AiV Successfully Deploys Deep Learning Inspection Solution for Wheel Nut Manufacturing at Poongkang

AiV Successfully Deploys Deep Learning Inspection Solution for Wheel Nut Manufacturing at Poongkang

AiV has successfully completed the deployment of its deep learning inspection solution for Poongkang, providing an AI-powered automated inspection system for automotive wheel nuts.

The solution supplied by AiV is designed to inspect wheel nuts for manufacturing defects using advanced deep learning technology. The system consists of proprietary AI inspection software, a customized optical system, and dedicated mechanical equipment.

The AI inspection software utilizes AiV's proprietary deep learning neural network developed specifically for manufacturing quality inspection. The model was trained using image datasets containing irregular surface defects found on wheel nuts. In addition, AiV collaborated with PKM, Poongkang's subsidiary, on the mechanical system design and integrated its AI technology with advanced optical engineering to develop and deliver a fully automated inspection solution.

Previously, Poongkang relied on conventional rule-based nut sorting equipment that could inspect standardized defects such as dimensional measurements and thread presence. However, plating-related defects—including yellow or colorless discoloration and residual plating solution—still required manual visual inspection. AiV's deep learning inspection solution overcomes these limitations by accurately detecting virtually all defect types that may occur on wheel nuts, including dents, compression marks, scratches, eccentricity, liquid residue, discoloration, and plating defects. As a result, the system delivers significantly higher inspection accuracy than both conventional sorting machines and manual inspectors, enabling a much stricter level of quality control.

The deployed solution achieved a 100% defect detection rate while maintaining a false positive rate below 0.5%, significantly improving both precision and consistency compared to manual inspection. Furthermore, by collecting highly accurate inspection data categorized by defect type, the system provides valuable insights for process optimization, resulting in improved product quality while reducing manufacturing costs.

To meet production requirements, AiV implemented a lightweight, high-speed deep learning inference algorithm capable of inspecting 120 parts per minute—equivalent to one product every 0.5 seconds. The company also incorporated proprietary mechanical and control software technologies to resolve production issues such as product jamming caused by residual plating solution.

Building upon this successful production deployment, AiV plans to expand its inspection solution business by focusing on manufacturers in the materials and components industry that require highly optimized, precision inspection systems. The company also intends to expand its AI inspection software business by offering its software platform as an MLOps-based service to major manufacturers and specialized equipment suppliers, where demand for industrial AI solutions continues to grow.

Yoon Hong-jun, Head of Quality at Poongkang, commented:

"The introduction of AiV's deep learning inspection solution has enabled us to reduce labor costs by minimizing the need for manual inspectors responsible for plating inspection. More importantly, defects that previously escaped detection due to variations in inspector condition have been virtually eliminated. Despite rising plating costs affecting profitability, the complete accuracy of the automated inspection process has significantly improved our manufacturing efficiency. We expect enhanced inspection reliability and higher customer satisfaction to strengthen our competitive advantage in the market. Given our confidence in the performance of AiV's deep learning technology, we are actively pursuing deployment of the solution across additional product lines."

Sung Min-soo, CEO of AiV, added:

"One of the key strengths of AiV's deep learning inspection solution is our ability to provide fully customized software and hardware systems optimized for each customer's smart factory environment. We will continue expanding into broader manufacturing sectors—including automotive components, display equipment, and other industrial applications—while accelerating our global expansion to become a leading deep learning machine vision company."