LAROMANCE6DInterpolation

commentnote

This is a custom NEML2 object for solid_mechanics. To enable it, configure MOOSE with libtorch and initialize the NEML2 submodule.

The documentation is for a NEML2 (Hu et al., 2024; Messner and Hu, 2023) implementation of the LAROMance surrogate model. The LAROMance surrogate model computes the inelastic effective strain rate used in an isotropic viscoplastic constitutive model update Tallman et al. (2020). The LAROMance surrogate model is fit to a database of polycrystalline crystal plasticity simulations. The material's microstructure, as described by the distribution of dislocation content, temperature, stress, and environment factor (precipitate phase fraction, neutron fluence, etc. ), varied in the crystal plasticity creep simulations used to create the database. The inelastic strain history and mobile and immobile dislocation evolution from the crystal plasticity simulations are then fit to an element-based LAROMance model (Ruybalid et al., 2024; Munday et al., 2025). The element-based input space provides an interpolation grid for the output state of the material. This version of the LAROMance model uses piecewise-continuous linear functions to interpolate the input grid. The six-dimensional (6D) input space for the LAROMance model, , corresponds to the variables varied in the LApx simulations—that is, temperature ( in units K), von Mises stress ( in units MPa), accumulated effective inelastic strain (), mobile or cell dislocation density ( in units ), immobile or wall dislocation density ( in units ), and environmental factor ( in units of number density; ). Interpolated outputs from the LAROMance model are inelastic effective strain rate (), and mobile and immobile dislocation density evolution rates ( and ). Transforms are applied to the input grid and output state to normalize the interpolation space, apply constraints on the output, and allow linear regression to fit a nonlinear relationship between inputs and outputs such as the logarithmic evolution of dislocation density due to accumulated effective inelastic strain Tallman et al. (2020). Additional information about the LAROMance models and a previous implementation is available here.

The LAROMance model implementation in NEML2 is based on multilinear interpolation on a regular nonuniform grid. The six dimensional interpolation grid contains , nodal output values , and transforms are contained in the JSON database file. Extrapolation is performed for outside the grid domain. See report for verification and additional details Munday et al. (2025).

Additional information about a previous version of the LAROMANCE model are documented in the LAROMANCE documentation.

Example Input File

An example input for a NEML2 LAROMANCE model is shown below for inelastic plastic strain rate interpolation, model_file_variable_name = 'out_ep'. The json file contains out_ep on a 6D grid defined by the variables (von_mises_stress, equivalent_plastic_strain, cell_dislocation_density, wall_dislocation_density, temperature, env_factor) which correspond to . The json file only contains interpolation grids for out_ep, out_wall, and out_cell which correspond to . In the below input file, NEML2 performs a nonlinear solve for , output_rate = 'state/ep_rate', which uses automatic differentiation to compute the derivative of the constitutive model with respect to out_ep, . The cell and wall dislocation density rates (out_wall and out_cell) are interpolated in seperate material blocks in the below input file. The dislocation densities are computed from the rates using forward Euler time integration and derivatives for out_wall and out_cell are never computed. The dislocation densities are not included in the nonlinear material solve and are lagged by a timestep.

[Models]
  [rom_ep]
    type = LAROMANCE6DInterpolation
    model_file_name = 'models/random_value_6d_grid.json'
    model_file_variable_name = 'out_ep'
    output_rate = 'state/ep_rate'
    # grid nodes
    von_mises_stress = 'state/s'
    equivalent_plastic_strain = 'state/ep'
    cell_dislocation_density = 'old_state/cell_dd'
    wall_dislocation_density = 'old_state/wall_dd'
    temperature = 'forces/T'
    env_factor = 'forces/env_fac'
  []
[]
(moose/modules/solid_mechanics/test/tests/neml2/laromance/models/laromance_matl_radial_return.i)

References

  1. Tianchen Hu, Mark C. Messner, Daniel Schwen, Lynn B. Munday, and Dewen Yushu. Neml2: a high performance library for constitutive modeling. Technical Report, Argonne National Laboratory (ANL), Argonne, IL (United States); Idaho National Laboratory (INL), Idaho Falls, ID (United States), 09 2024. URL: https://www.osti.gov/biblio/2440430, doi:10.2172/2440430.[BibTeX]
  2. Mark Messner and Tianchen Hu. NEML2 - The New Engineering Material Model Library, version 2. Technical Report, Argonne National Lab.(ANL), Argonne, IL (United States), 2023.[BibTeX]
  3. Lynn Munday, Daniel Schwen, Spencer, Benjamin, Tianchen Hu, and Andre Ruybalid. Laromance grade 91 model integration in neml2. Technical Report, Idaho National Laboratory (INL), Idaho Falls, ID (United States), 02 2025.[BibTeX]
  4. Andre Paul Ruybalid, Andrea Rovinelli, Kyle Gamble, Ryan Sweet, and Laurent Capolungo. Complete development of LAROMance-type surrogate models for Zircaloy, as well as initial integration and testing in MOOSE, which addresses industry needs for licensing advanced LWR fuels. Technical Report, Los Alamos National Laboratory (LANL), Los Alamos, NM (United States), 09 2024. URL: https://www.osti.gov/biblio/2467355, doi:10.2172/2467355.[BibTeX]
  5. Aaron E. Tallman, M. Arul Kumar, Christopher Matthews, and Laurent Capolungo. Surrogate modeling of viscoplasticity in steels: application to thermal, irradiation creep and transient loading in HT-9 cladding. JOM, 73(1):126–137, 10 2020. URL: https://doi.org/10.1007/s11837-020-04402-2, doi:10.1007/s11837-020-04402-2.[BibTeX]