Catch printing defects as they happen — before they become costly failures in your metal 3D printing production.
Name
Digital Twin AM Printing Process — AI-powered quality monitoring for additive manufacturing
Overview
This tool monitors metal 3D printing processes layer by layer, automatically detecting defects in powder distribution, melt pool quality, and laser performance. By combining computer vision with deep learning, it gives production engineers immediate visibility into print quality — enabling early intervention that prevents material waste and part rejection. The system keeps pace with industrial printing speeds.
Key Features
- Layer-by-Layer Quality Assessment: Evaluate every printed layer automatically, so defective builds are caught early — not after hours of wasted printing
- Multi-Sensor Defect Detection: Combine camera images and laser sensor readings for a complete picture of print quality, reducing blind spots that single-sensor systems miss
- Real-Time Monitoring: Track an active print session as it happens, giving operators time to intervene before defects propagate
- AI-Powered Anomaly Detection: Identify subtle quality deviations that human inspectors would miss, combining multiple sensor modalities for high-confidence results
- Part-Specific Quality Mapping: See quality scores mapped to individual components on the build plate, so you know exactly which parts are at risk
- Flexible Analysis Modes: Choose from statistical analysis (no training needed), self-supervised anomaly detection, or supervised classification depending on your data and needs
What Goes In / What Comes Out
Input
- Powder layer images: Camera photographs of each powder layer before melting (JPEG format)
- Melted layer images: Camera photographs of each layer after laser melting (JPEG format)
- Laser sensor data: Real-time laser power and performance readings captured during printing (text files)
- Build plan: The printer’s build configuration describing part placement and layer structure (CSV file)
- 3D geometry files (optional): STL models of the parts being printed, for spatial context
Output
- Quality reports: Per-layer and per-part quality scores in JSON, Markdown, or HTML format
- Defect heat maps: Visual maps showing where defects concentrate on the build plate
- Anomaly alerts: Real-time notifications when quality drops below acceptable thresholds
- Statistical summaries: Trend analysis across layers showing how print quality evolves over time
- Part-level assessments: Individual quality grades for each component on the build plate
How It Works
The system works like a quality inspector that never blinks. As each layer is printed, it captures powder and melt images alongside laser sensor data. First, traditional image analysis checks for obvious issues — uneven powder spread, irregular melt pools, or laser power anomalies. If something looks suspicious, a deep learning model (trained on thousands of printing layers) takes a closer look, combining all sensor data to confirm whether a real defect is present. This two-stage approach keeps analysis fast for normal layers while providing high-accuracy detection when it matters most.
Use Cases
-
Production monitoring: A manufacturing engineer starts a 12-hour metal print job containing 16 precision components. The system monitors every layer in real time, flagging a powder distribution anomaly at layer 450. The operator adjusts the recoater before the defect affects part quality, saving the entire build.
-
Post-build quality audit: After completing a batch of aerospace brackets, a quality manager runs the statistical analysis on the full dataset. The system generates a part-by-part quality report with defect heat maps, providing documented evidence for certification that specific components met quality thresholds.
-
Process optimization: A process engineer compares quality reports across multiple print sessions with different parameter settings. By analyzing where defects cluster and how laser performance correlates with melt quality, they fine-tune machine parameters to reduce scrap rates.
Partners
Montimage, CROOM
Pilot
Pilot 03 — Additive Manufacturing: This tool serves as the digital twin for the AM printing process, providing real-time quality monitoring and defect detection during metal 3D printing production runs.
License
TRL
TRL 7 — System prototype demonstrated in an operational environment. The tool includes comprehensive unit tests, CI/CD pipelines, pre-commit quality gates, and has been validated with real industrial printing datasets at production-relevant throughput.
