What will AI and data science bring to Ebara's manufacturing industry, which has more than 100 years of history?

As the evolution of AI is gaining attention, Ebara, a company with a long history of manufacturing, also has data scientists who use AI and the technology to analyze data. We have employees who use AI and data to solve various business and operational problems across the company, employees who visit customers and solve their problems by analyzing data from semiconductor manufacturing equipment, and even employees who use AI technology to improve Ebara's waste incineration facilities.We have people who use this technology in a variety of areas, and we are also nurturing young talent.

What these employees are conscious of on a daily basis is not the definite use of AI but rather how to clarify the issues at work and then thinking about how AI can contribute to their solutions. Sharing their opinions are three individuals involved with AI at EBARA: Keisuke Kawashima (Data Science Section, Corporate Data Strategy Team), Junya Machida (New Technology Development Section, Development Department, Core Technologies Group, EBARA Environmental Plant Co., Ltd.), and Yoshimasa Osone (Data Science Section, Equipment Division, Precision Machinery Company).

Kawashima: I belong to a team that solves company-wide problems based on AI and data strategies. In addition to promoting transformation using data in the manufacturing and after-sales service fields of various departments, we work together with departments to solve problems that exist between departments as well as issues specific to departments. Currently, I am working a lot with the Component Division of the Precision Machinery Company.

Ozone: I analyze data collected from Ebara's semiconductor manufacturing equipment used by our customers, and consider ways to improve our customers' productivity and work efficiency. For example, I work on abnormality detection by understanding operation patterns and trends in the equipment with data. Our goal is to combine semiconductors, AI, and software to improve our customers' manufacturing sites.

Machida: You work to improve the quality of Ebara's products and services using AI and ICT technologies such as machine learning, and are currently working to improve and streamline the operation and management of waste incineration facilities that Ebara has been contracted to operate and maintain by its customers. At EBARA Environmental Plant, we are promoting joint development with a AI start-up, and we are developing our own technologies for the areas we can handle. In addition, we developed a system where AI automatically detects signs and occurrences of falls based on camera images and video in response to the risk of people falling into pits where waste is stored at waste incineration facilities.

Kawashima: A major reason I chose to work at Ebara as a mid-career hire was that they have a team that widely promotes data utilization. I started working with AI and data when I worked for a manufacturer in my previous job. With the launch of a smart factory project, there were efforts to utilize the data collected from the plant, and I joined as a member. I was a beginner when it came to data science, but I had been doing chemical simulations up until university, so the psychological hurdle of handling data was small.

I then used the data to create a system for quality inspection and anomaly detection, and found it more rewarding than I expected, and became absorbed in it. I gradually wanted to work in a position that would help me make better decisions using data, and when I found out that Ebara had a specialized team, I decided to change jobs.

Ozone: I saw that Ebara was hiring data scientists for their new graduate positions, so I applied. I was working on using AI image processing technology to predict the next movement based on the physical condition of an opponent. When I mentioned this during my job interview, the Ebara employee who spoke to me, who is now my boss, told me, "That technology will be put to good use at Ebara," so I decided to join the company. It took a bit of courage to go from the sports industry to manufacturing, but those words made me decide.

Machida: Are there many people in your department who were involved in data science during their university days?

Ozone: Not at all. Some studied data science as a hobby or through activities in their private lives while working at other jobs. At the same time, since my generation, because there are data scientist positions for new graduates, there are some people who have had experience since they were students. People with a variety of career backgrounds work together.

Machida: I joined the company in 2017, but at that time there was no position for a data scientist, and I myself wanted to design equipment in the field of plant engineering. However, the year I joined the company, my current AI-related department was launched, and I was assigned there. It was unexpected, but like Kawashima, I began with AI image recognition and data analysis, which was a lot of fun right away. There wasn't as much information about AI out there as there is now, so I studied on my own and improved my knowledge and skills with the help of AI companies that were joint development partners.

Machida: In my case, the system I develop will be used by Ebara employees who are involved in the operation and management of waste incineration facilities, but rather than creating it unilaterally on our side as developers, I try to be conscious of implementing it only after the people on the ground are convinced that it is a system that is truly easy to use and that they understand that it will solve problems on the ground.

