You are on page 1of 19

A Simple Quantitative Method for Identification of Failure due to Fatigue Damage

GHAZI AL-KHATEEB*,** Department of Civil Engineering, Jordan University of Science and Technology, Irbid, 22110, Jordan AROON SHENOY** S.A.I.C.O. President, 1111 Arlington Blvd, Arlington, VA 22209, USA
ABSTRACT: A simple quantitative method is presented which is applicable to any type of fatigue testing that uses sinusoidal strain/stress input and which will work for experimentally identifying points of failure due to fatigue damage of any kind of material being tested. The present work utilizes strain-controlled bending beam fatigue test on asphalt mixtures to demonstrate the efficacy of this method. Distortions in the hysteresis loop or waveform are tracked to pinpoint the appearance of initial microcracks and final point of complete failure due to fatigue damage. Relationship between output signals for consecutive cycles with reference to initial stable cycle is used for computing R2. The R2 drops sharply from initial stable value of 1 to less than 0.5 and eventually to almost 0 with increasing loading cycles. The number of cycles determined from the fitted equation at R2 = 1 marks the point of first fatigue failure Nfff and R2 = 0 marks the point of complete fatigue failure Ncff. KEY WORDS: fatigue damage, repeated loading, hysteresis loop, sinusoidal waveform, asphalt mixtures.

INTRODUCTION

ATIGUE CRACKING IS

one of the major distresses that occur in asphalt pavements and considered in mechanistic-empirical design methods.

*Author to whom correspondence should be addressed. E-mail: ggalkhateeb@just.edu.jo **The opinions, findings, and conclusions expressed in this paper are those of the authors only and not necessarily of any organizations that the authors are affiliated with or consultants to in their professional capacity. Figures 411 appear in color online: http://ijd.sagepub.com

International Journal of DAMAGE MECHANICS, Vol. 20January 2011


1056-7895/11/01 000319 $10.00/0 DOI: 10.1177/1056789509343084 The Author(s), 2011. Reprints and permissions: http://www.sagepub.co.uk/journalsPermissions.nav

G. AL-KHATEEB

AND

A. SHENOY

Fatigue cracking in asphalt pavements occurs as a result of repeated traffic loading at intermediate temperature range of $ 1030 C. Substantial fatigue-related research has been done to develop theoretical models (Ramsamooj, 1991, 1999; Baburamani, 1999; Lee et al., 2000; Rodrigues, 2000; Zhang and Raad, 1999, 2001; Lundstrom and Isacsson, 2003) experimental laboratory testing techniques (Irwin and Gallaway, 1974; Porter and Kennedy, 1975) and data analysis methods to predict the fatigue performance of asphalt paving mixtures (Pell and Cooper, 1975; Van Dijk, 1975; Hopman et al., 1989; Pronk and Hopman, 1990; Tayebali et al., 1992, 1993; Rowe, 1993; Kim et al., 1997; Ghuzlan and Carpenter, 2000; Rowe and Bouldin, 2000; Al-Khateeb and Shenoy, 2004). Fatigue failure or the point of failure due to fatigue damage has been defined in various ways in the asphalt pavement literature. In some cases (e.g., Pell and Cooper, 1975; Van Dijk, 1975; Tayebali et al., 1992; Rowe, 1993), fatigue failure was defined arbitrarily; in other cases (e.g., Van Dijk and Vesser, 1977; Pronk and Hopman, 1990; Tayebali et al., 1992, 1993), it was based on the mode of loading; and sometimes energy methods (e.g., Hopman et al., 1989; Rowe, 1993; Kim et al., 1997; Rowe and Bouldin, 2000; Ghuzlan and Carpenter, 2000) were used to define fatigue failure. Thus, many researchers considered the reduction in initial stiffness of an asphalt mixture during fatigue testing as a measure of fatigue failure. Others observed crack propagation to track fatigue failure. Some others introduced dissipated energy concepts to define fatigue failure. Fatigue testing can be conducted using the constant stress mode of testing and the constant strain mode of testing. In the constant stress mode of testing, some researchers, (e.g., Pell and Cooper, 1975; Tayebali et al., 1992), defined fatigue failure as the complete fracture at the end of the fatigue test when the specimen fails due to tensile strains. Other researchers such as Rowe (1993) defined fatigue failure as occurring when the initial complex modulus has been reduced by 90%. Van Dijk (1975) defined fatigue failure as occurring when the initial strain doubled. In the constant strain mode of testing, since the strain stays constant and the stress decreases during the fatigue test, defining fatigue failure is more difficult. Several different fatigue failure definitions have been adopted in this mode. The most common and widely used definition for fatigue failure in the constant strain mode is the 50% reduction in the initial stiffness as defined by Pronk and Hopman (1990) and Tayebali et al. (1992, 1993). A 50% reduction in the initial modulus was also defined as fatigue failure by Van Dijk and Vesser (1977). Subsequently, the 50% reduction in stiffness was adopted to define the fatigue failure point by the AASHTO as a provisional standard TP8-94 (2002).

