Professional Documents
Culture Documents
A Mathematical Model For Helicopter Comprehensive Analysis
A Mathematical Model For Helicopter Comprehensive Analysis
Journal of
Aeronautics
Chinese Journal of Aeronautics 23(2010) 320-326 www.elsevier.com/locate/cja
Abstract
This article presents a new mathematical model for helicopter comprehensive analysis with the features of flexibility and
mathematical simplicity. The model synthesizes the rigid fuselage motion model with 6 degrees of freedom, coupled
flap-lag-torsion elastic rotor blade motion model, unsteady aerodynamics model with dynamic stall and high order generalized
dynamic wake model. A new blade structural operator with implicit form is formulated, and the components of the blade structure
model are independent of each other so that it is convenient to change or handle any component of blade structure without
changing the others. What is more, the entire model is developed in a strict state-space form to simplify the comprehensive
analysis. Finally, the UH-60 helicopter is taken as an example to predict the blade natural characteristics, the trim characteristics
including controls and fuselage attitudes as well as the airloads at blade section under the flight conditions of high speed with
moderate thrust and high thrust with moderate speed. The results are compared with UH-60 flight test data and those predicted by
two well-known comprehensive codes. The validity of the model presented in this article is verified.
the ground-fixed inertial coordinate system [eˆxg eˆ gy angular velocities of the rigid fuselage described by the
nonlinear Euler equations, and three attitude angles
eˆzg ]T , and CG is the helicopter center of gravity. The governed by the kinematic equations. The main rotor
origin of shaft coordinate system [eˆxs eˆsy eˆzs ]T is lo- states yR are the generalized displacements and gener-
cated at the hub center, eˆxs is aligned with the rotor shaft. alized velocities of the blade. The inflow states yI
iθ and iϕ are the longitudinal and lateral shaft tilt angles contain the main rotor and tail rotor inflow states, in
relative to the fuselage. The hub rotating coordinate which 12 harmonics with 91 inflow states are modeled
system [eˆxr eˆ ry eˆzr ]T rotates at angular velocity Ω with for main rotor to capture the major portion of the un-
respect to the nonrotating shaft reference frame. The steady inflow[10]. Unsteady aerodynamic states yA are
elastic deflections of the blade are defined within the modeled with 12 states capturing airfoil unsteady
undeformed coordinate system [eˆxu eˆ uy eˆzu ]T which aerodynamics and dynamic stall at each spanwise posi-
tion of blade. The states of each model can be origi-
forms a precone angle βP and a presweep angle ζP rela- nally expressed in state-space form except the states
tive to the hub rotating coordinate system. The vec- which describe the blade motion. The following sec-
tor eˆxu is along the undeformed elastic axis of the blade. tion will discuss the state-space form of blade motion.
The deformed coordinate system [eˆxd eˆ dy eˆzd ]T is at-
2.3. Equations of blade motion
tached to the deformed blade, with eˆxd being tangent to
the deformed elastic axis. u, v and w are the elastic For analysis, the blade is discretized into a number
displacements of point P on the blade elastic axis. x0 of beam elements. Each beam element consists of fif-
denotes the length along the elastic axis of the unde- teen degrees of freedom[2]. The finite element formula-
formed blade, while y0 and z0 denote the cross-sec- tion is based on the Galerkin approach of weighted
tional coordinates along the two directions being residuals. The equations of blade motion can be written
vertical to the elastic axis respectively. θG is the blade as
geometric pitch angle, which is the sum of the pitch PI ( y F , q, q , q , t ) + PS ( q , t ) +
control angle and the built-in twist.
PA ( yF , q , q , t ) + PD ( q , q , t ) = 0 (2)
where PI, PS, PA and PD are the vectors of the inertia,
structural, aerodynamic and lag damper nodal loads
(if there were) respectively, q is the vector of blade
generalized displacement. After special treatment of
the inertia loads presented later, the coupled
rotor/fuselage equations of motion can be expressed as
[ y FT qT ]T = E I−1 f R ( q , yF , q , t ) (3)
where EI is the inertia coupling matrix, f R a nonlinear
vector function.
(1) Blade structural loads
The main portions of the formulation of the struc-
tural operator are: the transformation between the de-
formed and undeformed coordinate systems, the cur-
vatures and twist of the elastic axis of the deformed
blade, the strain-displacement relationship, the stress-
Fig.1 Ground-fixed, fuselage and blade coordinate systems.
strain relationship and the stress-force relationship. All
of the ingredients are implemented independently in
2.2. Helicopter comprehensive analysis model
implicit formulation without adopting moderate de-
flection assumption and ordering scheme.
