Visual quality control in the foundry
Defect detection on high-temperature cast parts. Thermal and optical cameras at the line edge, MES integration and automated ISO compliance reports.
+38 %
accuracy on defect classification
ISO
QA reports generated automatically
Real-time
heatmaps for quality teams
Context
The problem
Inconsistent reject criteria
Visual inspection on high-temperature parts: operator judgment varies with fatigue, experience and lighting.
Thousands of defect types
The variability of casting defects requires a model trained on a large corpus: impossible to maintain by hand.
Manual compliance reports
Producing ISO QA reports is time-consuming and disconnected from the real-time production flow.
No immediate visual feedback
Quality teams had no aggregated view of recurring defect zones on the part.
Solution
Deployed architecture
Thermal + optical edge cameras
IR and visible coupling to capture surface defects and thermal signatures on parts that are still hot.
Multi-class defect classification
Trained on thousands of labeled examples: coverage of the foundry defect families observed in production.
MES and production line
Conforming / non-conforming decision fed back to the MES. No break in the existing PLC flow.
Heatmaps and ISO reports
Real-time visualization of defective zones. Automatic generation of QA reports for audits.
Results
Measured results
Measured in production conditions on the deployed scope.
+38 %
Classification accuracy
Gain measured vs human visual inspection on the targeted defect classes.
Auto
ISO QA reports
Systematic generation: an end to manual post-shift compilations.
Heatmaps
Real-time quality feedback
Quality teams immediately see at-risk zones on the part.
Edge
In-line decision
No cloud latency: inference as close as possible to the casting.
Lessons learned
What we took away
Get started
A similar case on your site?
Tell us about the process, the data available and your deployment constraints. We will assess feasibility within 48 hours.