Quantitative Fatigue Failure Criterion

Besides the fatigue failure definition of 50% stiffness reduction, some researchers have proposed energy-based failure concepts. In the SHRPA-404 Report (1994), dissipated energy was used for fatigue analysis. The dissipated energy per cycle is computed as the area within the stressstrain hysteresis loop. This energy decreases with an increasing number of load cycles in the strain-controlled fatigue test as the stress decreases; on the other hand, the dissipated energy per cycle increases as the number of load cycles increases for the stress-controlled fatigue test. Hopman et al. (1989) proposed the use of an Energy Ratio concept to define fatigue failure in the controlled strain fatigue tests. They plotted the energy ratio against the number of load cycles. A significant change in the slope of the curve occurred at a critical number of load cycles, which was considered the failure point on the curve. The critical number of load cycles corresponded to a 40% reduction in the complex modulus. Hopman et al. (1989) found evidence of crack initiation at the point where the significant change in slope occurred, i.e., deviated from a straight line. Kim et al. (1997) introduced the 50% reduction in pseudo stiffness as a failure point in fatigue testing, which was believed to be independent of mode of loading and stress/strain amplitude. Rowe (1993) and Rowe and Bouldin (2000) used the concept of Energy Ratio, which was proposed by Hopman et al. [14], to define fatigue failure. The fatigue failure for the controlled stress and the controlled strain modes was identified using an energy ratio that was defined as the ratio of the dissipated energy in the first cycle times the number of cycles (N) to the dissipated energy in the N-th cycle. Ghuzlan and Carpenter (2000) proposed a Dissipated Energy concept to define fatigue failure in asphalt mixtures. They used the ratio of the change in dissipated energy between two consecutive cycles (N, N + 1) to the total dissipated energy in the load cycle N. This new method was believed to define fatigue failure independent of the mode of loading in fatigue testing. Failure was selected as the point where this ratio increased rapidly after a consistent stable trend for this ratio with load cycles. Al-Khateeb and Shenoy (2004) presented a new distinctive fatigue failure criterion for asphalt paving mixtures that was based on observing the loaddeformation (or stressstrain) hysteresis loop or tracking the stress and strain waveforms during fatigue testing. They defined the point of first fatigue failure as that point at which the hysteresis loop or the response waveform started to show the first signs of distortion. The point of complete fatigue failure, on the other hand, was defined as the point at which the stress response was no longer dependent on the strain input due to the formation of interconnected fatigue cracks, and that could be identified by

G. AL-KHATEEB

AND

A. SHENOY

the extensive or almost complete distortion of the loaddeformation hysteresis loop or the response waveform.

OBJECTIVE The main objective of this study is to present a new simple quantitative method for identification of fatigue failure based on the changes to the loaddeformation (stressstrain) hysteresis loop or the response waveform. The quantitative method presented in this article defines the fatigue failure as the point at which the coefficient of determination (R2) value, for the relationship of the response (stress) signals for progressive loading cycles with reference to the response signal of initial stable cycle, decreases sharply.

MATERIALS Aggregate The aggregate consisted of 92% crushed diabase and 8% quartz and quartzite natural sand as shown in Table 1. The aggregate gradation met the 1991 Virginia Department of Transportation (VDOT) specifications for surface mixtures (SM-3). Binders Eleven binders were used in this study. They included a PG64-22 and a PG70-22 (unmodified binders), a PG70-28 (air-blown), and eight polymermodified binders of PG70-28, which consisted of the following polymers: Terpolymer (Elvaloy), StyreneButadieneStyrene Linear Grafted, StyreneButadieneStyrene Linear, StyreneButadieneStyrene Radial Grafted, EthyleneVinyl Acetate, EthyleneVinyl Acetate Grafted, EthyleneStyrene Interpolymer, and Chemically Modified Crumb Rubber. The PG numbers shown are based on the Superpave system description for Performance Grading. All the binders were from the same source, a Venezuelan crude (blend of Boscan and Bachaquero). The air-blown PG70-28 was obtained by noncatalytic air blowing of a PG52-28. The polymer-modified grades were obtained by adding various amounts of different polymers to the PG64-22, the PG52-28, or a mixture of the PG64-22 and the PG52-28 in appropriate proportions; the goal was to achieve a PG70-28 performance grade.

