Acta Mechanica Sinica

https://doi.org/10.1007/s10409-018-0782-z

RESEARCH PAPER

Lagrangian-based numerical investigation of aerodynamic

performance of an oscillating foil
Mengjie Zhang1 · Qin Wu1,2 · Biao Huang1 · Guoyu Wang1

Received: 26 December 2017 / Accepted: 12 April 2018

© The Chinese Society of Theoretical and Applied Mechanics; Institute of Mechanics, Chinese Academy of Sciences and Springer-Verlag GmbH Germany,
part of Springer Nature 2018

Abstract
The dynamic stall problem for blades is related to the general performance of wind turbines, where a varying flow field is
introduced with a rapid change of the effective angle of attack (AOA). The objective of this work is to study the aerodynamic
performance of a sinusoidally oscillating NACA0012 airfoil. The coupled k−ω Menter’s shear stress transport (SST) tur-
bulence model and γ −Reθ transition model were used for turbulence closure. Lagrangian coherent structures (LCS) were
utilized to analyze the dynamic behavior of the flow structures. The computational results were supported by the experiments.
The results indicated that this numerical method can well describe the dynamic stall process. For the case with reduced
frequency K = 0.1, the lift and drag coefficients increase constantly with increasing angle prior to dynamic stall. When the
AOA reaches the stall angle, the lift and drag coefficients decline suddenly due to the interplay between the first leading- and
trailing-edge vortex. With further increase of the AOA, both the lift and drag coefficients experience a secondary rise and
fall process because of formation and shedding of the secondary vortex. The results also reveal that the dynamic behavior of
the flow structures can be effectively identified using the finite-time Lyapunov exponent (FTLE) field. The influence of the
reduced frequency on the flow structures and energy extraction efficiency in the dynamic stall process is further discussed.
When the reduced frequency increases, the dynamic stall is delayed and the total energy extraction efficiency is enhanced.
With K = 0.05, the amplitude of the dynamic coefficients fluctuates more significantly in the poststall process than in the
case of K = 0.1.

Keywords Oscillating foil · Dynamic stall · Lagrangian coherent structures · Computational fluid dynamics (CFD)

List of symbols t Time step (s)

μ Dynamic viscosity (kg/m·s) M(t) Torque (N·m)
ui Velocity in the i-th direction (m/s) T Time period of one cycle (s)
x Position (m) wp (t) Pitching angular velocity (1/s)
k Turbulent kinetic energy (m2 /s2 ) ρ Density (kg/m3 )
ω Specific turbulent dissipation rate (1/s)
Re Reynolds number
w Angular velocity (1/s)
δ Lyapunov exponent 1 Introduction
ωz z-vorticity (1/s)
Q Second invariant of velocity gradient tensor (1/s2 ) With the development of renewable energy sources, wind
energy has attracted much attention as a clean, rich, and
widely distributed resource [1–3]. As an important wind
B Qin Wu energy capture device, wind turbines have been widely
[email protected]
applied [4–6]. However, the dynamic stall problem is closely
1 School of Mechanical Engineering, Beijing Institute of related to the general performance of wind turbine systems
Technology, 100081 Beijing, China in a varying flow field, such as yawed operation, sheared
2 Department of Thermal Engineering, Tsinghua University, inflow, gust conditions, etc., which are accompanied by a
Beijing 100084, China rapid change of the effective angle of attack [7–11]. Hence, it

123
 M. Zhang, et al.

is of significant importance to understand the dynamic inter- ing motion had a negative effect on the harvesting effi-
actions between transient wing motion and unsteady flow ciency.
structures. Moreover, as one of the primary flow energy con- Although much work has been carried out on the aerody-
version technologies, improved understanding of dynamic namic performance of oscillating wind turbine blades [32–
stall characteristics will help development of oscillating wing 36], significant uncertainty still exists regarding the influence
systems to enhance the energy harvesting performance of of unsteady flow features. The effect of the dynamic stall
wind or hydro energy. phenomenon on energy harvesting still cannot be explained
Many experimental and numerical studies have been car- clearly, and the influence of the reduced frequency on the flow
ried out on the aerodynamic performance of oscillating foils evolution and energy harvester efficiency still requires fur-
[12–17]. Ferreira et al. [18] used particle image velocime- ther investigation. The objective of the work presented herein
try (PIV) to visualize the transient flow in the operational is to analyze the flow vortex structures around an oscillating
regime. The results illustrated that the flow pattern is depen- foil using a Lagrangian-based numerical method, and provide
dent not only on the magnitude of the angle of attack (AOA) further insight into the interplay between the unsteady flow,
but also on the transportation and interaction of the shedding oscillatory motion of the foil, and aerodynamic performance.
vorticity. Wernert et al. [19] used laser-sheet visualization The numerical models are described in Sect. 2, followed by a
and PIV to investigate the unsteady flow around a pitching summary of the numerical setup in Sect. 3. In Sect. 4, detailed
airfoil. They found that the dynamic stall process can be analysis of the aerodynamic performance and flow structures
categorized into four stages: (1) attached flow, (2) develop- is presented. Finally, conclusions are drawn in Sect. 5.
ment of the leading-edge vortex (LEV), (3) poststall vortex
shedding, (4) flow reattachment. Lee and Gerontakos [20]
applied smoke flow visualization and hot-film sensors to
investigate the transient flow structures and dynamic stall
2 Numerical model
characteristics of an oscillating NACA0012 airfoil. They
illustrated that the boundary-layer transition and dynamic
2.1 Basic governing equations
stall point were delayed with increase of the reduced fre-
The flow field was simulated by solving the unsteady
quency. This conclusion has also been proved experimentally
Reynolds-averaged Navier–Stokes (URANS) equations, with
by Carr [21] and Ekaterinaris and Platzer [22]. Simpson
the continuity and momentum equations listed below:
et al. [23] experimentally studied the unsteady flow around
a sinusoidal heaving and pitching foil, focusing on the influ-
ence of the maximum AOA, Strouhal number, and aspect ∂u j
= 0, (1)
ratio on the energy extraction. The results revealed hydro- ∂x j
dynamic efficiency of 43% ± 3% with maximum AOA    
∂u i ∂(u i u j ) ∂p ∂ ∂u i
of 34.37◦ , Strouhal number of 0.4, and aspect ratio of ρ + =− + μ , (2)
∂t ∂x j ∂ xi ∂x j ∂x j
7.9.
With the development of computing equipment and tech-
niques, much attention has been paid to computational fluid where ρ is the fluid density (all flow conditions in this study
dynamics to better investigate dynamic stall characteristics being incompressible), t is time, u is the velocity, x is the
[24–28]. Hang et al. [4] numerically investigated the per- coordinate, p is the pressure, μ is the fluid viscosity, and
formance of offshore floating vertical-axis wind turbines subscripts i and j denote the directions of the Cartesian coor-
subjected to pitch motion. The results indicated that the dinates.
power output of the turbines and the range of aerody-
namic force variations were enlarged. The transient flow 2.2 Turbulence model
of an oscillating hydrofoil at different pitching rates was
studied numerically by Huang et al. [7,29]. The results The simulation solved the URANS equations by applying the
revealed that the pitching velocity had an important effect revised k–ω SST turbulence model, which couples the k–ω
on the hydrodynamic characteristics. The energy harvest- shear stress transport (SST) turbulence model [37] and the
ing performance of a fully activated flapping foil under γ −Reθ transition model [38–40]:
wind gust conditions was studied by Chen et al. [30].
Compared with the uniform flow condition, the energy  
harvesting efficiency was higher. Energy extraction effi- ∂ (ρk) ∂ ρu j k
+ = P̃k − D̃k
ciency with pitching motion was also investigated by Teng ∂t ∂x j
 
