Research Highlight: Kivanc Ekici
Dr. Kivanc Ekici, an associate professor in UT’s Department of Mechanical, Aerospace, and Biomedical Engineering, is intent on making turbines less prone to failure and capable of wringing ever more energy from the wind. He believes new designs for turbine blades hold the key to those gains.
A wind turbine idled by a blade failure can quickly become a $3- to $4-million white elephant. Cumulative annual losses across the nation’s entire wind-power network due to unplanned shutdowns measure in the billions of watts. Today’s high-speed computers have dramatically eased the design process through a discipline known as CFD, or computational fluid dynamics. CFD allows Ekici and other engineers to create virtual blades that display aerodynamic properties, chiefly lift and drag, and that also vibrate and deform as the wind flows around them.
Traditional CFD techniques calculate, in tiny time increments, aerodynamic effects of wind turbulence at thousands of test points on the computerized grid representing the turbine blade. These techniques can be highly accurate in evaluating the effects of design changes on blade performance, but they also come at “extremely high computational costs,” according to Ekici.
For instance, using a time-accurate CFD technique to model the effects of structural fatigue on a blade over two months of operation might take two months of processing time, making the technique impractical for most applications. To avoid this dilemma, Ekici is developing a new suite of software tools that may result in a 10- to 100-fold reduction in the time required to optimize blade design. Instead of solving the equations that model the fluid flow in the time domain, Ekici’s tools focus on solution techniques in the frequency domain, allowing measurement over much larger time increments.
“If we’re considering some kind of vibration of the blade, the frequency would correspond to that of the vibration,” Ekici explains. Data produced by these calculations will lead to better blade designs at a fraction of the computational time.