Artículos de revistas
An adaptive concurrent multiscale model for concrete based on coupling finite elements
Fecha
2018-01-01Registro en:
Computer Methods in Applied Mechanics and Engineering, v. 328, p. 26-46.
0045-7825
10.1016/j.cma.2017.08.048
2-s2.0-85029712154
2-s2.0-85029712154.pdf
7901652737291917
0000-0001-9004-7985
Autor
Universidade de São Paulo (USP)
Universidade Estadual Paulista (Unesp)
Zachry Department of Civil Engineering
Institución
Resumen
A new adaptive concurrent multiscale approach for modeling concrete that contemplates two well separated scales (represented by two different meshes) is proposed in this paper. The macroscale stress distribution is used as an indicator to identify critical regions (where the material is prone to degrade) with the explicit aim to enrich these zones with detailed mesoscale material information comprising three basic phases: coarse aggregates, mortar matrix and interfacial transition zone. Thus, the concrete initially idealized as a homogeneous material is gradually replaced and enhanced by a heterogeneous multiphase one. This technique is particularly powerful to handle cases where the region with nonlinear behavior is not easy to anticipate. Furthermore, the proposed approach does not require the definition of a periodic cell (or a RVE), and the meshes from distinct scales are totally independent. The new adaptive mesh technique is based on the use of coupling finite elements to enforce the continuity of displacements between the non-matching meshes associated with the two different scales of analysis. Besides that, mesh fragmentation concepts are incorporated to simulate the crack formation and propagation at the mesoscopic scale, without the need of defining complex and CPU-time demanding crack-tracking algorithms. The strategy is developed integrally within the framework of continuum mechanics, which represents an advantage with respect to other approaches based on discrete traction/separation-law. Numerical examples with complex crack patterns are conducted to validate the proposed multiscale approach. Furthermore, the efficiency and accuracy of the novel technique are compared against full mesoscale and standard concurrent multiscale models, showing excellent results.