AI-powered Defects Detection System for Alloy Wheel Manufacturer
ThirdEye Data successfully delivered an AI-powered defect detection system for a leading alloy wheel manufacturer. The intelligent system uses computer vision and machine learning models to identify surface, dimensional, and machining defects in real-time during production. The deployment led to significantly reduced defect rates, enhanced product quality, and streamlined quality assurance operations across all production lines.
Transforming Alloy Wheel Quality with AI-Driven Defect Detection
BUSINESS GOALS OR CHALLENGES
Business Goals
- Automate the defect detection process in alloy wheels to eliminate human errors.
- Meet the strict quality control benchmarks set by top global OEMs.
- Improve inspection efficiency, reduce cycle time, and maintain consistency in quality across batches.
- Scale the solution across all production lines while minimizing cost per inspection.
Understanding the Challenges:
- Manual inspections were inconsistent, error-prone, and resource-intensive.
- Meeting the precision and cosmetic standards of OEMs was difficult with human-based QA processes.
- Frequent production line stoppages due to false positives and missed defect identifications.
- Lack of real-time insights and traceability across inspection data.
Prerequisites and Preconditions:
A diverse and well-labeled defect image dataset, high-resolution industrial cameras, consistent lighting infrastructure, and scalable compute hardware were essential for training and deploying the AI models.
- Availability of high-resolution image datasets capturing a variety of defects in alloy wheels.
- Pre-established defect classification criteria from OEMs.
- Reliable lighting conditions (1000 ± 100 Lux) and optimal camera placement for image acquisition.
- High-performance computing infrastructure, both on-premise and edge-enabled.
- Collaboration with Customer’s quality engineering team for annotated data and continuous feedback.
THE SOLUTION
ThirdEye implemented a full-scale AI-based defect detection solution in a phased manner—starting from a PoC adhering to GM standards, followed by pilot deployment on a production line, and eventually scaling it across all lines. The solution utilizes deep learning models trained to detect a variety of structural and cosmetic defects with over 95% accuracy, providing real-time insights and automatic rejection mechanisms integrated with the manufacturing line.
High-level Solution Approach:
Image Acquisition System:
Installed high-res industrial cameras at multiple inspection points with controlled lighting for consistent image quality.AI-Powered Inspection Models:
Used CNNs and semantic segmentation models trained on a large dataset of annotated defects.
Applied transfer learning to speed up development and boost detection accuracy.
Real-time Inference & Decision System:
Deployed on edge devices for low-latency decision-making.
Integrated human-in-the-loop for complex case validation.
Categorized defects as critical, major, or minor, triggering automated rejections.
Continuous Learning:
Models retrained periodically using newly collected defect data to improve performance over time.
Technology Stack:
Computer Vision: OpenCV, YOLOv8, TensorFlow/Keras-based custom CNNs
Machine Learning: PyTorch, Transfer Learning, Semantic Segmentation Models
Hardware:
4K industrial-grade cameras
GPU-enabled edge devices (NVIDIA Jetson)
Lighting system: 1000 ± 100 Lux
Annotation Tools: CVAT, Labelbox
Deployment Infrastructure: Edge devices integrated with on-premise MES
Security: Encrypted data transfer, access control, and audit trails
VALUE CREATED
The system has been deployed 6 months back, we are sill in the phase of estimating the final ROI, Here are a few ROI calculations received from the customer based on the last 6 months performance:
- 95%+ defect detection accuracy achieved in real-time, reducing false negatives by 70%.
- 90% reduction in manual inspection efforts across all production lines.
- 30% improvement in production throughput due to faster inspection cycle times.
- 25% reduction in customer returns and rework associated with defective alloy wheels.
- Final ROI will be estimated in the next 9–12 months from full deployment based on savings in labor costs, reduction in scrap, and improved OEM compliance.
- Enhanced traceability and compliance with OEM audit requirements via automated defect logs.
Transforming Enterprises with
Data & AI Services & Solutions.
ThirdEye delivers Data and AI services & solutions for enterprises worldwide by
leveraging state-of-the-art Data & AI technologies.
