Capture workpieces under different positions and lighting, then augment small datasets with rotation, brightness and noise variations.
AI Welding Architecture
A welding system that sees, calculates and corrects
A closed loop connecting AI, server, simulation, robot, sensors and operator interface—reducing manual part finding, measurement, teaching and path correction.
Train and deploy YOLO/PyTorch models to locate parts, identify seams and calculate start and end points.
Distance sensors continuously measure parts, torch clearance and edge deviation without manual measurement.
Build a 1:1 model of the robot, external axes and tooling to calculate motion through inverse kinematics.
The server connects AI, PLC, laser and robot, exchanging position and correction data every 4 milliseconds.
Real-time seam and edge deviation dynamically adjusts the path for consistent welding.