carried out at 10C, where a factor of about 1.3 on the strain axis was found to apply between the tests. Furthermore, Figure 10 shows that the large difference between 30C and other temperatures in the trapezoidal test almost disappears in the uniaxial test. Whilst it is not the intention of this paper to attempt any detailed modelling, the following explanations for these effects are offered: a) In the uniaxial test, the entire cross section is subject to the same stress and strain level, whereas in the trapezoidal test it is only the outermost element which has the maximum stress or strain. Damage accumulation should therefore be much more rapid in the uniaxial test. If it is assumed that damage is proportional to strain to the power of 5.0 (reflecting the slopes of the characteristics shown in Figure 10), then the difference between the uniform strain distribution across the uniaxial specimen and the linearly varying distribution across the trapezoidal specimen should

b)

theoretically result in 6 times as much damage occurring per load application in the uniaxial case. This equates to a difference of a factor of 1.43 on the strain axis. Referring to Figure 10, this is probably a fair estimate of the difference at 10C and 20C but a significant underestimate at 30C. The additional difference at 30C is probably due to an accumulation of tensile permanent deformation during the uniaxial tests. Whilst this effect was very minor at lower temperatures, it may have influenced the result at 30C.

3.3 Beam on Elastic Foundation under Moving Wheel Load The results for this test series will be presented in the same way as for the other tests (Figures 11 and 12). However, there is a major difference in that the stress induced in the specimen changes during the test as the material loses stiffness and transfers increasing stress to the underlying rubber support. This leads to the pattern shown in Figure 11, where there almost appears to be a lower limit to stiffness reduction. In practice, the near constant stiffness shown on the right of Figure 11 was followed by an indeterminate phase, during which the strain gauges on the underside of the specimen either registered very high strain if a crack was forming within the gauge length or else registered very low strain if a crack was forming elsewhere. This point was usually quite sudden and very clear from the data.
1 0.9 0.8 Relative Stiffness 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0 0.2 0.4 0.6 0.8 1 Proportion of Life Used 10C 20C 30C

Figure 11: Average stiffness reduction pattern observed during beam tests Notice the quite different order of the lines shown in Figure 11 in comparison to those from the other two test types. The 30C characteristic now shows the least stiffness loss, whereas the 20 and 10C characteristics are similar. They all show a very large initial stiffness reduction, much greater than could possibly be explained by localised heating of the material. In terms of the number of load applications at a given strain level, Figure 12 shows another quite different pattern. Acknowledging the scatter, the 10C data lie very approximately on the trapezoidal characteristic. If the beam test can be taken to be representative of a pavement layer, this suggests that it may be approximately acceptable to use standard laboratory fatigue data and apply it directly in pavement analysis if the design temperature is 10C. However, at 20C, Figure 12 suggests that a shift factor of about 2 on the strain scale (about 30 on the load applications scale) is needed. At 30C, the shift factor rises to about 2.8 (170 on the load applications scale). The orders of magnitude of these shift factors certainly reflect those commonly used in practical pavement design. The increasing shift required at higher temperatures, reflects the increasing dominance of the rubber support.

1000 Trapezoidal - 10C Trapezoidal - 20C Trapezoidal - 30C Uniaxial - 10C Uniaxial - 20C Uniaxial - 30C Beam - 10C Beam - 20C Beam - 30C

Microstrain 100 1000

10000

100000

1000000

Num ber of Load Applications at Failure

Figure 12: Initial strain-life plot with beam on elastic foundation data included 4 CONCLUSION

The factors affecting asphalt fatigue can be clearly observed by carrying out different types of test and at different temperatures. However, the mechanism of damage accumulation appears to be unaffected by the very different stress distributions of different tests. The relationship between trapezoidal and uniaxial data supports the use of a power law for damage accumulation. Whilst no detailed analysis is offered in this paper, the data from direct wheel loading of a beam on an elastic foundation provides a way forward into analysis of a pavement structure. The commonly used factors converting from laboratory test data to fatigue characteristics for pavement design are reflected in the difference between results from beam testing and those from the other tests. The quite different factors found to apply at different temperatures may have significant implication for practical pavement design. REFERENCES Baaj, H., Di Benedetto, H. and Chaverot, P., Fatigue of mixes: an intrinsic damage approach, Proc. 6th Int. RILEM Symp. on Performance Testing and Evaluation of Materials, Zurich, 2003, pp 394-400. Bodin, D., de la Roche, C., Piau, J-M. and Pijaudier-Cabot, G., Prediction of the intrinsic damage during bituminous mixes fatigue tests, 6th Int. RILEM Symp. on Performance Testing and Evaluation of Materials, Zurich, 2003, pp 380-386. Di Benedetto, H., de la Roche, C., Baaj, H., Pronk, A. and Lundstrm, R., Fatigue of bituminous mixtures: different approaches and RILEM group contribution, Proc. 6th Int. RILEM Symp. on Performance Testing and Evaluation of Materials, Zurich, 2003, pp 15-38. Molenaar, A.A.A., Bottom-up fatigue cracking: myth or reality?, Proc. 5th Int. RILEM Conf. on Cracking in Pavements, Limoges, 2004, pp 275-282. Rowe, G. M., Application of the Dissipated Energy Concept to Fatigue Cracking in Asphalt Pavements, PhD Thesis, University of Nottingham, 1996. Thom, N.H., Choi, Y-K. and Collop, A.C., Top-down cracking, damage and hardening in practical flexible pavement design, Proc 9th International Conference on Asphalt Pavements, Copenhagen, 2002, 1: 10-1.

10

You might also like