1 of 40

VISION

Nonlinear Transonic Aeroelastic Analysis of a Cropped Delta Wing using Fluid-Structure Interaction and Dynamic Mode Decomposition

Langa Sheila C. F. & Takahashi Yusuke

Space Transportation System Lab.

Hokkaido University, Japan

CONFERENCE

24/10/2023

2 of 40

2

Outline

Background & Objective

1

2

FSI Model Analysis & Setup

3

4

Results & Comparison

Dynamic Mode Decomposition

3 of 40

3

2

Background & Objective

4 of 40

4

Computational

Fluid Dynamics

CFD

Computational Structural Dynamics

CSD

Fluid-Structure

Interaction

FSI

Fluid-Structure Interaction

Aeroelasticity

Structural vibrations

Elastic forces

(Solid Mechanics)

Aerodynamic forces

(Fluid Mechanics)

Dynamic Aeroelasticity

Stability and control

Static

Aeroelasticity

Inertial forces

(Dynamics)

Dowell, E. H., A Modern Course in Aeroelasticity, 6th edition, 2022

Wright, Jan R., Cooper, Jonathan E., Introduction to Aircraft Aerolasticity and Loads, 2nd Edition, John Wiley and Sons, 2015

5 of 40

5

Transonic Flutter

Self-excited

Vibration

Failure

Transonic Dip

Nonlinearities

Nonuniformities

Limit-Cycle Oscillation(LCO)

FSI

Bendiksen, Oddvar O., Review of Unsteady Transonic Aerodynamics: Theory and Applications, Progress in Aerospace Series, 47(2):135-167,2011

Dowell, E. H., A Modern Course in Aeroelasticity, 6th edition, 2022

https://www.youtube.com/watch?v=qpJBvQXQC2M&t=55s

6 of 40

6

Research Motivation

Most commercial aircraft fly in the transonic region, thus research on the range is essential

The primary sources of nonlinearities that cause the LCO are a current topic of debate

Recent configurations call for high flexibility and maneuverability and may be sensible to nonlinearities

7 of 40

7

Research Objective

to analyze transonic flutter and LCO through a fluid-structure interaction (FSI) model using open-source software (OSS)-based coupling.

Main Goal

by selecting wing models prone to flutter and/or LCO mechanism and ensuring FSI model validity while comparing to experimental data available.

Procedure

Why OSS

Limited multi-physics software available, high degree of freedom, flexibility, community, and others.

Objective

To contribute to research of transonic Flutter/LCO occurrence and mitigation mechanisms while discussing the merits or limitations of the proposed FSI model in capturing accurate transonic aeroelastic behavior.

8 of 40

8

2

FSI Model Analysis and Setup

9 of 40

9

Coupling Library

Communication

CFD/FVM Solver

CSD/FEM Solver

Data-mapping

Coupling Schemes

Time-interpolation

Bungartz, HJ., et al., preCICE – A fully parallel library for multi-physics surface coupling, 2016

10 of 40

10

Coupling Method

Aeroelastic Problem

Aeroelastic Solution

Fluid

Problem

Structural

Problem

Structural

Solution

Fluid

Solution

CalculiX

SU2

yield

preCICE

adapters

Bungartz, HJ., et al., preCICE – A fully parallel library for multi-physics surface coupling, 2016

11 of 40

11

FSI Analysis Setup – OSS

Calculix

Governing equation: Principle of Virtual Work

Time integration: alpha method

Strain tensor: Cauchy’s method

Deformation type: Nonlinear

Time step: 5E-4

SU2

Governing equation: Compressible NS

Time integration: Dual time stepping

Linear solver: FGMRES

Pre-conditioner: LU-SGS

Advection scheme: JST

Time-step: 5E-4

preCICE

Mapping: Nearest-neighbor

Scheme: parallel-implicit

Coupling type: Partitioned

Iterations: 50

Time-step: 5E-4

Structure Solver

Fluid Solver

Coupling Library

V2.15 & 2.20

V6.0.0 & 7.5.0

V1.61 & 2.5.0

12 of 40

12

Experiment by Schairer et al.

Attar et al.

Peng et al.

0.000889 m

Cropped Delta Wing Model

Attar, P.J., Gordnier R.E., Aeroelastic prediction of the limit cycle oscillations of a cropped delta wing, Structural Dynamics & Materials Conference, 18-21 April 2005.

Cui, P., Han, J., Numerical Investigation of the effects of structural geometrics and material nonlinearities on limit-cycle oscillation of a cropped delta wing, 2011

  • The cropped delta wing was chosen to provide validation of the present analysis model through comparison with the experiment and other computed results.
  • While Attar and Peng use structured and unstructured meshes with both Euler and N-S, the current simulation uses fully-unstructured and Compressible N-S equations.