Quantitative Fatigue Failure Criterion Table 1. Aggregate properties for the diabase.
Percent Passing (%) Sieve size (mm) 25.0 19.0 12.5 9.5 4.75 2.36 1.18 0.600 0.300 0.150 0.075 92% Diabase 100.0 98.6 73.9 58.7 39.5 27.2 19.4 14.4 11.1 8.2 5.7 8% Natural sand Blend 100.0 98.7 76.0 62.0 44.0 32.1 23.8 16.9 11.3 7.9 5.5 2.892 2.916 2.961 0.8

100.0 95.8 88.2 74.8 46.0 14.1 4.8 2.9 2.565 2.601 2.659 1.4

Specific gravity and percent absorption Bulk Dry SGa 2.933 2.956 Bulk SSDb Apparent SGa 3.002 % Absorption 0.8

Flat and elongated particles at 3-to-1 length-to-thickness ratio, percent by mass 21 NAc Los Angeles abrasion, percent loss by mass 14 Fine aggregate angularity 49
a c b

NAc 45

Specific gravity. Saturated surface dry. Not applicable.

Sample Preparation and Fatigue Testing The binders were heated to 163 C and mixed with heated aggregates in proportion to achieve a binder content of 4.85% by the total mass of mixture. All asphalt mixtures were short-term oven aged for 2 h at 135 C according to the AASHTO provisional practice PP2-00 (2000) and compacted, using a Slab-PakTM linear kneading compactor, into 180 500 50 mm slabs. Two beams, each 63 mm wide, 50 mm high, and 380 mm long with smooth faces, were then cut from each slab. The target air-void level was 7.0 0.5%. The air-void level in the sawed beam specimens was determined, and those beams having air voids outside the specified range were discarded. All asphalt mixtures were tested $ 48 h after compaction. Bending beam fatigue tests were performed according to the AASHTO protocol (2002). The tests were conducted in the strain-controlled mode at a

G. AL-KHATEEB

AND

A. SHENOY

strain level of 1000 microstrains and at a test temperature of 19 C. A vertical repeated sinusoidal displacement was applied at a frequency of 10 Hz with no rest periods. Three to five replicates were used for each asphalt mixture at the 1000-microstrain level. Two concentrated and symmetrical loads are applied on the fatigue beam specimen using a standard 4-point bending test, as shown in Figures 1 and 2. The beam specimen is forced back to its original position at the end of each load pulse after it is subjected to 4-point bending. Free rotation and horizontal translation are allowed at all load and reaction points, as shown in Figure 3.

Load Specimen clamp

Load Specimen

Reaction

Deflection

Reaction

Figure 1. Bending beam fatigue test schematic diagram using a standard 4-point bending test.

P/2 a = l /3 a

P/2

R Crack

l
Figure 2. Loading and geometry.

Quantitative Fatigue Failure Criterion

The data acquisition software (TestStarTM) of the device recorded the load and the deformation of the specimen. Tensile strains and stresses were calculated using the following equations: "t 12d h 3l2 4a2 1

where et = maximum tensile reference strain; d = maximum vertical deformation at the center of the beam; h = average specimen height; l = length of beam between outside clamps; and a = space between inside clamps = l/3. t lP bh2 2

where rt = maximum tensile reference stress; P = load applied by actuator; and b = average specimen width. Flexural stiffness (S) was calculated using the equation given below: S t "t 3

Applied strains, response stresses, and flexural stiffnesses were determined from the load and deformation amplitude, the geometry of the tested beam, and the distance between the beam supports. The fatigue data obtained on the various asphalt mixtures were then analyzed using three procedures: 1. The AASHTO TP8-94 50% reduction in stiffness (2002), 2. The Al-Khateeb and Shenoy criterion (2004), 3. The quantitative R2-based method presented in this study. The results for all tested asphalt mixtures are summarized in Table 2. In the AASHTO TP8-94 (2002) method, the number of load cycles to fatigue failure was determined using the 50% stiffness reduction point.

Figure 3. Freedom conditions of bending beam fatigue test.

10

G. AL-KHATEEB

AND

A. SHENOY

Table 2. Number of load cycles to fatigue failure.


