AI-enabled visual inspection tools are helping automotive Original Equipment Manufacturers (OEMs) to enhance quality and cut down on the cost of quality in each phase of the vehicle’s lifecycle, from product design and engineering to production, supply chain, and after sales and service.
AI-enabled Visual Inspection brings game-changing capabilities to transform the quality inspection process with fast ROI. Tech giants like IBM have already deployed such solutions at multiple OEMs which is resulting in immense benefits like:
- Immediate benefits – augments human senses and improves accuracy and consistency of quality inspection and error-proofing processes. Designed to detect vehicle defects, and miss-builds and ultimately reduce scrap, rework, warranty claims and recalls.
- Easy to deploy by using everyday (consumer) technology – leverages iOS device and its integrated camera. Enables visual inspections when devices are mounted in fixed locations (working stations) or are hand-held.
- Intuitive to use – The whole solution is designed for service technicians and shop floor professionals – to make image capturing, modelling and inferencing simple and straightforward. Contains various dashboards.
- Made for OEM service centres and Auto garages –
Successfully evaluated and deployed in numerous automotive shop floor scenarios. Supports various automotive triggers such as timestamps, VINs, or broadcast codes.
Typical Solutions
AI-enabled quality inspection can help in various scenarios. For eg. the IBM Visual Inspector enables to utilise AI at inspection stations like incoming parts section, weld quality inspection and repair, Sealant application, Laser brazing anomalies, Panel damage, Missing parts (fixtures, pins), paint quality, Engine (parts, hoses, electrical), Seats (parts, damage, wrinkles), Axles (parts, alignment, damage), Electrical connections (seated, soft seat, unconnected), Labels (wrong label, damage), Missing parts (bolts, clamps, hoses, fixtures), Incorrect part installation (alignment, tightness, clips, colour), Part damage (scratches, dents, wrinkles) etc.
The AI-enabled tool integrates with multiple shop floor tools. These are a modular set of digital tools designed to empower and upskill operators, technicians, and engineers on service lines. The solution applies AI, AR, IoT, and cognitive technologies to gain operational efficiencies, incl. higher workforce productivity, better OEE, and lower service cost per vehicle. These tools are designed in cooperation with major automotive OEMs and evaluated in real environments. IBM’s IAMA toolkit is modular, built for rapid deployment and offers high reusability and scalability to the automotive shop floor.
These solutions can work with wide portfolio of edge devices.
⮚ Improves inspection, failure detection and error-proofing processes with machine learning, and visual recognition.
⮚ Enables to build inspection systems based on various edge devices including optical cameras, infra-red cameras, lasers or mobile phones.
⮚ Contains an intuitive and user-friendly modelling environment.
⮚ Deployment & modelling methodology proven in a specific environment of automotive production lines (average accuracy >98%).
Clubbing the solution with a mobile device makes visual inspection easy to deploy and simple to use.
⮚ Easy to deploy by using everyday (consumer) technology – leverages iOS device and its integrated camera.
⮚ Intuitive to use – contains a native user-friendly iOS/iPadOS mobile app that makes image capturing and inferencing simple and straightforward.
⮚ Non-intrusive – mobile App that can even run exported models on Core ML, which enables local inferencing on-device without requiring network connectivity.
⮚ Enables visual inspections when devices are mounted in fixed locations (working stations) or are hand-held.
What are some challenges in adoption of a technology like this?
- Establish repeatable and smooth processes for rollout across garages and service centres.
- Ensuring People are a Part of the Transformation – Not a Target of it. Change Management as part of the programme to evangelise, enable, and co-create the solution.
- New Skills Required to Drive the Solution.
- Integration with Existing Systems (OT and IT) and Adaption to Plant Processes.
- Showing a deployed use case within the environment with a strong business case.
- Governance process to ensure best practices implemented for topics such as mounting techniques, model training, integration, and reporting.
What are the typical stages in adoption of a technology like this?
Each use-case will be implemented and evaluated following a structured methodology in 4 to 6 weeks timeframe.
Step 1 – Define the use cases
- Prioritise and select use-cases (business value vs feasibility) from the predefined
examples - Define metrics for business case and evaluation criteria
Step 2 – Install iOS device on station
- Download app from Apple App Store, configure and connect
- Mount iOS device in position
Step 3 – Setup image capture trigger
- Configure PLC message or App Visual Trigger
Step 4 – Capture good & bad samples
- Place the app in Auto-Capture mode for Image Capture
- Send trigger to capture images
- If possible, simulate defects in order to quickly capture both good and bad conditions
- App inserts images into Visual Insights dataset as configured in app inspection
Step 5 – Train visual model
- Using Visual Insights Studio, create labels, label objects and train models.
- Add new model to app inspection configuration
Step 6 – Operate and test
- Enable Auto Capture mode for image Inspect
- Enable image capture trigger messages
Step 7 – Evaluate and re-train
- Review model performance and repeat steps 4-6 as needed to refine model accuracy (up to 2 iterations)
Step 8 – Final evaluation of the use-case
- Success criteria: 8/10 issues correctly detected
Author
SANKALP SINHA
Business Development Executive
Automotive Aerospace & Defense