et al. [31]; the results showed that nonsinusoidal pitch- ∂ ∂k
+ (μ + σk μt ) , (3)
∂x j ∂x j

123
Lagrangian-based numerical investigation of aerodynamic performance of an oscillating foil

+ ( ∂ x jj ) = Cω Pω − βω ρω2
∂(ρω) ∂ ρU ω
∂t 
+ ∂∂xi μ + μ t ∂ω
σk ∂ xi (4)
+ 2ρ (1 − F1 ) σω2 ω1 ∂∂kxi ∂ω
∂ xi ,

where k and ω are the turbulent kinetic energy and spe-

cific turbulent dissipation, respectively, F1 is the blending
function, and P̃k and D̃k are the revised production and
destruction term, respectively, defined as

P̃k = γeff Pk , (5)

D̃k = min[max(γeff , 0.1), 1.0]Dk , (6)

where Pk is the original production terms, Dk is the original

destruction term, and γeff is the revised coefficient.

Fig. 1 2D fluid mesh and boundary conditions: a simulation domain

2.3 Lagrangian coherent structures and boundary conditions, b mesh distribution

In contrast to the Eulerian approach, the Lagrangian approach

considers the fluid flow as a particle-based dynamic system 3 Numerical setup and description
[41]. To characterize the separation rate of infinitely close
trajectories, the Lyapunov exponent is defined as 3.1 Numerical setup

  In this study, the NACA0012 airfoil was adopted with chord

1 |δx (x0 , t)|
σ = lim ln , (7) length of c = 0.15 m. According to the experimental setup
t→∞ t |δx0 | [20], the computational domain and boundary conditions
|δx0 |→0
are presented in Fig. 1a, containing two domains connected
by a sliding interface. The rectangular static domain has
where t is time and x0 is an arbitrary point in the dynamical
length of 18c and height of 6c, while the circular dynamic
system. Based on the Cauchy–Green deformation tensor, the
domain has diameter of 4c. The interaction between the static
separation rate can be obtained as
and dynamic domains is controlled using a CFX expres-
sion language (CEL) subroutine. The foil was located 7c
 T from the inlet. The pitching motion of the foil was simu-
∂ x (t0 + TLE ; t0 , x0 ) ∂ x (t + TLE ; t0 , x0 )
ΔtT0LE (x0 ) = , lated by setting the pitching motion of the dynamic domain,
∂ x0 ∂ x0
i.e., the cylindrical region around the 1/4-chord of the foil.
(8)
The dynamic domain is moving as a whole, while the grids
of the dynamic domain remain invariant. The inlet velocity
where TLE is the time interval and x(t0 + TLE ; t0 , x0 ) is and outlet pressure are set, and nonslip wall conditions are
the new position of point x0 after TLE . Then, the finite-time applied to the upper and lower flow boundary, as well as the
Lyapunov exponent (FTLE) during the time interval TLE is foil surface. Figure 1b shows the mesh distribution as well
defined as as the refined grids around the foil. A total of 450 nodes are
placed in the boundary layer, selected to meet the criterion
1  y + = yu τ /ν ≈ 1 [37].
σtT0LE (x0 ) = ln λmax ΔtT0LE (x0 ) , (9) Figure 2 shows the features of the pitching motion. The
|TLE |
sinusoidal pitching motion is defined as α = 10+15 sin(wt),
 where α is the AOA and w is the angular velocity, with the
where λmax ΔtT0LE (x0 ) is the maximum eigenvalue of the axis located 0.25c from the leading edge. In this numerical
Cauchy–Green deformation tensor. Based on the FTLE field, simulation, the reduced frequency K = wc/(2U∞ ) was set
the Lagrangian coherent structures (LCS) can be obtained as 0.05 or 0.1, as shown in Fig. 2b. The mean free stream
from the ridges of the FTLE field, and it has been proven that velocity U∞ was 14 m/s, and the turbulence intensity was
this is useful to capture the vortex boundary [42,43]. set as 0.08%, corresponding to a Reynolds number of Re =

