A massively parallel multiscale CAFE framework for the modelling of fracture in heterogeneous materials under dynamic loading. (January 2020)

• Record Type: Journal Article
• Authors: Hewitt, Sam; Margetts, Lee; Shterenlikht, Anton; Revell, Alistair
• Is Part Of: Advances in engineering software. Volume 139(2020)
• Publication Date: 2020-01
• Subjects: Multiscale -- Parallel -- Fracture -- Cellular automata -- Finite element; Computer-aided engineering -- Periodicals; Engineering -- Computer programs -- Periodicals; Engineering -- Software -- Periodicals; Periodicals; 620.0028553
• DOI: 10.1016/j.advengsoft.2019.102737
• Languages: English
• ISSNs: 0965-9978

Abstract

Highlights: A multiscale modelling mini-app for dynamic problems is presented. A mechanistic model for simulating fracture is coupled to a finite element library. A good representation of Mode I fracture in polycrystalline materials is seen. The mini-app achieved good scaling on approximately 24, 000 cores. Abstract: This paper presents a novel computational framework for modelling multiscale fracture that can be used to solve engineering problems subject to dynamic loading. The framework simulates, mechanistically, at the mesoscale, the physical processes that lead to brittle fracture. A homogenisation step is used to translate the accumulation of damage from the mesoscale to the macroscale (as a reduced stiffness in the corresponding region of the structure). In order to achieve this, the multiscale framework couples together two open source Fortran packages; the macroscale ParaFEM with the mesoscale CASUP. ParaFEM is a highly parallel finite element analysis library used to model structures at the continuum scale. CASUP is a package that uses cellular automata to simulate brittle fracture in polycrystalline materials. A simple test problem involving a vibrating cantilever beam is used to demonstrate the simulation of dynamic cyclic loading, leading to brittle cracking. In the cellular automata software, there are a range of parameters that can be adjusted, such as the fracture energy and grain size. These are explored to demonstrate how they might affect the predicted structural integrity of the cantilever beam. Parallel performance is investigated using a Cray XC30 supercomputer, showing that the software can make efficient use of tens of thousands of cores. This paper highlights that modelling the physical mechanisms that lead to damage and plasticity could be an attractive alternative to phenomenological constitutive models. This work will be of interest to researchers and practitioners needing more precise predictions or a better understanding of damage propagation under cyclic or impact loading. With further development, this type of framework will enable the insilico design and evaluation of new material microstructures; leading to improved performance of components and devices subject to extreme operating conditions.