Modeling Metallic Microstructure Using Moving Finite Elements.pdfVIP

May 1997 Los Alamos National Laboratory Theoretical Division Special Feature 7 9 9 T-1 1 Equation of State y a Mechanics of Materials M 8 Modeling Metallic Microstructure Using Moving Finite Elements 2 Andrew Kuprat and J. Tinka Gammel ] h p - To attack the general problem of evolution of metal- we have evolved an initial microstructure of randomly ori- p lic grain microstructure in 3D, we employ two different ented grains on a fixed lattice using a discrete effective m methods: using Monte Carlo (MC) techniques to anneal classical spin model originally developed to model mag- o an effective discrete model on a fixed lattice [1], and using netic domains (Potts model) [1]. For the first few time c Gradient Weighted Moving Finite Elements (GWMFE) steps, MC scales with the number of volume points. Ac- . s to evolve grain boundaries by mean curvature (local ve- tivity binning and N-fold or cluster flip MC techniques al- c locity proportional to the local curvature) – the simplest low significant speed-ups in the time evolution [1]. (Incor- i s continuum model for grain evolution [2,3]. These two poration of volume information may reduce the amount y h models reduce to each other in the appropriate limits. of these speed-ups.) We evolved both a regular fcc lat- p MC has the advantage of computational and algorithmic tice, where discreteness trapping is not a problem [1], and [ simplicity while GWMFE is more easily adaptable to in- unstructured grids resulting from the LANL X3D code, cluding coupled transport fields such as vacancies and where we see from Fig. 1 that 5th neighbor interactions 1 v impur