Report: Smart Factory Japan 2024 hosted by Nikkan Kogyo Shimbun
I visited Smart Factory Japan 2024, hosted by Nikkan Kogyo Shimbun, which was held at Tokyo Big Sight from February 20 to 22, 2024. It appeared that 86 companies participated, showcasing technologies and products related to production factories, including information management and processing systems that are indispensable for realizing smart factories on the manufacturing floor, as well as manufacturing equipment and devices. The event was also held online, with access available from February 14 to 29, 2024.
I went to the venue on the final day of the exhibition, February 22, and attended a lecture titled "'Non-stop' Edge Computing Supporting the Foundation of Manufacturing DX: Key Points from Implementation Examples" by Mr. Kazuki, Department Manager of the Business Development Department at Stratus Technologies Japan, Inc. Although I was not aware of it, Stratus Technologies is a company from the United States, and it seems they have been deploying the ztC Edge introduced this time since around 2018. This ztC Edge is said to be a zero-touch, secure, and highly automated edge computing platform designed for edge environments. The system is a rugged, compact computer that makes app deployment, management, and maintenance easy, saving time and effort. Furthermore, it is configured as a pair of two units, and through self-protection and self-monitoring functions, it significantly reduces unplanned downtime and realizes 24/7/365 high availability for business-critical applications. It was also stated that it provides a powerful platform for realizing high availability for business-critical applications by solving edge computing challenges through designs that enable remote application management, rapid and efficient multi-site deployment, and high availability for critical business operations and applications.
https://www.stratus.com/jp/solutions/platforms/ztc-edge/
In the lecture, it was suggested that if availability and reliability are ensured, it might be acceptable to have manufacturing equipment controlled by edge computers installed on the factory floor. Naturally, the approach to DX on the manufacturing floor involves digitization, labor saving, automation, and efficiency through the collection of on-site data and data visualization plus analysis/sharing, and the introduction of ztC Edge makes it possible to promote these at the edge, as well as to build more global operational configurations through connection with central control mechanisms or the cloud. However, the most important thing is the availability of the edge computer, and ztC Edge uses a redundant configuration that operates two units in parallel, constantly mirroring them so that if one fails, it can be backed up immediately. Also, it was stated that failed equipment can be replaced with a spare unit, and the backup system can be reconstructed immediately by swapping the two intranet terminal cords. In short, it was argued that there is sufficient value in introducing it because it eliminates the need for complex setting changes by IT engineers, and ultimately, it allows for loss cutting due to zero downtime and the reduction of fixed costs by eliminating the need for the aforementioned IT engineers.
As my own conclusion after listening to the lecture, I felt that it is undoubtedly a strong candidate for an edge computer to be introduced to the factory floor. Also, it is B4 size and has no cooling fan, so I judged that it is a device that can be installed in clean rooms like semiconductor manufacturing lines without any problems, which is another advantage. However, there is no doubt that it is necessary to scrutinize the computer's capabilities in terms of software capacity, the amount of data to be handled, and processing speed regarding what you want to do on-site.
The company operating the largest booth at the venue was SORACOM. SORACOM provides the "SORACOM" communication platform, which fuses the cloud with communication using dedicated SIMs to provide communication optimized for IoT/M2M in a reasonable and secure manner. Also, its business seems to be deployed in 140 countries around the world. What I heard about this time is a collaboration product with Amazon Monitron. Amazon Monitron is a machine learning-based end-to-end monitoring system that provides dedicated sensors to capture vibration and temperature data, and a gateway to automatically transfer the data to the AWS cloud. SORACOM provides the communication means to the cloud, but it seems they can be asked to build the total system. The dedicated sensors are completed at the size of an eraser, and it seems that data can be collected just by placing them on the monitoring target. By analyzing this data, ideally, it detects signs of potential equipment failure and enables notification of failure occurrence. The initial introduction cost is set at 143,000 yen, and since the monthly fee starts from 330 yen, it is considered that it can sufficiently handle the monitoring of machines with drive systems on small manufacturing lines.
https://soracom.jp/store/26949/
Next, the focus is on the introduction of local 5G. Although the communication speed is about the same as Wi-Fi, I believe that local 5G is an important technology for promoting digital transformation (DX) in manufacturing sites that require the use of large amounts of data because of its attractiveness in terms of communication range. This local 5G refers to a 5G network built independently within a company's premises or buildings, and can realize characteristics such as ultra-high speed, ultra-low latency, and massive simultaneous connections. This enables large-capacity wireless communication on the manufacturing floor, and acceleration of digitization is expected. This time, a local 5G service was proposed by NTT Business Solutions.
