基于多模态融合的焊接异常评估与缺陷定位方法及系统
By adopting a multimodal fusion-based method for welding anomaly assessment and defect localization, the problems of single perception and black-box assessment algorithms in the welding process are solved. This method enables precise localization and full lifecycle traceability of welding defects, improving the industrial reliability of assessment and the accuracy of localization.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- GUANGXI UNIV
- Filing Date
- 2026-05-08
- Publication Date
- 2026-07-17
AI Technical Summary
Existing technologies in the welding process of high-performance vehicle manufacturing suffer from problems such as limited sensing methods, high false negative rates, black-box evaluation algorithms, and lack of time-space conversion capabilities, making it difficult to accurately locate and repair welding defects.
A multimodal fusion method for welding anomaly assessment and defect localization is adopted. By acquiring multimodal physical signals, performing time synchronization processing and extracting features from a two-flow thermodynamic decoupling architecture, and combining cross-attention fusion, a multimodal temporal feature tensor is constructed. A temporal state monitoring model is used to identify transient anomalies, quantify weld depth and weld width, and achieve three-dimensional spatial localization by combining kinematic homogeneous transformation and molten pool solidification hysteresis compensation.
It achieves precise location and full lifecycle traceability of welding defects, breaks through the black box limitation of perception algorithms, improves the industrial reliability of assessment and the accuracy of defect location, breaks the separation between attribute assessment and location tracking, and realizes full inspection-free rework.
Smart Images
Figure CN122155541B_ABST