Professional Documents
Culture Documents
3 2023 Jae
3 2023 Jae
net/publication/360710740
High-efficiency hybrid trim method for CFD simulation of rigid coaxial rotor
Article in Proceedings of the Institution of Mechanical Engineers Part G Journal of Aerospace Engineering · May 2022
DOI: 10.1177/09544100221095369
CITATION READS
1 61
5 authors, including:
Haotian Qi
Nanjing University of Aeronautics & Astronautics
5 PUBLICATIONS 46 CITATIONS
SEE PROFILE
All content following this page was uploaded by Haotian Qi on 06 March 2023.
Abstract
In order to solve the trim problem of computational fluid dynamics (CFD) simulation for rigid coaxial rotor, a hybrid trim
model coupling CFD method and high-efficiency model (blade element theory) is established. In the trim process, the
Jacobian matrix is solved by the high-efficiency model, while the CFD solver is only called for rotor performances modifying
after each trimming step. The influences of pseudo time step, number of CFD revolution, and inflow model are investigated.
Validation cases of AH-1G and Harrington-1 rotors are carried out, and good agreements are obtained. Results show that
the trim efficiency can be significantly improved by saving the calculation of CFD for the Jacobian matrix. The trim accuracy
is guaranteed by the correction operation with the CFD solver at each step. Moreover, as only time-averaged rotor
performances are useful for trim, the efficiency can be further improved by adopting appropriately small pseudo time step
and CFD revolution. The hybrid trim model has high robustness. The accuracy of inflow model for the coaxial rotor affects
the convergence speed, but the final convergence can be achieved generally.
Keywords
helicopter, rigid coaxial rotor, trim method, CFD, inflow model
Date received: 25 August 2021; revised: 17 February 2022; accepted: 31 March 2022
Figure 1. Sketch of moving overset mesh for coaxial rotor. (a) C-O type blade mesh; (b) Overset mesh system.
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
According to the thrust distribution of a coaxial rigid rotor u0 12
u
in torque balance calculated by the CFD solver, the ratio of u λ∞ þ AAc λu
uBσClα C
λðr, λ∞ Þ ¼ u B C þ σClα θl r
t@ 16F
the upper and lower rotor thrusts in hover can be set as 1.2 A
for current cases. 2 8F
(8)
BEMT model for hover. Another aerodynamic model
λ∞ þ A
= !
considering the interaction of coaxial rotor has been σClα Ac
Figure 2. Sketch map of rotor blade element theory in forward flight. (a) Top view; (b) Side view.
where μ represents the rotor advance ratio. α indicates the Figure 3. The induced velocity distributions of the twin
angle between the free stream and the rotor disk plane. The rotors in forward flight can be expressed as
initial value of induced velocity is 2 3
The formulas of other rotor forces and moments can be þ A A’U VX2 þ VZ þ WU WU 5
easily deduced similar to this.
(19)
Modified Drees inflow model for forward flight. The non-
Z Z
uniform induced velocity in the rotor disk for a single rotor Nb ρ R 2π
Clα ðθL fL Þ
TL ¼ c UL2 drdψ
in forward flight can be approximately expressed as 4π R0 0 cosðfL Þ
a linear distribution formula
0 sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2
W ðr,ψÞ ¼ W ð1 þ κX r cos ψ þ κY r sin ψÞ (15) ¼ 2ρ@A’L VX2 þ VZ þ WL þ εðHÞWU WL
where W represents the average value of the induced sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1
velocity of the whole rotor disk. The expressions of
2
harmonic term coefficients κX and κY have been given by þ A A’L VX2 þ VZ þ WL WL A
Drees,25 which can called as Drees inflow model. The
original Drees inflow is proposed for single rotor. (20)
Valkov26 has applied it to the coaxial rotor according to the where
interaction characteristics of rotor wakes, as shown in
Qi et al. 145
Figure 3. Sketch map for wake interaction of a coaxial rotor.26 (a) Upper rotor; (b) Lower rotor.
Figure 4. Induced velocity distribution in rotor disk (μ = 0.1). (a) Upper rotor; (b) Lower rotor.