123
 M. Zhang, et al.

Fig. 3 Comparison of lift coefficient (Cl ) predicted using different grid

elements for reduced frequency K = 0.1, Re = 135,000, and U∞ =
14 m/s

process (α+ = − 5◦ − α+ = 25◦ ), the predicted lift coef-

ficient is approximately the same for all cases, increasing
approximately linearly with the AOA. In the downward rota-
tion process (α+ = 25◦ − α− = − 5◦ ), the numerical results
obtained using 2.2 × 105 and 5.0 × 105 gird elements are
approximately the same and agree well with the experiments.
Considering computational efficiency, 2.2 × 105 elements
Fig. 2 Features of pitching airfoil motion: a pitching airfoil, b sinu- were considered sufficient to yield grid independence.
soidal pitching motion
To further analyze the uncertainty of the solution, grid
convergence and numerical uncertainty were judged using
1.35 × 105 . For clarity, the upward stage of the oscillating the grid convergence index (GCI) method [44–49]. Accord-
cycle is denoted by α+ and the downward stage by α− . ing to Ref. [50], the representative grid size h, grid refinement
factor r , and apparent order p are defined as
3.2 Numerical verification
1/2
1 
N
Experimental data were obtained from Ref. [20],
hi = (Ak ) , i = 1, 2, 3, (10)
acquired using closely spaced multiple hot-film sensor N
k=1
arrays. In addition, the surface pressure distribution, thermal
ri j = h i /h j , i, j = 1, 2, 3, (11)
filament wake measurement, and smoke flow visualization  
1  (r21 −1)sgn (ε32 /ε21 ) 
p
technology were obtained to supplement the thermal film p= ln |ε32 /ε21 | + ln ,
data. According to Ref. [20], the hot line signal is sam- ln r21  (r32 − 1)sgn (ε32 /ε21 ) 
p

pled at 2 kHz. The surface pressure distribution is made up (12)

of 61 pressure joints, connected with seven quick-response
micropressure sensors and distributed on the upper and lower
surfaces of the model. With shutter speed of 1/1000 of a sec- where Ak represents the k-th element, N is the total element
ond, smoke flow visualization was carried out using a 60 Hz number, ε32 = f 3 − f 2 , ε21 = f 2 − f 1 , f i denotes the solution
camera. A potentiometer was used to measure the instanta- on the i-th grid, r21 = h 2 /h 1 , and r32 = h 3 /h 2 .
neous angle of attack of the wing with accuracy of 0.1◦ . The approximate and extrapolated relative error are
To obtain a grid- and time-independent solution, grid defined as
independence and temporal resolution validation were con-
 
firmed. Figure 3 shows the evolution of the lift coefficient  f1 − f2 
Cl = L/(0.5ρU∞ 2 sc) for the pitching foil with reduced fre- ea21 
= , (13)
f1 
quency K = 0.1, where L and s are the lift and the span  p 
 r21 f1 − f2 
length, respectively, obtained using three sets of grid config-  r p −1 − f 1 
 21 
urations with 9.1 × 104 , 2.2 × 105 , and 5.0 × 105 elements,
21
eext = . (14)
 r21p f1 − f2 
with the time step chosen as 1×10−4 s. In the upward rotation  p
r −1

21

123
Lagrangian-based numerical investigation of aerodynamic performance of an oscillating foil

Table 1 Discretization error and uncertainties for numerical results

r21 1.55
r32 1.51
f1 0.6744
f2 0.6594
f3 0.6622
p 3.7209
ea21 0.0222
21
eext 0.0054
21
GC Ifine 0.0067
32
GC Imedium 0.0352

Fig. 4 Comparison of lift coefficient (Cl ) predicted using different time

step sizes for reduced frequency K = 0.1, Re = 135,000, and U∞ =
The fine-grid convergence index is defined as 14 m/s

1.25ea21 the mean total power extracted to the total power available
21
GC Ifine = p . (15)
r21 − 1 in the oncoming flow passing through the swept area.
To quantify the energy extraction performance of an oscil-
The averaged lift coefficient is considered to be an impor- lating foil system, the nondimensional power coefficient
tant parameter for analysis of dynamic performance and was CPower and the energy extraction efficiency η are defined as
therefore chosen as the main parameter for uncertainty analy-
sis. The relative parameters were calculated and are presented PPower
CPower = , (16)
2 ρU∞ sc
1 3
in Table 1. Based on the error analysis results, the value of the
GCI for the average lift coefficient was found to be 0.67 % PPower = M (t) wp (t) , (17)
and 3.52%. As both uncertainty estimators lie in a reason-
PPower
able range, the medium grid (2.2 × 105 ) was selected for all η= , (18)
simulations in the present work.
1
2 ρU 3
∞ sd

Additionally, numerical results obtained using different 1 T
PPower = PPower (t)dt, (19)
time step sizes are presented in Fig. 4. Compared with the T 0
experimental data, the lift coefficient predicted with t =
1 × 10−3 s cannot reflect the transient lift evolution in the where PPower and PPower are the instantaneous and time-
downstroke, while when the time step was chosen as t = averaged power, respectively, extracted from the oncoming
1 × 10−4 s or t = 1 × 10−5 s, the predicted lift coefficient flow from the pitching contribution, M(t) is the torque about
remained almost the same. Hence, a time step of t = 1 × the pitching center, wp (t) is the transient angular velocity,
10−4 s was chosen for the computations, ensuring a Courant d is the vertical extent of the oscillating foil, and T is the
number of CFL x = U∞ × t/x ≈ 1 in the streamwise period of one pitching cycle.
direction and CFL y = V∞ × t/y ≈ 1 in the y-direction.

3.3 Extracted power and efficiency 4 Results and discussion

Improved understanding of dynamic stall characteristics will 4.1 Analysis of transient aerodynamic load and flow
help development of oscillating wing systems, which repre- features
sent one of the primary flow energy conversion technologies,
to enhance the energy harvesting performance for wind or Figure 5 shows the evolution of the lift coefficient Cl and drag
hydro energy. Hence, the instantaneous power coefficient coefficient Cd (Cd = D/(0.5ρU∞ 2 sc), where D is the drag)

CPower and the total energy extraction efficiency η were eval- with the AOA. The dynamic experimental results shown in
uated to quantify the energy extraction performance of the Fig. 5 were obtained from Ref. [20]. It is shown that, in the
oscillating foil system, also representing a valuable refer- upstroke from t1 (α+ = −5◦ ) to t7 (α+ = 25◦ ), the numeri-
ence for wind turbine blade designs. The instantaneous power cal results agree well with the experimental data, showing a
extracted from the flow comes from the pitching contribution, maximum predicted Cl very close to the measured value. In
and the energy extraction efficiency is defined as the ratio of the downstroke, the predicted lift coefficient presents small

