ENEOS Materials Corp. (formerly the elastomers business unit of JSR Corp.) and Yokogawa Electric Corp. (Tokyo) announce they have reached an agreement that Factorial Kernel Dynamic Policy Programming (FKDPP), a reinforcement learning-based AI algorithm, will be officially adopted for use at an ENEOS Materials chemical plant. This agreement follows a successful field test in which this autonomous control AI demonstrated a high level of performance while controlling a distillation column at this plant for almost an entire year. This is the first example in the world of reinforcement learning AI being formally adopted for direct control of a plant.
Over a 35 day (840 hour) consecutive period, from January 17 to February 21, 2022, this field test initially confirmed*3 that the AI solution could control distillation operations that were beyond the capabilities of existing control methods (PID control/APC) and had necessitated manual control of valves based on the judgements of experienced plant personnel. Following a scheduled plant shut-down for maintenance and repairs, the field test resumed and has continued to the present date. It has been conclusively shown that this solution is capable of controlling the complex conditions that are needed to maintain product quality and ensure that liquids in the distillation column remain at an appropriate level, while making maximum possible use of waste heat as a heat source. In so doing it has stabilized quality, achieved high yield, and saved energy.
In this field test, the autonomous control AI demonstrated the following four benefits:
- Year-round stability
The autonomous control AI maintained stable control of the liquid levels and maximized the use of waste heat, even in winter and summer weather, with external temperatures changes by about 40ºC. No problems were observed, and stable operation and high product quality was achieved throughout the field test.
- Reduced environmental impact
By eliminating the production of off-spec products, the autonomous control AI reduced fuel, labor, and other costs, and made efficient use of raw materials. While producing good quality products that met shipment standards, the autonomous control AI reduced steam consumption and CO2 emissions by 40%*4 in comparison to conventional manual control.
- Lightened workload and improved safety
The autonomous control AI eliminated the need for operators to perform manual inputs. This not only decreased workload and helped to prevent human error, it also reduced mental stress levels and improved safety.
- Robustness of the AI control model
Even after modifications were made at the plant during a routine shut-down for maintenance and repair, the same AI control model could remain in use.
ENEOS Materials found over the course of this one-year verification process that the autonomous control AI was a robust system that could achieve stable performance and optimize operations throughout the year, including in winter and summer. The company will look into applying this AI to other types of processes and plants, and will continue working to improve productivity and save energy by expanding the scope of autonomization.
To promote plant autonomization, on February 27 Yokogawa launched the provision of an autonomous control AI service for edge controllers*5, also a world first*6. In conjunction with this service, the company is offering customers who wish to achieve autonomous plant operations a global consulting service that covers everything from the identification of control issues to the investigation of optimum control methods and the calculation of cost-effectiveness, and includes safety, implementation, maintenance, and operation.
Going forward, ENEOS Materials and Yokogawa will continue to work together and investigate ways to carry out digital transformation (DX) through the use of AI for control and condition-based maintenance in plants.
Masataka Masutani, Division Director, Production Technology Division, ENEOS Materials Corporation, said: “Amidst severe challenges impacting the petrochemical industry such as the retirement of experienced personnel who help to ensure the safe operation of facilities, we are pleased with this demonstration of the use of AI to autonomously control processes that had previously been controlled manually. In addition to reducing operator workload, this test, which has continued for about a year, has demonstrated that this system can operate stably without being affected by seasonal changes or regular maintenance and repairs, and can save energy and reduce GHG emissions. Through smart production, we will continue to strive for safety and stability, decarbonize operations, and enhance competitiveness.”
Takamitsu Matsubara, Professor at the Nara Institute of Science and Technology, said:
“The key to reinforcement learning is how the reward function is designed. By closely incorporating process industry control knowledge in the reward function, it is possible to create an AI control model with a high level of reliability and validity that is able to achieve year-round stable operation. The fact that this field test confirmed the model’s ability to be applied as is even after the performance of regular maintenance and repair implies the robustness of the AI control model. I believe that FKDPP, a new control technology that can handle complex conditions, will make broad-ranging contributions to the development of industry around the world.”
Kenji Hasegawa, a Yokogawa Vice President and head of the Yokogawa Products Headquarters, said:
“I am very grateful to have been able to work alongside our customer to take up the challenge of this globally unparalleled autonomization initiative. Given the difficulty of controlling operations in actual plants due to the complex effects of physical and chemical phenomena, there are many areas where highly-experienced operators have still had to intervene. With a focus on products and consulting, Yokogawa will develop and expand the use of autonomous control AI, and work with our customers to drive their decarbonization, digital transformation, and autonomization efforts.”