Molded Part Integrity
Comprehensive Quality Control for Injection Molded Components
Our advanced AI vision systems ensure the highest quality standards for injection molded parts - detecting critical defects that could compromise functionality, appearance, and structural integrity before parts reach assembly.
Key Applications
Formation & Structural Defects (Critical Failures)
This category covers defects that often result in a scrapped part because they compromise its fundamental structure and function.
- Short Shots:When the molten material doesn't completely fill the mold cavity, resulting in an incomplete part. Your AI can instantly detect this by comparing the part's shape against a "golden template," immediately flagging any missing features.
- Flash:Excess material that seeps out of the mold, creating thin, unwanted fins on the part's edge. An AI system is excellent at spotting the fine, irregular outline of flash against the part's expected clean boundary.
- Weld/Knit Lines:The AI can be trained to identify the subtle lines or grooves that form where two fronts of molten material meet - determining if these lines are within acceptable cosmetic limits or represent a weak point that could lead to structural failure.
- Voids & Air Bubbles:For translucent or transparent parts, cameras can detect internal air bubbles that weaken the component. For opaque parts, they can identify surface-level voids and pits.
Surface & Cosmetic Defects (Quality & Finish)
These defects affect the part's appearance and can often be indicators of a problem in the molding process, such as temperature or moisture issues.
- Sink Marks:Small depressions or craters that appear on the surface when a section cools or shrinks improperly. Using specific lighting (like a dome light), cameras can highlight these subtle topographical changes that are difficult to see with the naked eye.
- Burn Marks & Discoloration:The system can easily detect discoloration on the part surface, which can indicate material degradation or problems with the mold's heating.
- Splay & Flow Marks:Your AI can identify the silvery streaks or wavy patterns caused by moisture in the material or improper flow into the mold, ensuring a perfect surface finish.
- Contamination:The camera can instantly spot foreign particles, such as black specks in a light-colored part, that were introduced by contaminated raw material or environmental factors.
Dimensional & Feature Verification (Fit & Function)
This ensures that the critical features of the part are present and correctly formed so it can be assembled properly.
- Broken or Deformed Pins/Clips:The system can verify the presence and integrity of every single clip, pin, or mounting boss - ensuring they aren't short, bent, or broken before assembly.
- Gate Vestige Inspection:The camera can check the spot where the plastic was injected (the gate) to ensure it has been trimmed correctly and doesn't have an excessively large or sharp remnant that could interfere with assembly.
- Warpage & Dimensional Accuracy:By analyzing the part's silhouette and key features, the system can detect warping or distortion, ensuring the part is flat and meets critical dimensional tolerances.
Why Manufacturers Choose Overview AI
Purpose-built AI for the unique challenges of injection mold quality control.
Golden Template Matching
Compare every part against a perfect reference, instantly flagging short shots, flash, and missing features.
Specialized Lighting Support
Works with dome lights and angled illumination to reveal sink marks and surface defects invisible to the naked eye.
All Defect Categories
Structural, cosmetic, and dimensional defects detected in a single inspection pass - no multiple systems needed.
From Install to Production in Hours
No months-long integrations. No dedicated vision specialists. Just fast, accurate inspection.
Install & Configure
Mount the camera, connect to your network via any browser - no software installs, no license keys. Full system operational in under 2 hours.
Train with Real Images
Capture as few as 5 production images, label defects in-browser, and train a production-grade model. Built-in augmentation generates 10x more training data automatically.
Deploy & Retrain
Push the model to the camera - no export steps, no IT tickets. Retrain on new edge cases from the production floor in under 1 hour, zero downtime.
Built for the Factory Floor
Enterprise-grade hardware and software designed for 24/7 industrial operation.
NVIDIA-powered on-device processing. No cloud dependency.
Store weeks of production data directly on each camera.
EtherNet/IP, Profinet, Modbus TCP - no middleware needed.
ISO 9001, ISO 27001 certified.
Node-RED on Every Camera
Visual programming for PLC triggers, reject mechanisms, data logging, and MES/ERP integration via REST APIs.
Custom Dashboards
Build operator, engineer, and management views with standard HTML/CSS/JS. Access from any browser on the network.
Haystack Discovery
Surface unknown defects and outliers automatically. Retrain to catch them - no vendor involvement required.
FAQ
Frequently asked questions
Which injection molding defects can vision actually catch?
The ones with a visible signature, which covers most of the common list: short shots and incomplete fill, flash at the parting line, sink marks and voids, burn and discoloration, splay, knit and weld lines, and warpage where it changes the silhouette. Internal voids with no surface consequence are the honest exception, since a camera sees the surface. In practice that is not the limitation it sounds like, because most molding defects that matter to a customer are visible ones.
Textured and glossy plastic are hard to inspect. How is that handled?
Both fight you, for opposite reasons. A glossy surface reflects its surroundings, so a highlight can look like a mark and a real defect can vanish into one. Texture adds its own busy pattern that a fixed threshold reads as noise. Controlled lighting deals with the gloss, usually diffuse or at an angle chosen so the specular return misses the lens. Texture is where a trained model earns its place, because it learns what the normal grain looks like instead of trying to describe it in a rule.
How much does the tool cavity matter?
A lot, and it is worth designing for. Multi-cavity tools produce parts that differ slightly and consistently by cavity, so a model trained on one cavity can reject another for being normal. Either train across all cavities so that variation is inside the accepted range, or track the cavity so results can be attributed. The payoff for tracking it is real: a defect trending on one cavity tells you which insert needs attention, which is more useful than knowing the part failed.
Ready to Automate Injection Mold Inspection?
See how Overview AI vision deployments drive real results in days, not months.