Figure 5. Induced velocity distribution in rotor disk (μ = 0.3). (a) Upper rotor; (b) Lower rotor.
" #
disk increases due to the interaction of the lower rotor.
UU ðr,ψÞ ¼ Vr þ VX sinðψÞ
(21) Correspondingly, the induced velocity of the lower rotor
UL ðr,ψÞ ¼ Vr VX sinðψÞ
behind the rotor disk is enhanced. When μ=0.3, the
interaction of rotor wakes is much weak. The induced
Through iteratively solving of equations (19) and (20),
velocity of the upper rotor is almost the same as the
uniform induced velocity coefficients of the twin rotors
original Dress inflow solution.
WU , WL can be obtained. Then, the distribution of induced
velocity in the rotor disk can be achieved according to
equation (16). Hybrid trim model
Figure 4 and Figure 5 show the induced velocity
The blade pitch of an isolated rotor for the coaxial system
distributions of a coaxial rotor at different advance ratio
can be expressed as
solved by the modified Drees inflow model. When μ=0.1,
the induced velocity of the upper rotor in the front of rotor θðψÞ ¼ θ0 þ θ1s sinðψÞ þ θ1c cosðψÞ þ / (22)
146 Proc IMechE Part G: J Aerospace Engineering 237(1)
where ψ represents azimuth angle, θ0 is collective pitch xðnþ1Þ ¼ xðnÞ þ ΔxðnÞ (29)
angle, θ1s and θ1c denote longitudinal and lateral cyclic
1
pitches, respectively. ΔxðnÞ ¼ J ðnÞ ½yðnÞ ytarget (30)
The control settings (x) and target variables (y) are as
shown below where ytarget is the target variables.
T
The pitch angles are updated through continuous
x ¼ ðθ0U ,θ1sU ,θ1cU ,θ0L ,θ1sL ,θ1cL Þ (23) iterations. At each step, they are adopted into the aero-
T
dynamic model to judge whether the target variables are
y ¼ CT ,CQ ,LOS,CMzU ,CMx ,CMzL (24) achieved according to the convergence criterion. If so, the
trim is completed. If not, it proceeds to the next iteration.
where CT and CQ represent the total thrust and torque
In this paper, the convergence criterion is
coefficients of the coaxial system. CMx and CMz are the
coefficients of rotor rolling and pitching moments. jεj < 1 × 102 (31)
At forward flight, when the rolling moments are bal-
anced, thrusts of the upper and lower rotors are not where ε is the iteration residual (error of y relative to
necessarily equal. If a separate upper (lower) rotor is used ytarget ), which can be expressed as
to define the lift offset, there will be some deviations. T
Considering this, average LOS is used to define the LOS ε ¼ εT ,εQ ,εLOS ,εMzU ,εMx ,εMzL (32)
of rigid coaxial rotor, which can be expressed as 2 ðCTU þ CTL Þ 3
εT ¼
ðCMxU CMxL Þ 6 CTtarget 7
LOS ¼ ¼ CTU þ CTL (25) 6 7
CT , CT 6 7
6 7
6 7
The relationship between control settings and target 6 εQ ¼ CQU þ CQL × 104 7
6 7
variables can be expressed by 6 7
6 7
6 7 (33)
6 εLOS ¼ LOS LOS Target 7
y ¼ f ðxÞ (26) 6 7
6 7
6 7
According to Newton iterative method, the above 6 εMzU ¼ CMzU × 104 7
6 7
formula can be expanded by Taylor series to the first order 6 7
6 εMx ¼ CMx × 104 7
and simplified. Then, the change of the control settings Δx 4 5
in iterating can be written as εMz ¼ CMz × 104
Figure 6. Flow chart of hybrid trim method for rigid coaxial rotor.
