A. T. Papagiannakis

Micromechanical analysis of viscoelastic properties of asphalt concretes

A methodology for relating the microstructure of asphalt concretes to their viscoelastic behavior is described. Imaging techniques are used to capture the asphalt concrete microstructure, and the finite element method (FEM) is used to model its stress-strain behavior in the time domain. Aggregates are modeled as linear elastic, and the binder is modeled through mechanistic models as either linear viscoelastic or nonlinear viscoelastic. The binder viscoelastic properties are input into the FEM algorithm by two methods: a built-in viscoelastic function and a user-specified material characterization subroutine. The latter handles non-linearity in an iterative piecewise linear fashion, whereby the mechanistic binder model parameters are updated as a function of the strain level. For each strain level, mechanistic models are fitted to describe binder viscoelastic behavior based on dynamic shear rheometer data. The two approaches used for specifying binder viscoelastic properties into the FEM algorithm were verified by comparing binder response predictions with direct measurements. Finally, the asphalt concrete micro-structure model was verified by comparing FEM predictions of dynamic shear modulus and phase angle with measurements obtained by using a Superpave® shear tester.

A volumetrics thresholding algorithm for processing asphalt concrete X-ray CT images

This paper presents an automated digital image processing technique for capturing the microstructure of asphalt concrete (AC) from X-ray computed tomography images. It applies to circular cross-section images of AC cores of known volumetrics. Its innovation is that it uses the volumetric properties as the main criterion for establishing grey-level thresholds for the boundaries between air–mastic and mastic–aggregates. The algorithm, implemented in MATLAB™, involves three stages. The first stage involves image pre-processing for contrast enhancement and noise removal. The second stage is the main thresholding routine accepting as input the enhanced images of the first stage and volumetric information for the AC. It consists of two components, namely volumetrics-driven thresholding and 3D representation/sectioning. The third stage further enhances particle separation through edge detection and image segmentation. Examples were provided in imaging actual AC cores of known volumetric properties. The algorithm was shown to produce realistic rendering of the microstructure of ACs. Their quality is suitable for input into numerical simulation.

An improved image processing technique for asphalt concrete X-ray CT images

This paper presents a comparative evaluation of two image segmentation techniques for processing asphalt concrete microstructure images obtained with X-ray computed tomography (CT). These are the adaptive enhancement-based thresholding algorithm (AETA) and the watershed segmentation embedded into the volumetric-based thresholding algorithm (VTA). Both these methods were used to process the X-ray CT images of nine asphalt concretes. These consisted of three mix types, each prepared with three aggregate types. The mix designs included a coarse matrix high-binder Type C mix, a gap-graded porous friction course mix, and a fine-graded Superpave Type C (Superpave) mix. The three aggregate types included hard limestone, granite, and soft limestone. All mixtures were prepared with a PG 76-22 modified binder. The comparison of the two methods was carried out both visually and quantitatively. The later was accomplished by comparing the gradation estimated from the images using purpose-designed software to the gradation obtained from mechanical sieving. The results show that the inherent over-segmentation problem with the VTA technique is effectively reduced using the AETA method. Overall, the AETA method outperforms the VTA watershed image segmentation method by producing better separation between connected and overlapping aggregates for the nine asphalt concretes included in the study. Its drawback is that it does not preserve the volumetric properties of the mixtures, as done by the VTA technique.

A wavelet interpretation of vehicle-pavement interaction

This paper utilizes a wavelet approach to interpret the interaction between truck dynamic axle loads and pavement roughness profile. The experimental data used was obtained from an instrumented 5-axle semi-trailer truck equipped with an air and a rubber suspension in the drive and trailer axles, respectively. Wavelet decomposes the original signal into a number of sub-band levels depending on its characteristics. The size of the dataset used allowed 11 levels of wavelet decomposition with test speed dependent frequencies ranging up to 4 cy/m. For dynamic load, the extent of variation in each of these wavelength sub-bands, was summarized through an energy metric effectively computed as the sum of the squares of the wavelet coefficients in each sub-band. Total energy was computed as the sum of the energy of all sub-bands and was normalized by dividing by the length of the test section. Relative energy was computed as the percent of total energy in each sub-band. The results were summarized through 3D plots of relative energy versus load frequency versus profile frequency. The profile frequencies mostly affecting dynamic loads depend on speed and range from 0.65 to 3.76 cy/m.

Pavement macro-texture analysis using wavelets

This paper utilises a wavelet approach to interpret the macro-texture data collected on asphalt and concrete pavement surfaces with a wide range of macro-texture properties. The experimental data were obtained using a circular track meter (CTMeter) device on pavements built at the Virginias Smart Road test facility. The size of the data-set allowed nine levels of wavelet decomposition with wavelengths ranging from 1.7 to 435 mm. The extent of macro-texture variation was summarised using the normalised wavelet energy metric defined as the sum of the squares of the detailed wavelet coefficients for the sub-bands that correspond to the macro-texture range of wavelengths divided by the length of the test section expressed in mm2/m. This metric was found highly correlated with the empirical mean profile depth measurements. Hence, the wavelet approach can be used to objectively analyse CTMeter measurements of pavement texture.

