press release

Fujikura Ltd. (Director, President and CEO Naoki Okada) conducted a proof of concept ^*1 of implementing deep reinforcement learning, a type of artificial intelligence (hereinafter referred to as AI), to control part of an actual manufacturing device, and confirmed its effectiveness. The results will be presented at NVIDIA GTC 2023 ^*2 on March 22nd.

【point】

・Accumulate AI technology and know-how independently and consider using AI in control devices in manufacturing process
・This technology has been shown to be faster and more accurate than humans or conventional programs.
・Presenting at a leading international conference on AI, which is expected to be applied to various manufacturing sites

＜ Summary of the Announcement ＞

We started researching AI in 2015, and in 2018, we application made a discriminative AI capable of recognizing images for visual inspection and introduced it into our group companies 's manufacturing process. This has enabled the Company to reduce inspection man-hours and improve production efficiency.
Furthermore, we have been considering the introduction of AI into automatic control of equipment as a candidate for AI utilization.

We conducted a conceptual verification in the form of implementing an AI program in the lens focusing device used in the manufacturing process, and confirmed that AI was able to operate faster and more accurately than human operators or conventional control programs.


Verification Results

There are few similar cases of AI-based equipment control, and it is expected that this technology will also be applied to other manufacturing processes. Therefore, the significance of confirming the effectiveness of application will be announced at NVIDIA GTC 2023.

In recent years, there have been concerns about a future decline in technical capabilities due to the aging of skilled engineers. Preventing this is an urgent issue for manufacturing industry, and we will continue to promote technological development aimed at solving social issues.
Please take a look at the results of this development.

"NVIDIA GTC 2023"
Announcement date and time: Wednesday, March 22, 2023, 10:00-11:00 a.m. (Japan time)
Speaker: Tatsuya Akiyama, Head of AI Development Group, Production System Development Department
URL: https://www.nvidia.com/ja-jp/

*1 Proof of concept
This refers to verification that clarifies, prior to development, whether a proposed new concept or technology is truly effective and feasible in achieving a goal or resolving a problem.

*2 NVIDIA GTC
This is a global AI conference for developers hosted by NVIDIA, a major US semiconductor manufacturer. The deep reinforcement learning training was carried out using the NVIDIA DGX system, and the inference was carried out using NVIDIA RTX, enabling rapid development and deployment.