Two engineering professors from West Virginia University have been awarded more than $100,000 from American Chemical Society’s Petroleum Research Fund to demonstrate a new method for building fast models of complex chemical reactions and processes.

David Mebane, assistant professor of mechanical and aerospace engineering, and Fernando Lima, assistant professor of chemical engineering, created dynamic discrepancy reduced modeling, which produces a set of fast models that may be useful in smoothing out the effects of disturbances on chemical processes.

The award will allow Mebane and Lima to take the project to the next phase: demonstrating it on Fischer-Tropsch, a large complex process that creates liquid fuels from gasified coal, biomass and natural gas.

“Reduced-order modeling methods like DDRM are important for building models over multiple length and time scales because of the need to encapsulate information about complex phenomena in a simplified form,” Mebane said. “Advantages of DDRM over other reduced-order models are its expressly dynamic character, which makes it better for modeling chemical reactions, and its ability to quantify uncertainty, which is useful for propagating error across model scales.”

Multi-scale models are particularly important in chemical engineering, Mebane said, where large-scale, industrial processes are designed around the properties of chemical reactions best described at the scales of atoms and sub-atomic particles.

“Control of chemical processes is important in chemical engineering because they usually have a relatively small window of optimum operation,” Mebane said. “Random disturbances can push the process out of this optimum window.

“Some controllers use a model of the process in order to make better, more effective corrections, but these models have to be very fast to evaluate because of the real-time nature of process control. Approximations are commonplace.”

DDRM is an approximation that is potentially better than others in this situation because the models it produces are guaranteed to be consistent with at least a limited set of process data. The multiple-model aspect of DDRM allows it to “learn” from the process in real-time, by swapping out models and retaining only those that are consistent with the real process behavior.

The two-year, $110,000 award comes from the Fund’s Doctoral New Investigator program, which provides grants for young faculty in the first three years of their tenure-track appointments.