Another set of specimens prepared from the base course of the trial road were subjected to a uniaxial tension-compression fatigue test with zero confinement, at a temperature of 20 °C in the controlled-strain mode at a loading frequency of 10 Hz. The test was performed twice on every specimen: (i) using a low strain amplitude (ε0 (L1) = 55 με) to determine mixture
properties in the linear viscoelastic range, and (ii) using a high strain amplitude (ε0 (L2) = 130
με) to obtain the nonlinear viscoelastic properties and prompt fatigue damage. However, both strain levels were, in effect, high enough to surpass the linear viscoelastic (LVE) region and cause damage. Therefore, one of the objectives of the fatigue analysis is to examine if the latest fatigue characterisation approaches can capture the damage that happened in the first fatigue test before starting the second one on the same specimen. Some specimens failed early, before the end of the first fatigue test, and did not complete the 200,000 cycles of the test. Consequently, it was decided to analyse each uniaxial T/C fatigue test independently. The dynamic modulus and phase angle values from the first cycle interval (N = 10) in each uniaxial T/C fatigue test (L1 and L2) were considered as the linear viscoelastic dynamic modulus (|𝐸∗|
𝐿𝑉𝐸) and the linear viscoelastic phase angle (𝜑𝐿𝑉𝐸), respectively.
Specimens prepared from the base course of trial section 3 (Limestone aggregate with unmodified 60-70 Pen bitumen) failed very early during the first T/C fatigue test (using strain amplitude of 55 με) and no data were collected. This could indicate that a significant amount
126 of damage was induced in the specimen during construction and/or the results were affected by segregation, which was earlier observed on samples from section 3.
Before analysing the data from both fatigue tests, the stress and strain amplitudes applied in this controlled-strain fatigue test using AMPT need to be investigated. Figure 62 shows an example of the strain and stress amplitudes applied on the specimens from the base course of field trial section 2 under strain L1 and L2 tests. In some cases of the conducted uniaxial T/C fatigue tests, the raw data after a certain number of loading cycles were discarded and neglected during analysis due to the discrepancy in the data points as shown in Figure 62.
(a) (b)
(c) (d)
Figure 62. Strain and stress amplitudes applied to the specimens of trial section 2 under both strain levels tests.
According to Figure 62 and the raw data of other trial sections, the strain amplitude was in most cases constant in each test but the target value (i.e. 55 µε and 130 µε) was not achieved
127 during both low and high strain amplitude fatigue tests (L1 and L2). This will show a fatigue behaviour for a different strain amplitude, not the target one. In a few cases, the strain increased rapidly and early in the test due to sudden cracking in the specimen. It is important to remember that these mixtures are stiff and this is expected to raise a concern about the durability and fatigue resistance, as mentioned earlier in the results for the |E*| test. Figure 63 shows the average strain amplitude applied to the tested specimens during the uniaxial T/C fatigue tests (L1 and L2). This will surely affect the reliability of the results, and emphasises the need to use a suitable characterisation approach (e.g. viscoelastic continuum damage approach).
Figure 63. Average strain amplitude applied to the tested field cores.
The main outputs of each fatigue test were dynamic modulus (|E*|) and phase angle (φ) against number of loading cycles (N). The results for each specimen of the field trial sections tested under strain amplitude L1 are presented in Figure 64.
128
Figure 64. Dynamic modulus and phase angle results for the field cores tested under strain amplitude L1 test.
129 According to the figures above, the replicate specimens of each mixture of the trial sections showed a different reduction rate of the modulus and increment of the phase angle. These preliminary results cannot be used alone to evaluate the performance against fatigue damage. In addition, damage that happened in the first fatigue test implies that this affected the specimen in the second fatigue test, which implies that the results of the second fatigue test cannot be quantified using the reduction (e.g. 50%) in dynamic modulus value only. A suitable fatigue characterisation approach should be implemented in order to assess each mixture and predict its fatigue life.
The results for each field cores’ specimen tested under strain amplitude L2 test (dynamic modulus (|E*|) and phase angle (φ) against the number of loading cycles) are presented in Figure 65.
Similar to the results of the strain level 1 test, the specimens of each trial base course showed a different reduction rate of the dynamic modulus and increment of the phase angle. Many specimens failed early in the second test under strain amplitude of 130 µε, as shown in Figure 65. Consequently, the preliminary results and the traditional interpretation (E* vs. N) of the uniaxial T/C fatigue test cannot be used directly to assess the performance against fatigue cracking.
130
Figure 65. Dynamic modulus and phase angle results for the field cores tested under strain amplitude L2 test.
131 It is worth mentioning that the interface between the two AC base layers in the specimens prepared from the field cores was checked using X-ray Computed Tomography (CT) at Texas A&M University at Qatar and no major discontinuity was found. Nevertheless, it is most likely that the interface did have an effect on the fatigue results since it is in the weakest plane in the test specimens. In order to avoid this problem in future projects, it is recommended to core the test specimens horizontally from the extracted field cores.