LSK iRobot Brings Vision AI to Vietnam Production Line, Inspecting 1.3 Million Trays Monthly

LSK iRobot subsidiary One Touch AI CEO Choi Jongwon, a professor at Chung-Ang University, demonstrates dome-shaped optical equipment. Photo provided by LSK iRobot.

LSK iRobot is moving to target the global manufacturing market by deploying a quality inspection automation system powered by vision artificial intelligence (AI) technology at a production site in Vietnam.

LSK iRobot said on October 27 that it had completed installation of a vision AI-based quality inspection system developed by its subsidiary OneTouch AI on the electrostatic discharge (ESD) carrier tray production line at Nano Chemtech Vietnam, its local production subsidiary. The system is scheduled to be introduced in stages starting in early to mid-November, for production of around 1.3 million trays per month.

This is expected to be the first time a vision AI-based inspection system has been applied to a large-scale production line in the tray industry. The system combines a vision AI algorithm developed by OneTouch AI with dome-shaped optical equipment. It can automatically detect scratches, cracks, foreign matter and changes in the shape of tray surfaces.

The core technology is Reflectance Transformation Imaging (RTI), which has been researched by the Visual Intelligence Laboratory (VILab) led by Professor Choi Jongwon of Chung-Ang University. OneTouch AI has adapted RTI for industrial inspection, comparing 3D reference data from normal products with actual observations to detect minute defects.

RTI analyzes images captured under lighting from multiple directions to identify an object's surface reflectance and fine details. OneTouch AI uses the technology to analyze three-dimensional features in 2.5D images, such as surface slopes and curves. It determines whether a defect is present by identifying areas that differ from the normal shape.

In particular, unlike conventional methods, the system reduces the burden of collecting large volumes of defective-product data in advance to train AI. Because it identifies anomalies by comparing products with 3D reference data from normal products, it can also flag previously unseen types of defects as potential anomalies.

Inspection speed and accuracy have also improved. According to the company, the takt time, or inspection time per part, has fallen from 10 seconds with the conventional manual inspection method to 0.5 seconds. Defect detection accuracy has also increased to 98%, compared with 70% to 95% for the previous method, depending on operator proficiency.

LSK iRobot expects the system to reduce the number of quality inspection workers by up to around 80% while significantly improving the consistency of quality inspections.

Data accumulated during inspections can also be used to improve production processes. Analyzing defect locations and recurring patterns in relation to production lots and equipment conditions can help identify signs of process abnormalities, such as recurring changes in shape in specific areas, and narrow down possible causes.

LSK iRobot plans to expand its vision AI-based quality inspection business, starting with the Nano Chemtech Vietnam production line. Drawing on data accumulated during local deployment, the company aims to broaden its business to include supplying vision AI optical equipment and software solutions, creating opportunities to supply global customers.

A representative of LSK iRobot said, "By working with our subsidiary OneTouch AI, which has AI technology, we have pursued both the automation of quality inspections and improvements in production efficiency at manufacturing sites." The representative added, "With the system's deployment in November as a starting point, we will demonstrate our quality competitiveness and seek opportunities to expand supplies to global customers."

This content was produced with the assistance of AI translation services.