A robot welding arc blow angle identification method, medium and device
By combining deep learning technology with auxiliary task learning and key point detection, real-time and accurate identification of arc blow angles during robotic welding is achieved, solving the problem of arc posture recognition in robotic welding and improving welding quality and automation level.
CN118840642BActive Publication Date: 2025-10-17SOUTH CHINA UNIV OF TECH
View PDF -1 Cites 0 Cited by
Patent Information
- Application Number
- CN202410847808.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-06-27
- Publication Date
- 2025-10-17
- Estimated Expiration
- 2044-06-27
Smart Images
Figure CN118840642B_ABST
Abstract
The present invention provides a method, medium, and device for identifying arc blow angles in robot welding. The method comprises: using an arc blow angle recognition model to identify welding arc images to obtain arc blow angle recognition results; obtaining the arc blow angle recognition model by obtaining welding arc image samples; establishing an arc blow angle recognition model comprising a backbone network, an angle recognition branch, and a key point detection branch; the model uses angle recognition as a primary task and key point detection as an auxiliary task, which is only involved in the training process; processing the welding arc image samples to construct arc blow angle labels and key point labels; and training, validating, and testing the arc blow angle recognition model to obtain a final arc blow angle recognition model. Based on an auxiliary task learning strategy, this method can improve recognition accuracy while maintaining a constant inference cost, achieving real-time and accurate identification of welding arc blow angles.
Need to check novelty before this filing date? Find Prior Art