In 2017 Yokogawa started developing AI for control applications. In 2018 FKDPP (Factorial Kernel Dynamic Policy Programming algorithm) was jointly developed by Yokogawa Electric Corporation and the Nara Institute of Science and Technology (NAIST). FKDPP was run on a simulator of a vinyl acetate manufacturing plant (distillation column, decanter). It was presented at the IEEE International Conference in August 2018 as a reinforcement learning technology that can be applied in plants (the world's first reinforcement learning technology to demonstrate high feasibility for use in plants).
Based on four valves and nine sensor values, FKDPP operated the valves to maximize the volume of products produced and ensured that quality values conformed to standards and safety standard values were within tolerances. Stable and optimized valve operation was achieved in just 30 learning trials.
In 2019 Yokogawa succeeded in an experiment using a control training device - a three tank level control system, via laptop PC. Although the system can be controlled with conventional PID control technology, it was shown that FKDPP can reduce the settling time by 50% to 70% while also preventing over-shoot.
In 2020 Yokogawa showed the potential for controlling an entire plant that was confirmed with a simulator. (Paper presented in April 2020).
The demonstration utilizing e-RT3 Plus (as a PLC) with FKDPP was unveiled in an exhibit at Measurement Exhibition 2020 OSAKA.
The experiment was successful and started practical use of minimized use of electricity and LPG for air conditioning control in semiconductor clean rooms in Komagane plant of Yokogawa, Japan.
In 2022, Yokogawa launches Autonomous Control AI Consultant Service and FKDPP controlled a chemical plant for 35 consecutive days (in a world first).
- Reinforcement learning AI can be safely applied in an actual plant
- Can control operations that have been beyond the capabilities of existing control methods (PID control/APC) and have up to now necessitated the manual operation of control valves based on the judgements of plant personnel.
In 2023 (in a world first), Yokogawa’s Autonomous Control AI (FKDPP) is officially adopted for use at an ENEOS materials chemical plant!
- Over two years of stable operation demonstrates this next-generation control technology can decrease environmental impact, achieve stable quality, and transform operations of control valves based on the judgements of plant personnel.
- Yokogawa’s Autonomous Control AI Algorithm (FKDPP) Takes Highest Honor in Japan Industrial Technology Awards!
In 2024 Yokogawa Launched Autonomous Control AI Service for Use with e-RT3
- Optimizes control to improve productivity and save energy
The purpose of the original field test in 2022 was to demonstrate that reinforcement learning AI (FKDPP: Factorial Kernel Dynamic Policy Programming algorithm) can be applied safely in plants where safety is an absolute necessity and also to demonstrate that reinforcement learning AI can be used to control areas that existing technology cannot automate.
Success in “Controlling" an Actual Plant with AI
Field Test Target Overview
AI directly controls the two heat sources (waste heat utilization) in the revoir section of the distillation column.
Reason for choosing the target : Until now, the target location could not be automated by PID control or APC, so it was manually controlled. (This target was requested by the production department.) Manual control at regular intervals 24 hours a day, 365 days a year is very difficult!
FKDPP operates the MV and controls it directly! No changing of the SV! New control method!
Generate AI control model with a plant simulator
- Plant model generated from design information for the relevant plant
- Reinforcement learning-based AI (FKDPP algorithm) learned and generated a control model
Comprehensively assess AI control model validity and reliability
- Checked with past operating data
- Was it stable?
- What kind of control was performed when problems occurred?
- Checked with real-time data
- Was it stable?
- Was product quality within spec?
- Were veteran operators satisfied with FKDPP control instructions?
Ensure safety, then control a real plant
- Ensured safety with existing interlocks and other safety functions
- Integrated with CENTUM™ VP integrated production control system, and incorporated into plant operations
- Ensured safety in operations (planned responses and established system for dealing with AI system malfunctions)
What Did FKDPP Actually Control?
In processes where existing control methods (PID control / APC) could not be applied:
- Separate substances with similar boiling points were controlled
- Maintain liquids in the distillation column at an appropriate level to ensure all products are compliant
- Achieved high quality outputs and increased energy savings while taking into account sudden external disturbances (rain, snow, etc.)
- Operate valves to maximize use of waste heat for heating
Value for Society
Autonomous Control
Areas that previously couldn’t be controlled with PID control and APC were autonomously controlled by reinforcement learning-based AI (the FKDPP algorithm)
Over Two Years
Managed and controlled with CENTUM™ VP integrated production control system
Safe Operation and Improved Productivity
Simultaneously achieved safe operation and improved productivity, with stable quality, high yield, and energy saving
Reduced Cost and Time Loss
Only high-quality products were produced, so losses in the form of fuel, labor costs, time, etc. that occur due to production of off-spec products were eliminated
It was proved that next generation control technology using reinforcement learning-based AI ( FKDPP) can greatly contribute to the autonomization of production, ROI maximization, and environmental sustainability.
Conclusion
Until now, there have been many parts of the plant that have not been fully automated.
The next generation control technology using reinforcement learning-based AI ( FKDPP) will autonomize areas that could not be automated with existing control methods while ensuring safety and improving productivity. FKDPP is a disruptive innovation that allows for a different dimension of control, particularly in such areas. This AI technology can be applied in the energy, materials, pharmaceuticals, and many other industries.
The daily monetary value of operations in large-scale plants is in the range of tens of millions of dollars. Autonomous control AI (FKDPP) can greatly contribute to the autonomization of production around the world, ROI maximization, and environmental sustainability, and will have a major economic impact.
In order for our customers to fully understand this value, Yokogawa welcomes customers who are interested in these initiatives globally. We aim to swiftly provide products and solutions that realize industrial autonomy.
Be sure to read our blog: Autonomous AI Control Service for Edge Controllers
Would you like to speak to one of our technical specialists to learn more? Just fill out the form below.