AVIQuality & Compliance
Automated Visual Inspection
Every Product Inspected. Every Result Recorded. AI visual inspection at line speed: defects found and rejected in real time, with every image and result linked to the batch and shared with quality, SPC and OEE.
The challenge
Inspection by eye is slow, tiring and inconsistent, and it leaves little evidence behind. Rule-based machine vision rejects good products that vary naturally in colour, texture or position, and every new product or defect type needs a specialist to reprogram it. Defects found late become rework, scrap, complaints or recalls.
Overview
LOGICAL SYSTEMS Automated Visual Inspection (AVI) is our AI inspection system for the production line. Inspection stations of industrial cameras, lighting and an edge computer capture every product as it passes. AVI judges each image beside the line at line speed and sends pass or fail to the PLC, so defective products are rejected automatically, with or without a network connection.
AVI learns from images of good products, so inspection can start before a library of defects exists, and it recognises known defect types as soon as examples are labelled. Models are trained, tested and approved in the AVI model workbench, released under change control, and monitored on the line for missed defects and false rejects.
Every result and image is recorded against the order, batch or serial number. Defects open deviations in QMS, defect rates appear on SPC charts and as quality losses in OEE, printed labels are checked against the approved version in Label Management, and images are kept as evidence in Traceability & Genealogy. AVI is built on the Inductive Automation Ignition platform and on open standards, from GenICam cameras to OPC UA for Machine Vision, so it fits the equipment you already run.
Capabilities
What AVI does
Inspection Stations
- Area-scan and line-scan cameras to GenICam, GigE Vision and USB3 Vision
- Lighting, optics and enclosures specified for each inspection task
- Triggering from photo-eyes, encoders or the PLC, in step with the reject
- Edge computers beside the line that keep inspecting when the network is down
AI Inspection
- Anomaly detection trained on images of good products only
- Defect classification once examples are labelled
- Presence, count, position and dimensional checks
- Optical character recognition and verification of printed text
Ready-Made Inspections
- Labels: presence, position, artwork and text against the approved version
- 1D and 2D codes read and verified, with print quality graded to ISO/IEC 15415 and 15416
- Closures, seals and fill levels
- Surface and cosmetic defects: scratches, dents, stains, contamination and colour
At the Line
- Pass or fail to the PLC over digital I/O, PROFINET or EtherNet/IP
- A missing or late result rejects the product, never passes it
- Review station where operators confirm uncertain results
- Live dashboards of defects by line, shift, product and defect type
Model Lifecycle & Validation
- Model workbench: label images, train, test and compare models
- Versioned models and image sets, released with electronic signatures
- Acceptance testing with defect test sets, following VDI/VDE/VDMA 2632-3 and, for injectables, USP <1790>
- Missed defects and false rejects monitored, with retraining from operator feedback
Quality Integration
- Defects open deviations and non-conformances in QMS
- Defect rates on SPC control charts and quality losses in OEE
- Images kept as evidence in Traceability & Genealogy
- Audit trail and data integrity to 21 CFR Part 11, EU Annex 11 and ALCOA+
Benefits
Why manufacturers choose AVI
Every product inspected
Consistent inspection at line speed on every shift, instead of samples checked by eye.
Defects stopped early
Defects are caught and rejected at the line, before they become rework, scrap or complaints.
Evidence for every batch
An image and a result for each inspected product, linked to the batch record and ready for audits.
Live sooner
Inspection starts from images of good products, so a new product does not wait for a collection of defects.
Fewer false rejects
AI models accept the natural variation in colour, texture and position that rule-based vision rejects.
Validated and audit-ready
Models are tested, approved and versioned under change control, with audit trails and electronic signatures for GMP.
Ask AVI in plain language
LS Assistant, the AI assistant in every Logical Systems application, answers questions from your live AVI records and shows the sources behind every answer. It keeps learning as your records, work instructions and resolved issues grow, and your data is never used to train AI models.
Answers with sources
Every answer cites the records and work instructions it used.
Your role, your site
It sees only what each user is allowed to see.
Asks before it acts
Every change it proposes needs a person's confirmation and is audited.
18 languages, voice on mobile
Questions, answers and content in your teams' languages.
Integration
Connects to the systems you run
Systems & protocols
- Industrial cameras (GenICam)
- Smart cameras and existing vision systems
- IP cameras (ONVIF, RTSP)
- PLC digital I/O, PROFINET, EtherNet/IP
- OPC UA
- MQTT
- REST API
- QMS, SPC, OEE, LMS and TXG
Standards
- GenICam
- OPC UA for Machine Vision (OPC 40100)
- 21 CFR Part 11
- EU GMP Annex 1
- GigE Vision
- USB3 Vision
- EMVA 1288
- VDI/VDE/VDMA 2632
- ISO/IEC 15415
- ISO/IEC 15416
- GS1
- USP <1790>
- EU Annex 11
- GAMP 5
- ALCOA+
- ISA-95
Walkthrough
See AVI working with scenarios from your own plant, guided by one of our engineers.
Talk to an engineerWorks with
Related solutions
See AVI in action
Book a demo with our engineers, using scenarios from your own plant.