"A world's first" is now routine – ENEOS Materials Successfully Achieves Autonomous Control at a Plant Using AI

ENEOS Materials Corporation logo
Download (559 KB)

Executive Summary

ENEOS Materials Corporation, a member of Japan's ENEOS Group, is a chemical company that researches, develops, manufactures, and markets synthetic rubber, thermoplastic elastomers, latex, and other materials. The company's high-quality products are widely used in automobiles, industrial products, footwear, and other daily necessities.

In a world's first, an AI technology has successfully controlled operations for an extended period at an industrial facility in Yokkaichi, Japan. This autonomous control AI solution from Yokogawa remains in use at the Yokkaichi Plant to the present day, controlling operations in a facility at this plant that previously had proven difficult to automate, which achieving both energy savings and maintaining product quality. This world's first AI solution is now a routine part of operations at this plant.

ENEOS Materials Corporation


About the Project

The aim of demonstrating that AI can control a plant

In 2020, Keiichiro Kobuchi, a head in Yokogawa Digital's DX Service Division, decided that the time was right for a test that would demonstrate whether autonomous control AI was up to the task of directly controlling operations at a plant. The AI solution under consideration was Factorial Kernel Dynamic Policy Programming (FKDPP), a reinforcement learning-based AI algorithm that had been jointly developed by Yokogawa and the Nara Institute of Science and Technology, and the plan was to conduct this test as part of an Industrial Safety Advancement Promotion Project that was subsidized by the Ministry of Economy, Trade and Industry (METI) to demonstrate the uses of smart industrial safety technology.

Mr. Kobuchi decided to approach JSR Corporation’s elastomers business unit (now owned by ENEOS Materials) with a proposal of partnering with Yokogawa in this demonstration. As this business unit had always taken a proactive stance in adopting new technology, this conversation proceeded smoothly. Mr. Kobuchi also was well acquainted with Masataka Masutani, General Manager of the company's Production Engineering Department, as the two had worked together on a committee that had formulated a "Guidelines for AI Reliability Evaluation in the Plant Safety Field" document for METI. When Mr. Masutani and his colleagues heard of Yokogawa's intent to conduct this autonomous control AI demonstration test, they immediately had ideas on where this could be applied: a butadiene production facility at the company's Yokkaichi Plant.

The goals for this project

In August 2020, Yokogawa and the members of Production Team 3 at the Yokkaichi Plant commenced this joint demonstration project. This team was selected because the distillation column for the production of butadiene, a raw material used in synthetic rubber, was their area of responsibility. And the team had a problem to solve in their manufacturing processes.

The level of liquids in this distillation column and the supply of waste heat and steam need to be tightly controlled to efficiently separate and extract high-purity butadiene from substances that have close boiling points. Fluctuations in the liquid level within the distillation column can negatively impact the quality of the extracted product, and the use of large amounts of steam wastes energy and is costly. Due to external factors such as weather-related changes in the ambient temperature, it had proven difficult for the plant’s distributed control system (DCS) and advanced process controls (APC) to optimize adjustments to two key valves and thereby keep the liquid in the column at a constant level, maximize quality, and conserve energy. As the company had been relying on its plant operators to control these valves manually, it was decided to conduct this test to determine whether these valve settings could be controlled by autonomous control AI. One of the objectives was to reduce operator workload.

For its part, Yokogawa wanted to prove through this test that an autonomous control AI could directly control a plant's operations, and even control processes where proportional-integral-derivative (PID) and APC cannot effectively apply.

Personnel from the Yokkaichi Plant and Yokogawa began working toward the goal of having autonomous control AI control even the manipulated variables (MV) for this facility. It should be noted that the two teams were unable to meet face-to-face at this time due to COVID-19-related restrictions.

Production unit at the Yokkaichi Plant
Production unit at the Yokkaichi Plant

Demonstration steps

Taking the following steps, the two companies used FKDPP to successfully control operations at the Yokkaichi Plant. The AI control model that was generated realized AI-powered operations for 35 days (January 17 to February 21, 2022) at this plant, a world’s first*1.
Press release:https://www.yokogawa.com/us/news/press-releases/2022/2022-03-22/

  • Step 1: Build a plant model of the actual plant on the plant simulator and let FKDPP self-learn on the simulator to generate an AI control model.
  • Step 2: Evaluate and refine the generated AI control model based on the customer's on-site knowledge and data. MV presented by the AI are input by the operators into the DCS to control operations at the plant.
  • Step 3: Ensure safety, then control a real plant.

Demonstration steps

Refining the AI control model

After the plant underwent a scheduled shutdown for periodic maintenance and repairs, a new version of the AI control model with even higher stability began operation in April 2022. In modifying this model, Yokogawa took into account information on the operational behavior of the autonomous control AI and the results of data analyses that had been provided by ENEOS Materials. Although this project had begun with the complex goal of "maximizing quality and energy savings," FKDPP demonstrated over the entire year that followed that it was able to operate the two valves efficiently using the self-learning AI control model. In so doing, ENEOS Materials set a brilliant record by being the first company to achieve long-term autonomous control of operations at one of its plants by using this autonomous control AI*2, a world's first.
Press release:https://www.yokogawa.com/news/press-releases/2023/2023-03-30/

Project outcomes

FKDPP remains in use at this plant to the present day, and it now plays a routine role in its operations. The automation of valve operations has brought about the following significant results:

  • Autonomous control reduces operator load
  • Stable production of high-quality butadiene through process stabilization
  • Reduction of steam usage by 40%, saving energy and reducing costs

Although the current AI control model is working well, plant personnel are currently making additional efforts to refine the model. The goal is to build a best-in-class AI control model that will take stabilization of the process to an ultimate level.


