AI-Powered Welding Inspection
Machine learning system integrated with 3D profile scanners for automated welding quality control with micrometer-scale defect detection.
The Challenge
A major German automotive manufacturer (154k+ employees) relied on manual welding inspection — a process that was time-consuming, error-prone, and couldn't scale with production demands. They needed an automated solution capable of detecting welding defects at micrometer precision.
Our Approach
We developed an AI-powered desktop application integrated with Wenglor 3D Profile Scanners, combining point cloud processing with machine learning for comprehensive weld analysis.
3D Profile Reconstruction
Raw scanner data was processed into detailed weld profile reconstructions using custom point cloud algorithms. This created a precise digital representation of each weld bead, capturing geometry, surface topology, and cross-sectional profiles.
ML-Based Defect Detection
Machine learning models were trained on a curated dataset of welding defects — cracks, porosity, misalignment, undercut, and excess penetration. The system classifies defects in real-time with micrometer-scale accuracy.
Operator Interface
A purpose-built desktop application provides operators with intuitive visualization of scan results, defect annotations, and pass/fail decisions with full traceability.
Technical Stack
- Hardware: Wenglor 3D Profile Scanner
- AI/ML: Custom defect classification models
- Processing: Point cloud reconstruction, Computer Vision
- Languages: C++
- Platform: Desktop application
Results
- Micrometer-precision welding error detection
- Significant reduction in inspection time
- Eliminated human error in quality assessment
- Full traceability and audit trail for compliance