Unsupervised Domain Adaptation Enables Cross-Process Weld Penetration Prediction in Laser and TIG Welding

Forum topic · 小凯 · 2026-06-26

Summary

Researchers Sen Li, Haichao Cui, and Chendong Shao propose an unsupervised domain adaptation (UDA) framework with a gradual source domain expansion (GSDE) strategy for weld penetration state classification in laser and TIG welding. Supervised deep learning models typically degrade under domain shift when transferred between welding processes with distinct physical mechanisms, such as from arc-dominated TIG welding to keyhole-based laser welding. The proposed method achieves 90.65% average accuracy on TIGFH and 90.72% on LSPS in same-process settings, surpassing supervised baselines by 35.83 and 38.87 percentage points. In cross-process transfer, it reaches 80.48% for TIG-to-laser and 81.13% for laser-to-TIG, improving over baselines by roughly 43 percentage points. UMAP visualization confirms the model learns domain-invariant features while maintaining discriminative class boundaries. The approach significantly reduces relabeling costs when deploying to new welding processes and improves the generalizability of intelligent monitoring across welding systems. Source: arXiv 2606.19223.

Paper Overview

Field: Computer Vision Authors: Sen Li, Haichao Cui, Chendong Shao Published: 2026-06-25 arXiv: 2606.19223

Abstract

Supervised deep learning has been widely used for weld penetration state classification; however, its performance often degrades significantly under domain shift, such as when transferring models between welding processes with distinct physical mechanisms—for instance, from arc-dominated tungsten inert gas (TIG) welding to keyhole-based laser welding. To overcome this limitation, the authors propose an unsupervised domain adaptation (UDA) framework integrated with a gradual source domain expansion (GSDE) strategy.

Key Results

Evaluated on dedicated TIG and laser welding datasets, the approach achieves high accuracy in both same-process and cross-process transfer tasks:

Same-process transfer: 90.65% average accuracy on TIGFH and 90.72% on LSPS, surpassing a supervised baseline by 35.83 and 38.87 percentage points, respectively.
Cross-process transfer: 80.48% accuracy for TIG-to-laser and 81.13% for laser-to-TIG, an improvement of 43.39 and 43.40 percentage points over the baseline.
Feature analysis: UMAP visualization confirms the model learns domain-invariant features while preserving discriminative class boundaries.

Significance

This method substantially reduces the relabeling cost required when deploying to new welding processes and enhances the generalizability of intelligent weld penetration monitoring across different welding systems.

--- *Collected automatically on 2026-06-26.*