123
 M. Zhang, et al.

accompanied by a low-pressure region at t3 (α+ = 20.4◦ ),

as shown in Figs. 6c and 7, which is responsible for the
sudden increase of the lift and drag coefficients. Until t4
(α+ = 23.5◦ ), the first LEV covers the entire suction sur-
face, as shown in Fig. 6d, and the corresponding lift and drag
coefficients approach their maximum.
During dynamic stall, at t5 (α+ = 24.4◦ ), the first
trailing-edge vortex (TEV) forms. Meanwhile, the first LEV
is detaching from the suction surface due to the interplay
between the first leading and trailing vortexes, resulting in
the sudden reduction of the lift and drag coefficients. Mean-
while, the leading edge develops a secondary vortex at the
same time, as shown in Fig. 6e, named the secondary LEV.
When the AOA reaches α+ = 24.8◦ (t6 ), the secondary LEV
extends to one-third chord of the suction side, accompanied
by a low-pressure region, as shown in Figs. 6f and 7. At
the same time, the lift and drag coefficients increase again.
Then, from t7 (α+ = 25◦ ) to t8 (α− = 24.7◦ ), the develop-
ment and attachment of the secondary LEV, as well as the
shedding of the first TEV, can be observed again, which cor-
responds to Cl and Cd approaching another peak value, as
shown in Figs. 5 and 6g, h. However, this peak value is lower
than the maximum value observed at t4 (α+ = 23.5◦ ). At
t9 (α− = 24.2◦ ), the secondary TEV forms and begins to
extrude the secondary LEV, as shown in Fig. 6i. When the
AOA reaches α− = 22.8◦ (t10 ), the secondary LEV sheds
away, resulting in a reduction of the slope of the lift and drag
coefficient curves, as shown in Figs. 5 and 6j.
During the poststall, the severe pressure gradient decreases
Fig. 5 Comparison of predicted a lift coefficient (Cl ) and b drag coef-
at t11 (α− = 21.8◦ ), and the suction surface forms a merged
ficient (Cd ) during the pitching process at reduced frequency K = 0.1
for Re = 135,000 and U∞ = 14 m/s vortex instead of the attached LEV, as shown in Figs. 6k and
7. At the same time, the secondary TEV has shed away. As
α decreases, formation and shedding of a third TEV occur
amplitude and low-frequency oscillating behavior from t7 from t12 (α− = 19.8◦ ) to t13 (α− = 18.1◦ ). The interac-
(α+ = 25◦ ) to t13 (α− = 18.1◦ ). This may be due to the inter- tion between the merged vortex and the third TEV results in
action between the induced secondary vortex on the leading small-amplitude oscillation of the lift and drag coefficients,
and trailing edge; detailed analysis of the flow features is as shown in Figs. 5 and 6l, m. When the angle of attack
presented in the following paragraphs. reaches t14 (α− = 10.5◦ ), the flow begins to transit from
The z-vorticity contours at 14 representative times and the turbulent to laminar, as shown in Fig. 6n, which causes the
predicted pressure coefficients are presented in Figs. 6 and dynamic forces to drop gently.
7. For sinusoidal pitching motion, the evolution of the flow In summary, the z-vorticity can be used to capture the evo-
structures and the corresponding aerodynamic responses can lution of the vortex structures in the transient flow field, and
be divided into three phases: (1) prior to dynamic stall, the extremum values of the z-vorticity represent the centers of
(2) dynamic stall, (3) poststall. circulation regions. However, a region of high vorticity (ωz )
Prior to dynamic stall, the flow is fully attached to the is not necessarily a vortex, as the shear flow near the wall
foil and the flow structure is quasisteady and laminar from t1 can also experience high shear. As shown in Fig. 6f, at t6
(α+ = − 5◦ ) to t2 (α+ = 17.5◦ ), as shown in Fig. 6a, b. The (α+ = 24.8◦ ), the approximate center position of the suction
lift coefficient Cl increases almost linearly, and the drag coef- face has high vorticity, but not a real vortex according to the
ficient Cd increases approximately as a quadratic function of instantaneous streamlines. Hence, there are some limitations
AOA, as shown in Fig. 5. At t2 (α+ = 17.5◦ ), the leading in using the z-vorticity ωz to distinguish between vorticity
edge induces a vortex, as it is subject to severe pressure gra- due to a vortex or shear. To provide insight into the vortex
dients, as shown in Figs. 6b and 7. The first LEV extends structures during the dynamic stall process, a new analytical
to approximately a semichord position of the suction side, method is applied in the next section.

123
Lagrangian-based numerical investigation of aerodynamic performance of an oscillating foil

a t1 (α+ = -5˚) b t2 (α+ = 17.5˚) c t3 (α+ = 20.4˚)

d t4 (α+ = 23.5˚) e t5 (α+ = 24.4˚) f t6 (α+ = 24.8˚)

g t7 (α+ = 25˚) h t8 (α- = 24.7˚) i t9 (α- = 24.2˚)

j t10 (α- = 22.8˚) k t11 (α- = 21.8˚ ) l t12 (α- = 19.8˚)

m t13 (α- = 18.1˚) n t14 (α- = 10.5˚)

Fig. 6 Contours of z-vorticity superimposed on instantaneous streamlines at representative times (K = 0.1)

123
 M. Zhang, et al.

(1) Prior to dynamic stall stage (α+ = 17.5◦ –23.5◦ )

