LSU researchers are using AI and machine learning to identify connections between continuously changing operating conditions and optimize production at BASF's Geismar chemical manufacturing plant. BASF, the world's largest chemical manufacturer, has a long-standing partnership with LSU related to workforce development, research, and diversity and inclusion programs.
– Image generated by AI. Published in the Fall 2022 issue of LSU's Working for Louisiana. The keyword was /imagine_chemical_tubes_pipelines_plant_interconnected.
BASF, the world's largest chemical manufacturer, is working with LSU chemical engineers to use artificial intelligence (AI) to better understand and predict the ups and downs of its production. This project strengthens LSU's ongoing partnership with BASF to develop new STEM talent across disciplines in Louisiana.
BASF's chemical manufacturing facility in Geismar, Louisiana, is one of the company's six largest integrated manufacturing sites across 80 countries. We supply products to a wide range of industries including agriculture, construction, energy and health. Chemicals such as solvents, amines, resins, adhesives, electronic grade chemicals, industrial gases, basic petrochemicals, and inorganic chemicals are produced at Geismar in approximately 30 interconnected production units. each containing its own subunits.
“Chemical manufacturing is complex,” says Kerr Wall, digitalization manager for BASF's Monomers division. “Operating conditions can change from minute to minute, and there is a large amount of data to collect. Big data presents an incredible opportunity to optimize processes and increase predictability to improve yields and utility usage. This increases energy efficiency and supports our global value of producing chemicals for a sustainable future.”
BASF is working with LSU to develop better data mining processes to organize data and more easily compare current operating conditions with historical data.
Wall and fellow digitalization colleague Eric Dixon have worked as production engineers in BASF's Intermediates Division since graduating from LSU in 2008, and LSU professor Jose Romagnoli has been working on the optimization of complex systems. I was reading about the research done on structuring and control. In particular, he uses AI and machine learning to derive new knowledge from disparate data.
“For us, it made perfect sense to work with LSU… Jose. [Romagnoli] has over 500 publications and a lifetime of experience in process control, machine learning, and other fields of understanding. We are trying to automatically determine the most important process and quality parameters in order to link current data to historical data and use this data as a predictive tool to enable process optimization. Masu. ”
car wall BASF Monomer Division Digitization Manager
Kerr Wall, an LSU Chemical Engineering graduate ('99), is the digitization manager for BASF's Geismar Monomer Division. He values the power of his AI and machine learning to create “soft sensors” that are fully data-driven inferences in places where physical sensors are impractical or impossible, such as chemical manufacturing plants. Masu.
Part of the project's goals are to devise optimized and automated workflows and develop what are called soft sensors, a machine learning term often used in the manufacturing industry.
“Soft sensors, for example, can tell you real-time quality parameters of a material just from the data, without having to run samples all day long in the lab,” Wall said. “Soft sensors help us estimate certain variables at any given time. This also helps us predict the final quality of the products we produce.”
Physical sensors are often unavailable or impractical in the extreme operating conditions of chemical manufacturing plants.
“Instead of waiting 12 hours for a lab sample to arrive or for the next shift to take a new sample, LSU gives you a way to predict what’s going on in your unit based on data and AI. ,” Wall said.
LSU researchers are using an unsupervised clustering approach to help BASF classify and label production data. Since the main goal of the project is to discover how a change in one production unit changes the operating conditions of other connected units, time is an important parameter.
“You can use flow rates. Materials going in and out of the plant,” Dixon said. “If one plant is operating at 50% capacity and a sister plant goes down, the feeding plant may have to reduce its rates by 20% until everything is resolved. By understanding how one event leads to another, we can make better decisions as events occur and evolve.”
“Unsupervised machine learning allows us to capture the unique behavior of a process and discover things we weren't necessarily looking for,” said Jose Romagnoli, LSU Cain Endowed Chair and Professor of Chemical Engineering. “Machine learning and AI give us better opportunities to optimize production.”
His colleague Xun Tang, assistant professor in the LSU Cain School of Chemical Engineering, is also involved in the project.
The unsupervised learning approach allows LSU's machine learning to “think outside the box” and discover production patterns and correlations that neither BASF nor LSU's chemical engineers would have thought of. Knowing that a change in one unit can force a change in other connected units is important for large companies like his BASF factory in Geismar, south of Rouge. Helps optimize operations of large integrated plants.
“BASF has a lot of production data, but understanding the underlying dynamics can be difficult,” Tan said. “Through this joint project, we can learn directly from the data to identify patterns. What we learn can then be used to predict new situations and optimize and maintain the operation of our systems. This will help BASF automate and optimize its plants, increasing product yield and quality while reducing costs.”
Prior to this partnership with BASF, Romagnoli and his team conducted a similar project with oil and gas company ExxonMobil. The first LSU student to work on this project and graduate with a Ph.D. is Gregory Robertson, who currently leads his application engineering in ExxonMobil's Automation and Innovation Division in Baton Rouge.
“Developing data-driven techniques for fault detection and diagnosis requires a significant amount of knowledge,” says Robertson. “In most cases, sensors are installed on critical variables to protect against abnormal events. However, in cases where fault conditions are difficult to define, the data-driven technology LSU is developing can help. It’s useful in your toolkit.”
BASF's long-standing partnership with LSU includes career and employment programs for students (scholarships, internships, mentoring, job shadowing, senior projects, etc.), campus-wide sustainability initiatives including the BASF Sustainable Living Lab, and research. collaboration, and diversity and inclusion programs in STEM fields to engage women, minorities, and veterans.
Kerr Wall and Eric Dixon are both long-time chemical engineering graduates from LSU (1999 and 2008, respectively). Mr. Wall is also currently a visiting professor at LSU. He has a PhD in bioinformatics and is currently conducting research on aging with Assistant Professor Alyssa Johnson in the LSU School of Biological Sciences.