Parameter Values
and Newton iterative formula. However, due to the model Figure 7. Temporal CT of the upper rotor for different
error, the rotor performances calculated by the high- pseudo time steps.
efficiency model may have deviation with the CFD
solver adopting same pitches. Therefore, it is necessary to
utilize the CFD solver to correct the rotor performances
after each iteration. In the trim model, the CFD solver does
Table 2. Time-averaged CT of upper rotor calculated with
not directly participate in the trim iteration, and only different pseudo time steps.
performs once after each trim iteration. Thus, the effi-
ciency can be greatly improved. Ncyc CTU Relative error
For distinguish, the final target variables to be achieved
5 0.00307 6.11%
are written as ytarget . And the target variables in the
9 0.00316 3.36%
trimming of high-efficiency model are marked as ytarget e ,
15 0.00326 0.30%
which are updated with the trim iteration to drive the
update of control settings. Figure 6 illustrates the flow 20 0.00327 Reference value
148 Proc IMechE Part G: J Aerospace Engineering 237(1)
Figure 8. Influence of calculation revolution on CT in hover (CT = 0.006). (a) Convergence history; (b) Time-averaged results; (c)
Temporal CT of the upper rotor; (d) Temporal CT of the lower rotor.
Figure 9. Influence of calculation revolution on CT in forward flight (μ = 0.15, LOS = 0.1). (a) Temporal CT of the upper rotor; (b)
Temporal CT of the lower rotor.
chart of the hybrid trim method. The specific trim pro- 4. According to the rotor performances calculated by the
cedure is shown below. CFD solver, the results calculated by the high-
efficiency model are modified through setting new
1. The target variables are assigned to the temporary temporary target variables, that is
target variables that need to be trimmed by the high-
efficiency aerodynamic model, that is, ðnÞ
targetð0Þ yetargetðnþ1Þ ¼ ytargetðnÞ
e þ ytarget yCFD (34)
ye ¼ ytarget .
targetð0Þ
2. According to ye , the trimmed pitches are ob-
tained by adopting the high-efficiency model and 5. According to the new temporary target variables
targetðnþ1Þ
Newton iterative method, which are the initial pitches ye , running the high-efficiency trim model
xð0Þ . again to obtain new pitches xðnþ1Þ .
3. The initial pitches are adopted into the CFD solver for 6. Applying the new pitches xðnþ1Þ into the CFD solver
some revolutions to obtain the rotor performances for a certain number of revolutions to achieve new
ð0Þ ðnþ1Þ
yCFD . rotor performances yCFD .
Qi et al. 149
Table 3. Trimmed collective pitches of the coaxial rotor under different setups (μ=0.15).
Trim setup θ0U (°) θ0L (°) Relative error of θ0U Relative error of θ0L
Figure 11. Trim histories for different inflow models in hover (CT = 0.009). (a) Collective pitches; (b) Residuals of rotor
performances.
Figure 12. Trim histories for different inflow models in forward flight (μ = 0.15, LOS = 0.1). (a) Uniform inflow; (b) Drees model; (c)
Modified Drees model; (d) Full CFD trim method.
revolutions is small. In comparison, the thrust of the lower revolution. The forward flight state converges more easily
rotor in the eighth revolution still shows obvious different than the hover state. This is because the initial velocity of
from the sixth revolution. This indicates that, the thrust of the flow field in hover is zero, and the development of
the lower rotor is more difficult to converge. This is be- rotor wakes absolutely relies on the induced flow gen-
cause the temporal thrust caused by the interaction from erated by the rotor rotating. In comparison, there is
the lower rotor reaches convergence only when the upper a certain initial velocity for the flow field at forward flight,
rotor wake has been fully developed. In view of this, in which makes the upper rotor wake moves obliquely
order to make the flow field converge, eight revolutions downward rapidly. On one hand, the rotor wake converges
are usually calculated for a coaxial rotor in hover for more easily. On the other hand, this weakens the rotor
detailed analysis. interaction, as the wake has a backward partial velocity.
Figure 9 shows the influence of revolution number on Therefore, the CFD revolution number for detailed
CT of the coaxial rotor in forward flight. Results of the analysis at forward flight is usually set as four. Consid-
fourth and fifth revolutions are pretty close, indicating that ering that the CFD calculation would also be carried out in
the temporal thrusts have converged at the fourth the following iterations, for the initial trimming step the
Qi et al. 151
Figure 13. Residuals of trim process for different inflow models (μ = 0.15, LOS = 0.1). (a) Drees model; (b) Modified Drees model.
Table 4. CFD revolutions and time consumptions for different trim methods.