In Fig. 8, the FTLE field and corresponding LCS at t2

(α+ = 17.5◦ ), t3 (α+ = 20.4◦ ), and t4 (α+ = 23.5◦ ) prior
to the dynamic stall stage are presented. In this study, based
on the velocity distribution in the flow field, the displace-
ment of each particle can be calculated from an integral of
the velocity, then the particle trajectory can be obtained dur-
ing the time interval TLE = 0.005 s (50 numerical steps).
Fig. 7 Comparison of predicted pressure coefficients at representative
The ridges of the FTLE field, LCSs, are used to define the
times (K = 0.1)
boundary of the vortex structures accurately. The LCSs can
be categorized as LCS A and LCS B in Fig. 8a, indicating
the flow region outside and inside the leading-edge vortex
4.2 Lagrangian-based analysis of flow structures in region, respectively. The pink particle inside LCS A exhibits
the dynamic stall process a clockwise trajectory during the time interval TLE = 0.005 s
with low value of FTLE. The brown particle outside LCS A
As the dynamic stall process plays a significant role in and the cyan particle inside LCS B move downstream due
energy capture by a wind turbine system, in this section, fur- to the quasilaminar flow during the time interval TLE . As α
ther analysis of the flow structures using the particle-based increases, it should be noted that, at t3 (α+ = 20.4◦ ), LCS A
Lagrangian approach is presented, with focus on the dynamic and LCS B become larger as the first LEV is growing. The
stall process from t2 (α+ = 17.5◦ ) to t11 (α− = 21.8◦ ). cyan particle trajectory inside LCS B remains clockwise dur-
According to the evolution of the LEV and TEV, the process ing the time interval TLE , which corresponds to the evolution
from t2 (α+ = 17.5◦ ) to t11 (α− = 21.8◦ ) can be divided of the small vortex at the leading edge. In Fig. 8c, LCS A
into two phases: (1) prior to dynamic stall stage, (2) dynamic has covered the whole suction surface. At the same time,
stall stage. LCS C, LCS E, and LCS D are newly formed, representing

Fig. 8 FTLE field and corresponding LCS during LEV developments stage (K = 0.1): a t2 (α+ = 17.5◦ ), b t3 (α+ = 20.4◦ ), c t4 (α+ = 23.5◦ )

123
Lagrangian-based numerical investigation of aerodynamic performance of an oscillating foil

the boundary of leading-edge vortexes with reverse rotational 4.3 Influence of reduced frequency on flow
direction and the first TEV. This can also be seen from the structures in dynamic stall process
purple particle inside LCS C and the pink particle inside
LCS D. To further investigate the effect of the oscillation fre-
quency on the aerodynamic performance, the transient flow
(2) Dynamic stall stage (α+ = 23.5◦ − α− = 21.8◦ ) structures and corresponding aerodynamic characteristics at
different reduced oscillating frequencies are discussed in this
To provide insight into the vortex structures in this stage, as section.
shown in Fig. 9, the FTLE field and corresponding LCS at Figures 10 and 11 show the evolution of the predicted lift
t5 (α+ = 24.4◦ ), t6 (α+ = 24.8◦ ), t7 (α+ = 25◦ ), t8 (α− = coefficient (Cl ) and power coefficient (CPower ) for different
24.7◦ ), t10 (α− = 22.8◦ ), and t11 (α− = 21.8◦ ) are presented. reduced frequencies (K = 0.05 and 0.1). Table 2 presents
At t5 (α+ = 24.4◦ ), LCS D rolls up and becomes larger in the energy extraction efficiency at the different reduced fre-
size with the development of the first TEV, accompanied by quencies. It can be observed that the total energy extraction
shedding of the first LEV. Meanwhile, LCS C forms outside efficiency at K = 0.1 was larger than the case of K = 0.05.
LCS B, which results from shedding of the first LEV and the The z-vorticity contours and the instantaneous streamlines at
interaction between the reverse-directional vortex structures. typical angles of attack for both cases are shown in Fig. 12,
During the time interval TLE , the black and purple particles in where t1 −t7 represent typical times for the case with K = 0.1
 
Fig. 9a bypass the circulation region and follow its boundary. and t1 −t7 for the case with K = 0.05. Comparisons of
LCS B indicates the boundary of the secondary LEV. After the predicted pressure coefficients for the cases with differ-
the time interval TLE , the green particle in Fig. 9a moves ent reduced frequencies (K = 0.05 and 0.1) are shown in
backward on the surface. Fig. 13.
When it reaches t6 (α+ = 24.8◦ ), LCS C becomes weaker. Prior to dynamic stall, it is found that the evolution of the
This means that the visibility of the vortex boundaries and the predicted Cl and the power coefficient (CPower ) are similar
vortex strength become weaker. LCS D becomes larger and for the different reduced frequencies, as shown in Figs. 10
sheds away. LCS B and E also become larger and attached and 11. The corresponding Cl increases almost linearly with
to the surface with the development of the secondary LEV. increasing angle, while the corresponding CPower remains
Meanwhile, LCS F is newly formed inside LCS E, as shown at low amplitude then increases sharply. At K = 0.05, the

in Fig. 9b. At t7 (α+ = 25◦ ), LCS B covers three-quarters formation of the first LEV occurs at t1 (α+ = 16◦ ), being
of the suction surface. The track of the gray-brown particle more specific than at K = 0.1, as shown in Fig. 12a, g.
inside LCS E indicates that the vortex center moves down- The larger adverse pressure gradient distributions at α+ =
stream. The blue particle inside LCS F rolls up and attaches 16◦ for K = 0.05 are responsible for the advanced for-
at the surface due to the suction force of the secondary LEV. mation of the first LEV, as shown in Fig. 13a. The first

When the angle of attack reaches α− = 24.7◦ (t8 ), LCS B has LEV begins to develop with the low-pressure region at t2
covered the whole suction surface with low-pressure region, (α+ = 17.1◦ ), resulting in the sharp increase of Cl and

