As IoT, AI, smart factories*, and Industry 4.0** are increasingly mentioned on a daily basis, more companies are announcing new business reform plans. How far have our customers implemented such plans? We interviewed consultants working on business projects under the theme of "actual situations in business reform, challenges to moving forward, and future directions."
* This is reproduced from an article in Chisounomori 2017 Vol. 5 published by Fujitsu Research Institute. (The interview was conducted on March 28, 2017.)
Haruyasu Ueda: Research Manager, Artificial Intelligence Research Center, Fujitsu Laboratories Ltd.
Yosuke Sawada: General Manager, Robot Business Unit, Engineering Planning Division, Product Planning Department, Denso Wave, Inc.
Kunimaro Suyama: Corporate Executive Officer, Fujitsu Research Institute
Noriaki Nakamura: VP, Executive Architect, Middleware Business Unit, Fujitsu Limited
Hiroyuki Kumagai: Principal Consultant, Manufacturing & Distribution Industry Sales Group, Fujitsu Limited
Masahiro Saiki: Managing Consultant, Fujitsu Research Institute
Yoshiyuki Ikeda: Group Leader, Digital Marketing Group, Fujitsu Research Institute
* Positions and affiliations are as of the interview.
Barriers to Business Reform Differ between Europe and in Japan
―――German Industry 4.0 has been much talked about recently. How does the Japanese equivalent differ?
Sawada: In Europe, there did not appear to be barriers to introducing the Internet in the Factory Automation (FA) industry. As of 2004, I often saw smaller companies performing remote maintenance for large enterprises. If I recall correctly, large enterprises were very positive on opening up their firewalls for this and had the idea of maintaining security as the top priority. They said that maintenance could be performed more efficiently over the Internet than by dispatching technicians, which showed how they were prepared for IT system integration from the start.
―――What efforts are you making in terms of ORiN (Note 3), which seems essential to business these days?
Sawada: One of ORiN's original goals was to achieve coexistence between the virtual and physical realms. The current version of ORiN was published in 2005, when the idea of CPS (Cyber Physical System) had in my opinion already been realized. The three concepts of ORiN are the factory monitoring system for collecting information from devices, the facility control system for controlling facilities via an integrated development environment on the IPC, and the embedded system that puts ORiN at the center of the software architecture as seen in our products.
The ORiN2 SDK was developed in 2006 based on the existing ORiN Ver2 released in 2005, but it was not welcomed in Japan because introducing PCs to factories was regarded as a terrible idea at that time. On the other hand, we successfully rolled it out in Europe. This was probably possible because system integrators have a strong presence in Europe, and many companies favored controlling all their facilities using the integrated development environment of their choice. In Japan, integrated facility control using ORiN became accepted after the Industry 4.0 boom, a five-to-six-year lag. Interestingly, it seems that ORiN caught attention more for its factory monitoring system than for its facility control system. Thankfully, ORiN Forum membership continues to increase by the month.
―――Japan has started to catch up with global trends.
Sawada: In Japan, the capability of retrofitting--or connecting old facilities to the latest apps, backbone systems, and IoT--is appreciated. ORiN's advantage is that nothing needs to be done for devices. Since the history of the Japanese automation industry is long and because supporting only new facilities would inhibit connecting existing facilities to IoT, an environment that allows for the coexistence of new and old facilities is a must. Especially in the FA industry, older standards remain in use while new standards continue to increase in number. ORiN enables gateways to be installed both on the application side and on the device side, which helps continuously create configurations that support new and old standards, a capability highly appreciated by customers.
Matching Needs with Seeds Using Digitization Technology
―――Please tell us about your efforts to facilitate digitization in factories.
Kumagai: Fujitsu's factories use PC-managed control to a much greater extent than those of ordinary companies, as exemplified by our use of a robot maker's development kit. Fujitsu joined the ORiN Forum to overcome poor efficiency and approach the ideal the forum aspires to. In Japan, control squads of automobile manufacturers are in charge of production technologies, whereas in Germany, external system integrators build car makers' production lines and thus can apply their system integration methods to various automakers. This is the largest difference between Japan and Europe.
