AI tool may fix chemical plants' blueprints

Sierre Ternoey, an industrial engineering student at Northeastern University, was tasked with addressing a common problem faced by chemical engineers worldwide. Chemical plants, used to produce a wide range of products, are designed using complex software that creates a chemical flowsheet, or blueprint, detailing how the factory will operate. However, this software often generates reports that indicate when something has gone wrong, but fail to explain why.

Ternoey set out to develop an AI tool that could read these reports, diagnose the issue, and then modify the flowsheet to produce a working simulation without altering the original design choices made by the engineer. Her skills in problem-solving and hands-on tools, acquired through her first-year course at Northeastern, were key to her success.

In the best-case scenario, the AI agent successfully diagnosed and resolved 26 out of 30 test cases, demonstrating its effectiveness and ability to make accurate predictions without interfering with the engineer's original intentions. Ternoey's experience, she said, fulfilled the goals of the course, which aimed to provide students with both practical skills and the ability to tackle complex problems independently.

Born in the German-speaking region of Switzerland, Ternoey found herself adjusting to life in Aachen, Germany, during the colder months. Despite initial concerns about fitting in and performing well at her co-op, she quickly became part of a welcoming community, participating in salsa lessons and taekwondo training. Her supervisor, Jan Pyschik, was impressed by her determination and willingness to take on the project, which was initially deemed too uncertain for a student thesis.

Pyschik and Ternoey hope that her successful AI agent will eventually lead to a scientific paper and a practical tool for engineers facing challenges with chemical process simulation software.