AI Transforms Chemical Industry’s Sustainability Practices
Key Takeaways
- Chemical companies use production planning software and AI to balance operational excellence, competitiveness and sustainability goals.
- Look for solutions that embed domain expertise in their models to provide trustworthy results within real-world constraints like chemistry and physics.
- The industry's varying levels of digital maturity require tailored solutions, from basic data historians to advanced AI/ML systems that integrate multiple data sources.
The chemical industry is driving some of the largest sustainability and net-zero initiatives in the world.
While the conversation around sustainability continues to shift, particularly given the current geopolitical and financial climate, few companies are relaxing their goals, noted Paige Morse, enterprise director of sustainability at Emerson’s Aspen Technology business in Washington, D.C.
“Rather, most chemical companies are instead focusing on how to enable operational excellence and competitiveness alongside them,” she said.
At the heart of these efforts are production planning software and AI that can help them better visualize and understand how operations impact profitability and emissions.
“Armed with this information, planners and plant operators have the insight for operational decisions based on local regulations, corporate goals, profits and other criteria. So chemical companies can better evaluate and prioritize their options based on variables such as feedstock quality or regulatory constraints and decide where to replace or build new assets to maximize operational excellence and lessen the environmental impact,” Morse added.
AspenTech Industrial AI is another product that claims it can bolster operations and sustainability goals.
“It directly addresses concerns our customers face around the transparency and explainability of ‘regular’ AI results, instead delivering AI embedded in models based on domain expertise,” said Heiko Claussen, chief technologist at Emerson’s Aspen Technology business, Bedford, Massachusetts.
“This establishes guardrails that address challenges of AI, thus providing robust and trustworthy results,” he added.
One application of Industrial AI is to close the simulation reality gap between a simulated and real asset using data from the field.
“This gap results from approximations in the model, deviations in construction or aging of the asset. By improving model accuracy, it is possible to further optimize efficiency, which is a key driver for both competitiveness and reaching environmental goals,” noted Claussen.
Similar effects can be achieved by modeling complex or custom processes that defy traditional strategies. “Significant efficiency gains can be achieved by optimizing these processes. In addition, guidance features help identify bottlenecks or unnecessary constraints in complex process models, thus driving efficiency, accelerating time to value and lowering the bar for new talent,” he added.
Having AI applications work with the framework of engineering fundamentals and domain knowledge provides specific guardrails for AI algorithms to keep results within real-world constraints such as chemistry, physics, electrical engineering and asset operating norms — enforcing material balances across a chemical process, for example — and stay within process safety envelopes.
This approach is important, emphasized Claussen, because it lets chemical companies benefit from AI’s agility, guidance and automation capabilities without the worry that it could accidentally propose infeasible solutions or incorrect recommendations that could be unsafe or cause environmental problems, such as increased emissions.
How AI Tools Benefit Real-World Projects
He cited two projects that have benefited from AI-based tools.
In one, a large European chemical producer needed to upgrade its advanced process control (APC) system to reduce energy use and increase yield predictability in ethylene oxide production. The company used AspenTech Industrial AI-based tools built into its DMC3 APC to rebuild, calibrate and tune the controllers.
Claussen said the upgrade to a new, more capable APC took under half the expected time, required no specialized technical experts and achieved an ROI in less than six months.
Continuing this approach, the company can further automate the control and adjustment of controllers with load and condition changes, reducing operator intervention and improving operational efficiency.
The second example is from a global chemical company using AspenTech Industrial AI-based hybrid models to profile the reactor unit performance for one of its high-value chemicals. The reactor always had problems with quality and yield. According to Claussen, the company is now leveraging process history data to build nonlinear relationships that accurately predict process performance. The result has been better prediction of actual performance and a 10% yield improvement for this process.
However, Claussen emphasized that whatever the scale of the enterprise, AI should be directly tied to business objectives to ensure maximum value, even with a limited budget. “Prioritize AI that will drive measurable improvements, such as improving efficiency, enhancing safety or meeting sustainability goals, rather than simply offering theoretical benefits,” he said.
SMBs might prefer hybrid models, where complex processes can be represented, calibrated and optimized without detailed modelling. They could also benefit from guidance features such as Aspen Virtual Advisor (AVA) for the DMC3 APC, which helps to identify opportunities for process optimizations. (Figure 1).
“Chemical processing companies run complicated, mission-critical operations. When choosing an AI partner, it is especially important to have trust in the partner’s domain expertise, depth of asset knowledge, industry knowledge, etc. This greatly improves the chances of AI implementations producing reliable results and meeting business objectives,” said Claussen.
“While there is no one-size-fits-all strategy to ensure a competitive edge while meeting environmental impact goals, digital technologies are playing a significant role in ensuring companies are armed with the most comprehensive data so they can choose the best path forward,” concluded Morse.
Addressing Differing Digital Maturities
Stephen Reynolds, global industry principal for chemicals for AVEVA, Chicago, also emphasizes that the chemical industry’s wide spectrum of digital maturity means that solutions offered by vendors must be tailored to each application.
“We have to find out where a customer is on their digital journey; what they are looking to achieve. It’s always a case of finding the right tool for the right solution,” he said.
A chemical engineer who has used digital solutions for over 20 years, he believes that the key is to use one that can contextualize data and enable analytics within the data historian.
“With AVEVA’s PI System we can group data by plant, by unit, by equipment. We can run reports based on just emissions or maintenance assets or process batches, for example, depending on how we organize our database,” he explained.
