WVU researcher trains next-gen AI for chemicals industry to take a quantum leap forward

WVU researchers Yuhe Tian and Austin Braniff want to make the energy-hungry chemical manufacturing industry more sustainable, so they're developing artificial intelligence that will run on a lightning-fast quantum computing platform and have the capacity to innovate new, outside-the-box designs for manufacturing processes. (WVU Photo/Nathaniel Godwin)

A West Virginia University engineer is creating powerful, unconventional artificial intelligence tools that can reimagine the sustainability of chemical manufacturing. 

Story by Micaela Morrissette, WVU Research Communications
Photos by WVUToday

Benjamin M. Statler College of Engineering and Mineral Resources

Yuhe Tian said she believes innovations to current chemical manufacturing processes, including transitions focused on energy sustainability, can be driven by “quantum AI,” or machine learning on a cutting-edge quantum computer that uses subatomic particles to store information and solve problems. With $240,000 from the National Science Foundation, Tian is launching a two-year project that will harness quantum intelligence to innovate the design of environmentally friendly chemical plants.

“My team will take a first-of-its-kind approach to using quantum machine learning to identify novel, optimal, sustainable chemical process designs,” said Tian, assistant professor of chemical and biomedical engineering at the WVU Benjamin M. Statler College of Engineering and Mineral Resources. “We want to find out how the manufacture of chemicals like hydrogen and ammonia can be greener, primarily in terms of reducing carbon dioxide emissions and energy consumption. Instead of heavily relying on human expertise, our framework will systematically generate outside-the-box process design solutions.”

Austin Braniff, a chemical engineering doctoral student from Mineral Wells, will work with Tian on the fellowship project, which is supported by the EPSCoR RII Track-4 Research Fellows Program and hosted by the Cornell University AI for Science Institute.

According to the U.S. Energy Information Administration, bulk chemical production accounted for 33% of industrial energy consumption in 2020, making it the largest energy user in the domestic industrial sector, with resulting greenhouse gas emissions of 274 million metric tons.

“The global chemical market is highly competitive, so it’s imperative that manufacturers find cleaner but economically viable pathways to chemical production,” Tian said.

“Computer-aided process design” is a tool that screens possible process technologies and set-ups, evaluating trade-offs between various solutions and yielding significant cost savings for manufacturers. Recently, computer-aided process design has heavily relied on AI. However, because chemical manufacturing design is so complicated, involving multiple different large units working in combination — reactors, heat exchangers, separators, etc. — standard AI can take a long time to deliver solutions.

“The question is how to integrate emerging technologies for chemical manufacturing into existing processes in a way that is more sustainable but ensures the industry remains economically competitive,” Tian said. “That requires both the speed advantage of quantum computing and the intelligent discovery of AI.”

Her approach to leveraging quantum AI is based on a unique framework. Most computer-aided process design starts with the human user specifying a particular design or set of possible designs for evaluation and optimization. Tian’s “generalized modular representation framework,” on the other hand, doesn’t create designs by putting different pieces of equipment together in different configurations to form connected units. Instead, the “Legos” for the designs are physical laws and fundamental phenomena such as heat or mass transfer, Tian said. The AI and quantum computing algorithms will combine those physical phenomena in different orders and combinations to produce design solutions.

“When a process synthesis approach uses units — reactors, separators, mixers — as its basic building blocks, it is hindered from generating brand-new unit operations or integrating multiple tasks into a single unit. The design will also be limited by the engineers’ expertise and their pre-formed ideas,” Tian said. “However, when the building blocks are not units but phenomena like reaction, separation or mixing, then structural combinations can vary and we can discover unexpected, counterintuitive designs.”

By shifting that process-driven framework to a high-speed intelligent quantum computing space, Tian’s approach may help accelerate the discovery and development of process designs that are innovative, efficient and sustainable.

“We have promising new technologies that offer chemical manufacturers enhanced energy efficiency and environmental sustainability, but assessing how they would scale up for operations in a chemical plant and how to integrate them with existing plant processes and equipment is challenging,” she said. “Our mission is to drive systematic innovation by applying artificial intelligence and quantum computing to chemical production.”

---