Summary
Our latest paper in Additive Manufacturing introduces a physics-informed machine learning approach for efficient and accurate bead geometry prediction in Wire Arc DED. The hybrid PINN framework balances physical fidelity and computational cost, showing promise for scalable, data-efficient modeling in metal AM.
We’re excited to announce the publication of our recent work in the journal Additive Manufacturing titled:
“Bead Geometry Prediction in Wire Arc Directed Energy Deposition using Physics-Informed Machine Learning and Low-Fidelity Data” by Asif Rashid, Farzad Vatandoust, Akshar Kota, and Shreyes N. Melkote.
Read the full paper here
This work presents a coupled Physics-Informed Neural Network (PINN) framework that integrates process physics with limited experimental data to accurately predict bead geometry in Wire Arc DED. The approach offers a scalable and computationally efficient alternative to high-fidelity simulations, achieving comparable accuracy with significantly reduced training time.
We are also thrilled to share that this paper was recognized as a “Recent Top Paper” on SSRN’s Polymers & Soft Matter Physics eJournal downloads list!