Another is the ability to create alerts and distribute them to the right people.
Reynolds illustrated the importance of this with the example of a plant he worked at where process temperature was a key variable. Thermocouples would fail but not be addressed for days or sometimes even weeks.
“We used PI to detect these failures and send an immediate notification to our maintenance system for replacements. So, it would go straight on the books without requiring human intervention. That kind of data integration across different applications is what’s important today, plus its communication, notification and visualization in the same place through the same platform,” he added.
While the company believes that cloud and AI technologies are the future, a host of tools are available prior to that level.
“To each tool its best purpose, and they build on each other. This is important because not every customer is racing to the cloud or has the necessary data to use AI yet,” he added.
Coming up as a chemical engineer, Reynolds would pull all his data into Microsoft Excel, look for a trend and go and do his job. “Many customers still operate with that level of sophistication. We still talk to customers who have a DCS or HMI SCADA system in their plant, but no data historian and no way of visualizing outside of the control system,” Reynolds noted.
The first step is moving away from those spreadsheets by automating some of the watchdogs that all companies keep track of — KPIs, for example. “That’s all done with PI, so think of it as foundational condition monitoring — whether it’s process monitoring, equipment monitoring or environmental monitoring, we can get that first line of defense and troubleshoot the issues and report out at that stage.”
Next is building simulations based on first-principle models of the process, which will show where a plant should be operating rather than where it is operating.
Then you get into the AI, machine learning (ML) stage where you are combining all this data with non-realtime information — perhaps bringing in lab analytics, weather data, cost and performance information from an ERP system.
“Now you’re building a data warehouse or data lake out of a big hodgepodge of data, which AI and ML can process quickly. We can interrogate it, flex it a little and come up with that final 20%: the cherry on top of the sundae. That’s the AI bit. Today, we would use it for that last little piece because we’ve exhausted all other tools getting there,” he explained.
For example, Covestro has been using PI System and AVEVA’s Process Simulation tool to investigate process limits and energy targets across multiple facilities as part of a long-term project. As process understanding has improved, Covestro has reduced the energy footprint of these plants by 30%.
Then there’s Solvay’s chemical plant in São Paulo, Brazil, which wanted to install real-time equipment performance monitoring.
The PI System was the foundation for this, while custom hybrid AI models calculated performance indicators and identified failures in 173 pieces of equipment, including pumps, centrifugal compressors, liquid ring compressors, fans, a cooling tower and a turbo generator.
PI System tools were used to model the equipment variables, build KPIs, configure notifications, notify the maintenance team about potential abnormal conditions and present the information through user-friendly dashboards. The company also noted a better integration between maintenance, operations and engineering teams.
“Solvay also has had some success using PI System as a demonstration of compliance, so it was a key element of their environmental/emission management program and reporting that they submitted to their local environment agency,” said Reynolds.
Another customer is looking to modularize green ammonia plants.
“Basically, they are trying to put an ammonia plant in a freight container and use our simulation engine to scale down their ammonia process to fit. All these new processes that are coming out of R&D can use this modelling to scale up and move forward from pilot testing to commercial-scale production,” said Reynolds.
“Our bread and butter continues to be optimization and operability of plants — being able to balance production and cost versus emissions and carbon capture,” he added.
“I think in the future that’ll start pushing down levels as the AI capability gets better and better. Then, generative AI can start thinking ahead about what we are trying to achieve. As that gets more powerful, some of these other tools might get supplanted. We’re not there yet,” Reynolds said.
What is nearer at hand, however, is the need for chemical companies to look at data across the boundaries of their whole business — integrating a carbon value chain, for example.
“The data community is coming together to securely share specific data sets with your customers, your suppliers, your process operators etc. It’s an interesting opportunity, particularly as companies try to tackle Scope 3 emissions. That’s a really difficult, nebulous kind of thing to try and track,” he concluded.
Probing the Product Portfolio
Siemens Digital Industries Software, Plano, Texas, also is incorporating more AI tools in its products. One plan, for example, is to use generative AI as a co-pilot to assist customers throughout the product portfolio.
Toby Hallitt, pre-sales solutions consultant, also urges customers to understand the importance of training in these products.
Siemens gives all customers access to a self-paced learning platform with bespoke training modules designed to get them quickly up to speed with all the company’s digital solutions. Specialized support is available from dedicated subject matter experts.
“Overall, customers should prioritize solutions which are critical to the company’s business objectives. However, adopting a fully digital approach by incorporating different portfolio elements early on may yield long-term benefits,” he concluded. ⊕
About the Author
Seán Ottewell
Editor-at-Large
Seán Crevan Ottewell is Chemical Processing's Editor-at-Large. Seán earned his bachelor's of science degree in biochemistry at the University of Warwick and his master's in radiation biochemistry at the University of London. He served as Science Officer with the UK Department of Environment’s Chernobyl Monitoring Unit’s Food Science Radiation Unit, London. His editorial background includes assistant editor, news editor and then editor of The Chemical Engineer, the Institution of Chemical Engineers’ twice monthly technical journal. Prior to joining Chemical Processing in 2012 he was editor of European Chemical Engineer, European Process Engineer, International Power Engineer, and European Laboratory Scientist, with Setform Limited, London.
He is based in East Mayo, Republic of Ireland, where he and his wife Suzi (a maths, biology and chemistry teacher) host guests from all over the world at their holiday cottage in East Mayo.