as shown in Fig. 8, which leads to the secondary increase of CPower , as shown in Figs. 10 and 11. When it reaches t3
the lift coefficient. (α+ = 19.5◦ ), the first LEV has developed and covered
As α decreases in the downstroke phase, the secondary the entire suction surface, as shown in Fig. 12b, c. As α
LEV sheds away and the secondary TEV is induced. At t10 increases, the low-pressure region moves to the trailing edge
(α− = 22.8◦ ), LCS G, which represents the boundary of the with the development of the first LEV, significantly enhanc-
secondary TEV, extrudes LCS B, as shown in Fig. 9e. The ing the lift and power prior to the dynamic stall process.
gray-brown particle is attracted to leave the foil by the suction The adverse pressure gradient distributions at the same angle
force. Because of the interaction of the secondary LEV and are larger for K = 0.05 than K = 0.1, as shown in
TEV, the blue and cyan particles are forced to attach to the Fig. 13a–c. This is because the pitching velocity is slow at
surface. This results in a decline of the lift coefficient again. K = 0.05, and the formation of the first LEV can develop
When it reaches t11 (α− = 21.8◦ ), LCSs weaken gradually. adequately.
As α decreases, LCS F constantly merges with the free stream During dynamic stall, the stall point is again more spe-
flow and sheds away on account of the interaction between cific and the maximum value of Cl and CPower becomes
the merged vortex and the third TEV, which is responsible low for K = 0.05, as shown in Figs. 10 and 11. This sig-
for the small-amplitude oscillating behavior in the lift and nificantly affects the total energy extraction efficiency of
drag coefficient curves. the oscillating foil. With increase of the angle of attack, as
shown in Fig. 13d, e, the adverse pressure gradient distri-
butions at the suction surface for K = 0.05 become low
with low variations instead of the complex pressure coef-

123
 M. Zhang, et al.

a t5 +
) b t6 +
) c t7 +
)

FTLE
field

LCS

d t8 -
) e t10 -
) f t11 -
)

FTLE
field

LCS

Fig. 9 FTLE field and corresponding LCS during poststall vortex shedding stage (K = 0.1)

123
Lagrangian-based numerical investigation of aerodynamic performance of an oscillating foil

Fig. 10 Evolution of predicted Cl with different reduced frequencies (K = 0.05 and 0.1) for Re = 135,000: a full range, b zoomed view of
dynamic stall process

Fig. 11 Evolution of power coefficient (CPower ) with different reduced frequencies (K = 0.05 and 0.1) for Re = 135,000: a full range, b zoomed
view of dynamic stall process


ficient distributions at K = 0.1. At t4 (α+ = 21.7◦ ) in Table 2 Energy extraction efficiency at different reduced frequencies
the upstroke phase, the first LEV has already shed away (K = 0.05 and 0.1)
and three vortices align alongside the upper surface com- K 0.05 0.1
pared with t4 (α+ = 24.4◦ ), which is responsible for the
η (%) 32.18 55.46
decline of Cl and CPower , as shown in Figs. 10, 11, and 12d.
Compared with K = 0.1, the small-amplitude oscillating
behavior becomes more severe for the case with K = 0.05

from t5 (α+ = 23◦ ), as shown in Figs. 10 and 11. More- the dynamic stall process, the evolution of the dynamic per-
over, Cl and adverse pressure gradient distributions are lower formance further results in a decrease of the total energy

than at K = 0.1, as shown in Figs. 10 and 13g, h. At t5 extraction efficiency compared with the case of K = 0.1.

(α+ = 23◦ ), the vortex near the center clearly shrinks while When the angle of attack increases to t7 (α+ = 24.5◦ ),
another pair of vortices merge, as shown in Fig. 12e, which the third TEV is induced on the trailing edge, as shown in
is responsible for the increase of Cl again. However, com- Fig. 12f. As α declines in the downstroke phase, TEV for-
pared with the formation of the attached secondary LEV at t5 mation, interaction, and shedding are repeated many times,
(α− = 24.7◦ ), the merged vortex interacts with the secondary being responsible for the decreasing amplitude and high-
TEV at K = 0.05. It can be observed that no attached sec- frequency oscillating behavior of the flow. This is mainly
ondary LEV was induced at K = 0.05. This is responsible for attributed to the merged vortex and TEV having sufficient
the difference in the adverse pressure gradient distributions time to develop and interact for K = 0.05. This oscillat-
for different reduced frequencies, as shown in Fig. 13e–g. At ing behavior may result in the decrease of the total energy
t6 (α+ = 24◦ ), development and shedding of the secondary extraction efficiency of the oscillating foil. For K = 0.1,
TEV result in the decline of Cl and CPower 10 and 12f. In the pitching velocity is too fast for the vortex to develop and
interact completely.

123
 M. Zhang, et al.

K=0.05 K=0.1
a t’1 (α+ = 16˚) h t1 (α+ = 17.5˚)

b t’2 (α+ = 17.1˚) i t2 (α+ = 20.4˚)

c t’3 (α+ = 19.5˚) j t3 (α+ = 23.5˚)

d t’4 (α+ = 21.7˚) k t4 (α+ = 24.4˚)

Fig. 12 Contours of z-vorticity superimposed on instantaneous streamlines at different angles of attack (K = 0.05 and 0.1)

123
Lagrangian-based numerical investigation of aerodynamic performance of an oscillating foil

e t’5 (α+ = 23˚) l t5 (α- = 24.7˚)

f t’6 (α+ = 24˚) m t6 (α- = 22.8˚)

g t’7 (α+ = 24.5˚) n t7 (α- = 21.8˚)

Fig. 12 continued

5 Conclusions leading- and trailing-edge vortex, which is responsi-

ble for the reduction of the lift and drag coefficients.
The coupled k−ω SST turbulence model and γ −Reθ tran- With increase of the angle of attack, the secondary LEV
sition model were used to simulate the dynamic stall phe- repeats this formation, interaction, and shedding pro-
nomenon for an oscillating NACA0012 airfoil. The numeri- cess, leading to a secondary rise and fall of the lift
cal results were validated by comparison with experimental and drag coefficients. During poststall, the leading edge
results. The main conclusions are as follows: of the suction surface forms a merged vortex, which
interacts with the TEV, corresponding to large-scale
(1) During dynamic stall, the first LEV is forced to shed low-frequency load fluctuations. Meanwhile, the LCSs
away due to the strong interaction between the first gradually fade away with shedding of the merged vortex.

