This repository contains the Python code for the paper "A PINN–EWMA framework with uncertainty quantification for high-fidelity temperature field modeling and early defect warning in directed energy deposition of thin-walled structures"
WARNING: These codes are written only for the purpose of demonstration and verification. While the correctness has been carefully checked, the quality such as standardability, clarity, generality, and efficiency has not been well considered.
Our study proposes an enhanced PINN–EWMA framework specifically designed for high-fidelity temperature field prediction and early defect warning in the Directed Energy Deposition (DED) process of thin-walled high-entropy alloys. The enhanced PINN model integrates temperature-dependent material properties, entropy and energy-based regularization, and infrared sensor uncertainty modeling, enabling high-fidelity temperature field predictions in DED processes.
The main function of the code is implemented in (main_PINN.py), while the training function is provided in (train.py). The underlying FNN architecture is defined in (model.py), and the sampling strategy is specified in (util.py).