To keep the current infrastructure running while the working population is declining, automation and efficiency using AI and other technologies are inevitable. That said, each person working on site has experience and thoughts that they have developed in their respective jobs. Automating these things may at first glance make people on the front lines feel like their workload will decrease.

We need to make sure that everyone understands that we are creating a system that will reduce the workload on the front lines and contribute to improving productivity, and that they can use it with confidence.

Ozone: This is similar to what you just said, but when we visit our customers' sites, even if the data has already been collected and we can foresee that a certain function can be created, we always try to think again about whether the customer really needs it. Ideally, if we were to create a new function, we would work out in concrete terms what results it would produce and how much it would contribute to the business, and then align our views with the customer.

To achieve this, rather than just talking about hypotheticals, we sometimes first create a system of a smaller scale and cost, and then consult with the customer on whether it is necessary based on its track record.

Kawashima: I think what you just said is really important. Because there are times when people approach us and ask things like “Can you do something with the data we have?” or “Can we do something with AI?” but this is not the correct way to move forward. The first thing we want to clarify is the purpose of what problems we have and how we want to solve them. Once that purpose is established, AI and data come into play as a means to solve the problem.

Expectations for AI have been rising recently, but this technology cannot do everything. If you have a large amount of data without having decided on the problem, it can be difficult to know where to start, and if you don't start with the problem, you may have a large amount of data but not the numbers needed to solve it. Ozone: In that sense, it would be ideal to start by asking people in the field to list the problems they are facing, rather than starting with AI and data, and then think about what we can do with AI. Kawashima: That's right. I think it would be nice to have our team as a place to consult, so I would like to have people start by reaching out to us and saying, I have this problem. Even an email would be fine. We would then like to work with those on the ground to resolve the issue. Although we may be knowledgeable about data, of course the people on the ground have more in-depth knowledge of the business and the issues involved.

Ozone: By working together, the people on the ground will also accumulate knowledge of data science, which will be beneficial for both sides.

Ozone: In my field, I think it would be a great advantage for someone who is familiar with the equipment to also acquire knowledge of AI and data. As Kawashima mentioned earlier, expectations for AI are sometimes too high. A practical system is the most likely to produce results, so if it becomes widespread, it will become one of Ebara's strengths.

Machida: I think what's interesting about Ebara is that we can think about how to use AI to address the issues our customers and society face. For example, a company that specializes in AI development may mainly create AI models and systems requested by customers, but we don't just create models; we can start by thinking about how to use AI to solve problems that need to be solved.

In particular, I am working on improving waste incineration facilities, so I am thinking about implementing AI to solve social issues such as decarbonization and environmental problems. I think there is a real thrill in being able to work on something big. Kawashima: As you can tell from what you two have said, the great thing about Ebara is that it allows you to be involved in a variety of fields through AI and data. In addition to semiconductors and the environment, there are of course pumps and many new businesses. Data scientists are full of intellectual curiosity and enjoy analyzing a wide variety of data, so I think the breadth of the scope of this company is appealing.

Kawashima: I would like to be able to track the traceability of shipped Ebara products such as pumps and blowers, and also to be able to obtain data even after shipment. We have already started working on this, and we believe that we can build better products and sales systems based on operational data.

Ozone: I would like to be able to explain to customers in an easy-to-understand way how AI and data science work, as well as their benefits. In particular, AI is an unknown for customers, and there are times when we are asked to explain how it works in detail. But on the other hand, too much simplification will not lead to acceptance or trust. I want to develop the skills to convey information that is easy to understand and persuasive.

Machida: As for garbage and waste, while efforts to decarbonize and address environmental issues will become increasingly important in the future, we cannot continue with our current methods for the next 20 or 30 years. I want to create a new type of waste incineration facility by incorporating not only AI but also new technologies and ingenuity. Currently, we are focusing on our own facilities, but we would be happy if we could create good products and spread them outside the company, and through that, we would like to contribute to solving social issues.