A Unified Analytical and Hybrid Physics-Informed Neural Network Framework for Inverse Design of Multi-layer Impact Welding with Multi-modal Validation: A Case Study on 3-Layer Vaporizing Foil Actuator Welding of Dissimilar Materials

Deepak Kumar

doi:10.1016/j.jmatprotec.2026.119324

A Unified Analytical and Hybrid Physics-Informed Neural Network Framework for Inverse Design of Multi-layer Impact Welding with Multi-modal Validation: A Case Study on 3-Layer Vaporizing Foil Actuator Welding of Dissimilar Materials

2026-05-04

0

PRE

AI

D

Deepak Kumar

J

Jungyu Choi

H

Hyeonbeom Lim

J

Jeong-Suk Kim

H

Hyuckmin Kwon

J

Junyeong Jeong

Y

Yu Mao

A

Anupam Vivek

T

Taeseon Lee *

DOI：10.1016/j.jmatprotec.2026.119324

Original

Original request for help

Save

Abstract

En 中文

• Hybrid physics-informed neural network via physics-filtered data enables inverse design of 3-layer spot impact welding. • A novel analytical model captures recursive shocks & thermal evolution in a true 3-layer system. • Two timing-thermal screening metrics define localized weldability limits at both interfaces. • Validated by meshfree simulations, in situ velocimetry, lap-shear, and metallography. • Deployable design tool enables rapid, physics-consistent process window selection.

Keywords:

Inverse design

Physics-informed neural network

Multi-layer impact welding

Thermal evolution

Weldability limits

A Unified Analytical and Hybrid Physics-Informed Neural Network Framework for Inverse Design of Multi-layer Impact Welding with Multi-modal Validation: A Case Study on 3-Layer Vaporizing Foil Actuator Welding of Dissimilar Materials

Abstract

Journal

Organization

Cited Papers

Citing Papers