JohnnyMercu |
June 23, 2021 04:39 |
Dakota parallel
Hi everyone,
I hope my post won't be off topic, here is my question :
I aml currently using Dakota 6.13 with OpenFoam 2012 in order to optimize parameters, and I am trying to build a surrogate-based optimization. The thing is I want to run several OpenFOAM in parallel to reduce the calculation time. Does anyone have already done something similar ? The OpenFOAM simulation I use runs on 6 processors and here is my dakota input file :
Code:
environment
tabular_data
method
id_method = 'SBGO'
surrogate_based_global
model_pointer = 'SURROGATE'
method_pointer = 'SOGA'
max_iterations = 15
replace_points
output verbose
method
id_method = 'SOGA'
soga
seed = 10983
max_function_evaluations = 30000
#initialization_type unique_random
output silent
model
id_model = 'SURROGATE'
surrogate global
dace_method_pointer = 'SAMPLING'
#correction additive zeroth_order
gaussian_process dakota
method
id_method = 'SAMPLING'
sampling
samples = 50
seed 531
sample_type lhs
model_pointer = 'TRUTH'
model
id_model = 'TRUTH'
single
interface_pointer = 'TRUE_FN'
variables
active uncertain
uniform_uncertain 4
descriptors 'came_start' 'stiffness1' 'stiffness2' 'stiffness3'
lower_bounds -210 200 1200 2600
upper_bounds -150 400 1600 3000
interface
id_interface = 'TRUE_FN'
analysis_drivers = 'dakota3.sh'
fork
work_directory named 'workdir/run'
directory_tag directory_save
parameters_file 'params.in'
results_file 'results.out'
link_files 'dakotaParameter.orig' 'transfert.py' 'ecriture_xml_v3.py'
responses
objective_functions=1
descriptors 'res'
no_gradients
no_hessians
sense 'max'
Best regards,
Johnny
|