The equations of motion for the entire model are The transformation between the deformed and un-
expressed as a system of standard first order differen- deformed coordinate systems is given by
tial equations
y = f ( y, u, t ) (1) e d = T du e u (4)
where t is time; u the control vector; the state vector
eˆzd ]T , e = [eˆx
u u
where e d = [eˆxd eˆ dy eˆ uy eˆzu ]T , and
y contains the fuselage states yF , main rotor states yR ,
T du=[sij]3×3 is the transformation matrix. If the elastic
inflow states yI and unsteady aerodynamic states yA . displacements u, v, w and φ are assumed to be known,
The fuselage states yF include the translational and the elements of [sij]3×3 and their derivatives sij,x can be
· 322 · Li Pan et al. / Chinese Journal of Aeronautics 23(2010) 320-326 No.3
calculated, then the curvatures κy and κz, and twist τ Thus, the force which acts on the unit area of the
can be calculated exactly[8] by Eq.(A1) in Appendix A. cross section of the deformed blade is
The strain-displacement relationship is the key in- t =σ xxGx + τ xy G y + τ xz G z (10)
gredient of the implicit formulation of the structural using Eq.(9) and Eq.(10), we obtain
operator. Under the Bernoulli-Euler hypothesis, the T
position vector ρ of an arbitrary point P on the blade ⎡ μ T Ci μ μT a i1 ⎤
relative to the hub center is given by ⎢ ⎥
t = ⎢ μ T Ci μ μ T a i 2 ⎥ e d (i = 1, 2,3) (11)
⎢ ⎥
⎡ x0 + u ⎤ ⎡τϕ ( x0 , y0 , z0 ) ⎤
T T T
⎡e ⎤ ⎢⎣ μT Ci μ μ T a i 3 ⎥⎦
ρ = ⎢ 0 ⎥ e + ⎢ v ⎥ e + ⎢⎢
⎢ ⎥ r ⎢ ⎥ u
y0 ⎥ ed
⎥ (5)
where a ij is the jth column of matrix Ai. In Eq.(11), the
⎢⎣ 0 ⎥⎦ ⎢⎣ w ⎥⎦ ⎢⎣ z0 ⎥⎦
summation convention is adopted. The resultant force,
where e r
=[e xr e ry e zr ]T , e is the offset of the blade which acts on the cross section of the deformed blade,
is obtained by integration
flap, lag or pitch hinge (if there were) measured from T
⎡ f iT a i1 ⎤
the hub center. ϕ ( x0 , y0 , z0 ) is a cross-sectional warping ⎢ ⎥
function. Thus, the base vectors of the point P after F = [T V y Vz ]e d = ∫∫ tdy0 dz0 = ⎢ f iT a i 2 ⎥ e d (12)
deformation are
A ⎢ ⎥
⎢⎣ f iT a i 3 ⎥⎦
⎛ ⎡1 + u ⎤ T ⎡τ , xϕ + τϕ, x ⎤ ⎟⎞ ⎫⎪
T
⎡τϕ ⎤
T
where
⎜⎢ ,x
⎥ ⎢ ⎥ ⎟ d⎪
⎢ ⎥
G x = ⎜ ⎢ v, x ⎥ T ud + ⎢ y0 ⎥ T xd + ⎢ 0 ⎥ ⎟e ⎪ f iT = [tr( D1Ci ) tr( Dy0 Ci ) tr( Dz0 Ci )
⎜⎢ ⎥ ⎢ ⎥ ⎟ ⎪
⎜ w ⎢
⎣ 0⎦
z ⎥ 0
⎝ ⎣ ,x ⎦ ⎣ ⎦ ⎠ ⎬ tr( Dϕ Ci ) tr( Dϕ, x Ci ) tr( Dϕ, y Ci ) tr( Dϕ, z Ci )] (13)
⎪
G y = [τϕ, y 1 0]e d ⎪ The operator tr(·) means the trace of a matrix, and ma-
⎪
trix D(*) is defined as D(∗) = ∫∫ (∗) μ μ d y0 d z0 , and A
T
G z = [τϕ, z 0 1]e d ⎪⎭ A
∫∫ ϕ μ μ
ud du T xd T
where T = (T ) , and T is the transformation ma- d y0 d z0 . The elements of the matrix D(*) are
A
trix between the triads e d and e ,xd , which is given by the blade section properties.
