Vous êtes sur la page 1sur 18

Low-cost numerical

methods for unsteady


flows
Kiran Ramesh

Features of unsteady flows

Trailing-edge wakes

Flow separation

loss in lift, pressure drag

Large added-mass

result in downwash on airfoil

dominant force in high-frequency motions

Leading-edge vortices

increase lift and cause variation in pitching


moment

Responsible for high-lift flight in insects


2

Overview of modelling approaches


Approach

PROS

CONS

Theoretical

fast, improve
understanding of
physics

limited by
assumptions such as
small-amplitudes,
attached flow

Experiment, CFD

reliable, improve
understanding

expensive/slow,
unsuitable for real-time
simulation and design

Semi-empirical

depend on several
fast, can be used for parameters, do not
simulation and design provide insight into
physics

My approach: Start with theoretical modelling - eliminate


assumptions, augment theory to account for viscous phenomena
3

Large-angle unsteady thinairfoil theory

Airfoil (body frame) starts at inertial


frame and moves along any path

Airfoil vorticity is represented as a


Fourier series

Discrete-time method: at each timestep, one vortex shed from trailing-edge

Assumptions: inviscid, incompressible


flow

Amplitudes, frequencies may be


arbitrarily large

Large-angle unsteady thinairfoil theory

Solves for the vorticity on airfoil and trailing-edge vortex

Time-stepping method: airfoil moves along any


arbitrary path

pitch

plunge

variation in freestream velocity

Typical simulation of 500 time-steps: 3 sec

Large amplitudes, nonplanar wakes, added mass are modelled

Flow separation and LEV formation are not modelled


5

Results from theory

Motion (grey line, right axis): pitch up to


25 deg, hold, pitch down to zero

High pitch rate: typical of MAVs, insect


flight

Theory fails after point D

Most important flow


feature : leading-edge
vortex

Trailing-edge separation
is negligible
Flow visualisation from experiment, CFD at four instants
6

Criterion for LEV formation

Suction results at leading edge when airfoil


is at an angle of attack, because of flow
turning around the leading edge

Hypothesis - Start of LEV formation can be correlated with the


leading-edge suction exceeding a critical value

Suction at leading-edge from thin-airfoil theory: Proportional to


first Fourier coefficient

Leading Edge Suction Parameter (LESP) =

A0 (t)

Verification of LESP
hypothesis

SD7003 airfoil, Re=100,000

Upper surface
LEV formation

Start of LEV formation is


determined from CFD

LEV formation always occurs


when LESP > critical LESP (0.14),
so long as airfoil and Reynolds
number are same
8

Lower surface
LEV formation

Control of LEV occurrence using LESP


50

0.2
Critical LESP (U.S.)

(Baseline)

SD7003, Re=100,000

40

0.1
LESP

30

Baseline motion: pitch-up to 45 deg,


LEV forms on airfoil upper surface

(deg)

Baseline

20
0.1

10
0

t*

0.2

Critical LESP (L.S.)

t*

By reverse-calculating suitable plunge,

50

No LEV formation

0.2
Critical LESP (U.S.)

40

30

1.5

LESP

(deg)

h/c

20

10

0.5

2.5

LEV formation on lower surface

t*

50

2.5

0.1

Baseline
Combined

0.1

0.2

Critical LESP (L.S.)

t*

0.2
Critical LESP (U.S.)

40

30

1.5

LESP

(deg)

h/c

20

10

0.5

t*

0.1

Baseline
Combined

0.1

0.2

Critical LESP (L.S.)

t*

LESP based discrete-vortex


method (LDVM)

Vortex is shed from the leading-edge


when LESP > critical LESP

Strength of vortex calculated


such that LESP = critical LESP

Typical run time for 500 time-steps: 25s

Salient feature: Models intermittent vortex shedding

Fortran code based on this model :

Available open-source at http://sourceforge.net/projects/unsflow

Outputs forces on airfoil and video of flow-visualisation as well as


fundamental quantities such as pressure distribution, velocity and vorticity
on airfoil