Initial step Iteration Solution of Jacobian Total number of CFD Total computation
Trim method (rev) steps (rev) matrix (rev) revolution (rev) time (h)
Figure 14. Trim history of the coaxial rotor in high-speed flight (μ = 0.4, LOS = 0.1). (a) Pitch angles; (b) Residuals.
revolution number is set as four for hover and three for Figure 10 shows the trim history of cyclic pitches for
forward flight in this paper. different Ncyc. Both of them meet the trim criterion in six
According to the basic shape of rotor wake, in hover steps, showing that the pseudo time step has little effect on
and low-speed flight the aerodynamic interactions for the trim efficiency. The difference of θ1c is more obvious
coaxial rotor are much strong. There are only two pitches than θ1s . θ1s corresponds to the lateral asymmetry of flow
in hover, while there are six in forward flight. Therefore, field, which is mainly determined by the advance ratio, so
low-speed flight is a typically difficult trim state for the the pseudo time step has little effect on it. θ1c corresponds
coaxial rotor. So, it is more meaningful to analyze the to the longitudinal asymmetry. The interaction of the
influence of trim parameters at this state. The state of μ = coaxial rotor further causes the lateral asymmetry of flow
0.15 is investigated by adopting different pseudo time field. This interaction can be simulated more accurately by
steps (Ncyc) and CFD revolutions (Nrev). The target var- adopting larger pseudo time. Therefore, the trim results of
iables are CT = 0.01, LOS = 0.1, and the other four target θ1c meet more obvious differences.
values are all zero. Table 3 shows the trimmed collective In summary, although different pseudo time steps have
pitches for different setups. The results of Ncyc = 9 are impact on the rotor temporal thrusts, the difference of
similar, and the results of Ncyc = 15 are similar. It can be time-averaged values is small. Therefore, smaller pseudo
seen that the influence of pseudo time step is greater than time steps can be selected to improve the efficiency. For
the CFD revolution number. each trim step after the initial calculation, one revolution
152 Proc IMechE Part G: J Aerospace Engineering 237(1)
Figure 15. Sectional Cn at different spanwise sections. (a) 0.6 R; (b) 0.75 R; (c) 0.97 R; (d) 0.99 R.
Table 5. Trimmed pitch angles for AH-1G rotor. The convergence processes of the full CFD method and
modified Drees model are similar, both within six steps.
Pitch angles θ0 (°) θ1s (°) θ1c (°)
Table 4 gives the time consumptions for the two trim
Flight test 6.0 5.5 1.7 methods. The cases are carried out on a 24-core computer.
Reference11 6.1 5.1 1.3 For the full CFD method, after each step, the CFD solver
Hybrid model 6.2 4.8 1.4 needs to be called for six revolutions to solve the six
columns of the Jacobian matrix. The total number of CFD
revolution is 38. For the hybrid trim method, the time
of CFD solver can meet the requirement for trimming. In consumption of the high-efficiency model can be ignored
this paper, the pseudo time step for trimming is set as 15 in compared with the CFD solver. So, the calculation amount
the state of strong aerodynamic interaction (μ < 0.2). And of CFD in each step can be saved by six revolutions. Using
at a larger advance ratio (μ ≥ 0.2), the rotor interaction is the modified Drees model, only eight revolutions are
weak, and the pseudo time step can be set as 9. needed to complete the trim. The total time consumption
has been saved by about 78%. This illustrates that, the
Influence of inflow model. Figure 11 compares the trim hybrid trim method established in this paper can signifi-
histories of different inflow models in hover. The uniform cantly improve the trim efficiency.
inflow model converges after seven steps, while the BEMT Figure 14 shows the trim history of the coaxial rotor,
model converges after five steps. The trimmed collective when μ = 0.4, CT = 0.01, LOS = 0.1. The difference of
pitches are nearly same. This shows that the trim efficiency trimmed pitches of the upper and lower rotors are small.
can be improved by using the BEMT model, as the in- And the trim quickly meets the convergence criterion in
teraction of the coaxial rotor is estimated more accurately. only five steps. This is because the rotor interaction is
Figure 12 and Figure 13 show the inflow model effect much weak at a large advance ratio. Note that the induced
on the trimming of the low-speed flight. The state is μ = velocities calculated by the modified Dress inflow model
0.15, CT = 0.01, LOS = 0.1. Here, Drees model means that and the original Dress inflow model are almost same, as
the original Drees model without rotor interaction is di- mentioned above. So, there is no need to use the modified
rectly applied to the coaxial rotor. History of full CFD trim Dress inflow model at this state.
method is also given for comparison. In order to ensure the
comparability of trimming, the full CFD method adopts
Validation cases
the same initial pitches with the modified Drees inflow.
