Aircraft Stability Derivative Estimation from Finite Element Analysis
Abstract
Two processes for the computational determination of aerospace vehicle stability derivatives have been developed. The first process is a forced oscillation parameter identification technique. The second method decouples the position and velocity boundary condition to find the quasi-steady forces that result from each independently of the other. For the first method, the dc-chirp signal was used to excite the aircraft's motion for its ability to fully excite velocity and maintain small displacements. The ARMA model is used to fit the generated data because it does not assume any form for or neglect the unsteady terms, and allows the model to better form to the data through changing the model order. Once the model order and coefficients are found, simple summations and combinations of the model coefficients produces the stability derivatives of the aircraft. For the second method, the non-inertial boundary condition equations and was altered to eliminate the consistent boundary condition assumption and allow for the independent specification of position and velocity. The transpiration boundary condition equations did not assume consistent boundary conditions and as such did not require alteration. Stability derivatives were estimated by allowing the forces at various conditions to converge, and then approximating the derivative with a finite difference equation between the quasi-steady forces at different states. The results of the first method worked well to capture dominant terms, but was not robust for predicting lesser terms. This method also required a great deal of time to generate the data and find the best model size. The results of the second method were quite good. The decoupled boundary condition specification method compared well with theoretical, empirical, and experimental data over a range of Mach numbers and geometric complexities including result of the F-18 in the transonic regime. This method is much faster as a steady state solver can be used and a model is not required. By separating the effects of position and velocity, each can be predicted to a higher degree of accuracy more rapidly.
Collections
- OSU Theses [15752]