The HVS tests were simulated using an incremental-recursive program known as CalME (Ullidtz et al., 2007). Data from each HVS test were imported into a CalME database. The data comprised information on loads (time of application and load level), temperatures at different levels, RSD results, MDD resilient and permanent deformations and pavement profiles.

The backcalculated layer moduli from the last FWD test before commencement of the HVS loading were used as the initial layer moduli (for asphalt layers at the reference temperature of 20 ºC). Layer moduli were backcalculated using CalBack. For asphalt layers the master curve was obtained from frequency sweep tests on beams in the laboratory, with the exception of the original DGAC layer where the master curve was based on FWD backcalculated moduli. For the subgrade the change in stiffness with changing stiffness of the pavement layers and with changing load level was obtained from FWD backcalculated values. These parameters were used with the response model (LEAP, Symplectic Engineering Corporation, 2004) to calculate stresses, strains and deflections in the pavement structure. The strain in the overlay over an existing cracked asphalt layer was calculated using the reflection cracking model described below.

To predict the pavement performance, in terms of cracking and permanent deformation, a number of models were used. Parameters for prediction of asphalt fatigue damage were obtained from controlled strain fatigue tests on beams. Repeated Simple Shear Tests at Constant Height (RSST-CH) were used to determine the parameters for predicting permanent deformation in the asphalt layers. A crushing model was developed for the self-cementing base layer, consisting of recycled building waste material with a high content of crushed concrete. Cracking at the pavement surface was calculated from the reflection damage to the surface layer, using a model developed based on previous simulations of HVS tests and the WesTrack experiment, with coefficients modified based on the results of the present experiment. For the details of these and other models used in CalME see Ullidtz, et al. 2008a and 2008b.

An incremental-recursive process was used to simulate the performance of the test sections. The time increment used was one hour. For the first hour of the simulation the program would read the temperatures from the database and calculate the moduli, for a constant wheel speed of 9.6 km/h, the approximate speed of the HVS wheel. The number of loads during the first hour, as well as the load level and the tire pressure, were also read from the database. The modulus of the subgrade would be adjusted to the stiffness of the pavement layers and to the load level. If the test had wheel wander, five different positions of the wheel would be considered. For the first wheel position the stresses and strains at the center line of the test section were calculated and used to determine the decrease in moduli and the increase in permanent deformation of each of the pavement layers. The output from these calculations were used, recursively, as input to the calculation for the next wheel position. Because of the changes to moduli, response, damage, and permanent deformation the “time hardening” procedure was used (Deacon et al. 2002).

The first step in the simulation is to make sure that the calculated pavement response is reasonably close to the actual pavement response during the test. The calculated pavement response is used to predict the pavement performance (damage and permanent deformation). Therefore, if the calculated response is not reasonably correct it would be futile to try to use it for calibration of the performance models. For the HVS tests used for this paper, response measurements were available in the form of resilient MDD deflections and/or RSD deflections.

Once the resilient deflections are predicted reasonably well during the simulations, it is possible to calibrate the performance models so that the permanent deformation of each layer, the decrease in layer moduli and the observed surface cracking, are reasonably well predicted.