10

Illustration of LDVM
Motion (grey line): pitch up to 25 deg, hold, pitch down to zero; leading-edge pivot

Validation against CFD and experiment

LEV and vortex interactions are successfully modelled by LDVM


0.3

30
LESP

20

LESP

(deg)

0.1

30
Exp

0.2

0.1

LESPcrit

CFD

(deg)

10

10

10

30
CFD

20

0.4

Cd

(deg)

Cm

0.5

10

4
t*

11

10

30
CFD

0.8

LDVM

4
t*

1.5

0.5

20

Cl

t*

Exp

LDVM

LDVM

20
(deg)

0.4

10

0.8

10

4
t*

10

Illustration of LDVM
Motion (grey line): Smooth pitch-up to 90 deg
and pitch down to zero; trailing-edge pivot

Validation against CFD

3 instances of LEV formation


0.4

90
LESP

90

LESPcrit

CFD

0.3
LESP

75
(deg)

LDVM

75
(deg)

Cl

0.2

60

60

0.1

45

45

30

30

0.1

15

15

0.2

t*

t*

90
CFD

90

LDVM

CFD

75
(deg)

Cd

LDVM

0.5
Cm

75
(deg)

60

60

45

0.5

45

30

30

15

1.5

15

4
t*

t*

12

Perching manoeuvres
Motion : pitch-up to 45 deg, simultaneous deceleration to zero velocity
0.6

Reynolds number is not


constant - critical LESP
changes

Theory
CFD

Theory

1.8

CFD
0.4

1.2

Cl

0.2

0.6

Critical LESP is
calibrated against Re at
leading edge and used
in simulation
1

0
0.2

CFD

Calibration data from CFD


Spline fit

0.8
0.6
LESPcrit

0.4

LDVM

0.2
0

log10(ReLE)

13

t*

t*

Hovering maneuvers
Motion : plunge up and down, zero freestream velocity

Freestream velocity is zero

Theory
CFD

0.4

Induced velocity at leading


edge is however not zero
Critical LESP is calibrated
against Re at leading edge
and used in simulation

Cd
0

0.4

0.4

0.8

Calibration data from CFD


Spline fit

0.8
0.6
LESPcrit

0.4

LDVM

0.2
0

0.4

Cl

CFD

Theory
CFD

log10(ReLE)

14

t*

0.8

t*

Application to FSI - Flow


past a 2DOF airfoil

Airfoil constrained by springs

Degrees of freedom - pitch, plunge

LESP based discrete-vortex


method is coupled with the
structural model

If freestream velocity is greater than


flutter velocity, airfoil extracts
energy from the freestream

15

Self-sustained limit-cycle
oscillations

30

At speeds below flutter velocity:

(deg)

15

At speeds above flutter


velocity:

15
30

Divergent oscillations
bounded by LEVs

300

600
*
t

900

1200

300

600
t*

900

1200

30
15

Limit-cycle oscillations

(deg)

15

Structural and aerodynamic


parameters may be optimised
for small-scale power
generation

30

16

Parametric studies of LCOs

0.5
0

0.5

U*/U*

1.5

2.5

No LCOs

50

(deg)

1.5

40
30
20

0.5

U*/U*

Other parameters: Airfoil shape, centre of gravity, spring


stiffnesses, pivot location
Detailed studies are made possible by low-order model

17

1.5

Divergent

60

Multipeak LCOs

Divergent

Each data point is a


simulation for 1200
convective times 80,000 time steps

Multipeak LCOs

No LCOs

E.g., Variation in LCO


characteristics with
freestream velocity

2.5

Significant outcomes

Fast method for prediction and design - based on a theory,


augmented with numerical procedures (discrete vortex
shedding) to model nonlinearities

Excellent balance between fidelity and cost - Typical run time


on a normal PC is < 1 min

Handles any arbitrary motion kinematics; predicts forces and


flowfield while retaining fundamental quantities (unlike semiempirical methods)

18

Vous aimerez peut-être aussi