Schairer, E.T., Hand, L.A., Measurements of unsteady aeroelastic model deformation by stereo photogrammetry, Journal of Aircraft, 1999.

13 of 40

13

Solid Meshes - Overview

Coarse Mesh 122,626 nodes

Refined Mesh 270,762 nodes

14 of 40

14

Fluid Meshes - Overview

Coarse Mesh 682,546 cells

Refined Mesh 5,217,900 cells

Same Domain 4x2x2

15 of 40

15

Computational Conditions

Case

Exp.

q (psi)

Exp. Mach

Exp.

Re x10E6

Sutherland Re x10E6

Constant

Re x10E6

1

2.58

0.879

2.70

2.23

2.70

2

2.78

0.878

2.93

2.40

2.93

3

2.98

0.874

3.15

2.57

3.15

4

3.15

0.872

3.36

2.72

3.36

5

3.33

0.869

3.56

2.88

3.56

6

3.45

0.860

3.56

2.98

3.56

7

3.88

0.860

3.72

3.35

3.72

8

4.41

0.860

---

3.81

---

9

4.93

0.860

2.93

4.26

2.93

10

5.46

0.860

---

4.72

---

In order to match the experimental Reynolds number, SU2’s constant viscosity model =1.4E-5 was used.

in kPa: 20.54, 21.72, 22.96, 23.79, 26.75

Schairer, E.T., Hand, L.A., Measurements of unsteady aeroelastic model deformation by stereo photogrammetry, Journal of Aircraft, 1999.

Coarse

Refined

16 of 40

16

2

Results and Comparison

17 of 40

17

Eigenfrequency Analysis

Schairer, E.T., Hand, L.A., Measurements of unsteady aeroelastic model deformation by stereo photogrammetry, Journal of Aircraft, 1999.

Coarse Mesh

Refined Mesh

Coarse Mesh

Mode 1. 26.70 Hz

Mode 2. 89.25 Hz

Mode 3. 133.44 Hz

Experiment

Mode 1. 26.7 Hz

Mode 2. 88.2 Hz

Mode 3. 131.8 Hz

Refined Mesh

Mode 1. 26.70 Hz

Mode 2. 89.27 Hz

Mode 3. 133.47 Hz

18 of 40

18

Eigenfrequency Analysis

First Bending

First Torsion

Second Bending

Upper : Coarse Mesh

Lower : Refined Mesh

19 of 40

19

FSI Results – LCO and Flow Distribution

Refined mesh at 3.33 psi, 22.96 kPa

Coarse mesh at 3.33 psi, 22.96 kPa

Trailing edge

Leading edge

20 of 40

20

FSI Results at 3.33 psi – Mach and Cp

  • Difference in Reynolds number from about 2.88E6 and 3.56E6.

  • Leading motion of the wing is the bending motion.

  • We observe a vortex breakdown in the Refined model.

  • Fluid mesh refinement technique seems to play an important role on the FSI analysis.

Upper : Coarse Mesh

Lower : Refined Mesh

3.33 psi = 22.96 kPa

21 of 40

21

FSI Results – 2.98 - 3.33 psi

Cost for the Coarse Mesh Model: about 10 – 12 hours

Cost for the Refined Mesh Model: about 5-10 days

50.8 Hz

46.9 Hz

50.8 Hz

54.9 Hz

50.8 Hz

54.9 Hz

22 of 40

22

FSI Results - Comparison

Attar, P.J., Gordnier R.E., Aeroelastic prediction of the limit cycle oscillations of a cropped delta wing, Structural Dynamics & Materials Conference, 18-21 April 2005.

Cui, P., Han, J., Numerical Investigation of the effects of structural geometrics and material nonlinearities on limit-cycle oscillation of a cropped delta wing, 2011

Schairer, E.T., Hand, L.A., Measurements of unsteady aeroelastic model deformation by stereo photogrammetry, Journal of Aircraft, 1999.

  • Refined model results are closer to Attar’s simulation, especially in terms of displacement.
  • It is necessary to study the phenomena that maintains a constant frequency between 3.15 to 3.45 psi (21.72 to 23.79 kPa)

23 of 40

23

2

Dynamic Mode Decomposition

24 of 40

24

DMD – Parameters and Setup

Parameters

Value

Timestep

5.0E-4

Variable

DISP (Fluctuation)

Dynamic Pressure

2.98 – 3.33 psi

Used timesteps

0 - 999

Interval

1

Time range

0 – 5 s

Max Number of samples

1000

Mesh Case

Coarse Mesh

  • The DMD is performed on the coarse mesh case to understand if the frequency from the FSI analysis can be captured and if the range remains constant.
  • DMD analysis performed using Revun V0.80 written by Prof. Yusuke Takahashi.

