Enhanced nonlinear analysis of 3D concrete structures

Abstract

Although numerical simulation of concrete has a significant background in the framework of simplified one- and two-dimensional elements, a full triaxial description of the structural behaviour of this material is still subject to active research. High fidelity modelling has only been enabled once the required computational capacity achieved an appropriate threshold, and it is precisely of such computational nature that there are diverse drawbacks the material model has to overcome.

For concrete, an existing model combining plasticity and isotropic damage is chosen in this work, and this choice over multi-surface plasticity is duly justified. Additionally, an extension to anisotropic damage is proposed. Focus is set on a series of algorithmic enhancements that significantly increase robustness in stress evaluation, in particular from stress states that pathologically associate to a singular Jacobian matrix and stress-returns that lead towards sensitive areas of the failure surface in principal stress space, where plastic flow is undefined.

Reinforcing steel is modelled as embedded bars inside the corresponding concrete parent elements, with solely axial stiffness. An arbitrary orientation inside the concrete elements is allowed but otherwise the discretised bars share the parent element morphology, order and degrees of freedom, resulting in a perfect bond interaction. An improved and systematic linearising procedure is presented to track the intersections of each bar segment with its embedding parent element, which can be readily applied to any element type and order. This facilitates an accurate calculation of this constituent’s contribution to the parent element’s stiffness matrix and nodal force vector.

The robustness of the enhanced material model is verified by means of numerical tests, highlighting the convergence ratio, and validation ensues via simulations of established benchmark tests. Finally, some case studies are presented to illustrate the performance of the model at structural level, with insight into various issues of computational nature.