numerics.yaml#
The algorithms used in DAMASK can be configured via a YAML file that is typically called numerics.yaml.
This optional file needs to be specified with the –numerics option when using the DAMASK grid or mesh solvers, or located in the current working directory and named numerics.yaml to be included in a simulation with commercial FEM solvers.
Using damask.Config is the recommended way for the generation and modification of numerics.yaml.
Hand-written files should be checked for correctness, e.g. with yamllint or on www.yamllint.com.
Reference Examples#
All available parameters are usually listed in the file examples/config/numerics.yaml.
The content of this file is shown below.
numerics#
# Default values of all available numerical parameters
# Please note that keys are case sensitive
solver:
grid:
N_staggered_iter_max: 10 # max number of field-level staggered iterations
N_cutback_max: 3 # maximum cutback level (0: 1, 1: 0.5, 2: 0.25, etc)
damage:
N_iter_max: 100 # maximum iteration number
eps_abs_phi: 1.0e-2 # absolute tolerance for damage evolution
eps_rel_phi: 1.0e-6 # relative tolerance for damage evolution
phi_min: 1.0e-6 # residual integrity
thermal:
N_iter_max: 100 # maximum iteration number
eps_abs_T: 1.0e-2 # absolute tolerance for thermal equilibrium
eps_rel_T: 1.0e-6 # relative tolerance for thermal equilibrium
mechanical:
N_iter_min: 1 # minimum iteration number
N_iter_max: 100 # maximum iteration number
eps_abs_div(P): 1.0e-4 # absolute tolerance for fulfillment of stress equilibrium
eps_rel_div(P): 5.0e-4 # relative tolerance for fulfillment of stress equilibrium
eps_abs_P: 1.0e3 # absolute tolerance for fulfillment of stress BC
eps_rel_P: 1.0e-3 # relative tolerance for fulfillment of stress BC
update_gamma: false # update Gamma-operator with current dPdF (not possible if FFT: memory_efficient == true)
PETSc_options: -snes_type ngmres -snes_ngmres_anderson # PETSc solver options
FFT:
memory_efficient: true # precalculate Gamma-operator (81 doubles per point)
divergence_correction: size+grid # use size-independent divergence criterion {none, size, size+grid}
derivative: continuous # approximation used for derivatives in Fourier space {continuous, central_difference, FWBW_difference}
FFTW_plan_mode: FFTW_MEASURE # planning-rigor flags, see manual at https://www.fftw.org/fftw3_doc/Planner-Flags.html
FFTW_timelimit: -1.0 # time limit of plan creation for FFTW, see manual on www.fftw.org. (-1.0: disable time limit)
alpha: 1.0 # polarization scheme parameter 0.0 < alpha < 2.0 (1.0: AL scheme, 2.0: accelerated scheme)
beta: 1.0 # polarization scheme parameter 0.0 < beta < 2.0 (1.0: AL scheme, 2.0: accelerated scheme)
eps_abs_curl(F): 1.0e-10 # absolute tolerance for fulfillment of strain compatibility
eps_rel_curl(F): 5.0e-4 # relative tolerance for fulfillment of strain compatibility
mesh:
N_cutback_max: 3 # maximum cut back level (0: 1, 1: 0.5, 2: 0.25, etc)
N_staggered_iter_max: 10 # max number of field level staggered iterations
p_s: 2 # order of displacement shape functions
p_i: 2 # order of quadrature rule required
bbarstabilization: false
mechanical:
N_iter_max: 250 # Maximum iteration number
eps_abs_div(P): 1.0e-10 # absolute tolerance for mechanical equilibrium
eps_rel_div(P): 1.0e-4 # relative tolerance for mechanical equilibrium
Marc:
unit_length: 1.0 # physical length of one computational length unit
homogenization:
mechanical:
RGC:
eps_abs_P: 1.0e+4 # absolute tolerance of RGC residuum (in Pa)
eps_rel_P: 1.0e-3 # relative ...
eps_abs_max: 1.0e+10 # absolute upper-limit of RGC residuum (in Pa)
eps_rel_max: 1.0e+2 # relative ...
Delta_a: 1.0e-7 # perturbation for computing penalty tangent
relevant_mismatch: 1.0e-5 # minimum threshold of mismatch
viscosity_exponent: 1.0e+0 # power (sensitivity rate) of numerical viscosity in RGC scheme
viscosity_modulus: 0.0e+0 # stress modulus of RGC numerical viscosity (0: without numerical viscosity)
# suggestion: larger than the aTol_RGC but still far below the expected flow stress of material
dot_a_ref: 1.0e-3 # reference rate of relaxation (about the same magnitude as straining rate, possibly a bit higher)
dot_a_max: 1.0e+0 # threshold of maximum relaxation vector increment (if exceed this then cutback)
Delta_V_max: 1.0e-5 # maximum allowable relative volume discrepancy
Delta_V_modulus: 1.0e+12
Delta_V_exponent: 5.0
phase:
mechanical:
r_cutback_min: 1.0e-3 # minimum (relative) size of step allowed during cutback in phase state calculation
r_cutback: 0.25 # factor to decrease size of step when cutback introduced in phase state calculation (value between 0 and 1)
r_increase: 1.5 # factor to increase size of next step when previous step converged in phase state calculation
eps_rel_state: 1.0e-6 # relative tolerance in phase state loop (abs tol provided by constitutive law)
N_iter_state_max: 10 # state loop limit
plastic:
r_linesearch_Lp: 0.5 # factor to decrease the step if Lp calculation fails to converge
eps_rel_Lp: 1.0e-6 # relative tolerance in Lp residuum
eps_abs_Lp: 1.0e-8 # absolute tolerance in Lp residuum
N_iter_Lp_max: 40 # stress loop limit for Lp
f_update_jacobi_Lp: 1 # frequency of Jacobian update of residuum in Lp
integrator_state: FPI # integration method (FPI = Fixed Point Iteration, Euler = Euler, AdaptiveEuler = Adaptive Euler, RK4 = classical 4th order Runge-Kutta, RKCK45 = 5th order Runge-Kutta Cash-Karp)
eigen:
r_linesearch_Li: 0.5 # factor to decrease the step if Li calculation fails to converge
eps_rel_Li: 1.0e-6 # relative tolerance in Li residuum
eps_abs_Li: 1.0e-8 # absolute tolerance in Li residuum
N_iter_Li_max: 40 # stress loop limit for Li
f_update_jacobi_Li: 1 # frequency of updating the Jacobian of residuum in Li
generic:
random_seed: 0 # fixed seeding for pseudo-random number generator (0: use random seed)
Note
The default YAML parser used in DAMASK does not support advanced features, such as references and line continuation statements, unless DAMASK was compiled with libfyaml.