Wavelet-based characterisation of aggregate segregation in asphalt concrete X-ray computed tomography images

Aggregate segregation, defined as the non-uniform distribution of fine and coarse aggregates within asphalt concretes (ACs), is recognised as a serious pavement construction problem that reduces pavement life. This paper describes a mathematical approach for quantifying the degree of directional aggregate segregation in ACs. It utilises a wavelet approach for separating coarse from fine aggregates and the normalised wavelet energy to quantify aggregate concentration (i.e. white grey scale intensity). A directional segregation index (SI) is defined as the ratio of fine-to-coarse aggregate normalised wavelet energy. This index is shown to be a good indicator of the degree of segregation observed in X-ray CT images of AC samples. The approach demonstrated using laboratory compacted samples can be applied to quantify segregation in 2D images of composite materials.

Wavelet Analysis of Energy Content in Pavement Roughness and Truck Dynamic Axle Loads

A wavelet approach is used to study the interaction between pavement roughness profile and heavy-truck axle dynamics. Profile and load variation are quantified in terms of their sum of squares or energy, in units of millimeters squared per meter and kilonewtons squared per meter, respectively. The dynamic axle load and pavement roughness profile data are decomposed into distinct frequency subbands and relative or percent energy is computed. The results are presented in the form of three-dimensional plots of relative dynamic load energy, pavement roughness energy, and pavement roughness frequency. They demonstrate the drastically different response of two types of suspensions to roughness excitation. Energy also was normalized by dividing by the length traveled, resulting in units of millimeters squared per meter and kilonewtons squared per meter for the profile and the dynamic load, respectively. The normalized energy results show a much higher dynamic load activity for the rubber suspension than for the air suspension. The normalized energy of load shows a high correlation with the normalized energy of the pavement roughness. The normalized energy of the pavement roughness profile shows a high correlation with the conventional international roughness index.

Energy Harvesting from Roadways

This paper presents a preview of an ongoing study to develop an energy harvesting system based on piezoelectric elements embedded into the pavements structure. The system development involved designing and testing a number of prototypes in the laboratory under controlled stress conditions. In addition, it involved numerical modeling of the stress distribution in the power generation module and economic analysis of the value of the electric power generated, under a given traffic composition scenario. The results available to date suggest that this technology shows promise in powering LED traffic lights and wireless sensors embedded into pavement structures.

Interpreting asphalt concrete creep behaviour through non-Newtonian mastic rheology

This paper explores the use of non-Newtonian mastic models in interpreting asphalt concrete creep behaviour. Three non-Newtonian models were considered, namely the Bingham model, the Herschel-Bulkley model and the Casson model. They were fitted to dynamic shear rheometer (DSR) strain sweep data on mastics. A total of nine mastics were included involving three asphalt mix designs and three aggregate types (hard limestone, granite, and soft limestone). A PG 76-22 binder was used for all of the mastic specimens. Unconfined creep tests were performed on the asphalt concrete mixtures that constitute the mastics using a compressive stress of 207 kPa. The mastic and asphalt concrete mixture specimens were tested at 60°C. The characteristics of the non-Newtonian mastic models (i.e. yield stress, plastic viscosity and consistency) were found to be highly dependent on mastic compositions. Furthermore, the secondary creep curves of the asphalt concrete mixture that constitute the mastics were also found to be correlated to the constants of the mastic models. The plastic viscosity and consistency explained over 75% of the observed variation in asphalt mix creep slope, while the yield stress explained approximately 50% of the observed variation in the asphalt concrete creep intercept.

Effect of Traffic Load Input Level on Mechanistic-Empirical Pavement Design

This study investigated the effect of the input level of axle load spectra on performance predictions obtained with the Mechanistic–Empirical Pavement Design Guide (MEPDG). Traffic monitoring data collected from 50 weigh-in-motion (WIM) stations distributed across the state of Ohio were analyzed to obtain site-specific and regional axle load spectra. The site-specific axle load spectra were calculated for each axle group (single, tandem, tridem, and quad) and truck class (Classes 4 through 13). The regional axle load spectra were developed on the basis of statewide averages and cluster analysis. The statewide average was calculated with information from all sites, whereas the cluster averages were created on the basis of Class 9 tandem axles. Baseline designs for new flexible pavements were defined in the MEPDG for each of the WIM sites having available data, and site-specific, statewide average, cluster average, and MEPDG default axle load spectra were used to assess the impact of the input level of axle load spectra on the performance of the pavement structure. The statewide average and the cluster averages were found to yield similar predictions of pavement performance, with the cluster averages being slightly closer to the site-specific predictions of pavement performance. In addition, the MEPDG default axle load spectra were found to underestimate the pavement service life significantly and, if used for design purposes, would result in overly conservative pavement layer thicknesses. Therefore, the authors recommend the use of site-specific axle load spectra when possible in the design of flexible pavements. However, if site-specific data are not available, statewide average axle load spectra should be used.