Injection Molding Inspection AI

Add a real-time inspection layer to every shot on the line with Vision AI for injection molding inspection. Whether you're molding medical housings and diagnostic consumables, automotive connectors, consumer goods, or glass-filled structural parts across one plant or five, Roboflow extends your QC coverage to every unit on the line, on the cameras and inspection stations your facility already runs, validatable under ISO 9001, IATF 16949, ISO 13485, and FDA 21 CFR Part 820.

Part defects on every shot:

• Detect short shots, flash, sink marks, weld lines, burn marks, splay, and voids on the part as it leaves the press or lands on the outfeed conveyor, including subtle cosmetic defects on textured and glossy surfaces.
• Flag contamination, black specks, and color shift against a golden reference so a bad regrind batch or a dirty hopper is caught in minutes, not at the end of the run.
• Pair detection with a dimensional check to catch warpage and out-of-spec features on the same camera.

Mold and cavity condition between cycles:

• Inspect the open mold each cycle for stuck parts, broken cores, residue buildup, and worn or damaged cavity surfaces before the platen closes.
• Confirm inserts and overmold components are seated in every cavity before injection so a missing or misloaded insert never becomes a scrapped shot or a damaged tool.
• Trend cavity-by-cavity defect rates so tooling engineers can schedule maintenance on the cavity that is drifting instead of pulling the whole tool.

Assembly guidance and final verification at the press:

• Guide operators through press-side assembly steps (clips, seals, labels, secondary components) with live confirmation that each step is complete before the part moves on.
• Verify finished assemblies for presence, orientation, and correct variant, and log the result to the part or lot record.
• Deploy the same models across multiple plants from one platform, with local edge inference so inspection keeps running if the network does not.

Bring intelligence to every shot today.

‍

Frequently asked questions

What is injection molding inspection with Vision AI?

Injection molding inspection with Vision AI uses cameras at the press, on the robot, or on the outfeed conveyor and a computer vision model trained on your parts to find defects such as short shots, flash, sink marks, weld lines, burn marks, warpage, and contamination on every shot. The same approach inspects the open mold for stuck parts, residue, and worn cavities, confirms inserts are loaded, and guides assembly steps that happen at the press. Every result is logged with its image so the inspection can be documented under ISO 9001, IATF 16949, and, for medical molding, ISO 13485 and FDA 21 CFR Part 820.

Can Vision AI detect sink marks and short shots on textured or glossy parts?

Sink marks and marginal short shots on textured, glossy, or dark parts are among the hardest inspection tasks in molding, because the defect is a shallow change in surface geometry rather than a clear edge, and part-to-part variation in gloss and color makes fixed thresholds unreliable. A deep-learning model trained on your parts learns what a good surface looks like across that variation, and consistent lighting (often low-angle or diffuse dome lighting) makes the shallow depression visible to the camera.

Does injection molding inspection support ISO 13485 and FDA 21 CFR Part 820?

Yes. ISO 13485 (the quality management standard for medical device manufacturing) and FDA 21 CFR Part 820 (the Quality System Regulation for medical devices, including production and process controls and the design history file) both expect validated inspection processes with documented acceptance criteria and traceable records. FDA 21 CFR Part 11 adds requirements for electronic records and signatures. Roboflow serves as the inspection engine: it runs validated models, records every image and decision with a timestamp and model version, and supports IQ/OQ/PQ documentation. Your quality, regulatory, and validation teams keep ownership of the acceptance criteria and the validation protocol.

Can it integrate with our press controllers, MES, and quality system?

Yes. Inspection results are available over REST, MQTT, and OPC UA and can be written directly to a database, so pass/fail and defect data flow into MES and quality platforms such as SAP, Plex, Ignition, AVEVA, MasterControl, and Veeva Vault QMS. For press-level control, results can drive a PLC output or a signal to the machine controller (Engel, Arburg, Husky, and others) to hold the cycle, reject the shot, or stop the robot when a stuck part or missing insert is detected. For regulated molding, every decision carries an FDA 21 CFR Part 11 audit trail.

‍