Related Experiment Video
Updated: Dec 26, 2025
07:58
Automatic Laser-based Geometry Capture for Finite Element Analysis of Weld Beads
Published on: July 25, 2025
663
Real-Time Weld Quality Prediction Using a Laser Vision Sensor in a Lap Fillet Joint during Gas Metal Arc Welding.
Kidong Lee1, Insung Hwang1, Young-Min Kim1
1Joining R & D Group, Korea Institute of Industrial Technology, 156 Gaetbeol-ro, Yeonsu-Gu, Incheon 21999, Korea.
Sensors (Basel, Switzerland)
|March 19, 2020
Summary
This study introduces a laser vision sensor and deep neural network models for real-time nondestructive testing in gas metal arc welding. The system predicts internal weld quality and tensile strength from external bead shapes, enhancing welding robustness.
Area of Science:
- Materials Science and Engineering
- Robotics and Automation
- Quality Control and Nondestructive Testing
Background:
- Real-time quality monitoring is crucial for robust gas metal arc (GMA) welding.
- Nondestructive testing (NDT) methods are essential for ensuring weld integrity and process control.
- Complex robotic welding motions pose challenges for accurate sensor data acquisition.
Purpose of the Study:
- To develop and implement a laser vision sensor (LVS) system for real-time weld monitoring.
- To create robust image processing algorithms for precise laser line extraction on welds.
- To establish deep neural network (DNN) models for predicting internal weld quality and tensile strength.
Main Methods:
- Design and fabrication of a specialized laser vision sensor (LVS).
- Development of an image processing algorithm for laser line extraction.
- Implementation of a gyro sensor-based camera calibration for robotic motion compensation.
- GMA welding experiments under varied conditions to collect data.
- Development of deep neural network (DNN) models using external bead features and welding parameters.
Main Results:
- Successful extraction of precise laser lines from welds using the developed algorithm.
- Accurate camera calibration achieved despite complex welding robot movements.
- DNN models demonstrated capability in predicting internal bead shapes.
- DNN models effectively predicted the tensile strengths of welded joints based on input data.
Conclusions:
- The integrated LVS and DNN approach enables effective real-time NDT for GMA welding.
- The developed system enhances weld quality prediction and ensures joint robustness.
- This technology offers a pathway to improved automation and quality assurance in welding processes.