Nakamura: One of my customers is a major steel-frame construction company. They had the problem that steel-frame parts made according to drawings were often deemed unusable at construction sites because even experienced personnel had failed to detect small deviations. To solve this problem, we used Augmented Reality (AR) to superimpose virtual parts made using 3D CAD software based on drawings over actual parts, which allowed the customer to detect small deviations from the drawings just by looking, thereby realizing zero defects and satisfying the customer. We fused what the customer wanted to achieve (needs) with digital technologies (seeds), and the reaction helped solve the problem. I thus realized that co-creation with customers creates new possibilities and raises the value of technology, so I want to spread this approach, which I think will become a universal trend.
Saiki: What my customers have recently mentioned is use of "audio information." Images only show the surface, whereas sounds provide information about the internal structure and potential damage. Customers think damage may be able to be detected by checking sounds during the final product inspection and preceding production processes. This kind of new possibility resulted from operational insights, and this area may have large application potential.
Sawada: We also regard 3D simulation as the first instance in which digitization has been accepted. We unified the real and virtual realms by enabling our virtual robot controllers (VRC) to be connected with commercially available simulation software as well as connecting simulators and physical machines.
Kumagai: The combination of 3D simulation of the final product with use of AR technology enables superimposition over what actually exists already. We can do what must be done only by combining various technologies.
How to Fill the Gap between Technology's Rapid Evolution and Slow Practical Application?
―――Customers will appreciate suggestions on how to introduce new technology in response to their needs.
Kumagai: For example, the shipbuilding industry started to pay attention to various IT technologies at an early stage and has actively attempted to practically apply technologies that are still in the research stage. In some cases, they have rejected our suggestions, saying that they had already tried them 10 years ago. Technologies evolve every two to three years, making what had been unusable practical. This time lag caused the customer to exclude from their options what they had already tried too soon. Technology always changes and advances, which must be examined by careful observation because it will soon become obsolete.
Nakamura: Technologies are replaced every quarter to half year, making it difficult to continue using the same technology for a long time. The point is to synchronize our ways of working with the technology evolution cycle.
Ueda: Deep learning, which two years ago was estimated to take 10 years to defeat people in Go matches, soon succeeded in beating top-tier professional players and surpassing human beings, suddenly attracting much attention since 2012. In the meantime, more variations have been created. Currently, I pay close attention to a learning algorithm known as Generative Adversarial Networks (GAN), which allows learning of both judgment and deception. This approach has begun to facilitate creative applications, such as complementing missing parts of images and reproducing colors in decolorized images.
―――What are the latest successful applications of AI-related technology?
Ueda: Deep Tensor is the most successful technology. Regular machine learning algorithms enable learning from arrays of data such as images but have problems that make it difficult to learn from connections in graph structures including bonded compounds, reference relations between documents, and friendship relationships on social networks. To solve this problem, we launched Deep Tensor, which adopts the approach of decomposing the connections between data sets into a unified representation to enable learning. In the drug discovery field, for example, this helps find connection patterns among proteins that even experts failed to discover by entering the proteins' chemical formulas and having it learn whether the medicine is effective or not. The latest applications include the intrusion detection that focuses on heavy traffic and distinguishing risky loans by paying attention to potential borrowers' credit histories.
Kumagai: Fujitsu is conducting in-house PoC testing in advance and attempting to apply image-related machine learning algorithms to the final inspection process in factories in order to replace human inspectors with machines. The progress of one technology stimulates development of surrounding complementary technologies; combining several such technologies makes it possible to do what was previously impossible, which sometimes leads to interesting developments.
Waves of IoT and Standardization Coming to Factories
―――When a customer toured the Shimane Fujitsu factory with me, I was asked which parts connected to IoT. Since everything is automated through sensors, each of which is part of IoT, we may need to explain more to visitors by showing them a list of IoT-connected parts.
Saiki: Since semiconductor factories involve different manufacturers in each process and data is accumulated in various formats, when we try to help, there are always problems with linking different data sets. For this reason, the ORiN Forum should also standardize data formats.
