How Close Is the Chemical Industry to True Autonomy?

On the road to autonomous operations, industry partnerships  like the Yokogawa-ENEOS collaboration could take on increasing importance. In March, Evonik and Siemens announced plans to collaborate on the development of next-generation autonomous technology for chemical plants. Henrik Hahn, chief digital transformation officer for Evonik Industries, says the companies view the road to autonomy as “a progressive, structured journey rather than a ‘big bang’ transformation.” 

“The autonomy vision becomes achievable only through strong partnerships that combine deep operational know‑how with automation, digitalization and AI expertise,” asserts Hahn. “This is exactly why Evonik and Siemens are pursuing this work together: Autonomy requires the integration of technology, process knowledge and organizational readiness.”

“At this point, semi-autonomous capability is achievable, and we’re already seeing elements of it in practice,” said Ravande, who also is the founder of the Houston-based provider of AI-powered operational software. “I describe it as supervised autonomy — essentially autopilot with human oversight. You can think of it in terms of self-driving cars today. You still need a person in the driver’s seat, paying attention and ready to intervene if the system makes a mistake.”

Ravande’s description of “supervised autonomy” reflects a broader industry reality: Autonomy in chemical operations isn’t an all-or-nothing proposition. At the same time, the definition of autonomy — and what progress looks like — could vary across organizations. 

Evonik and Siemens define autonomy in chemical production as “the capability of a plant to run, optimize and adapt its operations safely and efficiently with minimal human intervention, supported by dynamic, self‑learning digital systems,” according to Hahn. 

“In practical terms, autonomy means that a plant can independently manage routine tasks, anticipate and correct deviations and optimize key performance parameters while operators move into a supervisory, expertise based role,” Hahn said.

To define the progression from automation to autonomy and help manufacturers understand where they are in the journey, ARC Advisory Group developed what it calls an Autonomous Operations Maturity Model. The five-level autonomy index is summarized below: 

• Level 0 (no autonomy) – Operations rely on humans to make all decisions and perform all functions. 
• Level 1 (operations assistance) – Basic automation technology supports human decision-making.
• Level 2 (regulatory automation) – Basic control systems maintain process variables automatically using technologies such as PID loops.
• Level 3 (advanced regulatory) – Automation can recognize defined abnormal conditions and take corrective actions within programmed limits.
• Level 4 (select autonomy) – Systems can make certain operational decisions independently, such as switching to standby equipment in the event of a pump failure, while notifying human operators.
• Level 5 (full autonomy) – Facilities function as largely self-directing operations with minimal need for human intervention.

A Schneider Electric report published in late March concluded that companies in the energy and chemicals sector are making more progress toward autonomy than expected. 

Based on a survey of 400 executives across 12 countries, the report stated that organizations in the sector are operating at an average autonomy level of 3.52 out of 5 on the ARC Autonomous Operations Maturity Model. On average, energy and chemical companies aim to achieve Level 4.02 (80% autonomy) by 2030. While North America lags other regions in terms of progress, nearly 90% of executives from North American companies rank autonomy as a high priority. 

“Autonomy is rapidly becoming the new operating model of industry,” said Gwenaelle Avice Huet, an executive vice president at Schneider Electric.”

In terms of the most realistic starting points for implementing autonomous technology, Axel Lorenz, CEO of process automation at Siemens, said chemical facilities should look for opportunities to shift personnel away “from manually coordinating routine operational decisions toward supervising increasingly automated process control and optimization.”
 
“Operators still spend significant amounts of time monitoring different screens, acknowledging alarms, investigating root causes and manually identifying and implementing relevant SOPs,” Lorenz says.

According to Lorenz, a “high-impact candidate for early-stage autonomy” is state-based and procedural automation, integrated with digital twins and shared data layers. 

The goal of these systems is to recognize defined operating conditions, such as startup, steady-state operation or abnormal events and automatically execute the appropriate response based on established procedures. 

For example, instead of requiring an operator to interpret alarms and manually initiate corrective actions during a process upset, the system would identify the condition and follow preapproved workflows to stabilize the process within defined safety limits.

“This technology lays the foundation for tighter closed-loop integration with optimization and control technologies,” Lorenz says. “For example, rather than optimizing only setpoints within a fixed operating mode, future systems will increasingly be able to move plants safely between operating modes to realize greater benefits.” 

Barriers to Autonomy: Data, Integration and the Human Factor 

While AI tends to grab most of the headlines, it’s only one part of a broader technology stack that organizations are leveraging to make autonomous operations a reality. 
Evonik and Siemens say they’re developing a layered “digital plant” architecture that will bring together existing plant control systems, digital twins, edge computing, data connectivity and AI-driven analytics into a unified operating environment. 

A vital component of that digital plant technology stack is “seamless data connectivity, both vertically from ERP to the shop floor and horizontally across systems,” Lorenz says.

That level of data connectivity remains aspirational for many chemical facilities, according to Ravande, who asserts that slow progress on the road to digitalization could be the industry’s biggest hurdle to autonomy.

“Many customers that we deal with are still on paper, still using walkie-talkies, still relying on binders,” Ravande says. “Their OT and IT stacks are disconnected, and the front line is still operating with little to no [digital] technology.”

Other notable challenges include the complexities of integrating advanced technologies into legacy systems, as well as the increased cybersecurity risks that come with expanding connectivity across plant operations. But not all the hurdles are related to technology. Autonomy will dramatically change the role of the operator in a chemical facility, reshaping how work is structured and how decisions are made on the plant floor. That’s more of a cultural sea change than a technological shift.     

“Operators will work more closely with digital systems, focus more on exception handling and specialized workflows and supervise a broader operational scope than today,” Lorenz says. “As autonomy expands, digital literacy and cross-functional operational understanding will become increasingly important.”

The Road Ahead

Despite the challenges, real-world deployments are beginning to demonstrate what’s possible.

FKDPP, the AI protocol that enabled autonomous control of a distillation column at the ENEOS facility in Yokkaichi, Japan, is particularly promising. The FKDPP algorithm can learn optimal control directly from operational data instead of relying on pre-built process models, making it highly adaptable to chemical operations with hard-to-control dynamics.

Also noteworthy is FKDPP’s rapid learning capability. The algorithm can build robust control models in just 30 learning trials.

This past December, Chemical Processing recognized Yokogawa with a Vaaler Award for its role in developing FKDPP and implementing the technology at the ENEOS plant’s distillation column. In an episode of the Chemical Processing “Distilled” podcast, Yokogawa solution consultant Karthik Gopalakrishnan asserted that autonomous control “is something [that] is eventually going to be part of our life.” 

But he also emphasized that there will always be some level of human involvement.

“At the end of the day, you still need a human in the loop to make sure those AI [controls] are not going haywire and going toward AI hallucination, which is a real thing,” Gopalakrishnan said.

But in terms of the industry’s overall progress on implementing autonomous operations, he added: “We are absolutely there, and we are going to get there more and more.”