Number of load cycles to fatigue failure AASHTO TP 8-94 Quantitative failure Qualitative failure criterion (2002) method criterion [present work] Al-Khateeb and Shenoy (2004) Visual distortion of waveform or hysteresis loop 10,000 4,000 8,400 217,000 20,000 30,000 10,000 36,700 16,700 36,700 8,400

Asphalt mixture PG 64-28 PG 70-22 Air-Blown Elvaloy SBS LG SBS L SBS RG EVA EVA G ESI CMCRA

50% Stiffness reduction 5,323 3,144 7,614 97,389 9,911 8,774 12,372 5,905 7,183 10,301 4,158

Initial point of sharp change in R2-Value 7,292 3,623 7,434 197,531 15,801 12,012 13,144 12,654 11,840 34,937 3,631

980

Flexural stiffness (MPa)

735
Stiffness = 107.1 Ln(N) + 1673.7 R = 0.99
2

490

245

0 0 50,000 100,000 150,000 200,000 250,000 300,000 Number of load cycles (N)

Figure 4. Flexural stiffness vs load cycles for Elvaloy.

The actual number of load cycles to fatigue failure can be established graphically from interpolation of fatigue data or calculated after fitting a curve to the data; the natural logarithm function (ln) of the load cycles in this case was used to fit the stiffness versus load cycles data (Figure 4).

Quantitative Fatigue Failure Criterion


500 400 300 200 Stress (kPa) 100 0 100 200 300 400 500 0.0012 0.001 0.0008 0.0006 0.0004 0.0002 Strain (m/m)
Figure 5. Stressstrain hysteresis loop.

11

Al-Khateeb and Shenoy (2004) defined a distinctive fatigue failure criterion through visual observations of the loaddeformation (stressstrain) hysteresis loop or waveform during fatigue testing. A slight distortion in the hysteresis loop or the response sinusoidal waveform indicated the first fatigue failure and an extensive (or almost complete) distortion in the hysteresis loop or the response waveform pointed to the complete fatigue failure. This failure criterion presented in Al-Khateeb and Shenoy (2004) is, however, a qualitative criterion that is based on visual observation of the hysteresis loop or the response waveform during fatigue testing. In a fatigue test, when a sinusoidal strain is applied to the sample, a sinusoidal stress response is expected even for a heterogeneous material like the asphalt concrete. This is valid as long as the test sample is integral and undamaged. The stress paths are all well defined at that time, and the loaddeformation (or stressstrain) hysteresis loop is also well defined and smooth, as shown in Figure 5. With repeated fatigue loading (strain in this case), the material starts to fatigue and microcracks are induced in the system. These microcracks introduce discontinuities in the stress paths, and the stress response starts to distort. This can be clearly seen by observing the loaddeformation (or stressstrain) hysteresis loop. Initially the distortion of the hysteresis loop is slight and indicates the onset of the induced defect due to the repeated strain. By observing the hysteresis loop for each loading cycle after that point, the progression of the fatigue damage is actually monitored. After the initial

12
(a) 500 300 Stress (kPa) 100 100 300 500 0.002
90,000 110,000 130,000

G. AL-KHATEEB

AND

A. SHENOY

150,000

170,000

0.002

0.004 Strain (m/m) Before failure

0.006

0.008

0.01

(b)

300
190,000 210,000 230,000

Stress (kPa)

100

100

300 0.008

0.01

0.012 Strain (m/m) First failure

0.014

0.016

(c)

300 200
250,000 270,000

Stress (kPa)

100 0 100 200 300 0.014 0.016 Strain (m/m) Complete failure

0.018

Figure 6. Stressstrain hysteresis loop through the progress of fatigue failure for Elvaloy.

slight distortion of the stressstrain loop, the shape buckles slightly, followed by a radical change in the shape of the loop with the progress of the fatigue failure. The progression of the fatigue failure and the progressive distortions in the stressstrain hysteresis loops are shown in Figure 6. Basically, the stressstrain curves are scaled with the loaddisplacement curves. Hence, the same information can be obtained from the observation

Quantitative Fatigue Failure Criterion


(a) Load / deformation (kN / mm) 0.8
Load = response

13

0.4 0 0.4 0.8 517.1

Deformation = input

517.2

517.3

517.4 Time (s) Before failure

517.5

517.6

517.7

(b) Load / deformation (kN / mm)

0.8
Load = response

0.4 0 0.4 0.8 1017.6

Deformation = input

1017.7

1017.8

1017.9

1018

1018.1

Time (s) First failure (c) Load / deformation (kN / mm) 0.8
Load = response

0.4 0 0.4 0.8 2018

Deformation = input

2018.1

2018.2

2018.3 Time (s)

2018.4

2018.5

2018.6

Complete failure

Figure 7. Loaddeformation relationship through the progress of fatigue failure for EVA grafted.

of the sinusoidal waveform. The input (strain or deformation) signal is smooth and well defined and stays so from the start to the end of the fatigue test. However, the response (stress or load) sinusoidal waveform starts smooth and well defined, but soon distorts and finally becomes almost flat; the latter indicates that the stress response is no longer dependent on the strain because of the discontinuities that have emerged from the cracking (Figure 7).