Kumagai: The Industrial Machinery Steering Committee of the Robot Revolution Initiative (RRI) annually publishes the results of their activities and has been trying to prepare Japanese standards mainly with Japanese machining tool manufacturers under the lead of the Industrial Machinery Division of the Ministry of Economy, Trade, and Industry. Since the Japanese machining tool industry recognizes themselves to be the world's best tool manufacturers with the best technologies, they must keep up the industry so that Japanese companies can thrive and position themselves to lead the rest of the world. That is why the country is attempting to unify and standardize specifications that currently differ from one machining tool manufacturer to another. Although the perspective of business is opposite that of users, users' opinions will be given more respect and they will demand systems that are easy to use and operate, free from manufacturer-imposed constraints. That will draw a line between standardized parts and manufacturers' proprietary parts.
―――If foreign machining tool manufacturers launch standard-compliant products, they may dominate the market.
Sawada: System integrators will probably play a more important role. When clients in Italy introduced the first monitoring system involving ORiN, it was system integrators, not end users, that connected the backbone systems and linked devices, which is the norm in system integration in the food industry. System integrators adopted ORiN because maintaining everything independently was difficult. In that sense, Europe prefers standards.
Nakamura: In Japan, we have built systems tailored to each customer. In Europe, operational tasks have been standardized under the Information Technology Infrastructure Library (ITIL), which caused a company called ServiceNow to take on all operational tasks, and this type of company is now coming to Japan. I think it is important to use open technologies from the perspective of total optimization and to actively work on standardization.
What Must Be Improved?
―――Please share your thoughts on how these technologies should be used and any obstacles to their realization.
Ueda: IoT enables data to be accumulated steadily; it is important to determine which analyses should be conducted on such data and what to use the analysis results for. The biggest obstacle to realization is to build a system that enables problem-conscious people to conduct analysis while increasing the number of data scientists.
Sawada: Regarding AI and robots, which have become buzzwords, it is unrealistic for a single robot maker to provide everything if robots become the output devices for AI. By making connections to ensure openness, intelligence for autonomy will also be required to prevent robots from becoming mere machines to be controlled by AI.
Kumagai: Some customers have a problem in that data is not accumulated or data is accumulated but not in a usable format. Basic IoT usage is to use accumulated information in one's tasks. Human intelligence and veterans' know-how are key because visualization alone does not improve anything.
Nakamura: In the West, it is common to carry out business reform using digital technology, whereas in Japan, the percentage of companies actually working on such reform is less than 20% according to a report. If we can clarify what customers want to do and associate it with the necessary technologies, we can make progress in use of digital technology. This requires co-creation with customers, and we must further strengthen partnerships, look into business activities, and share challenges.
―――Since it may be difficult to obtain the correct answers by asking customers what they want to do, a more appropriate approach would be to ask them what their biggest problems are.
Ikeda: According to the Ministry of Internal Affairs and Communications, the IoT penetration rate will double in 2020 compared to 2016. At that time, efforts to create new value by making full use of information over IoT and so forth will be necessary. There are an increasing number of instances in which we help customers look into their existing data and then consider how to use it. This reveals their problems and what kinds of data should be obtained, stored, and made available. In addition, the Fujitsu Group will work to achieve automatic visualization and digitization of existing business processes because processes that are not digitized present obstacles to many companies.
―――Based on new knowledge about IoT and AI as well as Japan's direction which we obtained and confirmed today, we will continue to serve our customers in the most helpful way.
*: Smart factory: A factory with an environment that produces the maximum profits by connecting every machine to the Internet to grasp and accumulate detailed activity logs of the machines and achieve efficient operation throughout the entire factory based on such information.
**: Industry 4.0: A German national project promoted by the country's government. The main objective is to advance the manufacturing industry; it is also called the fourth industrial revolution because it will fundamentally change the picture of the manufacturing industry in the 21st century.
***: ORiN: (Open Resource interface for the Network) A communication interface that provides unified access and representations for devices in factories regardless of manufacturer or model.