Use this URL to cite or link to this record in EThOS:
Title: Asphalt fatigue failure analysis and modelling : experimental studies and theoretical formulation
Author: Ahmed, Taher
ISNI:       0000 0004 6059 4074
Awarding Body: University of Liverpool
Current Institution: University of Liverpool
Date of Award: 2016
Availability of Full Text:
Access through EThOS:
Access through Institution:
This thesis focused on the review and background theory of previous studies on fatigue failure criteria, experimental work and results’ analysis. Several techniques are used for testing fatigue performance for both hot mix asphalt (HMA) and fine aggregate matrix (FAM), such as two-, three- and four-point bending, indirect tension and uniaxial tests. In recent years, a new technique has been introduced using the Dynamic Shear Rheometer (DSR). This technique is based on applying a sinusoidal deformation or loading onto small cylindrical samples, 12 mm diameter and 50 mm height, and the response is analysed to obtain phase angle and deformation data under any given circumstances, such as temperature, frequency, etc. The DSR is limited to test bitumen and fine aggregates matrix FAM samples only; nevertheless, no research efforts have been found that use a DSR to study the performance of full HMA samples. In this work, a successful trial was proposed using a DSR for fatigue testing of HMA under controlled strain and stress modes. Two types of aggregates, limestone and granite with two binder grades, 40/60 and 160/220, were employed to prepare four different mixes of hot rolled asphalt (HRA) and dense bitumen macadam (DBM). A technical procedure was adopted to prepare the DSR samples (12 mm in diameter and 50 mm in height) and a statistical procedure based on histograms and modes for the bulk density of the DSR samples was used to select the samples to be tested for fatigue. An approach was developed based on sweep strain/stress amplitude to arrive at a suitable strain and stress amplitude at the damage region to be used in the fatigue test. A new fatigue index (FIR) parameter was derived from the dissipated pseudo-strain energy for the stress-pseudo-strain relationship to be used for evaluating fatigue performance. Results showed that there is a plateau value for FIR which can be used to evaluate fatigue performance, and this value increases when the normalised shear modulus decreases to less than 0.35 and 0.20 for strain and stress test modes respectively. In addition, the FIR results were in agreement with the results from other reliable approaches that have been used for evaluating fatigue performance, such as the energy ratio (ER) and the traditional approach (TA). A two-point bending (2PB) test for trapezoidal samples was used to verify the DSR technique using FIR, TA and ER approaches; the analysis of results revealed the same conclusions as the DSR technique. The variance in the results of the tested samples was studied using error bars in terms of standard of error for all approaches in both techniques: DSR and 2PB. This variance analysis revealed that FIR has low variation in comparison with the TA and ER approaches in both test techniques. A computational model based on artificial neural networks (ANNs) was used in this work for developing models to predict the fatigue performance of hot mix asphalt (HMA). The fatigue performance was defined according to the criteria of the TA, ER and FIR approaches. The results revealed an excellent correlation between the predicted and experimental data. Bias analysis for ANN models involving average error, intercept and slope showed that the strain test mode was more accurate than the stress test mode. A fracture mechanistic approach was also used to evaluate the fracture performance of HMA tested in DSR. A simple fracture model was developed based on a modified Paris’ law and fatigue test parameters represented by relaxation test coefficient (m) and the dissipated pseudo-strain energy to calculate an internal damage parameter, namely the fracture damage index (FIc). The analysis of the results for FIc was in agreement with the FIR, ER and TA approaches; also, it showed better performance analysis in terms of variation than the TA and ER approaches. The fracture model, FIc, was used as a base for developing a model for predicting fatigue life of HMA in terms of number of cycles for strain and stress test modes. The bias analysis revealed that the strain model’s prediction accuracy was better than that of the stress model. Hysteretic behaviour represents the nonlinear relationship of the stress–strain response for HMA under cyclic loading during fatigue testing. In this regard, a successful trial introduced for modelling the hysteresis loops using Bouc-Wen model for HMA samples tested for fatigue under controlled strain mode using DSR. The nonlinear least squares algorithm was used to estimate the seven parameters for the Bouc-Wen model using experimental results for hysteresis loops of the HMA samples tested in DSR; these parameters control the shape and slope of the degraded hysteresis loops. The outcome of this work confirmed that there is a good agreement between the modelled and experimental hysteresis loops. Due to the variation in the fatigue performance of HMA samples as a result of their properties, the Bouc-Wen model was not able to fully simulate the degradation for different samples when there were changes in the parameters. To improve the Bouc-Wen model’s simulation performance, an ANN technique was used to develop models to predict its parameters; this technique improved the Bouc-Wen model’s performance in Phase I, while its performance in Phase II-III was still poor, despite the degradation simulation being clear. This work revealed the feasibility of using the DSR technique in evaluating the fatigue performance of full HMA according to the developed approaches for preparing and selecting DSR samples and performing fatigue tests. In addition, the work confirmed that limestone has a better fatigue performance for both HRA and DBM mixes than granite. On the other hand, the theoretical part included developing several models based on an ANN and constitutive equations showed the efficiency of these models in predicting the fatigue life of HMA samples tested in the DSR.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available