About XFLR5 calculations and experimental measurements
XFLR 5
Revision 1.1 – Copyright A. Deperrois – October 2009
The experiment – General comments The experiment has been set up and carried out by Matthieu Scherrer's team at the CEAT in Toulouse, France, beginning of 2008 – thanks to them all Details can be found at http://sailplane-matscherrer.blogspot.com/ The predictions published at the address above had been provided before the measurements were available Francesco Meschia used XFLR5 V3.21 / VLM – his results are referred to as "FMe" – thanks, Francesco The author used XFLR5 V4.00, which unfortunately was finished in a hurry and was not totally reliable at the time – and it's an understatement Since then, the code has been debugged and improved, the new results with comments are provided in the following slides The validity of the measurements has not been questioned
Revision 1.1 – Copyright A. Deperrois – October 2009
The test sailplane
Revision 1.1 – Copyright A. Deperrois – October 2009
The model
The analysis has been run with and without the body, using either LLT, 3D panels or VLM methods Revision 1.1 – Copyright A. Deperrois – October 2009
The sign conventions Cl>0
CL>0 CD>0 α>0 Cm>0 Cy>0 β>0
Cn>0
CL CD Cm Cy Cl Cn
Revision 1.1 – Copyright A. Deperrois – October 2009
ref area ref length SWing SWing SWing MAC SWing SWing MAC SWing MAC
axis system stability axis stability axis stability axis A/C axis A/C axis A/C axis
Lift Curve – No sideslip Lift curve Measurement vs prediction - V=20m/s
Lift curve Measurement vs prediction - V=40m/s
1.25
1.25
1.00
1 FMe Meas ure V=40 Meas ure V=40 0.75 Cl - VLM2 - XFLR5_V4.09 Cl - Panels with body - XFLR5_V4.09 Cl - Panels No Body - XFLR5_V4.09 0.5
0.75
CL
CL
0.50
0.25
0.25 FMe Meas ure V=20 Meas 5 ure V=20 10 Cl - VLM2 - XFLR5_V4.09 Cl - Panels with body - XFLR5_V4.09 Cl - Panels No Body - XFLR5_V4.09 Cl - LLT - XFLR5_V4.09
0.00 -5
0 -0.25
-0.50
0 -5
0
5
10
-0.25
-0.5 α (°)
α (°)
o All methods LLT, VLM and Panels predict correctly the value of the zero-lift angle, in this case ~-1.25° o The LLT is the method which fits best the non-linearity of the lift curve o All methods tend to underestimate the decrease in lift at high a.o.a. ; the LLT is the method which gives the most realistic trend
Revision 1.1 – Copyright A. Deperrois – October 2009
Drag Polar – No sideslip Drag polar Measurement vs prediction - V=40m/s
Drag polar Measurement vs prediction - V=20m/s 1.25
1.25
1.00
1
0.75
0.75
CL 0.25
0.00 0.000
0.025
0.050
0.075
0.25
0 0
0.100
-0.25
-0.25
-0.50
-0.5
CD
FMe V=40 V=40 V=40 Cl - VLM2 - XFLR5_V4.09 Cl - Panels with body - XFLR5_V4.09 Cl - Panels No Body - XFLR5_V4.09
0.5 CL
FMe V=20 V=20 V=20 Cl - VLM2 - XFLR5_V4.09 Cl - Panels with body - XFLR5_V4.09 Cl - Panels No Body - XFLR5_V4.09 Cl - LLT - XFLR5_V4.09
0.50
0.025
0.05
0.075
0.1
CD
o All methods, LLT, VLM and Panels tend to underestimate the total drag o It is difficult to tell which of the induced or viscous drag is underestimated, but my guess would be that it's the viscous part o This could be due to several causes : the conditions in the wind tunnel are not as laminar as expected, the flow transitions from laminar to turbulent at some point along the wing's chord inadequate values for NCrit are used in XFoil when building the foil polar mesh The 3D interpolation of 2D viscous results underestimates the viscous drag Revision 1.1 – Copyright A. Deperrois – October 2009
Pitching moment – No sideslip Pitching moment curve Measurement vs prediction - V=40m/s
Pitching moment curve Measurement vs prediction - V=20m/s 0.10
0.10
FMe Meas ure V=20 Meas ure V=20 Cm - VLM2 - XFLR5_V4.09 Cm - Panels with body - XFLR5_V4.09 Cm - Panels No Body - XFLR5_V4.09
0.05
0.05
-5
0
5
-5
10
0
5
10
-0.05 CM
-0.05 CM
α (°)
0.00
0.00
-0.10 -0.15
-0.10 -0.15 FMe Meas ure V=40-0.20 Meas ure V=40 Cm - VLM2 - XFLR5_V4.09 Cm - Panels with body - XFLR5_V4.09 -0.25 Cm - Panels No Body - XFLR5_V4.09
