What Defects Can a Vision Inspection Machine Detect in Fasteners?

Byinfo@visionsortpro.com

Ask a supplier “what can your machine detect” and you’ll usually get a confident “everything.” The useful answer is more specific. Vision inspection for fasteners works by photographing every part from multiple angles and comparing what the cameras see against models trained on your parts — so the real question is which defect types the cameras and algorithms can reliably catch.

Based on our inspection cases across screws, bolts, nuts, pins and stamped parts, here are the ten defect families a vision inspection machine handles — with the accuracy you can expect.

1. Dimensional Defects

Length, height, head size, bore and outer diameter. These are the quiet defects: a bolt 0.2 mm too long still threads in — until it doesn’t, at your customer’s assembly station. Our machines measure to ±0.005 mm with telecentric optics and high-resolution cameras, checking dimensions on every part rather than on a sample. Typical cases: screws checked for height and head size, precision bolts for length, concentricity parts for bore and outer diameter.

2. Thread Defects

Thread presence or absence, damage, wrong pitch. Threads hide from top-down views — a nut can look perfect while its threads are stripped or missing entirely. That’s why thread checking uses side imaging in the same cycle: our tube-nut inspection case catches thread presence and absence automatically, and special screws are checked for threading and end-face defects. Thread problems are among the most common screw defects and among the easiest for a human to miss at line speed.

3. Surface Defects

Plating defects, scratches, dents. Surface flaws rarely affect function on paper, but they’re what your customer’s customer sees first — and in export business, appearance complaints cost shipments. Detection here depends on lighting as much as cameras: programmable LED ring lights, backlights and coaxial illumination reveal plating irregularities, uniformity and coverage that a single overhead light would wash out. Our stamping cases include side-wall plating defects and surface scratches and dents.

4. Head Defects

Head geometry — flanged, countersunk, hex, malformed. A head formed out of spec may still drive in, but it won’t sit flush or hold torque, and it will jam automated drivers. Head checks compare geometry against the trained model: our precision bolt case inspects head geometry alongside length and threading. Because heads are captured from multiple angles in one pass, a cracked or deformed head shows up from the side view even when the top looks fine.

5. Burrs

Burrs come from worn dies and forming wear, and they grow gradually — which is exactly why they slip through sampling. A batch checked Monday passes; by Thursday the die has worn and burrs appear. 100% inspection catches the drift the day it starts. Our cases cover burrs on stamped profiles, rubber buckles and outer diameters of turned parts.

6. Cracks

Cracks and material fatigue are the defects with the highest consequence — a cracked bolt fails under load, and there’s no second chance. Detection uses high-resolution imaging from multiple angles; our stamping inspection case explicitly covers strip breakage, cracks and material fatigue. No vision system can promise to catch every sub-surface crack — but at 99.99% sorting accuracy, the machine catches what human eyes miss at speed, every part, every shift.

7. Missing Features

Missing material, incomplete forms, absent features. A part can pass dimensional checks and still be wrong — a cold-formed feature that didn’t fill, a pin that sheared off. Our cases include missing material on formed shells and stamping profiles. Deep-learning models trained on good parts flag anything that deviates from the learned geometry, including features that are hard to write a rule for but obvious to see.

8. Wrong Parts

Orientation errors and marking defects — a logo mix-up, a part fed the wrong way. Wrong orientation in an automated assembly line stops the picker or the driver; wrong marking on a branded fastener is a customer complaint waiting to happen. Our logo-tag case catches logo mix-ups and surface dents in the same pass, and complex fasteners are checked for orientation alongside threading and geometry.

9. Mixed Parts

Mixed parts — two SKUs in one bin — are usually a logistics failure, not a production one, which is why they surface at your customer’s receiving dock instead of your own. Because each product runs on a stored recipe, the machine knows exactly what the current part should look like; anything outside that recipe routes to the NG bin. If your worry is a single machine handling several part numbers, that’s what recipe changeover on the touchscreen HMI is for.

10. How Vision Inspection Detects These Defects

The same four steps cover all ten defect families:

  1. Feed and separate — a vibrating feeder and rotating disc present parts one at a time.
  1. Multi-angle capture — up to 24 industrial cameras photograph top, bottom and side surfaces in one cycle, with programmable lighting matched to the defect type.
  1. AI judgment — deep-learning models trained on your parts classify OK/NG in milliseconds, measuring to ±0.005 mm.
  1. Sorting and records — ejectors separate parts into bins; every result and image is stored for SPC export and traceability.

The models are self-training: as your production runs, the system keeps learning your parts — so detection of fastener defects improves with use instead of degrading.

One honest note on limits: what a machine can detect depends on how the defect presents to the camera. That’s why every project starts with samples — send us your parts and your inspection requirements, and our engineers evaluate them free of charge and tell you exactly which defects the machine will catch on your parts, with a recommended configuration. No cost, no commitment.

FAQ

What defects can a vision inspection machine detect on fasteners?

Ten main families: dimensional errors, thread defects, surface and plating flaws, head geometry defects, burrs, cracks, missing features, orientation errors, marking defects and mixed parts — trained on your specific parts.

Can it detect defects inside threads?

Yes. Thread presence, absence and damage are checked via side imaging in the same inspection cycle — no second handling required.

How small a defect can it detect?

Machines measure to ±0.005 mm with telecentric optics and high-resolution cameras; detection limits for visual defects depend on part size and surface finish, which is what sample evaluation confirms.

Will the machine learn new defect types?

The AI models are self-training and improve on your production data. New defect types are added by training on samples of the new defect — handled by our engineers remotely.

Can I send samples to verify detection on my parts?

Yes — send samples with your inspection requirements. Our engineers evaluate them free of charge and return a recommended machine configuration showing which defects will be caught. No cost, no commitment.

See What It Catches on Your Parts

Reading about defect families is one thing; seeing your own parts sorted is another. Send samples and your inspection requirements — we’ll run the evaluation and return the recommended configuration with the defects it will catch, in writing.

Related reading: Common Stamping Strip Defects and How to Detect Them

WhatsApp: +86 173 2844 2829 | Email: info@visionsortpro.com | https://visionsortpro.com

Similar Posts