Eq.(A2). In order to use the implicit approach[11], the The resultant structural moment about the point
base vectors should be written in terms of the (y0=0, z0=0) can be obtained by integration
cross-sectional coordinates and warping function, that M = [ M x M y M z ]e d = ∫∫ d × tdy0 dz0 (14)
is A
where s11 = 1 s11 . Through a series of partial integra- fuselage coordinate system. To complete the above-
tions, the structural nodal loads vector for the ith ele- mentioned operation with numerical method and avoid
ment with length li is given as any algebraic expansion, every component of y c must
be written as, for example
⎡ p1sx H u , x ⎤
⎢ 2 ⎥ w = [0 0 1 0 0 0 0 0 0 0 0 0 0 0] y c
li ⎢ psy H v , xx + psy H v , x ⎥
1
Psi = ∫ ⎢ 2 ⎥ dxe (17) (24)
⎢ sz w, xx + psz H w, x ⎥
0 p H 1
the inertia load per unit span is given by where c is blade chord.
The angle of attack α and pitch rate q are the inputs
pI = − ∫∫ ρ a P dy0 dz0 (20)
A of the Leishman-Beddoes unsteady and dynamic stall
where ρ is the mass density of the blade. Similarly the model in state-space form[13]. Leishman-Beddoes
inertia moment is given by model can adequately capture unsteady phenomena,
flow separation, dynamic stall, and transonic com-
qI = − ∫∫ ρ d × a P dy0dz0 (21) pressibility effects. This aerodynamic model consists
A
of three distinct parts: 1) an attached potential flow
Then the inertia nodal loads vector for the ith element
formulation for linear unsteady airloads described by 8
is given as
states, the state-space formulation can be written as
PIi = ∫ [ pIx Hu pIy Hv pIz Hw qIx Hφ ]T dxe ⎫⎪
li
(22) x = Ax + B[α q ]T
0
⎬ (29)
Two special treatments for aP are conducted here. [C C ] = Cx + D[α q ] ⎪⎭
P
N
P T
M
T
Firstly, aP is written in terms of the cross-sectional where x is the state vector, matrices A, B, C and D can
coordinates as the form of Eq.(7) to satisfy the re-
be found in Ref.[13], and CNP and CMP are respectively
quirement for the section integration of Eq.(20) and
Eq.(21). Secondly, in order to express the equations the total potential normal force coefficient and pitching
of motion in the form of state-space, the accelera- moment coefficient; 2) a separated flow formulation
tion-dependent terms in Eq.(19) must be identified for nonlinear unsteady airloads described by 3 states;
and written as 3) a dynamic stall formulation for vortex induced air-
loads described by one state. Additionally, the general-
a P = Ey c + gi ( yc , t ) (23)
ized dynamic wake theory is used for inflow analysis,
where yc = [u v w p q r Ω u u, x v v, x w w , x φ] ; u ,
T in which the effects of the shed wakes (Theodorsen’s
function effect[14]) are included. To avoid duplicating
v , w and p, q, r are respectively the translational and the blade shed wake effect, the corresponding elements
angular velocity components of the fuselage in the in the matrix B (Eq.(29)) must be set to zero[6].
· 324 · Li Pan et al. / Chinese Journal of Aeronautics 23(2010) 320-326 No.3
In order to verify the model mentioned above, the To verify the developed structure operator, the ro-
UH-60 helicopter is taken as an example to predict the tating natural frequencies of the blade predicted by the
natural frequencies of rotor blade as well as the trim present formulation are compared with those obtained
characteristics and corresponding airloads of blade by UMARC. The rotating speed of the blade is 27.0
section under two critical flight conditions. The results rad/s. Table 1 summarizes the results of comparison,
are compared with those predicted by the existing which shows that there is a good agreement, even for
comprehensive analysis codes UMARC[2] , CAMRAD high frequency modes.
