Perpustakaan Digital ITB

Where non linear aerodynamics sets a precedent in the governing flow, to which the transonic Mach regime is concerned, producing aeroelastic models under such conditions have taken its toll on the balance between computational cost and accuracy. This thesis investigates the utilization of the Support Vector Machine algorithm to produce a surrogate model/response surface based on estimations of the damping coefficients of aeroelastic models in order to predict the flutter boundary for a binary flutter system based on a NACA 64A010 airfoil. A single fidelity aeroelastic model, in which the aerodynamics is governed by the Euler equations and the aeroelastic equations of motion solved in the time domain, is studied across a parameter space. Whereby, the proposed space is bounded and varied in the Mach range of 0.7-0.9 and the Flutter Speed Index of a range 0.4-2.0, wherein, the transonic dip is successfully demonstrated. The Prony series based Matrix Pencil, as a system identification method, is used to estimate the damping coefficients, in which the inputs are the Pitch and Plunge responses of the aeroelastic system. With the damping estimates, a single Support Vector Classification along with two Support Vector Regression models, Epsilon-SVR and Least Squares-SVR (LS-SVR), were used to generate the response surfaces. With the regression models indicating promising results, two new sampling plans based on an Augemented Latin Hypercube of 75 samples was proposed with the aim to further reduce the computational cost. Results indicate that the Epsilon-SVR model with its satisfactory generalization capabilities produced the best predicitions for the flutter boundary, however, the results also suggests the difficulty posed when dealing with training and test distributions in the samples.