123
 M. Zhang, et al.

Fig. 13 Comparisons of predicted pressure coefficients at same geometric angles of attack for different reduced frequencies (K = 0.05 and 0.1)

123
Lagrangian-based numerical investigation of aerodynamic performance of an oscillating foil

(2) The reduced frequency significantly affects the flow 4. Hang, L., Zhou, D., Lu, J., et al.: The impact of pitch motion of
structures and energy extraction performance in the a platform on the aerodynamic performance of a floating vertical
axis wind turbine. Energy 119, 369–383 (2017)
dynamic stall process. At K = 0.1 and 0.05, the evolu- 5. Mckenna, R., Leye, P.O.V.D., Fichtner, W.: Key challenges and
tion of Cl and CPower is approximately similar prior to prospects for large wind turbines. Renew. Sustain. Energy Rev. 53,
the dynamic stall process. However, compared with the 1212–1221 (2016)
case of K = 0.1, the dynamic stall point is advanced 6. Lu, K., Xie, Y., Zhang, D., et al.: Systematic investigation of the
flow evolution and energy extraction performance of a flapping-
and no attached secondary LEV is generated when airfoil power generator. Energy 89, 138–147 (2015)
K = 0.05. Additionally, the small-amplitude oscillat- 7. Huang, B., Wu, Q., Wang, G.Y.: Numerical simulation of unsteady
ing behavior of the dynamic curve becomes more severe cavitating flows around a transient pitching hydrofoil. Sci. China
for the case of K = 0.05. These phenomena result in Technol. Sci. 57, 101–116 (2014)
8. Lee, T.: Effect of flap motion on unsteady aerodynamic loads. J.
the oscillating behavior of the power coefficient (CPower )
Aircr. 44, 334–338 (2015)
and affect the total energy extraction efficiency. The total 9. Birch, D.M., Lee, T.: Tip vortex behind a wing undergoing deep-
energy extraction efficiency is higher for K = 0.1 than stall oscillation. AIAA J. 43, 2081–2092 (2015)
K = 0.05. 10. Liu, T.T., Huang, B., Wang, G.Y., et al.: Experimental investigation
of the flow pattern for ventilated partial cavitating flows with effect
(3) LCSs defined by ridges of the FTLE field were utilized
of Froude number and gas entrainment. Ocean Eng. 129, 343–351
to investigate transient flow structures. Compared with (2017)
the Eulerian approach, e.g., using the z-vorticity (ωz ), 11. Wang, Y.W., Xu, C., Wu, X.C., et al.: Ventilated cloud cavitating
such Lagrangian-based analysis of flow structures in the flow around a blunt body close to the free surface. Phys. Rev. Fluids
2, 084303 (2017)
dynamic stall process can effectively avoid overpredic- 12. Long, X.P., Cheng, H.Y., Ji, B., et al.: Large eddy simulation and
tion of vortex structures. The dynamic behavior of the Euler–Lagrangian coupling investigation of the transient cavitating
flow structures was effectively identified using the FTLE turbulent flow around a twisted hydrofoil. Int. J. Multiph. Flow 100,
field. 41–56 (2018)
13. Wang, G.Y., Wu, Q., Huang, B.: Dynamics of cavitation–structure
interaction. Acta Mech. Sin. 33, 685–708 (2017)
14. Huang, B., Zhao, Y., Wang, G.Y.: Large eddy simulation of
In the future, the three-dimensional effect [51–54] and its turbulent vortex-cavitation interactions in transient sheet/cloud
interactions with turbulence and energy harvesting are wor- cavitating flows. Comput. Fluids 92, 113–124 (2014)
15. Choudhry, A., Arjomandi, M., Kelso, R.: Methods to control
thy of further investigation, so LCS and particle tracking dynamic stall for wind turbine applications. Renew. Energy 86,
techniques will be applied to three-dimensional flow fields 26–37 (2016)
to present more details of the transient flow structures. To 16. Hameed, M.S., Afaq, S.K.: Design and analysis of a straight bladed
determine the spatial and temporal variation of the turbu- vertical axis wind turbine blade using analytical and numerical
techniques. Ocean Eng. 57, 248–255 (2013)
lent structures more accurately, direct numerical simulations 17. Huang, B., Young, Y.L., Wang, G.Y., et al.: Combined experi-
(DNS) and large-eddy simulations (LES) will be added in mental and computational investigation of unsteady structure of
future work. In addition, the effect of the pitching amplitude sheet/cloud cavitation. J. Fluids Eng. Trans. ASME 135, 071301
on the flow evolution and energy harvesting performance will (2013)
18. Ferreira, C.S., Bussel, G.V., Kuik, G.V.: 2D CFD simulation of
also be discussed further in the future. dynamic stall on a vertical axis wind turbine: verification and vali-
dation with PIV measurements. In: 45th AIAA Aerospace Sciences
Acknowledgements This work was supported by the National Post- Meeting and Exhibit (2006)
doctoral Program for Innovative Talents (Grant BX201700126), the 19. Wernert, P., Geissler, W., Raffel, M., et al.: Experimental and
China Postdoctoral Science Foundation (Grant 2017M620043), the numerical investigations of dynamic stall on a pitching airfoil.
National Natural Science Foundation of China (Grants 51679005 and AIAA J. 34, 982–989 (1996)
91752105), and the National Natural Science Foundation of Beijing 20. Lee, T., Gerontakos, P.: Investigation of flow over an oscillating
(Grant 3172029). airfoil. J. Fluid Mech. 512, 313–341 (2004)
21. Carr, L.W.: Progress in analysis and prediction of dynamic stall. J.
Aircr. 25, 6–17 (1988)
22. Ekaterinaris, J.A., Platzer, M.F.: Computational prediction of air-
foil dynamic stall. Prog. Aerosp. Sci. 33, 759–846 (1997)
References 23. Simpson, B.J., Hover, F.S., Triantafyllou, M.S.: Experiments in
direct energy extraction through flapping foils. in: the Eighteenth
1. Melício, R., Mendes, V.M.F., Catalão, J.P.S.: Transient analysis International Offshore and Polar Engineering Conference, 2008
of variable-speed wind turbines at wind speed disturbances and a 24. Shehata, A.S., Xiao, Q., Saqr, K.M., et al.: Passive flow control for
pitch control malfunction. Appl. Energy 88, 1322–1330 (2011) aerodynamic performance enhancement of airfoil with its appli-
2. González, L.G., Figueres, E., Garcerá, G., et al.: Maximum-power- cation in wells turbine—under oscillating flow condition. Ocean
point tracking with reduced mechanical stress applied to wind- Eng. 136, 31–53 (2017)
energy-conversion-systems. Appl. Energy 87, 2304–2312 (2010) 25. Tseng, C.C., Hu, H.A.: Dynamic behaviors of the flow past a pitch-
3. Karbasian, H.R., Esfahani, J.A., Barati, E.: The power extraction ing foil based on Eulerian and Lagrangian viewpoints. AIAA J. 54,
by flapping foil hydrokinetic turbine in swing arm mode. Renew. 712–727 (2016)
Energy 88, 130–142 (2016)