-0.20 -0.25 -0.30
-0.30 α (°)
o All methods, LLT, VLM and Panels predict correctly the moment coefficient Cm 0 at zero lift, and the lift coefficient Cl0 at zero-moment except for the model which includes the body o Except for the Panel method with body, all methods give an adequate trend for the slope Cm = f(α) o The modeling of the body seems to generate considerable numerical noise ; this could be due to the difficulty to model connections between wing and body Revision 1.1 – Copyright A. Deperrois – October 2009
Notes about sideslip The simulation of sideslip has been introduced in XFLR5 v4.09 The order in which a.o.a. and sideslip are applied has its importance In XFLR5, sideslip is modeled by rotating the model about the z-axis The resulting model is analyzed using the conventional VLM and panel methods This method has been preferred because it is simple to implement, however the usual convention is to apply the angle of attack first, then the sideslip rotation As a result, the model's position is not exactly the same at high a.o.a. or sideslip angles than it is in the experiment
The rolling moment, yawing moment and lateral force coefficients are issued from the non-viscous part of the VLM and Panel analysis, hence are the same for all speeds; experimentally though, a difference has been measured which would tend to show that the viscosity influences the distribution of pressure forces
Revision 1.1 – Copyright A. Deperrois – October 2009
Results for sideslip – lateral force Jibe 2 pre diction vs M eas ure m ent Lateral coe fficie nts at α =2
Jibe 2 pre diction vs M e as ure m e nt Late ral coe fficie nts at α =6
0.08
0.08
0.06
0.06 0.04
0.04
0.02 0.00
0 -20
-10
0
10
20
Cy
Cy
0.02 -20
-10
Measure Cy V=20 Measure Cy V=40 Cy - VLM2 Cy - 3D Panels
-0.06 -0.08 β ( °)
0
10
-0.04
-0.02 -0.04
-0.02
Measure Cy V=20 Measure Cy V=40 Cy - VLM2 Cy - 3D Panels
-0.06 -0.08 -0.10 -0.12 β (°)
o Lateral force prediction is satisfactory although not as precise as lift coefficient prediction
Revision 1.1 – Copyright A. Deperrois – October 2009
20
Results for sideslip – Rolling moment Jibe 2 prediction vs M e as ure m e nt Late ral coe fficie nts at α =2
Jibe 2 pre diction vs M e as ure m e nt Late ral coe fficie nts at α =6 0.15
0.08 0.06
0.10
0.04 0.05
0.02
-20
-10
0.00
Cl
Cl
0 0
10
20
-20
-10
-0.02 Measure Cl V=20 Measure Cl V=40 GRm - VLM2
-0.04 -0.06
GRm - 3D Panels -0.08 β (°)
0
10
20
-0.05 Measure Cl V=20 Measure Cl V=40 GRm - VLM2 GRm - 3D Panels
-0.10
-0.15 β ( °)
o Sideslip generates a rolling moment ; this is the basis of 2 axis rudder-elevator flight o For this particular plane with no dihedral, this moment is low and thus difficult to predict Revision 1.1 – Copyright A. Deperrois – October 2009
Results for sideslip – yawing moment Jibe 2 pre diction vs M e asure m e nt Late ral coe fficients at α =6
Jibe 2 pre diction vs M eas ure m e nt Late ral coe fficients at α =2° 0.35 Measure Cn V=20
0.30
Measure Cn V=40
0.25
GYm - VLM2
0.20
GYm - 3D Panels
0.40 Measure Cn V=20 Measure Cn V=40 GYm - VLM2 GYm - 3D Panels
0.30
0.20
0.15 Cn
Cn
0.10 0.05
0.10
0.00 -20
-10
-0.05
0
10
-0.10
20
0.00 -20
-10
0 -0.10
-0.15 -0.20 β ( °)
-0.20 β (°)
o Yawing moment predictions give the correct trend – no more
Revision 1.1 – Copyright A. Deperrois – October 2009
10
20
General conclusions The VLM analysis is precise enough for most applications LLT is useful where precise lift curves are required, especially to account for viscous effects The 3D Panel method does not improve notably the accuracy of the results All methods tend to underestimate the drag – probably its viscous part The simulation of the body is more a nuisance than a help
Revision 1.1 – Copyright A. Deperrois – October 2009
In the hope that this helped !
Revision 1.1 – Copyright A. Deperrois – October 2009