25 of 40

25

DMD – Frequency oscillation

Mode

Frequency (Hz)

Coincidence

9, 10

50.75, -50.75

0.8942

15, 16

100,84, -100.84

0.9272

22, 23

49.05, -49.05

0.9641

20, 21

118.36, -118.36

0.9788

FSI Frequency at 2.98 is 50.8 Hz

FSI Frequency at 3.15 is 54.9 Hz

FSI Frequency at 3.33 is 54.9 Hz

  • The modes are detected using the Greedy Algorithm-based mode-sensing approach.

Mode

Frequency (Hz)

Coincidence

13, 14

54.75, -54.75

0. 6433

11, 12

109.22, -109.22

0.7201

18, 19

50.64, -50.64

0.8164

7, 8

163.93, -163.93

0.9964

Mode

Frequency (Hz)

Coincidence

16, 17

55.41, -55.41

0.7439

13, 14

110.68, -110.68

0.7870

18, 19

49.73, -49.73

0.9688

11, 12

166.04, -166.04

0.9843

26 of 40

26

DMD – Displacement and Reconstruction

Though DMD was able to capture the FSI Frequency ranges, it was not able to capture the LCO trend nor displacement.

27 of 40

27

DMD – Spatial Modes at 2.98 psi

Mode 10, F=50.75 Hz

Mode 20, F=118.36 Hz

Mode 22, F=49.05 Hz

FSI Frequency at 2.98 psi (20.54 kPa) is 50.8 Hz

Modes 10 & 20 resemble 1 and 3 from the eigenfrequency analysis : Bending.

Imaginary

Real

28 of 40

28

DMD – Spatial Modes at 3.15 psi

Mode 13, F=54.75 Hz

Mode 11, F=109.22 Hz

Mode 18, F=50.64 Hz

FSI Frequency at 3.15 psi (21.72 kPa) is 54.9 Hz

Modes 13 & 18 resemble 1 and 3 from the eigenfrequency analysis : Bending.

Imaginary

Real

29 of 40

29

DMD – Spatial Modes at 3.33 psi

Mode 16, F=55.41 Hz

Mode 13, F=110.68 Hz

Mode 18, F=49.73 Hz

Modes 16 & 18 resemble 1 and 3 from the eigenfrequency analysis : Bending.

FSI Frequency at 3.33 psi (22.96 kPa) is 54.9 Hz

Imaginary

Real

30 of 40

30

2

Summary & Future Work

31 of 40

31

Summary

  • An aeroelastic model for a cropped delta wing was analyzed and compared to experimental data as well as numerical studies

  • Fluid mesh refinement techniques proved to affect the LCO results more than the solid mesh and vortex breakdown around the leading edge was observed

  • The Refined mesh model using constant viscosity approach yielded displacement results that are close to Attar’s model

  • A DMD study was performed on the coarse mesh model. Bending motion was confirmed, and the DMD analysis showed a difference in frequency stagnation

32 of 40

32

Future Work

  • Studying the phenomena that maintains a constant frequency between 3.15 to 3.45 psi

  • With the recent change to the use of supercomputers, the simulation will be rerun with different iterations and timesteps

  • Understanding the impact of timesteps and total simulation time in the FSI analysis & Implementing different refined meshes for the fluid case to understand the vortex breakdown phenomena

  • Continue to analyze the displacement data through DMD and perform a DMD analysis for the Refined model as well.

33 of 40

33

2

Thank you for your attention!

Questions and/or Comments?

34 of 40

34

2

Ongoing Optimization Techniques

35 of 40

35

Optimization Techniques – Grid Study

Same Fluid mesh, different Solid mesh

Same Solid mesh, different Fluid mesh

  • Grid study confirmed that the fluid mesh has a greater impact on the LCO analysis.

Solid Grid Study

Fluid Grid Study

36 of 40

36

Optimization Techniques – Timestep

1E-4, LCO Frequency = 52 Hz

  • Changing the timestep size alone showed no difference in the LCO amplitude and a slight difference in the LCO Frequency.

5E-4, LCO Frequency = 51.9 Hz

5E-5, LCO Frequency = 52 Hz

37 of 40

37

Optimization Techniques – Total Time

LCO Frequency = 53.9 Hz

  • Changing the maximum simulation time slightly reduced the LCO Frequency. LCO occurrence is seen in 0.175s case though not as fully developed.

LCO Frequency = 51.9 Hz

LCO Frequency = 51.3 Hz

38 of 40

38

Optimization Techniques – Iterations

50 inner iterations, no LCO reached

100 inner iterations, LCO occurs

  • Inner Iteration study when applied to different timesteps revealed a difference in LCO occurrence.

39 of 40

39

Optimization Techniques – Refinement

40 of 40

40

Optimization Techniques – Reconstruction

Full reconstruction at 3.15 psi

Full reconstruction at 3.33 psi