The trimmed pitches of the four setups are almost same, as AH-1G single rotor. At present, there is still limited ex-
each of them is verified by the same convergence criterion. perimental data about trim for coaxial rotor. The flight test
The Drees and the uniform inflows both converge within of AH-1G rotor 28 conducted by NASA Ames Research
eight steps. The modified Drees inflow model converges Center has been widely used for the trim validation of
faster, as it has higher accuracy. single rotor. The flight state is μ=0.19, Matip=0.65,
Qi et al. 153
Pitch angles θ0U (°) θ1sU (°) θ1cU (°) θ0L (°) θ1sL (°) θ1cL (°)
revolution is suggested to be four for hover and three 8. Lyu W, Wang S and Yang A. Some improvements of hybrid
for forward flight for current cases. For the following trim method for a helicopter rotor in forward flight. Aerosp
trim steps, it can be just one, as each change of control Sci Technol 2021; 113: 106709. DOI: 10.1016/j.ast.2021.
settings is much small, and the simulation can fast 106709.
9. Yuan Y, Chen R and Li P. Trim investigation for coaxial rigid
converge. Smaller pseudo time step can be used for
rotor helicopters using an improved aerodynamic in-
trim at high advance ratio (such as μ ≥ 0.2), as the
terference model. Aerosp Sci Technol 2019; 85: 293–304.
interaction turns weak. DOI: 10.1016/j.ast.2018.11.044.
3. The hybrid trim model for rigid coaxial rotor has high 10. Chung K, Hwang C, Jeon W, et al. Numerical predictions of
robustness. The accuracies of the high-efficiency rotorcraft unsteady air-loadings and BVI noise by using
aerodynamic models are different when adopting a time-marching free-wake and acoustic analogy. Firenze,
different inflow models. A more accurate high- Italy: 31st European Rotorcraft Forum, 2005, pp. 1071–1075.
efficiency model makes the trim converge faster. 11. Kim JW, Park SH and Yu YH. Euler and navier-stokes
However, they all can achieve the convergence by simulations of helicopter rotor blade in forward flight using
coupling with the CFD solver. an overlapped grid solver. In: AIAA Computational Fluid
Dynamics. San Antonio, 2009.
12. Ma Y, Chen M, Wang Q, et al. Main helicopter rotor
Declaration of conflicting interests
trimming using computational fluid dynamics method in
The author(s) declared no potential conflicts of interest with forward flight. P I Mech Eng G-J Aer 2018; 232: 169–179.
respect to the research, authorship, and/or publication of this DOI: 10.1177/0954410016673394.
article. 13. Zhao JG and He CJ. A viscous vortex particle model for rotor
wake and interference analysis. J Am Helicopter Soc 2010;
Funding 55: 12007. DOI: 10.4050/JAHS.55.012007.
The author(s) disclosed receipt of the following financial support 14. Ye Z, Xu G, Shi Y, et al. A high-efficiency trim method for
for the research, authorship, and/or publication of this article: CFD numerical calculation of helicopter rotors. Int J Aeronaut
This work was supported by the National Natural Science Space 2017; 18: 186–196. DOI: 10.5139/IJASS.2017.18.2.23.
Foundation of China (No. 12102154), the Foundation of Key 15. Lakshminarayan VK and Baeder JD. Computational in-
Laboratory of Rotor Aerodynamics of China Aerodynamics vestigation of microscale coaxial-rotor aerodynamics in
Research and Development Center (No. RAL202103-2), and hover. J Aircr 2010; 47: 940–955. DOI: 10.2514/1.46530.
