AI used to control process manufacturing operations

Following the announcement of the successful conclusion of a field test, which saw artificial intelligence (AI) used to autonomously control the valves of a distillation column at the ENEOS Materials Corporation chemical plant in Japan for 35 days, Control Engineering Europe spoke to Dr. Hiroaki Kanokogi, general manager in the Yokogawa Products Headquarters who is in charge of AI development.

He explained AI control for industrial process plants is not yet as far advanced as the AI technology required for predictive maintenance. In many processes external atmospheric temperature changes can have a big effect on the process and controlling a process to take into account these unexpected disruptions requires complex control of temperature, pressure, and flow rate due to the potential unwanted chemical or physical reactions that can result.

In the process world, the proportional-integral-derivative (PID) control mathematic-based algorithm is used to control processes. However, it is not able to adequately handle the unexpected disturbances caused by continually changing weather patterns which means that human intervention is often still required in situations where short-term changes that occur, due to varying atmospheric temperature, may have an adverse effect on a process. In this field test, the AI is able to autonomously control this kind of complex process which previously had required manual operation.

Although control of distillation columns is well-established, optimum control for truly stable and high-efficiency production remains a challenge for many process sectors. Traditionally, knowledge is handed down from engineer to engineer but, due to a growing engineering skills shortage more plant operators are now looking to autonomize these challenging operational areas within their plants.

A key asset

Because distillation columns are a key asset in so many process industries this was where the Yokogawa/ENEOS Materials AI control project was focussed. Some of the valves on the distillation column at the ENEOS Materials chemical plant were being manually controlled to ensure both product quality and energy saving are always optimized.

“ENEOS Materials wanted to utilize waste heat within the distillation column to help make the plant more environmentally friendly and to benefit from energy savings. However, waste heat is not a stable commodity and is also affected by atmospheric temperature changes,” Kanokogi said. Because product quality was not an acceptable trade off to gain energy savings, the company had previously looked for a suitable control solution. PID control could not adequately control the system and neither could advanced process control (APC) solutions, so the organization still had to rely on operators to manually control the distillation column valves.”

This is why the company was keen to work with Yokogawa to find an autonomous control solution.

The live trial at ENEOS Materials confirmed that reinforcement learning AI can be safely applied in a real-life plant and has demonstrated that the technology can control operations that have been beyond the capabilities of existing process control methods.

In the field test, the AI solution successfully dealt with the complex conditions needed to ensure consistent product quality and to maintain liquids in the distillation column at an appropriate level while making maximum possible use of waste heat as a heat source. In so doing it was able to stabilize quality, achieve high product yield, and also increase energy savings.

While rain, snow, and other weather conditions were significant factors that could disrupt the control state by causing sudden changes in the atmospheric temperature, throughout the trial period, the end-product adhered to quality standards and has since been shipped.

The 35-day live trial has proved that next generation control technology, using the FKDPP reinforcement learning-based AI protocol, can greatly contribute to the autonomization of production, ROI maximization, and environmental sustainability. “The results of the trial found that a consistent and high-quality product was produced and that any losses in the form of fuel, labour costs and time that can occur due to production of off-specification products were eliminated,” Kanokogi said.

Safety first

Safety was obviously high on the list of priorities for the trial. “To ensure safe operation during the trial period we initially used the FKDPP protocol on a plant simulator to create a control model. Then we assessed the AI behavior – checking AI data against past operating data from the distillation column to ensure it was stable, that product remained within specification, and that veteran operators were satisfied with the AI control instructions,” Kanokogi said.

When all the data had been checked, the FKDPP AI protocol was finally connected to the plant. “To ensure safety remained paramount during the trial we integrated the FKDPP algorithm into the existing CENTUM VP production control solution so that operators were always able to see the AI status on the CENTUM HMI display. Because we integrated the AI protocol into the CENTUM control system the engineers would be able to deploy any of the safety interlocks, or other safety functions in place around the plant, at any time during the trial if needed,” Kanokogi said.

