Improvement of 3D mean field models for capillarity-driven grain growth based on full field simulations

Abstract

International audienceIn the present study, mean field models of grain growth (Hillert and Burke–Turnbull models) are compared with 3D full field simulations considering an isotropic grain boundary energy and mobility and under the absence of second-phase particles. The present 3D full field simulations are based on a level set description of the grain interfaces within a finite element framework. The digital initial microstructures are generated using a coupled “Voronoï–Laguerre/dense sphere packing” algorithm. Based on full field simulation results, new formulations of Burke–Turnbull and Hillert models are proposed. In contrast with classical formulations, the new ones account for the possible heterogeneity of the initial grain size distribution

A novel highly efficient Lagrangian model for massively multidomain simulation applied to microstructural evolutions

“…if quality Q mean (N j , N k ) will improve by swapping then 15: [71,72,67,73] or Zener Pinning (ZP) [74,49,50,51]. In [53] isotropic GG was used to compare the TRM model to other approaches (LS-FE [72,25,56]). The base model used to represent this mechanism is commonly known as curvature flow, where the velocity of an interface is proportional to its local mean curvature.…”

2D and 3D simulation of grain growth in olivine aggregates using a full field model based on the level set method

“…The level set framework used here for the full field modeling of the microstructure supports implicit description of grain boundaries by level set functions in a FE framework. Initially designed for metallurgic considerations, as microstructure evolutions in industrial processes, this model has already been used in 2D or 3D to simulate dynamic [20] or static [32] recrystallization and grain growth [19,22] with possible inert second phase particles [33] in some metallic materials. As for the phase field approach, the main weakness of this procedure is its computational cost, related to the large number of LS functions needed to describe the microstructure.…”

“…where α and n are dependent on initial grain size distribution [19], grain shape or presence of secondary phases, pores or melt [11]. In the case of a monophasic system without pores or melt, values of α and n can be plotted, for log-normal distributions, against the σ µ ratio [19,22]. As an example, α and n are respectively close to 1 2 and 1 (leading to eq.14) for initial grain size distributions which are log-normal with a σ µ ratio equal to 0.45 considering 2D grain growth [19] and may be equal to 0.35 considering 3D grain growth [22].…”

“…The latter is the one chosen for this work. The numerical approach presented in [20,21,22,19] is used to describe microstructural evolutions of materials (essentially metals) during industrial processes. This approach deals with a LS description of interfaces in a FE formulation.…”

Modeling of dynamic and post-dynamic recrystallization by coupling a full field approach to phenomenological laws

“…Since many decades, phenomenological recrystallization models have been proposed like the well-known Avrami relationship aiming to describe the global recrystallized fraction during hot deformation [7,8,9]. Many years later, new recrystallization models emerged with the intention to implicitly describe the microstructure by considering spherical grains [10,11,12,13,14]. In these models, each physical mechanism is described by a physical or phenomenological law.…”