A cross-process welding penetration status prediction algorithm based on unsupervised domain adaptation in laser and TIG welding

2026-06-24

AI Summary

This study addresses the significant performance degradation of weld penetration state classification models when transferred across different welding processes—such as TIG and laser welding—due to domain shift. To tackle this challenge, the work introduces unsupervised domain adaptation (UDA) into cross-process welding monitoring for the first time, proposing a novel UDA framework integrated with a Gradual Source Domain Expansion (GSDE) strategy. By progressively expanding the source domain data and optimizing feature alignment, the method effectively bridges the distributional gap between processes, enabling high-accuracy penetration state prediction without requiring labeled target-domain data. Experimental results demonstrate that the approach achieves average accuracies of 90.65% (TIG) and 90.72% (laser) in within-process settings, and substantially improves cross-process transfer accuracy to 80.48% and 81.13%, surpassing supervised baselines by over 43% and markedly reducing annotation costs for new welding processes.

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, we propose an unsupervised domain adaptation (UDA) framework integrated with a gradual source domain expansion (GSDE) strategy. Evaluated on dedicated TIG and laser welding datasets, our approach achieves high accuracy in both same-process and cross-process transfer tasks. Specifically, it attains average accuracies of 90.65% on TIGFH and 90.72% on LSPS in same-process settings, surpassing a supervised baseline by 35.83% and 38.87%, respectively. More notably, in cross-process scenarios, it reaches 80.48% for TIG to Laser and 81.13% for Laser to TIG, improving upon the baseline by 43.39% and 43.40%. UMAP visualizations verify that the model learns domain-invariant features while maintaining discriminative class boundaries. This method considerably lowers the relabeling cost for new welding processes and enhances the versatility of intelligent monitoring across different welding systems.

Problem

• weld penetration status prediction
• domain shift
• cross-process transfer
• laser welding
• TIG welding

Innovation

• unsupervised domain adaptation
• cross-process welding
• weld penetration prediction
• gradual source domain expansion
• domain-invariant features