Artificial Intelligence Autonomously Controls JSR Chemical Plant

Aug. 4, 2022

Using reinforcement learning AI developed by Yokogawa, along with other Yokogawa automation technologies, a distillation column at a JSR Corp. plant was autonomously operated for 35 days.

David Greenfield, editor in chief

In a field test to determine the viability of artificial intelligence (AI) to control a chemical plant, automation technology supplier Yokogawa Electric Corporation and chemical manufacturer JSR Corporation successfully operated a distillation column at the JSR plant autonomously with AI for 35 days. The test ran from January 17 to February 21, 2022, for a total of 840 hours of operation.

Yokogawa and JSR say the test confirmed that reinforcement learning AI can control operations beyond automated PID and Advanced Process Control (APC) applications that typically require the judgement of plant personnel and manual operation of control valves.

The specific AI technology used at the JSR plant—the Factorial Kernel Dynamic Policy Programming (FKDPP) protocol—was developed by Yokogawa and the Nara Institute of Science and Technology in 2018. IEEE recognized this technology as being the first reinforcement learning-based AI suitable for use in plant management. In addition to the FKDPP protocol’s ability to handle tasks beyond PID and APC, Yokogawa says specific strengths of this AI technology include an ability to deal with conflicting targets, such as handling the dual needs of maintaining high quality and achieving energy savings.

See how your peers across industry are coupling sensor data with artificial intelligence.

Other Yokogawa automation technologies used in the test included:

OmegaLand plant simulator;
Centum VP integrated production control system;
Exaopc OPC interface package (OPC-compliant software that enables management of databases used in the processing industries); and
GA10 data logging software.

According to Yokogawa and JSR, specific achievements of the AI technology during its 35 days of autonomously operating the plant include:

Ensuring product quality while eliminating excess fuel, labor, and time costs associated with the production of off-spec products.
Maintaining liquids in the distillation column at an appropriate level while making maximum possible use of waste heat as a heat source.
During inclement weather conditions that could disrupt the control state due to temperature changes, products produced during those periods still met the plant’s standards.

Masataka Masutani, general manager of production technology at JSR

Though the test proved that complex chemical plant operations can be successfully controlled autonomously, Yokogawa and JSR point out that “there are still many situations where veteran operators must step in and exercise control.”

The two companies plan to continue exploring the application of the FKDPP protocol to other processes and plants to improve productivity.

“We have heard comments [from operators in the field] that not only has the burden on operators been reduced [with this AI technology], but the very fact that we have taken on the challenge of [using] this new technology and succeeded is motivation for moving forward with digital transformation,” said Masataka Masutani, general manager of production technology at JSR.

Kenji Hasegawa, vice president at Yokogawa Electric, added that the success of this test suggests that AI can significantly contribute to autonomous production operations, maximization of ROI (return on investment), and environmental sustainability around the world. “With our gaze fixed firmly on a world of autonomous operation that forms the model for the future of industries, we are now promoting the concept of IA2IA (Industrial Automation to Industrial Autonomy), which considers the impact of differences in humans, machines, materials, and methods in industry,” he said.

Food producer uses AI to automate visual inspection of food tray components.