14

G. AL-KHATEEB

AND

A. SHENOY

The transitional point when the sinusoidal stress waveform goes from the smooth profile to the first level of distortion indicates the initiation of the first level of cracking (first failure). In the next few cycles, it is this initial level of cracking along with the applied strain that controls the response stress and maintains the distorted shape of the sinusoidal stress waveform. However, when the material fatigues further and the crack intensity rises, the applied strain can no longer drive the stress response and a flattened stress or load waveform is obtained, as shown in Figure 7 (complete or ultimate fatigue failure).

QUANTITATIVE METHOD FOR FATIGUE FAILURE IDENTIFICATION R2 Statistic The quantitative method for determination of fatigue failure criterion is based on the R2 statistic for the relationship between the output response stress signals for consecutive loading cycles with reference to initial stable cycle in a strain-controlled test or for the relationship between the output response strain signals for consecutive cycles with reference to initial stable cycle in a stress-controlled test. In the present strain-controlled test, R2 value for the relationship between the stress signals for consecutive cycles with reference to the initial stable cycle was computed. The change in the R2 value with the progressive cycles was monitored as shown in Figure 8. The R2 value at the beginning of the fatigue test as expected was found to be approximately 1.0 or close to 1.0. With the progress of the fatigue test as the sample starts to fatigue and experiences microcracks, the R2 value starts to decrease sharply as shown in Figure 8. A linear relationship was fitted to the data and two resulting linear lines were obtained: the first line represents the beginning of the fatigue test before fatigue failure when the R2 value is approximately equal to 1.0, and the second line represents the stage when fatigue failure starts to take place. In Figure 8, at the intersection point between the two lines, the first fatigue failure point is identified. This point is determined by solving the fitted equation shown in Figure 8 at R2 value of 1.0. The complete fatigue failure point is determined by solving the same equation at R2 value of 0.0 or in other words at the intersection point with the x-axis. At an R2 value of 1.0, the relationship between the response (output) signal (waveform) of progressive cycles with respect to the response signal (waveform) of the first cycle is ideal. No microcracking or failure is taking place at

Quantitative Fatigue Failure Criterion


1.00 0.80 R 2-value 0.60 0.40 0.20 0.00 1000

15

10,000 N

100,000

1,00,0000

Figure 8. R2 vs N on semi-logarithmic scale for EVA asphalt mixture.

this stage. When the R2 value starts to show a sharp decrease as shown in Figure 8, this is an indication of fatigue failure occurrence. The same result can be obtained if the data are plotted on a semi-log scale as shown in Figure 9. The logarithmic relationship of the second set of data (after the R2 value starts to decrease sharply) looks linear on the semi-log scale, however, the intersection of this linear-looking line with the first horizontal line (Figure 9) identifies the first fatigue failure point. At the point when the second line intersects with the x-axis, is when the complete fatigue failure takes place. An algorithm for sensing the onset of fatigue failure can be performed and only requires the computation of the R2 value described above. The general form of the fitted equation of the second line in Figure 9 is: R2 Value A log N B 4

Equating both sides of the equation to 1.0 results in Equation (5) below: A log N B 1:0 and therefore,
1B Nfff 10 A

The Nfff represents the number of load cycles at first fatigue failure and that is happening when R2 value = 1.0. The complete fatigue failure point is determined by equating both sides of Equation (4) to 0.0 as shown in the equation below: A log N B 0:0 7

16
1.20 1.00 0.80 R 2-value 0.60 0.40 0.20 0.00 1000
2

G. AL-KHATEEB

AND

A. SHENOY

10,000 N

100,000

Figure 9. R vs N for air-blown asphalt mixture.

