TRIART Co., Ltd. (located in Iizuka City, Fukuoka Prefecture CEO: Kentaro Imazu, hereinafter to be referred as TRIART) and Toyota Motor Kyushu Co., Ltd. (located in Miyawaka City, Fukuoka Prefecture, CEO: Nagata RI, hereinafter to be referred as Toyota Kyushu) started trial operation for [Defective Product Analysis and Sensing System] based on the Lexus production line of Kyushu Miyata Factory. This system uses the cooperation of workers and AI to detect defective products that only experienced workers can detect by feeling.
This is to inspect the shape, elongation and other aspects of the inland steel plate produced under pressure through the thermal energy camera installed inside the press machine. This is the main reason for product cracks, and it is also an important indicator that affects the product qualification rate. The inspection of these products has traditionally been judged by experienced workers based on their senses. It is a field with high demands on technology and experience.
TRIART uses its proprietary technology, Composite AI “4CAS”*1, which has many achievements in sensory information processing based on image data, ① to generate a master image that will be the standard shape of the panel from the image of the thermos camera. ② By calculating the differences (expansion points, contraction points, etc.) between the produced panel and the master image, we have made it possible to evaluate all the panels that are produced every 5 seconds. As a result, it is possible to determine the rule of “what kind of case will cause the defect to occur”, making it possible to achieve the accuracy of experienced workers level to detect failure signs more efficiently. Moreover, the accuracy of this evaluation is not only self-learning by AI through the template ①, but the staff selects the approximate target area in the process, and then uses AI to conduct self-learning to ensure the overall accuracy of the project.
DX (Digital Transformation) that utilizes AI tends to be imaged as “fully automated by AI”, but in the skills that workers have already acquired, or in the work where human judgment is clearly superior Utilizing human power and designing flows that complement each other with AI will speed up the resolution of many issues. This is a very flexible and cost-effective system in manufacturing plants with multiple commodity production lines. This proposal (Composite AI [4CAS]) is not to generate a black box completely controlled by AI for detection but is composed of multiple control AI. This system, which was jointly developed by Toyota Kyushu and TRIART, can be said to be a good example of how humans and AI work together on an equal footing in the physical workplace of the manufacturing industry.
Toyota Kyushu is a world-renowned Lexus production plant with countless experienced skilled workers and produces some of the world’s highest quality products*2. At the same time, the plant is also developing the latest production technologies. For example, the technology products announced this time can bring more room for advancement and self-development to experienced workers. We can also expect that there will be more and better products in the future.
*1Composite AI, which has many achievements in sensory information processing based on image data, TRIART also implement this technology in the difference comparison function of the ” Image comparison system MIIDEL”.
*2The Miyata Factory has won the world’s leading platinum award five times in the quality survey (“IQS” automobile initial quality survey) in the factory category by the US market research company J.D. Power.
[Image Detection Support System], which uses the camera of the mobile phone to check product labels, has started trial operation
The image comparison system [MIIDEL6] is now on sale
[Defective Product Analysis and Sensing System] developed with Toyota Kyushu which using human and AI cooperation to inherit the experienced workers
MIIDEL can now also be used for label inspection work
The image comparison system [MIIDEL5] is now on sale
Introduction of MIIDEL has started at Fukuoka Prefectural Office
Support for error detection with document synthesis