https://www.nttbizsol.jp/service/local5g/
It seems that the company keeps the total necessary expenses low by outsourcing the "5G core," which controls the communication essential for local 5G services, to its own data center. Also, regarding the introduction of communication networks, they support everything from preliminary procedures (obtaining radio station licenses, etc.) to construction and operation so that customers can feel at ease even without specialized knowledge, and they have prepared two pricing plans, a subscription type and a lump-sum payment type, according to requests, and it seems there is an introduction plan starting from an initial cost of 2.02 million yen and a monthly fee of 350,000 yen. Due to these features, NTT Business Solutions' local 5G service is said to be able to contribute to the acceleration of customers' digital transformation (DX) and the resolution of regional issues.
With the introduction of local 5G communication, for example, it is expected that real-time situational awareness will become possible, as it will be possible to handle images and video using drones and AGV cameras, in addition to the current situation where numerical data is acquired by sensors attached to production equipment to grasp the situation on the factory floor. Also, in worker support using AR glasses, it is expected that appropriate work instructions can be given remotely in real time, allowing even inexperienced workers to work efficiently hands-free. As described above, I believe that the introduction of local 5G has the potential to lead to a major step up in realizing operational efficiency and productivity improvement in promoting DX on the manufacturing floor.
Next, I focused on the product quality monitoring system using images. First, I listened to an explanation about Zukei Elmic's FA Finder.
https://www.elwsc.co.jp/app-package/
This is a platform for mutual cooperation between video and IoT data/equipment, and it is said that it can easily link different types of IoT devices, AI, and the user's existing equipment with camera images. It acquires images with cameras using events as triggers. Using this image data, it determines whether the product or manufacturing equipment is good or defective. At the venue, a device was exhibited that observes the quality of textile products transported on a conveyor belt using an NEC Platforms line camera. Also, the obtained image data was to be judged for quality by a machine learning judgment program generated by Cross Compass (https://www.cross-compass.com/). This is expected to expand the scope of application of image inspection and promote the spread of inspection automation. I was convinced that this would replace the quality judgment by the human eyes of skilled workers, and could serve not only as labor saving but also as a countermeasure for the depletion of skilled workers. As a side note, Cross Compass is a company I have worked with before, and they provide AI solutions that support digital transformation (DX) in the manufacturing industry. Their core product, "MANUFACIA," is provided as an AI generation tool that promotes digital transformation of production lines and solves manufacturing industry issues. In other words, it generates diagnostic programs for anomaly detection and predictive maintenance functions, and since it can provide AI programs that do not require a GPU, it has very high compatibility with existing systems.
Next, I listened to Adacotech. The visual inspection AI provided by the company provides unique anomaly detection AI technology that utilizes HLAC (Higher-order Local Auto Correlation) feature quantities, which is one of the image analysis technologies.
https://techblog.adacotech.co.jp/entry/Introduction-to-hlac
HLAC feature quantities are feature quantities calculated by simple processing such as neighborhood operations and image summation operations in a manner similar to performing filter processing on image pixel data. Specifically, it seems to calculate the co-occurrence features of pixel values at positions within the target image area. Using the data obtained in this way, it defines a normal model from good product images and judges images with different shape features that deviate significantly from that distribution as abnormal. It seems to be particularly good at surface inspection of sheets (mesh sheets, metal mesh, carbon sheets, leather grain, etc.), and it seems that this is also good for products that are continuous in a straight line, such as textile products. Also, since the calculation is simple, a GPU is not required, and it can be applied to existing image processing systems. This is expected to promote the spread of inspection automation through the use of normal PC-based systems.
Summary
I saw and heard at Smart Factory Japan 2024 this time, and I learned that in promoting DX at general manufacturing sites, it is necessary to digitize vibration and temperature for sensor management of manufacturing equipment, and means for that are provided, and by automating quality judgment through product visual inspection, it is possible to automate the monitoring ability of skilled workers, thereby achieving business continuity and labor saving at the same time. In this way, by comparing the equipment operating status with the product quality results, it is expected that quality maintenance/improvement will also be possible. Also, although these are just a few examples from this exhibition, I learned that these smart factories can be realized even on small manufacturing lines with investments that are not that large.