In this case, Ncff 10 A


B

The Ncff represents the number of load cycles at complete fatigue failure and that is occurring when R2 value = 0.0. The R2 values were calculated in the same way and plotted against the number of load cycles for all replicates of the eleven asphalt mixtures used in the study. The Nfff and Ncff values were determined as well and averaged over the number of replicates. These values were compared with the values obtained using the AASHTO TP8-94 (2002) method (the 50% stiffness reduction) and the values obtained using the qualitative fatigue failure criterion for Al-Khateeb and Shenoy (2004) as shown in Table 2. The associated stress response values at the three stages for all replicates of all asphalt mixtures were also computed. In other words, the stress response was calculated when R2 = 1.0, R2 started to decrease sharply (first fatigue failure point = Nfff), and when R2 = 0.0 (complete fatigue failure point = Ncff), respectively (Figure 10). In addition, the stress reduction relative to the stress of the initial stable loading cycle was calculated as the R2 decreased (Figure 11). In this case, since the fatigue test was conducted in the strain-controlled mode (i.e., the strain input was constant throughout the test), the stress reduction provided values for the stiffness reduction at first failure and complete failure as well. Figure 10 shows that the stress decreased sharply at the initial stable loading cycle until the first fatigue failure point and after that the stress reduction rate was lower till the point of complete fatigue failure. The initial

Quantitative Fatigue Failure Criterion


Associated stress response (kPa)

17
PG 64-28 PG 70-22 Air-blown Elvaloy SBS LG SBS L SBS RG EVA EVA G ESI CMCRA

1500 1200 900 600 300 0

R2 = 1

First failure, R2 decreases sharply

Complete failure, R2 = 0

Figure 10. R2 and associated stress response values for all asphalt mixtures.
PG 64-28 PG 70-22 Air-blown Elvaloy SBS LG SBS L SBS RG EVA EVA G ESI CMCRA

0 Stress response reduction (%) 20 40 60 80 100

R2 = 1

First failure, R2 decreases sharply

Complete failure, R2 = 0

Figure 11. R2 and associated stress or stiffness reduction values for all asphalt mixtures.

stress response or stiffness values were naturally different from mixture to mixture, and so the stress response or stiffness values at the point of first fatigue failure were also different from material to material. At the end, when the material failed completely (i.e., at the point of complete fatigue failure), the stress response or stiffness values were minimal. Figure 11 also provided similar information, but in this case about the reduction in stress response or stiffness values relative to the initial stable loading cycle. At the beginning, the stress response or stiffness was stable (0.0% reduction). However, with repeated fatigue loading cycles, the

18

Table 3. Stress and strain values with associated change in R2 for EVA asphalt mixture-replicate no. 3.

Stress values Stress values for a stable for a cycle before stable cycle failure (kPa) (kPa)

Stress values Actual strain Actual strain Actual strain for the values for the values for a Stress values Stress values values for the Stress values cycle at cycle at for the first cycle first cycle at for a stable cycle for a at first first failure stable cycle complete complete before failure stable cycle (kPa) failure (kPa) failure (ke) (ke) (kPa) failure (kPa) (ke)

G. AL-KHATEEB
AND

273.1 90.5 13.8 229.1 405.1 574.7 646.5 653.0 620.4 516.0 372.5 183.4 44.9 240.5 403.6 462.3 527.5 547.1 547.1 442.7 279.7

273.1 90.5 13.8 229.1 405.1 574.7 646.5 653.0 620.4 516.0 372.5 183.4 44.9 240.5 403.6 462.3 527.5 547.1 547.1 442.7 279.7 R2 = 1.0

1099.5 1036.4 947.1 863.6 699.9 521.9 347.2 217.2 134.5 93.7 109.1 165.6 277.5 444.2 601.7 759.7 900.1 1028.1 1103.8 1145.9 1124.6

273.1 90.5 13.8 229.1 405.1 574.7 646.5 653.0 620.4 516.0 372.5 183.4 44.9 240.5 403.6 462.3 527.5 547.1 547.1 442.7 279.7

129.7 5.7 118.2 203.0 261.7 281.2 300.8 353.0 353.0 307.3 242.1 137.7 33.4 44.9 110.1 155.7 201.4 214.4 221.0 188.4 123.1 R2 = 0.98

1122.7 1080.6 994.0 862.3 708.6 545.9 389.1 256.7 161.0 108.1 101.6 141.6 233.7 367.5 527.4 693.4 849.1 980.2 1070.0 1115.4 1116.5

273.1 90.5 13.8 229.1 405.1 574.7 646.5 653.0 620.4 516.0 372.5 183.4 44.9 240.5 403.6 462.3 527.5 547.1 547.1 442.7 279.7

85.6 59.5 72.5 53.0 59.5 66.0 118.2 85.6 0.8 26.9 13.8 46.4 59.5 46.4 46.4 13.8 5.7 33.4 118.2 92.1 92.1 R2 = 0.0

868.1 709.4 543.7 386.0 247.4 151.9 108.6 102.0 139.2 225.9 357.8 514.6 682.3 846.1 980.3 1076.1 1119.0 1123.0 1085.9 989.7 859.3

