German Aerospace Center (DLR) AEROGUST M30 Progress Meeting 23-24 November 2017, Bordeaux Presented by P. Bekemeryer / J. Nitzsche With contributions of C. Kaiser 1, S. Görtz 2, R. Heinrich 2, J. Nitzsche 1, M. Ripepi 2, and M. Widhalm 2 1 Institute of Aeroelasticity, Göttingen, Germany 2 Institute of Aerodynamics and Flow Technology, Braunschweig, Germany
Contents WP3 Task 3.2 DLM Updating with CFD Data WP4 Task 4.1.2 - Nonlinear unsteady LSQ-ROM approach
Computational Fluid Dynamics (CFD) Doublet-Lattice-Method (DLM)
Switch to NASA CRM for D3.3
linear vs. nonlinear CFD Prediction of gust loads with fast, time-linearized methods retaining the fidelity of RANS Application of an aeroelastic reduced order model (ROM) Validation of the time-linearized methods by comparison to nonlinearly obtained gust responses with small amplitudes Linearized aeroelastic gust response of a transport aircraft configuration
Gust load Possible scenario Linearized CFD / Corrected DLM (Step 1) DLM Nonlinear CFD Gust amplitude
Gust load Possible scenario Linearized CFD / Corrected DLM (Step 1) DLM DLM generally agrees quite well with CFD Nonlinear CFD Gust amplitude
Gust load Possible scenario Linearized CFD / Corrected DLM (Step 1) DLM Nonlinear ROM (Step 2) Nonlinear CFD Gust amplitude
12M Meeting Paris: LFD gust extension CREAM improvements
Linear aerodynamic transfer function: f ω = A ω q gust (ω) Aerodynamic forces are projected onto the generalized coordinates yielding the generalized aerodynamic forces (GAF) Application of the inverse Fourier Transform to obtain the linearized time-domain results Interpolation in the frequency-domain here: 33 samples
NASA Common Research Model (CRM) hybrid unstructured/structured CFD grid half model in flight shape Parameter Value Number of grid nodes 3.7 10 6 Surface nodes 10 5 Mean aerodynamic chord 7m Transonic flight state Parameter Value Mach number 0.86 Angle of attack 1.641 Free-stream velocity 260.7m/s Reynolds number 56.3 10 6
Gust encounter of small amplitude 1-cos gusts as defined by CS-25 but with only 5% of the amplitude Gust signal at the aircraft's nose
Fixed Aircraft! Aero only! No fluidstructure coupling yet! 213.36m 91.44m
Fixed Aircraft! Aero only! No fluidstructure coupling yet! 213.36m
Fixed Aircraft! Aero only! No fluidstructure coupling yet! 213.36m Transfer function of gust forces real part imaginary part
Linear coupling by aeroelastic feedback loop in terms of generalized coordinates in the frequency domain: Linear aeroelastic transfer function: f ω = I A ω S ω 1 A ω q gust (ω) Linearized time domain results by applying the inverse Fourier Transform
condensed dynamic model (FERMAT model) here: reduction to 3 degrees of freedom two rigid body modes first flexible mode splined onto the aerodynamic surface Parameter Value Total mass 2.6 10 5 kg Moment of inertia 2.58 10 6 kgm 2 Eigenfrequency 1.057Hz Heave Mode Pitch Mode BendingMode
Elastic European Aircraft! Union Fluid-structure coupled solution! forces 91.44m motion
Elastic European Aircraft! Union Fluid-structure coupled solution! forces 213.36m motion
Conclusion Time-linearized aerodynamic and aeroelastic ROM presented Validation for gust encounter of a transport aircraft (CRM) with the LFD Quasi-steady CFD-Corrected DLM shows promising aerodynamic results, but for aeroelastic applications a correction with single pitching motion may not be sufficient A good aerodynamic gust response for fixed aircraft is not sufficient for predicting the aeroelastic response A perfect time-linearized ROM can not predict the nonlinear gust response