123
 M. Zhang, et al.

26. Ducoin, A., Astolfi, J.A., Deniset, F., et al.: Computational and 41. Haller, G., Yuan, G.: Lagrangian coherent structures and mixing in
experimental investigation of flow over a transient pitching hydro- two-dimensional turbulence. Phys. D 147, 352–370 (2000)
foil. Eur. J. Mech. B Fluids 28, 728–743 (2009) 42. Wu, Q., Huang, B., Wang, G.: Lagrangian-based investigation of
27. Bhat, S.S., Govardhan, R.N.: Stall flutter of NACA 0012 airfoil at the transient flow structures around a pitching hydrofoil. Acta
low Reynolds numbers. J. Fluids Struct. 41, 166–174 (2013) Mech. Sin. 32, 64–74 (2016)
28. Wang, S., Ingham, D.B., Ma, L., et al.: Numerical investigations 43. Tseng, C.C., Liu, P.B.: Dynamic behaviors of the turbulent cavi-
on dynamic stall of low Reynolds number flow around oscillating tating flows based on the Eulerian and Lagrangian viewpoints. Int.
airfoils. Comput. Fluids 39, 1529–1541 (2010) J. Heat Mass Transf. 102, 479–500 (2016)
29. Huang, B., Ducoin, A., Young, L.Y.: Physical and numerical inves- 44. Wang, Z.Y., Huang, B., Zhang, M.D., et al.: Experimental and
tigation of cavitating flows around a pitching hydrofoil. Phys. numerical investigation of ventilated cavitating flow structures with
Fluids 25, 102109 (2013) special emphasis on vortex shedding dynamics. Int. J. Multiph.
30. Chen, Y.L., Zhan, J.P., Wu, J., et al.: A fully-activated flapping foil Flow 98, 79–95 (2018)
in wind gust: energy harvesting performance investigation. Ocean 45. Roache, P.J.: Quantification of uncertainty in computational fluid
Eng. 138, 112–122 (2017) dynamics. Annu. Rev. Fluid Mech. 29, 123–160 (2003)
31. Teng, L., Deng, J., Pan, D., et al.: Effects of non-sinusoidal pitching 46. Gohil, P.P., Saini, R.P.: Effect of temperature, suction head and flow
motion on energy extraction performance of a semi-active flapping velocity on cavitation in a Francis turbine of small hydro power
foil. Renew. Energy 85, 810–818 (2016) plant. Energy 93, 613–624 (2015)
32. Lai, J.C.S., Platzer, M.F.: Jet characteristics of a plunging airfoil. 47. Roache, P.J.: Verification of codes and calculations. AIAA J. 36,
AIAA J. 37, 1529–1537 (2015) 696–702 (2012)
33. Gharali, K., Johnson, D.A.: Dynamic stall simulation of a pitching 48. Kwasniewski, L.: Application of grid convergence index in FE
airfoil under unsteady freestream velocity. J. Fluids Struct. 42, 228– computation. Bull. Pol. Acad. Sci. Tech. Sci. 61, 123–128 (2013)
244 (2013) 49. Kleinhans, M.G., Jagers, H.R.A., Mosselman, E., et al.: Procedure
34. Guo, Q., Zhou, L., Wang, Z.: Comparison of BEM-CFD and full of estimation and reporting of uncertainty due to discretization in
rotor geometry simulations for the performance and flow field of a CFD applications. J. Fluids Eng. 130, 078001 (2008)
marine current turbine. Renew. Energy 75, 640–648 (2015) 50. Hunt, J.C.R., Wray, A.A., Moin, P.: Eddies, stream, and conver-
35. Wu, Q., Wang, Y.N., Wang, G.Y.: Experimental investigation of gence zones in turbulent flows. Center for Turbulence Research
cavitating flow-induced vibration of hydrofoils. Ocean Eng. 144, Report CTR-S88. pp. 193–208 (1988)
50–60 (2017) 51. Choudhry, A., Leknys, R., Arjomandi, M., et al.: An insight into
36. Wu, Q., Huang, B., Wang, G.Y., et al.: Experimental and numerical the dynamic stall lift characteristics. Exp. Therm. Fluid Sci. 58,
investigation of hydroelastic response of a flexible hydrofoil in 188–208 (2014)
cavitating flow. Int. J. Multiph. Flow 74, 19–33 (2015) 52. Martinat, G., Braza, M., Hoarau, Y., et al.: Turbulence modeling of
37. Menter, F,R.: Improved two-equation k − ω turbulence models for the flow past a pitching NACA0012 airfoil at 105 and 106 Reynolds
aerodynamic flows. NASA Tech. Memo. 34, 103975 (1992) numbers. J. Fluids Struct. 24, 1294–1303 (2008)
38. Langtry, R.B., Menter, F.R., Likki, S.R., et al.: A correlation-based 53. Velkova, C., Todorov, M., Dobrev, I., et al.: Approach for numerical
transition model using local variables-part I: model formulation. J. modeling of airfoil dynamic stall. in: Proceedings of BulTrans-
Turbomach. 128, 413–422 (2006) 2012, 26–28 September, Sozopol, 1–6 (2012)
39. Langtry, R.B., Menter, F.R., Likki, S.R., et al.: A correlation-based 54. Tseng, C.C., Cheng, Y.E.: Numerical investigations of the vortex
transition model using local variables-part II: test cases and indus- interactions for a flow over a pitching foil at different stages. J.
trial applications. J. Turbomach. 128, 423–434 (2006) Fluids Struct. 58, 291–318 (2015)
40. Menter, F.R., Langtry, R.B., Völker, S.: Transition modeling for
general purpose CFD codes. Flow Turbul. Combust. 77, 277–303
(2006)

123