II[1] or the flight test data. The parameters and aerody- Table 1 Comparison between blade natural frequencies
namic data used for computation are taken from
Ref.[12] and Ref.[15]. Natural frequency
Blade mode UMARC[18]/ Calculation/ Relative error/%
(rev−1) (rev−1)
3.1. Trim scheme
1st lag 0.270 0.268 0.7
2ed lag 4.670 4.710 0.9
The calculation of the trim state of the helicopter 3rd lag 12.420 12.600 1.4
dynamic system (Eq.(1)) consists of four coupled sets 1st flap 1.030 1.035 0.5
of nonlinear algebraic equations. The first set of equa- 2ed flap 2.820 2.820 0
tions used in the trim procedure represents the equilib- 3rd flap 5.170 5.270 2.0
rium of the fuselage undergoing a level flight or a 4th flap 7.900 7.980 1.0
steady, coordinated turn. If the n/rev periodic motion 1st torsion 3.880 3.960 2.0
of the fuselage in trim state is allowed, the X-force 2nd torsion 12.410 12.970 4.3
equilibrium along the fuselage axes must satisfy
2π
To verify the model, two steady level flight condi-
∫0 udψ = 0 (30) tions, one is corresponding to high speed with moder-
ate thrust (μ = 0.37, Cw/σ = 0.084, μ is advance ratio,
Similar conditions can be applied to v , w , p , q Cw is weight coefficient, σ is rotor solidity) and the
other corresponding to high thrust with moderate
and r as well as Euler rates of the fuselage, θ , φ and
b b speed(μ = 0.24, Cw/σ = 0.130), are examined to predict
(ψ b − ψ bp) , where ψ bp is the prescribed turning rate the trim characteristics and airloads of typical blade
section. Table 2 presents the comparison between con-
and assumed being constant and it defines the turning trols and fuselage attitudes predicted by present analy-
flight condition. In addition, two kinematic relation- sis at trim condition and those obtained from UMARC
ships between attitude angles and fuselage aerody- and flight test data[18] under the flight condition of high
namic angles should satisfy the same condition as that speed with moderate rotor thrust. The accuracy of the
for Eq.(30). The second set of equations is derived predictions is good. The discrepancy of the lateral cy-
from the system of dynamic inflow equations. The clic pitch exists in both UMARC and present analysis.
time derivatives of the inflow states for main rotor and
tail rotor are also required to satisfy the same condition Table 2 Comparison between predicted controls and
as that for u in Eq.(30). The third set of equations ob- fuselage attitudes obtained from present
analysis and UMARC and flight test data
tained from the nonlinear ordinary differential equa- (μ = 0.37, Cw/σ = 0.084)
tions of motion of the blade by using the global
Galerkin method[16] to transform into algebraic equa- Value
Parameter
tions. The fourth set of equations is derived from the Flight test UMARC[18] Calculation
equations governing unsteady aerodynamics, and also θ0/(°) 13.21 12.84 12.57
using the Galerkin method. However, the equation θ1c/(°) 6.56 4.30 5.06
describing the leading-edge vortex airloads involves θ1s/(°) −9.07 −9.70 −8.45
Boolean calculations, which is not suitable to be θ0t/(°) 8.70 7.22 7.98
solved by the Galerkin method. Therefore, this equa- θb/(°) −6.98 −7.84 −5.30
tion is formulated in the discrete form[17] here. Finally, φb/(°) 0.20 −2.51 0
the four sets of nonlinear algebraic equations are then
Note: θ0—Collective pitch, θ1c—Lateral cyclic pitch, θ1s—Longitudinal
solved through the application of a standard nonlinear cyclic pitch, θ0t—Tail rotor collective, θb—Fuselage pitch angle,
function solver. φb— Fuselage roll angle
For a given flight condition, defined by gross weight,
Figs.2-3 show the predicted airloads of blade section
flight speed, flight path angle and turning rate, the trim
at x0/R=0.865 (R is the rotor radius) together with the
solution can give the control inputs of main and tail
flight test data[19] and results predicted by CAMRAD
rotors, fuselage attitudes, airloads at blade section and
II[19] under the both flight test conditions. There is a
so on.
fair agreement among the flight test data and the theo-
No.3 Li Pan et al. / Chinese Journal of Aeronautics 23(2010) 320-326 · 325 ·
References
[1] Johnson W R. Technology drivers in the development
Fig.3 Nondimensional airloads comparison (μ = 0.24, Cw/ of CAMRAD II. AHS Aeromechanics Specialists Con-
σ = 0.130, x0/R = 0.865). ference. 1994.
[2] Bir G S, Chopra I. University of Maryland advanced
retical results predicted by present model and CAM- rotorcraft code (UMARC) theory manual. UM-AERO
RAD II. The main discrepancy of the airloads between Report 94-18, 1990.
the test data and theoretical results in Fig.2 is the phase [3] Rukowaki M J, Ruzicka G C, Ormiston R A, et al.
lag of negative normal force. The primary contribution Comprehensive aeromechamics analysis of complex
to the discrepancy comes from the coupling of the rotorcraft using 2GCHAS. AHS Aeromechanics Spe-
transonic effects and 3D flow effects (for example, cialists Conference. 1994.
