yash-n-p / akhil-cr-workshop

Framework

The Library is to develop a methodology to use machine learning models as a surrogate for finite element human body models.

FE

This library conatin Finite Elemental class which performs various operations to Run FE simulations and post processing the simulation results. The FE simulations are run using the sampled data set using stanadard sampling methods for FE-parameters input data provided by the user. The simulation results are processed into post processing software named META to obatin required results. The results from this will used as the raw data for training Machine learning Algorithms.

Flowchart of FE:

Input files:

This is a user defined files which is specified by the user in settings.yaml

# Input settings file for FE simulations to get the user input:
Newfolder_name: 'akhil3t_ne'
Runs: 1
key: 'run_main_upd.key'
config: 'config.yaml'
# LS Dyna Run settings
LS_Run_path: 'abc.exe'
NCPU: 4
# Add Meta
meta_exec : 'qwe.bat'
# meta session parh
ses_path : 'K:/New/'
# meta sesion file name
ses_file : 'file.ses'

The config.yaml defined by the user with DOE values:

parameters:
  'delta velocity' :
        type :  dynaParameter
        parameter : DV
        default : 45
        max : 60
        min : 40
        distribution: Uniform

Design of Experiments

Using pyDOE library to sample data based on Latin hypercube method. Each parameter sampling is done by Uniform distribution.

Machine Learning

under development...