This project aimed to develop a surrogate model to predict residual stress of a single weld bead on a plate using arc travel length, travel speed and net power input rate.
Tensorflow and Keras are required to be installed
- pip3
pip3 install tensorflow
Clone the repo to your local directory
git clone https://github.com/leoxiaoyuan/ANN_for_welding.git YOUR_DIRECTORYFolder Abaqus_scripts:
contains scripts that used for creating thermal and mechanical analysis input files and collecting longitudinal stresses on B-D line. Subroutine used for defining the heat source model is stored in Thermal_inp_files folder. The shell scripts for submitting the thermal and mechanical job arrays to CSF are stored in the relevant folders.
Folder ANN_results:
contains the final tuned ANN model and predictions on the test dataset.
Folder extracted_data:
contains 205 sets of data of sampled welding simulations.
Folder data_analysis:
contains scripts for predicting residual stress.
Folder Surrogate_model_modeling:
contains two scripts which built a ANN and a Gaussian Process model.
🚩The use of the developed ANN surrogate model
:one: Open 'ANN_results\saved_model\Predict.py'
:two: Change the welding parameters to those to be predicted
:three: Run the python script, and the predicted stresses will be printed out
🚩 The use of Abaqus python scripts to generate simulations
:one: Open Abaqus command terminal
:two: Change the directory to '~/Abaqus_scripts/'
:three: Run the command 'Abaqus cae noGUI=Create_input_files.py -- a b c d e f'
a, b, c, d, e, f refer to specimen length(m), width(m), height(m), welding length(m), welding speed(m/s), net energy input rate(W) respectively
:four: Submit thermal analysis simulation by running Run_Thermal
:five: Submit mechanical analysis simulation by running Run_Mechanical
🚩The use of Abaqus python scripts to collecting data form simulations
:one: Open Abaqus command terminal
:two: Change the directory to '~/Abaqus_scripts/Mechanical_inp_files/'
:three: Run the command 'Abaqus cae script=extract_data.py'