Next steps - Test/Develop methodology to include unsteady CFD/LFD training data (towards D3.7)
WP 4 - Adapting the loads process for nonlinear and innovative structures (lead by UCT) Task T4.1 New ROM developments for gusts (by INRIA, DLR, UNIVBRIS, Optimad and NUMECA) Subtask T4.1.2 Extension of Linear Frequency Domain (LFD) solver, ROM of LFD solver and nonlinear unsteady leastsquares ROM approach to gusts, and comparison of the three methods. 1) Extension of LFD method for gusts (already done) 2) Nonlinear unsteady least-squares (LSQ) ROM approach for gusts 3) Gust simulations with TAU full-order model (reference) 4) Comparison of the three methods WP 4
M24 - outcomes Nonlinear LSQ ROM predictions Correction of a bug in the accelerated greedy MPE procedure Set-up of the LSQ-ROM approach for the application to discrete gusts Nonlinear high-fidelity full order model with TAU FFAST Crank Airfoil 2D test case gust responses for different gust lenghts and amplitudes, with the disturbance velocity approach (DVA) M30 - outcomes Extension of LFD method for Gust Implemented an analytical derivation of the right-hand side forcing term Nonlinear LSQ ROM predictions Fixed an issue with the reconstruction of the training signal for discrete gusts WP 4 Nonlinear high-fidelity full order model with TAU Adjusted the investigated case to FFAST Crank Airfoil 2D test case D Publications Submitted an abstract to the AIAA Aviation 2018 for the AeroGust special session
2D Test Case: FFAST crank airfoil Gust condition Half Gust Length (ft) Amplitude (ft/s) (EAS) 30 37.18 50 40.49 70 42.82 100 45.45 150 48.62 200 51.01 250 52.95 300 54.58 350 56.00 Load alleviation factor Fg = 1.0 Case Altitude (ft) Flow condition Mach number Amplitude scaling D 35000 0.754 0.74662 I 35000 0.86 0.74662 air density: ρ = 0.37968 kg/m 3 Reference chord: c ref = 8 m Aerodynamic model only Disturbance velocity approach (DVA) Gust response for: AoA = 0.0 deg (Case I) AoA = 1.4 deg (Case D) V Moment Lift Weight Drag WP 4
T4.1.2 - Nonlinear unsteady LSQ-ROM approach WP 4
Building the POD subspace for discrete gusts Collecting snapshots coming from the all/some (clustering) gust unsteady simulations Gust perturbations disturbance velocity approach, rigid aircraft, no motion Gust Amplitude Ag Training input (Lg, Ag) X Training output C L (Lg, Ag) 1 (Lg, Ag) 2 (Lg, Ag) 3 time ONLINE ROM prediction Gust Length Lg OFFLINE w t i w t j Global POD (+ MPE) (Lg, Ag) X Flow field output time history Flow field output time history! WP 4
2D Test Case: FFAST crank airfoil Grid (62272 nodes) Case I Steady state @ Mach 0.86, AoA 0.0 deg WP 4
Nonlinear LSQ-ROM Results for Gust Validation of the LSQ-ROM for discrete gusts: Using the snapshots of a selected gust (here the H=30 ft one), build the POD-ROM and reproducing the same output Open Questions: What is the optimal training signal? How will the MPE affect the accuracy of the ROM? What is the decrease in computational cost? Is this increase in accuracy as well as cost really necessary when considering a full aircraft (comparison LSQ-ROM to LFD)? WP 4
Next steps Use unsteady least-squares ROM approach for predicting discrete gusts responses Compute gust simulations for the NASA CRM full aircraft and apply the nonlinear LSQ-ROM approach Include the greedy MPE selection in the ROM prediction Recompute the ROM predictions using Numpy/Scipy with Intel Math Kernel Library (MKL) compiler D4.2 Report on comparison of gust ROMs with gusts computed with Linear Frequency Domain solver and TAU (M33) WP 4