Customer Satisfaction

Q. What were your thoughts when you first heard of the plan to use AI at a plant for the first time?

"I thought it sounded very interesting. Our corporate culture is very positive, actively adopting new technologies. We've been working on IT and DX for a while now, and we've made a lot of improvements, such as doing facility inspections with drones and using image data to detect abnormalities in synthetic rubber products. AI is our latest innovation."

"I was excited to be selected for this project, and I'm grateful for the opportunity to work with the latest autonomous control AI technology."

Mr. Fukami (Manager of Production Team 3 Production Department), Mr. Yada (Production Team3 Production Department)
Mr. Fukami (Manager of Production Team 3 Production Department),
Mr. Yada (Production Team3 Production Department)

Q. Is it difficult to control those valves?

"We had been trying to automate the operation of those valves with the DCS and APC, but were unable to do so. The operators had to control them manually, while taking into account the effects of their actions on the level in the distillation column, and they also had to frequently intervene manually if disturbances occurred. They had a heavy workload. Everyone wanted to automate this and also save energy and cut costs."

Q. Did you have any concerns about AI controlling actual plants?

"We had no worries at all. Besides, our plant is equipped with safety instrumentation, and if things go wrong, we can simply stop the AI and go back to normal DCS operation."

"During the 35 days of continuous operation, I received several calls on my cell phone from operators asking if it was OK to stop the AI. The operators were a little nervous at first because the AI could do things with the valves that they had not thought of. Now they feel comfortable letting the FKDPP do the operation."

Q. How is the operation status with FKDPP?

"We use the AI control model that operates in the plus/minus 5% range for the target level, and the control is executed accordingly."

"A human cannot operate two valves at the same time, but an FKDPP can. The FKDPP sometimes performs valve operations that humans would never think of. We humans can gain new know-how by watching this."

Q. What is your level of confidence in the autonomous control AI?

"The reliability of the FKDPP, and the current AI control model, is 100%. We are in the process of building a new model, but if it doesn't work, we can go right back to our current model. No problem."

Q. You achieved an amazing "world's first". Did you get any feedback from within your company?

"Thanks to you, we received the President's Award."

"Employees of ENEOS Group sometimes ask me, 'Are you from "that" Production Team 3 at the Yokkaichi Plant?' I feel that we have attracted a lot of attention within the group." 

"I held an AI workshop for younger colleagues at the plant. When I asked, 'Are there any areas you would like FKDPP to control?' there were many more answers than I expected. I feel that our employees are becoming more aware that autonomous control AI can automate areas that humans can’t handle."

Mr. Suzuki (Deputy Manager of Production Team 3 Production Department), Mr. Tachi (Production Team3 Production Department)
Mr. Suzuki (Deputy Manager of Production Team 3 Production Department),
Mr. Tachi (Production Team3 Production Department)

Q. What are your comments and expectations for Yokogawa?

"We were only able to meet online because of COVID-19, and our first face-to-face meeting was for the final debriefing session. Had we been able to meet in person right from the start, we would have been able to share information more easily, but we managed alright with weekly online meetings and the online exchange of data, and were able to successfully complete the project."

"Despite these circumstances, we made steady progress with the project. We stayed in close touch with our Yokogawa counterparts, and they paid close attention to the data that we provided, commenting on it daily."

Q. What are your future challenges?

"The distillation column remains in a satisfactory state, but we would like to increase the stability of the column level to about plus or minus 2%. If we can do that, operator workload will be further reduced. In addition to stabilizing column level, we would like to improve the quality of the synthetic rubber that we produce and improve our energy savings."

"We would like to apply this autonomous control AI solution at our downstream polymerization plant. The polymerization process will be a difficult challenge, but we want more than anything to replicate the distillation column success in other parts of our company."

"We had thought that we had reached the limit of the improvements that could be made, but we have realized with the introduction of new technologies like AI that there is more that can be done."

Messrs. Fukami, Yada, Suzuki, Tachi (Yokkaichi Plant), Mr. Kobuchi (Yokogawa)
Messrs. Fukami, Yada, Suzuki, Tachi (Yokkaichi Plant), Mr. Kobuchi (Yokogawa)

Department name and job titles are as of interview.

*1:Based on a February 2022 Yokogawa market survey for cases involving the use of AI to directly change manipulative variables (MV) at chemical plants.
*2:Defined by Yokogawa as being an AI that has a high level of robustness, allowing it to derive an optimal control method and respond autonomously to a certain degree even in situations that are encountered for the first time.

FKDPP demonstration system is on display at the Yokogawa headquarters showroom.
FKDPP demonstration system is on display at the Yokogawa headquarters showroom.



  • Base Chemical

    Yokogawa has been serving the bulk chemical market globally and is the recognized leader in this market. With products, solutions, and industry expertise, Yokogawa understands your market and production needs and will work with you to provide a reliable, and cost effective solution through the lifecycle of your plant.

    See More

Related Products & Solutions

  • AI Product Solutions

    We have been solving customers’ issues in many industries using our AI. Utilizing our analysis experience and elemental technologies, we can offer a lineup of easy-to-use AI products. AI functions such as predictive detection and future prediction can be easily applied, and operation efficiency is improved.

    See More