Following a regular maintenance operation, ENEOS Materials has stated that it intends to continue using AI to control the operation. It believes that the duration of the trial was sufficient to confirm that reinforcement learning AI can be safely applied to the plant. ENEOS Materials believes that the demonstration shows potential of AI for addressing control challenges that previously could not be resolved at chemical plants and will be investigating its wider application in other processes and plants with the aim of achieving further improvements in productivity.

Going forward, the two companies will also continue to work together and investigate other ways of using AI in plants.

Commenting on the project, Masataka Masutani, division director of the production technology division at ENEOS Materials, said, “The petrochemical industry is under strong pressure to improve safety and efficiency in its production activities by utilizing new technologies, such as IoT and AI. In this experiment, we took on the challenge of the automation of plant process control using AI control technology. We verified that AI is able to autonomously control the processes that were previously performed manually on the basis of operators’ experience, and we are firmly convinced of the usefulness and future potential of AI control. From those in the field, we have heard comments saying that not only has the burden on operators been reduced, but the very fact that we have taken on the challenge of this new technology and succeeded is motivation for taking forward digital transformation. Henceforth, we shall expand operations controlled with AI, and work to enhance chemical plant safety, stability, and competitiveness.”

Takamitsu Matsubara, associate professor at NAIST, who jointly developed the FKDPP protocol, said: “I am very glad to hear that this field test was successful. Data analysis and machine learning are now being applied to chemical plant operations, but technology that can be used in autonomous control and the optimization of operations has not been fully ready until now. The reinforcement learning AI FKDPP algorithm was jointly developed by Yokogawa and NAIST in 2018 to realize autonomous control in chemical plants.

Despite having to refer to a large number of sensors and control valves, the AI can generate a robust control policy in a limited number of learning trials. These features helped to improve the efficiency of the development process and led to the achievement of autonomous control for a long period of 840 hours during the field test. I think this very difficult achievement of autonomous control in an actual distillation column and the fact that the level of practical application has been raised to the point where the entire production process and safety are integrated into one system have great significance for the entire industry. I look forward to seeing what happens next with this technology.”

The next step

Going on to discuss the next steps in Yokogawa’s AI journey, Kanokagi explained that the company’s vision for autonomous control – IA2IA (industrial automation to Industrial autonomy) – is to help processors achieve their smart manufacturing goals. “We feel that the success of the ENEOS Materials project is a very tangible step towards IA2IA, where plants will start to think for themselves about how to improve product quality or KPIs.”

In 2021, Yokogawa surveyed 534 decision makers at 390 manufacturers and 42% believed that in the next three years the application of AI to plant process optimization will have a significant impact on industrial autonomy.

The final piece of the AI puzzle for the process sector will be the development of an improvement point discovery AI algorithm which will aim to discover potential improvement points in multiple processes and will define issues by looking at big data to autonomously identify problems.

When this is added to problem analysis AI – which analyses problems based on sensor data and defines factors or creates new KPIs for improvements – and autonomous control AI, such as the FKDPP algorithm, which continually searches for the optimum control model in a process, it will be possible to create an autonomous plan-do-check-act (PDCA) loop for continual process optimization.

“The timeline to create a working version of discovery AI technology will hopefully be quicker than for the FKDPP algorithm which we believe to be the most difficult. The loop could be completed within a few years,” Kanokogi said. “The process of improving operations to achieve key performance indicators can take six months to a year when performed manually. With AI, operation improvement can be done autonomously, 24-hours a day, 365 days a year. Until now, it has not been possible to fully automate some process plants. The next generation control technology, using reinforcement learning-based AI, will autonomize areas that cannot be automated using existing control methods while also ensuring safety and improving productivity.”

– This originally appeared on Control Engineering Europe’s website. Edited by Chris Vavra, web content manager, Control Engineering, CFE Media and Technology, cvavra@cfemedia.com.

Do you have experience and expertise with the topics mentioned in this content? You should consider contributing to our CFE Media editorial team and getting the recognition you and your company deserve. Click here to start this process.