Implications of spatial variability characterization in discrete particle models

Abstract

Characterization of the inherent spatial variability of the mechanical properties of advanced composite materials as well as of standard concretes is among the primary concerns of the presented paper. In the context of discrete particle models the random field concept is frequently adopted in order to account for the fluctuations of material characteristics on scales independent of the geometrical characterization of the mesostructure as represented by particular particle configurations mimicking the aggregate placement. The specific goal of this paper is to introduce spatial variability into structural reliability by providing a sample reduction strategy for the discrete framework. Presented simulations of a classical experiment utilize the well-established Lattice Discrete Particle Model (LDPM) and demonstrate the general potential of the proposed strategy towards current state-of-the-art in stochastic mechanics and some relating open problems.

Discrete element framework for modeling tertiary creep of concrete in tension and compression

“…Hence, the numerical scatter partially captures the experimental scatter resulting from the material heterogeneity. A more detailed discussion on the particles placement and the related influence on the numerical scatter can be found in[65,66].…”

Modeling Adhesive Anchors in a Discrete Element Framework

“…As observed in the experiments, the random LDPM aggregate placement leads to a non-axisymmetric cone. Work is currently being performed to further improve the realism concerning the LDPM particle placement by mimicking the gravity effects during casting or introducing correlation between particle placement and local concrete properties (see [ 33 , 34 ]).…”

Random and Gradient Based Fields in Discrete Particle Models of Heterogeneous Materials

“…This approach captures most microstructural effects of concrete very well, when compared to the continuum framework, however introducing higher order spatial variability enables to control and interpret the response scatter [6], [9].…”

“…This may lead to a conflict if the governing probability distribution, used for the random field generation, is to be maintained. Otherwise, truncated distributions may be used or the realization of random field can be rescaled to fit admissible range [6], [19], [20]. This is a quite popular abstraction level, especially for continuum models, but stipulates full independence between microstructure, i.e.…”

“…This has prompted research into Monte Carlo (MC) based small-sample simulation methods [3], [4], where, despite the capacity of current computers, and in particular in the context of spatial variability, practical utilization requires the availability of an effective sampling strategy that would dramatically reduce the number of required realizations while maintaining accurate estimates of the response characteristics (low-probability large-consequence events) [5]. Recent attempts addressing the sampling strategy for spatial variability in the MC simulation framework include [6], [7], which is based on the original work of [8], where critical samples of stochastic processes are identified.…”