High-tech Research Key laboratory of Zhenjiang (SS2018002). 16. Qi H, Xu G, Lu C, et al. A study of coaxial rotor aero-
dynamic interaction mechanism in hover with high-efficient
ORCID iDs trim model. Aerosp Sci Technol 2019; 84: 1116–1130. DOI:
10.1016/j.ast.2018.11.053.
Haotian Qi https://orcid.org/0000-0001-8718-0558 17. Qi H, Xu G, Lu C, et al. Computational investigation on
Liangquan Wang https://orcid.org/0000-0003-0592-1071 unsteady loads of high-speed rigid coaxial rotor with high-
efficient trim model. Int J Aeronaut Space 2019; 20: 16–30,
References Article. DOI: 10.1007/s42405-018-0133-0.
1. Yuan Y, Thomson D and Chen R. Investigation of lift offset 18. Pomin H and Wagner S. Navier-stokes analysis of helicopter
on flight dynamics characteristics for coaxial compound rotor aerodynamics in hover and forward flight. J Aircr
helicopters. J Aircr 2019; 56: 2210–2222. DOI: 10.2514/1. 2002; 39: 813–821. DOI: 10.2514/2.3001.
C035190. 19. Roe PL. Approximate Riemann solvers, parameter vectors,
2. Bagai A. Aerodynamic design of the X2 technology dem- and difference schemes. J Comput Phys 1981; 43: 357–372.
onstrator main rotor blade. Canada: 64th American Heli- DOI: 10.1016/0021-9991(81)90128-5.
copter Society Annual Forum. MontréalAHS, 2008, pp. 20. Van Leer B.. Towards the ultimate conservative difference
29–44. scheme. V. A second-order sequel to Godunov’s method.
3. Burgess RK. The ABCÔ Rotor – A historical perspective. J Comput Phys 1979; 32: 101–136. DOI: 10.1016/0021-
Baltimore, MD: 60th American Helicopter Society Annual 9991(79)90145-1.
Forum, 2004. 21. Spalart P and Allmaras S. A one-equation turbulence model
4. Ferguson K and Thomson D. Performance comparison for aerodynamic flows. In: 30th Aerospace Sciences Meeting
between a conventional helicopter and compound helicopter and Exhibit. Reno, NV: AIAA, 1992, pp. 1–22.
configurations. P I Mech Eng G-J Aer 2015; 229: 22. Yoon S and Jameson A. Lower-upper symmetric-gauss-
2441–2456. DOI: 10.1177/0954410015577997. Seidel method for the Euler and Navier-stokes equations.
5. Zhao QJ, Zhao GQ, Wang B, et al. Robust Navier-Stokes AIAA J 1988; 26: 1025–1026. DOI: 10.2514/3.10007.
method for predicting unsteady flowfield and aerodynamic 23. Johnson W. Helicopter theory. NY, USA: Dover Pub-
characteristics of helicopter rotor. Chin J Aeronaut 2018; 31: lications, 1994, pp. 903–920.
214–224. DOI: 10.1016/j.cja.2017.10.005. 24. Leishman JG and Ananthan S. An optimum coaxial rotor
6. Zhang K, Zhao Q, Zhang X, et al. Analysis of rotor aero- system for axial flight. J Am Helicopter Soc 2008; 53:
dynamic response during ramp collective pitch increase by 366–381. DOI: 10.4050/JAHS.53.366.
CFD method. P I Mech Eng G-J Aer 2021; 235: 994–1007. 25. Drees JM. A theory of airflow through rotors and its ap-
DOI: 10.1177/0954410020965404. plication to some helicopter problems. J Helicopter Assoc
7. Romani G and Casalino D. Rotorcraft blade-vortex in- Great Britain 1949; 3: 79–104.
teraction noise prediction using the lattice-Boltzmann 26. Valkov T. Aerodynamic loads computation on coaxial
method. Aerosp Sci Technol 2019; 88: 147–157. DOI: 10. hingeless helicopter rotors. In: 28th Aerospace Sciences
1016/j.ast.2019.03.029. Meeting. Reno, NV, USA: AIAA; 1990.
Qi et al. 155