A. SHENOY

Quantitative Fatigue Failure Criterion

19

reduction in stress response or stiffness became more marked and increased rapidly till the point of first fatigue failure. The reduction then continued at a lower rate, until the point of complete fatigue failure. The reduction in stress response or stiffness at the point of first fatigue failure ranged from 48.1% to 82.9%. This result actually pointed to a significant finding that the stiffness reduction should not be an identification or definition of fatigue failure point due to the fact that each material responds differently to fatigue and the initial stiffness is by default different from material to material. Therefore, the procedures followed in this article and the criterion described in Al-Khateeb and Shenoy (2004) are indeed the proper way to identify the fatigue failure points for asphalt mixtures and other materials. Table 3 shows the actual applied strains and the stress responses when R2 = 1.00, when first failure started to take place (R2 = 0.98), and at complete failure when R2 = 0.00 (close to zero). The table provides an idea about the change in the stressstrain relationship as R2 value changes with the progress of fatigue test until complete fatigue failure.

FINDINGS AND CONCLUSIONS This article shows that there is no need to go through any level of calculations based on formulae for the dissipated energy or energy ratio, or rely on a somewhat arbitrary 50% reduction in stiffness in order to identify the fatigue failure points. The fatigue failure is revealed directly from raw data by observing the distortion of the loaddeformation hysteresis loop or the response waveform at the onset of the first crack appearance. Before occurrence of fatigue failure, the stress and strain signals are strongly correlated (R2 = 1) whereas after failure, they are no longer correlated (R2 = 0). This marks a very significant and clear definition of the points of fatigue failure. The first fatigue failure point is identified as occurring when the shape of the stressstrain (loaddeformation) hysteresis loop starts to show the first signs of distortion from the original smooth oval-or-elliptical shape. The initially distorted shape typically lasts for a long period before a very irregular shape of the distorted waveform shows up marking the point of the complete fatigue failure and the ultimate fatigue failure of the sample is reached. In this article, the R2 value for the relationship of each of the response (stress) signals for progressive cycles with reference to the response signal of a stable cycle was computed. The change in the R2 value with the progressive cycles was monitored as shown in Figure 8. The R2 value at the beginning of the fatigue test is expected to be approximately 1.0. With the progress of the fatigue test as the sample starts to fatigue and experiences microcracks, the

20

G. AL-KHATEEB

AND

A. SHENOY

R2 value starts to decrease sharply as shown in Figure 8. A linear relationship was fitted to the data and two resulting linear lines were obtained: the first line represents the beginning of the fatigue test before fatigue failure when the R2 value is approximately equal to 1.0, and the second line represents the stage when fatigue failure starts to take place and eventually goes to complete fatigue failure. The first fatigue failure is obtained at the intersection of the two fitted linear lines when R2 value is equal to 1.0. On the other hand, the complete fatigue failure is obtained at the intersection of the second linear line with the x-axis when R2 value equals 0.0. This new fatigue failure identification method is based on quantitative measures of the R2 value of the relationship between the response signal or waveform (sine wave) of progressive loading cycles and the response signal or waveform of initial stable cycle. This new quantitative method accurately describes the true point of fatigue failure based on fundamental loaddeformation (stressstrain) relationship of the material and hence only raw data of fatigue testing are needed to determine the points of first and complete fatigue failure. Though the method was developed and demonstrated for fatigue testing of asphalt mixture, it is ubiquitously applicable to any type of fatigue testing that utilizes a sinusoidal strain or stress input, and will work for identification of the points of fatigue failure of any kind of material that is being tested. In conclusion, the method bypasses the need to perform complicated calculations for the viscoelastic response of the material and provides a simple quantitative method to determine fatigue damage. REFERENCES
AASHTO PP2-00 (2000). Standard Practice for Mixture Conditioning of Hot-Mix Asphalt (HMA). AASHTO Provisional Standard. AASHTO TP8-94 (2002). Method for Determining the Fatigue Life of Compacted Hot-Mix Asphalt (HMA) Subjected to Repeated Flexural Bending. AASHTO Provisional Standards. Al-Khateeb, G. and Shenoy, A. (2004). A Distinctive Fatigue Failure Criterion, Journal of the Association of Asphalt Paving Technologists, 73: 585622. Baburamani, P. (1999). Asphalt Fatigue Life Prediction Models A Literature Review. Research Report ARR 344, Australia. Ghuzlan, K.A. and Carpenter, S.H. (2000). Energy-derived, Damage-Based Failure Criterion for Fatigue Testing, Transportation Research Record: Journal of the Transportation Research Board, 1723: 141149. Hopman, P.C., Kunst, P.A.J.C. and Pronk, A.C. (1989). A Renewed Interpretation Method for Fatigue Measurements, Verification of Miners Rule. In: 4th Eurobitume Symposium in Madrid, Vol. 1, pp. 557561. Irwin, L.H. and Gallaway, B.M. (1974). Influence of Laboratory Test Method on Fatigue Results for Asphaltic Concrete. Fatigue and Dynamic Testing of Bituminous Mixtures, ASTM STP, 561: 1246.

