[Edaily Reporter Park Jung-Soo ] WINTEC Co.,Ltd.(320000)has increased the inspection speed of its six-sided multilayer ceramic capacitor (MLCC) visual inspection equipment—which utilizes the company’s proprietary artificial intelligence (AI) inspection platform—to 13,000 units per minute.
HanWool Semiconductor announced on the 1st that it achieved an inspection speed of 13,000 units per minute and a defect detection accuracy of approximately 98% based on 0603-sized MLCCs in its internal testing. The company is currently conducting actual performance evaluations of these results through an external specialized agency.
This performance test was conducted by capturing images of all six sides of ultra-small MLCCs moving at high speed, with AI determining the presence of defects in real time. The company explained that it increased defect detection accuracy from approximately 95% to approximately 98% while maintaining an inspection speed of 13,000 units per minute.
The 0603-spec MLCC is an ultra-small component measuring approximately 0.6 mm × 0.3 mm. During the inspection process, images must be captured from multiple angles while the product is being transported at high speed to detect micro-cracks, chipping, foreign matter, and electrode abnormalities. Minimizing false negatives and false positives while maintaining throughput is considered a core technology for inspection equipment.
This equipment incorporates “HawAIe,” an AI inspection platform currently under development by HanWool Semiconductor, and the AI analysis engine “AlohaNet.” HawAIe is a hybrid approach that combines traditional rule-based image inspection with AI deep learning analysis.
The system is structured so that defects with consistent locations and shapes are identified using conventional image processing methods, while areas that are difficult to assess—such as microcracks or irregular surface defects—are analyzed by AI. It is designed so that accumulated inspection data is used for AI training, enabling the system to enhance its classification performance based on the customer’s product specifications and defect types.
On-device AI technology is also being implemented. The system is designed to enable AI inference within the device itself without an external network connection, thereby maintaining inspection speed while enhancing the security of production data.
HanWool Semiconductor is currently conducting further development with the goal of increasing the inspection speed—currently at 13,000 units per minute—to 15,000 units per minute by the end of the year. The company also plans to gradually expand the scope of application to its own MLCC post-process inspection equipment, including composite measuring instruments, indentation testers, and ultrasonic non-destructive testing equipment.
A HanWool Semiconductor official stated, “This performance enhancement is significant because it not only increases the inspection speed to 13,000 units per minute but also improves AI defect detection performance under actual high-speed production conditions,” adding, “We will objectively verify inspection speed and accuracy through evaluations by external specialized institutions and enhance equipment reliability.”
The official continued, “With the goal of developing next-generation equipment capable of inspecting 15,000 units per minute by the end of this year, we will further enhance performance and expand the scope of ‘Hawaii’ to grow into an AI inspection platform company covering the entire MLCC back-end process.”