Generalised Kelvin contact models for DEM modelling of asphalt mixtures: Fid-Bau Portal

Generalised Kelvin contact models for DEM modelling of asphalt mixtures

Câmara, Gustavo / Azevedo, Nuno Monteiro / Micaelo, Rui / Silva, Hugo

Rigid particle models based on the discrete element method (DEM) have been adopted to model creep, fracture, and the viscoelastic behaviour of asphalt mixtures considering an irregular micro-structure and particle contacts. Within a DEM framework, the Burgers contact model, which is known to have a narrow frequency and temperature range, is usually adopted to model viscoelastic properties. In this study, a new explicit three-dimensional generalised Kelvin (GK) contact model formulation for the DEM model is proposed for asphalt materials. The model is implemented following two different methodologies (GK $_1$ $_{1}$ and GK $_2$ $_{2}$ ). The models are validated in uniaxial tension-compression sinusoidal tests for predicting the dynamic modulus ( $\vert E^{\ast }\vert$ $| E^{*} |$ ) and phase angle (ϕ) of these composites at a frequency range of 1–10 Hz at $20\, ^{\circ }$ $20^{\circ}$ C. Four mixtures are investigated based on the modelling of their mastic. The influence of the GK contact parameters on the dynamic response of mastics is assessed. Results show that $\kappa _m$ $κ_{m}$ has an important influence on both rheological properties and that $\eta _m$ $η_{m}$ can be used for small adjustments focussing on the predicted phase angle. Moreover, it is shown that a viscoelastic contact model should be adopted to simulate aggregate-to-mastic contacts in mixtures. As expected, the response obtained for both GK models is the same but the simulations with the GK $_1$ $_{1}$ are 14% faster. In addition, the response predicted with the proposed GK contact model is compared with the response predicted with a Burgers contact model. The DEM predictions obtained for the GK model are more accurate. For mastics, the average errors for $\vert E^{\ast }\vert$ $| E^{*} |$ and ϕ when adopting the GK model (Burgers) are 2.40% (3.46%) and 3.64% (4.17%), respectively. For mixtures, the average errors for $\vert E^{\ast }\vert$ $| E^{*} |$ for the GK model (Burgers) are 7.00% (7.92%). The proposed contact model greatly enhances the DEM ability to simulate the viscoelasticity of asphalt materials.