Blog
The Procunit Blog
Technical writing on machine vision inspection, defect escape rate, false positive analysis, PLC integration, and what it actually takes to train a model on a small defect dataset. Written by quality engineers and ML engineers who work on production lines.
All articles
Why Manual Visual Inspection Fails Above 100 Parts Per Minute
The physiological limits of human visual inspection — and what the data says about escape rates on high-speed lines.
Training Defect Detection Models with 50 Samples
Why few-shot defect detection works in manufacturing — and how to curate a 50-image set that generalizes.
Surface Defect Types in Stamped Metal Components
Edge cracks, burrs, scoring, and dimensional drift — a field reference for quality engineers on stamping lines.
Integrating AI Vision with PLC Reject Systems
Ethernet/IP, Modbus TCP, and discrete I/O — comparing PLC interface methods for AI vision reject signals.
The Real Cost of False Positives in Automated Optical Inspection
How every false reject adds cost — rework, operator time, and eroded confidence in your inspection system.
Dimensional Drift Detection vs Go/No-Go Gauges
When inline AI vision outperforms contact gauges — and when it doesn't. A comparison for machined parts.
Assembly Error Detection: Connector Orientation on PCBs
Connector polarity errors are common and expensive. Here's how AI vision catches them before board test.
Lighting Setup for Machine Vision Inspection Lines
Ring lights, dome lights, dark-field — choosing the right illumination for your defect type and part geometry.
Measuring Defect Escape Rate in Production
How to calculate escape rate, what sampling strategy to use, and how to benchmark before an AI pilot.
Procunit Raises Angel Funding and Shares Product Roadmap
We closed our angel round to accelerate field deployments and build multi-camera support. Here's what's coming.
Reducing Scrap Rate with Inline AI Inspection
Scrap rate is a lagging indicator. Inline inspection turns it into a leading one — here's the math.
Camera Placement Geometry for Cylinder Bore Inspection
Field angle, working distance, and coaxial illumination — the geometry of inspecting bores without a vision PhD.
ROI Analysis: AI Visual Inspection in Discrete Manufacturing
A worked example with real cost categories — scrap, rework, warranty, and inspection labor — for a stamping line.
FPA Metrics Quality Engineering Teams Should Track
False positive analysis isn't just about the reject rate — it's about understanding model confidence over time.
Connector Pin Absence Detection on High-Volume Harness Lines
Missing terminal pins are hard to catch visually at speed. This is how AI handles it at 200+ connectors per minute.