Quantitative Fatigue Failure Criterion

21

Kim, Y.R., Lee, H.J. and Little, D.N. (1997). Fatigue Characterization of Asphalt Concrete Using Viscoelasticity and Continuum Damage Theory, Journal of the Association of Asphalt Paving Technologists, 66: 520569. Lee, H.J., Daniel, J.S. and Kim, Y.R. (2000). Continuum Damage MechanicsBased Fatigue Model of Asphalt Concrete, American Society of Civil Engineering (ASCE) Journal of Materials in Civil Engineering, 12: 105112. Lundstrom, R. and Isacsson, U. (2003). Asphalt Fatigue Modelling using Viscoelastic Continuum Damage Theory, Road Materials and Pavement Design, 4: 5175. Pell, P.S. and Cooper, K.E. (1975). The Effect of Testing and Mix Variables on the Fatigue Performance of Bituminous Materials, Journal of the Association of Asphalt Paving Technologists, 44: 137. Porter, B.W. and Kennedy, T.W. (1975). Comparison of Fatigue Test Methods for Asphalt Materials. Research Report 183-4, Project 3-9-72-183, Center for Highway Research, University of Texas at Austin. Pronk, A.C. and Hopman, P.C. (1990). Energy Dissipation: The Leading Factor of Fatigue. Highway Research: Sharing the Benefits. In: Proceedings of the Conference, the United States Strategic Highway Research Program, London. Ramsamooj, D.V. (1991). Fatigue Cracking of Asphalt Concrete Pavements, Journal of Testing and Evaluation, 19: 231239. Ramsamooj, D.V. (1999). Prediction of Fatigue Performance of Asphalt Concrete Mixes, Journal of Testing and Evaluation, 27: 343348. Rodrigues, R.M. (2000). A Model for Fatigue Cracking Prediction of Asphalt Pavements Based on Mixture Bonding Energy, Internation Journal of Pavement Engineering, 1: 133149. Rowe, G.M. (1993). Performance of Asphalt Mixtures in the Trapezoidal Fatigue Test, Journal of the Association of Asphalt Paving Technologists, 62: 344384. Rowe, G.M. and Bouldin, M.G. (2000). Improved Techniques to Evaluate the Fatigue Resistance of Asphaltic Mixtures. In: Proceedings of 2nd Eurasphalt & Eurobitume Congress, Barcelona, Spain. SHRP-A-003A Report (1992). Fatigue Response of Asphalt-Aggregate Mixes, Part 1 Test Method Selection by Tayebali, A.A., Deacon, J.A., Coplantz, J.S., Harvey, J.T. and Monismith, C.L. Strategic Highway Research Program, National Research Council, Washington, D.C. SHRP-A-404 Report (1994). Fatigue Response of Asphalt-Aggregate Mixes.Strategic Highway Research Program, National Research Council, Washington, D.C. Tayebali, A.A., Rowe, G.M. and Sousa, J.B. (1992). Fatigue Response of AsphaltAggregate Mixtures, Journal of the Association of Asphalt Paving Technologists, 61: 333360. Tayebali, A.A., Deacon, J.A., Coplantz, J.S. and Monismith, C.L. (1993). Modeling Fatigue Response of Asphalt-Aggregate Mixes, Journal of the Association of Asphalt Paving Technologists, 62: 385421. Van Dijk, W. (1975). Practical Fatigue Characterization of Bituminous Mixes, Journal of the Association of Asphalt Paving Technologists, 44: 3872. Van Dijk, W. and Visser, W. (1977). The Energy Approach to Fatigue for Pavement Design, Journal of the Association of Asphalt Paving Technologists, 46: 137. Zhang, T. and Raad, L. (1999). Numerical Methodology in Fatigue Analysis: Basic Formulation, Journal of Transportation Engineering, 125: 552559. Zhang, T. and Raad, L. (2001). Numerical Methodology in Fatigue Analysis: Applications, Journal of Transportation Engineering, 127: 5966.

You might also like