shock relief by yaw flow[18]) in high speed flight, and it [4] Benoit B D, Dequin A M, Kampa K, et al. HOST, a
general helicopter simulation tool for Germany and
is difficult to predict the pitching moment as well as
France. Proceedings of the 56th annual AHS forum.
the blade elastic torsion near the tip. The major differ- 2000.
ence in Fig.3 is the phase shift of the second large [5] Wang H W, Gao Z. Unsteady rotor airloads prediction
negative pitching moment caused by the second dy- using a comprehensive aeroelastic analysis. Journal of
namic stall cycle in the fourth quadrant. In addition, in Nanjing University of Aeronautics & Astronautics
the theoretical calculation the flow does not reattach 2003; 35(3): 268-272. [in Chinese]
· 326 · Li Pan et al. / Chinese Journal of Aeronautics 23(2010) 320-326 No.3
⎡0 0 0 τ 0 0 0⎤
Society 2005; 50(2): 195-205. T
A12 = ⎢ 0 1 0 0 0 0 0 ⎥
[10] He C J. Development and application of a generalized
(A5)
dynamic wake theory for lifting rotors. PhD thesis, ⎢ ⎥
Georgia Institute of Technology, 1989. ⎣⎢ 0 0 1 0 0 0 0 ⎦⎥
[11] Celi R. Helicopter rotor blade aeroelasticity in forward
⎡0 0 0 τ , x τ 0 0 ⎤
T
flight with an implicit structural formulation. AIAA
Journal 1992; 30(9): 2275-2282. A13 = ⎢⎢0 0 0 0 0 0 0 ⎥⎥ (A6)
[12] Howlett J J. UH-60 Black Hawk engineering simula-
tion program: Volume I—mathematical model. NASA ⎢⎣0 0 0 0 0 0 0 ⎥⎦
CR-166309, 1981.
⎡0 0 0 0 0 τ 0 ⎤
T
[13] Leishman J G, Crouse G L. State-space model for un-
steady airfoil behavior and dynamic stall. AIAA- A2 = ⎢1 0 0 0 0 0 0 ⎥ (A7)
⎢ ⎥
1989-1219, 1989.
[14] Leishman J G. Principles of helicopter aerodynamics.
⎣⎢0 0 0 0 0 0 0 ⎦⎥
⎡0 0 0 0 0 0 τ ⎤
T
1st ed. New York: Cambridge University Press, 2000:
309-313, 410-412.
A3 = ⎢ 0 0 0 0 0 0 0 ⎥ (A8)
[15] Davis S J. Predesign study for a modern 4-bladed rotor ⎢ ⎥
for the RSRA. NASA CR-166155, 1981. ⎣⎢1 0 0 0 0 0 0 ⎦⎥
[16] Celi R. Effects of hingeless rotor aeroelasticity on
helicopter longitudinal flight dynamics. Journal of the
⎛ ⎡1 0 " 0⎤ ⎞
⎜ ⎢0 ⎟
0 " 0 ⎥⎥ ⎟
B1 = ⎜⎜ A1 A1T − ⎢
American Helicopter Society 1991; 36(2): 35-44. 1
(A9)
#⎥ ⎟
[17] Leishman J G, Beddoes T S. A generalized method for
2 ⎢# #
unsteady airfoil behavior and dynamic stall using the ⎜ ⎢ ⎥ ⎟
⎜ ⎣0 0 " 0 ⎦ 7×7 ⎟⎠
indicial method. 42nd Annual Form of the AHS. 1986. ⎝
[18] Sitaraman J. CFD based unsteady aerodynamic model-
ing for rotor aeroelastic analysis. PhD thesis, Univer- B2 = ( A1 A2T ) / 2 (A10)
sity of Maryland, 2003.
B3 = ( A1 A3T ) / 2 (A11)
[19] Postsdam M, Yeo H, Johnson W. Rotor airloads predic-
tion using loose aerodynamic/structural coupling. For a linear elastic and isotropic material
Journal of Aircraft 2006; 43(3): 732-742.
⎡E 0 0 ⎤
Biographies: Q = [qij ] = ⎢ 0 2G 0 ⎥⎥
⎢ (A12)
Li Pan Born in 1981, he received B.S. degree from Nan- ⎢⎣ 0 0 2G ⎥⎦
jing University of Aeronautics and Astronautics (NUAA) in where E is elastic modulus and G the shear modulus of
2004, and now he is a Ph.D. candidate of NUAA. His main the material.
research interests include helicopter aerodynamics, flight Ci = qij B j (i, j = 1, 2,3) (A13)
dynamics and aeroelasticity.
E-mail: lipan@nuaa.edu.cn The expressions of structural operators are