Editor’s page: The expanding role of AI in process operations
| By Scott Jenkins, senior editor
Respondents to a Q4 2024 survey of industrial leaders conducted by the ARC Advisory Group (Dedham, Mass.; www.arcweb.com) ranked artificial intelligence (AI) first, by a relatively large margin, among the digital technologies that they expect to be the most impactful for industrial operations over the next five years. At the recent ARC Forum event in February in Orlando, Fla., there was indeed much discussion of how a range of AI tools can be deployed in industrial applications.
There seemed to be widespread recognition that effective AI tools require a foundation of structured and contextualized data on which to base AI applications. The idea of data as an enabler of AI was expressed, among others, by keynote speaker Nico Duursema, CEO of energy transition company Cerilon (Calgary, Alta.; www.cerilon.com). His company is currently building a greenfield gas-to-liquids facility in North Dakota. The new facility provides an opportunity to create a “born-digital” plant that focuses on “data centrality” and is enabled by machine-learning AI, Duursema explained.
Recognition also seemed to coalesce around the idea that getting the most out of industrial AI offerings and capabilities means developing specific AI tools for each skill required and for each use case. Causal AI, which focuses on cause-and-effect relationships in complex processes, could aid in root-cause analyses and improved predictive capabilities. Agentic AI refers to systems where multiple AI “agents,” each with a specialized capability, are trained on different data and utilize different AI techniques. These AI agents then work together toward a mutually agreed objective.
The use of generative AI was also a prominent topic at the event, with several examples of AI-powered natural language models (NLMs) being deployed as a new generation of user interface for plant personnel. An example of this comes from Celanese Corp. (Irving, Tex.; www.celanese.com). Celanese digital manager Ibrahim Al-Syed stressed the importance of human-centered design when talking about progress of AI systems at the company’s Clear Lake, Tex. manufacturing facility. Through a partnership with industrial software developer Cognite (Oslo, Norway; www.cognite.com), Celanese plant personnel are now able to use an NLM-based user interface to access plant and process data. Cognite is using an agent-based AI approach, where chat-based NLMs are connected to AI agents that are trained on a facility’s data and that are designed for specific tasks or roles, such as troubleshooting, simulation, and others.
Among the drivers for incorporating AI technologies are the workforce challenges facing the chemical process industries (CPI). It is here that generative AI may have an even deeper impact. As experienced workers retire, taking decades of knowledge with them, generative AI systems can serve to capture, organize and transfer this knowledge to newer workers in a readily available form. “Facilities are seeing diminishing experience overall,” says Chris Stogner, senior director of offer management at Schneider Electric (Boston, Mass.; www.se.com), “so we have to find ways to put AI to work.” Schneider is exploring how to engage AI in an area where it has not had much penetration: process safety. At the ARC event, Schneider announced a patent on the use of AI technology to generate process hazards analyses (PHAs) and layers-of-protection analyses (LOPAs) for industrial processes. Stogner says humans can introduce biases into these types of analyses, and use of AI is a way of interjecting objectivity back into the process of assessing hazards. The company is looking for industrial partners to co-innovate and test the capability, which involves AI agents tasked with identifying potential process dangers and solutions. “It’s essentially a dual AI agent, where one agent uses plant design information to search for possible ways to create problems and dangers in a process, while another agent attempts to alleviate those hazards,” Stogner explains.
Chemical Engineering
Editor’s page: The expanding role of AI in process operations
| By Scott Jenkins, senior editor
Respondents to a Q4 2024 survey of industrial leaders conducted by the ARC Advisory Group (Dedham, Mass.; www.arcweb.com) ranked artificial intelligence (AI) first, by a relatively large margin, among the digital technologies that they expect to be the most impactful for industrial operations over the next five years. At the recent ARC Forum event in February in Orlando, Fla., there was indeed much discussion of how a range of AI tools can be deployed in industrial applications.
There seemed to be widespread recognition that effective AI tools require a foundation of structured and contextualized data on which to base AI applications. The idea of data as an enabler of AI was expressed, among others, by keynote speaker Nico Duursema, CEO of energy transition company Cerilon (Calgary, Alta.; www.cerilon.com). His company is currently building a greenfield gas-to-liquids facility in North Dakota. The new facility provides an opportunity to create a “born-digital” plant that focuses on “data centrality” and is enabled by machine-learning AI, Duursema explained.
Recognition also seemed to coalesce around the idea that getting the most out of industrial AI offerings and capabilities means developing specific AI tools for each skill required and for each use case. Causal AI, which focuses on cause-and-effect relationships in complex processes, could aid in root-cause analyses and improved predictive capabilities. Agentic AI refers to systems where multiple AI “agents,” each with a specialized capability, are trained on different data and utilize different AI techniques. These AI agents then work together toward a mutually agreed objective.
The use of generative AI was also a prominent topic at the event, with several examples of AI-powered natural language models (NLMs) being deployed as a new generation of user interface for plant personnel. An example of this comes from Celanese Corp. (Irving, Tex.; www.celanese.com). Celanese digital manager Ibrahim Al-Syed stressed the importance of human-centered design when talking about progress of AI systems at the company’s Clear Lake, Tex. manufacturing facility. Through a partnership with industrial software developer Cognite (Oslo, Norway; www.cognite.com), Celanese plant personnel are now able to use an NLM-based user interface to access plant and process data. Cognite is using an agent-based AI approach, where chat-based NLMs are connected to AI agents that are trained on a facility’s data and that are designed for specific tasks or roles, such as troubleshooting, simulation, and others.
Among the drivers for incorporating AI technologies are the workforce challenges facing the chemical process industries (CPI). It is here that generative AI may have an even deeper impact. As experienced workers retire, taking decades of knowledge with them, generative AI systems can serve to capture, organize and transfer this knowledge to newer workers in a readily available form. “Facilities are seeing diminishing experience overall,” says Chris Stogner, senior director of offer management at Schneider Electric (Boston, Mass.; www.se.com), “so we have to find ways to put AI to work.” Schneider is exploring how to engage AI in an area where it has not had much penetration: process safety. At the ARC event, Schneider announced a patent on the use of AI technology to generate process hazards analyses (PHAs) and layers-of-protection analyses (LOPAs) for industrial processes. Stogner says humans can introduce biases into these types of analyses, and use of AI is a way of interjecting objectivity back into the process of assessing hazards. The company is looking for industrial partners to co-innovate and test the capability, which involves AI agents tasked with identifying potential process dangers and solutions. “It’s essentially a dual AI agent, where one agent uses plant design information to search for possible ways to create problems and dangers in a process, while another agent attempts to alleviate those hazards,” Stogner explains.
For more from the ARC Forum, see: https://www.chemengonline.com/first-open-process-automation-opa-control-system-at-a-commercial-manufacturing-facility/ . ■