Probabilistic data-driven melt pool models

This is the official implementation of the paper "Probabilistic Data-Driven Modeling of a Melt Pool in Laser Powder Bed Fusion Additive Manufacturing" published on IEEE Transactions on Automation Science and Engineering (TASE).

Please feel free to ask any questions about this work!

Introduction

The widespread adoption of laser powder bed fusion (LPBF) additive manufacturing is hampered by process unreliability problems. Modeling the melt pool behavior in LPBF is crucial to develop process control methods. While data-driven models linking melt pool dynamics to specific process parameters have shown appreciable advancements, existing models often oversimplify these relationships as deterministic, failing to account for the inherent instability of LPBF processes. Such simplifications can lead to overconfident and unreliable predictions, potentially resulting in erroneous process decisions. To address this critical issue, we propose a probabilistic data-driven approach to melt pool modeling that incorporates process noise and uncertainty. Our framework formulates a problem that includes distribution approximation and uncertainty quantification. Specifically, the Gaussian distribution with higher order priors, aided with variational inference and importance sampling, is used to approximate the probability distribution of melt pool characteristics. The uncertainty inherent in both LPBF process data and the modeling approach itself are then decomposed and approximated by using Monte Carlo sampling. The melt pool model is improved further by using a novel grid-based representation for the neighborhood of a fusion point, and a neural network architecture designed for effective feature fusion. This approach not only refines the accuracy of the model but also quantifies the uncertainty of the predictions, thereby enabling more informed decision-making with reduced risk. Two potential applications, including LPBF process planning and anomaly detection, are discussed.

Installation

It is suggested to use Conda to configure Python environments. First git clone the project:

git clone https://github.com/qihangGH/probabilistic_melt_pool_model.git

Then change the directory to probabilistic_melt_pool_model:

cd probabilistic_melt_pool_model

Create the virtual environment:

conda env create -f prob_mp.yaml

Activate the environment:

conda activate prob_mp

Data

Download the melt pool monitoring dataset released by the National Institute of Standards and Technology (NIST) at https://data.nist.gov/od/id/85196AB9232E7202E053245706813DFA2044, including In-situ Meas Data.zip and Build Command Data.zip. A dataset description article is provided at https://nvlpubs.nist.gov/nistpubs/jres/124/jres.124.033.pdf.

In Build Command Data.zip, unzip XYPT Commands folder, and in In-situ Meas Data.zip, unzip the Melt Pool Camera folder.

Preprocessing

Extract melt pool size

python mp_feature_extraction.py \
    --data_dir <dir_to_Melt_Pool_Camera> \
    --save_filepath <filepath_to_save_results>

Prepare the grid-based neighborhood dataset

First run

python datasets/grid_based_data.py \
    --original_data <dir_to_Build_command_data> \
    --melt_pool_feature_path <filepath_to_mp_feature> \ 
    --save_path <path_to_save_results>

You can change the layers to be processed in datasets/grid_based_data.py. Then generate the dataset by

python datasets/generate_grid_based_datasets.py \
    --root_save_dir <root_dir_to_save_data> \
    --patch_data_path <patch_data_generated_by_grid_based_data.py>

where root_save_dir is the folder to save the dataset, and patch_data_path is the same as the save_path set in grid_based_data.py.

Training

python train_mp_models.py \
    --data_dir <directory_of_the_dataset> \
    --save_dir <directory_to_save_training_results> \
    --model_name <model_name>

where data_dir is the directory of the dataset, which is the same as root_save_dir set in generate_grid_based_datasets.py, save_dir is the directory to save training results, and model_name can be non_bayesian, gaussian, or student. The non_bayesian indicates a deterministic model, gaussian indicates a probabilistic model with the Gaussian distribution, and student is the Gaussian distribution with higher order prior, resulting in a Student's t-distribution. There are many other settings about the model structure, learning rate, etc. You can refer to training_mp_models.py for more details.

Testing

python test_mp_models.py \
    --save_dir <directory_where_training_results_saved>

where save_dir is the same as the one set in train_mp_models.py.

Citation

If you find our work and this repository helpful, please consider citing our papers:

@article{fang2022process,
  author={Fang, Qihang and Xiong, Gang and Zhou, MengChu and Tamir, Tariku Sinshaw and Yan, Chao-Bo and Wu, Huaiyu and Shen, Zhen and Wang, Fei-Yue},
  journal={IEEE Trans. Autom. Sci. Eng.}, 
  title={Process Monitoring, Diagnosis and Control of Additive Manufacturing}, 
  year={2024},
  volume={21},
  number={1},
  pages={1041--1067},
  doi={10.1109/TASE.2022.3215258}
}

@article{fang2024probabilistic,
  author={Fang, Qihang and Xiong, Gang and Zhao, Meihua and Tamir, Tariku Sinshaw and Shen, Zhen and Yan, Chao-Bo and Wang, Fei-Yue},
  journal={IEEE Trans. Autom. Sci. Eng.}, 
  title={Probabilistic Data-Driven Modeling of a Melt Pool in Laser Powder Bed Fusion Additive Manufacturing}, 
  year={2024},
  pages={1--18},
  doi={10.1109/TASE.2024.3412431}
}