Non-Linear Lifting-Line Solver

S. J. Hulshoff, Faculty of Aerospace Engineering, Delft University of Technology

([email protected])

This page describes a non-linear lifting-line method which can predict the lift, drag and pitching moment
coefficients of wing and tail configurations at low speeds. The method makes use of experimental section
data, and can be used to estimate maximum lift conditions.

Method Description
Limitations
Running the Program
Interpreting the Results

Method Description
The lifting-line concept, developed by Prandtl in the 1910's, has seen wide application to the prediction of
finite-wing characteristics. Although since superseded by far more elaborate numerical methods, it remains
useful in the conceptual design stage, particularly for predicting the maximum lift of wings using non-linear
experimental data. The non-linear lifting line concept has been extended here for use with multiple lifting
surfaces by using a simplified implementation of the vortex-lattice method.

The approach is illustrated in figure 1. Each lifting surface is replaced by a single row of horseshoe vortices,
extending along the quarter-chord line. In a traditional vortex-lattice method, a flow-tangency condition is
applied at a three-quarter-chord control point aft of each horseshoe vortex, leading to a system of linear
equations for the unknown vortex strengths. In the current program, this flow-tangency condition is modified
to include the deviation of experimental section data from the theoretical thin-airfoil lift curve. As this
deviation depends on the local angle of attack, which in turn depends on the velocity field induced by the
horseshoe vortices, the resulting problem is non-linear, and is solved iteratively.
 Figure 1: Vortex-lattice method applied to a lifting line

In general the method can predict total forces quite accurately, with the exception of the induced drag, which
tends to be less accurate at both low and high aspect ratios. Some non-linear lifting-line methods use a
detailed integration of spanwise induced angle of attack to partially overcome this. Although this approach is
not used here, the current method is general enough to be applied to swept or non-planar lifting lines, and can
account for interactions occurring between multiple lifting wings.

Limitations
The vortex-lattice method used to predict the induced velocity field is in essence a singularity-superposition
technique for solving Laplace's equation for incompressible, inviscid flows. Thus the estimated local angle of
attack values predicted are inaccurate for Mach numbers greater than 0.3, and do not take into account any
additional disturbances arising from large separated regions. As the method is applied to lifting lines, it also
becomes inaccurate for wings with aspect ratios below 4.

The prediction of maximum lift relies on the choice of experimental data, for which the airfoil section, its
roughness and Reynolds number should be as close to the anticipated situation as possible. Spanwise load
distributions computed beyond the stall of the first airfoil section will likely be unrealistic, as wing stall
includes highly three-dimensional processes which cannot be simulated here. In spite of this, the prediction
of CLmax is normally pretty good, and the reasonable agreement of post-stall force coefficients shows that at
least some of the features of post-stall stall behaviour can be approximated with the non-linear lifting-line
technique.

Two components of drag are predicted, profile drag (CDp on the results page) and induced drag (CDi on the
results page). The profile drag is estimated using the provided section Cd data, evaluated at the local angle of
attack computed with the vortex-lattice method, and then integrated along the span. The induced drag is
computed via force integration, which is not always reliable for a line of horseshoe vortices. Although a
correction has been applied, be sure to check the computed induced drag value with other sources.
Running the Program
Running the program should be as simple as filling out the html form, pressing "Submit Run" and waiting for
the results to return. If the run was successful, they will include plots of the geometry, residual history,
integrated lift coefficients versus angle of attack, and the spanwise load distributions for the last angle of
attack requested. The form input data is described below:

Section 1: Global Data

# Alpha: Number of angles attack for the run (Maximum 15)

Alpha min: Starting angle of attack

Alpha max: Ending angle of attack

Reference Area: Factor used to non-dimensionalize the total force and moment coefficients
(normally equal to the main wing area)

Reference Length: Factor used to non-dimensionalize the total moment coefficient (normally
equal to the main wing MAC)

Reference Point: Used to compute moments. Normally the wing MAC quarter-chord or the
centre of gravity

[v]: The y=0 Symmetry plane is set on by default. It can be deactivated in order to consider an
asymmetric geometry by clicking in the checkbox.

[ ]: The Ground plane can be activated to simulate ground effect. This works best when the
lowest lifting surface is more than 0.2 semi-spans above the z-location specified.

Section 2: Numerical Parameters

Relaxation Factor: Controls iterative update of circulation values. (Normally set to values
below 0.1, particularly when computing above stall or with larger numbers of panels)

Tolerance: Acceptable error in the values of circulation. Used to stop the iteration process

Max iterations: The maximum number of iterations per angle of attack

k2: Second order-dissipation coefficient. (Can be set to zero for positive-lift-curve-slope range.
When computing near stall, values near 1.0 are appropriate. Otherwise, set to zero or a small
value. Excessive dissipation can result in inaccurate solutions or lack of convergence.

k4: Fourth-order dissipation coefficient. (Less effective than k2, but also has less impact on
solution accuracy. Keep to zero if possible.)

Section 3: Lifting Surface Parameters (Repeated)

[ ]: Click in the checkbox to activate the surface, otherwise it will be excluded from the
calculation

Name: Name for the surface

 # panels: Number of spanwise panels (horseshoe vortices)

dist: Panel distribution:

0: Constant
1: .35 cosine
2: .40 cosine
3. .45 cosine
4. half-cosine
10: full-cosine

x,y,z c/4: Root and tip quarter-chord coordinates

chord: Chord length at root and tip

twist: Twist at root and tip sections (degrees)

flap: Flap deflection at root and tip (degrees) (Note that this must be in the range available for
the airfoil section data)

Section: Airfoil section name at root and tip, e.g.:

ifplate:
Inviscid flat plate. No stall or profile drag. 0.2c flap, deflection 0-30 degrees

NACA4415:
NACA 4415, Experimental data, Re=4x10^6. 0.2c flap, deflection 0-60 degrees,
flap Cm&Cd is estimated

NACA0012C:
NACA 0012, Computed with XFoil, Re=6x10^6, flap deflection -30 to 30 degrees

Section is repeated allowing you to specify multiple lifting surfaces. It is assumed that these surfaces
represent the right half of the aircraft. The left half will be represented in the calculation using a symmetry
condition (the coordinate system is: x = streamwise; y = spanwise; z = up). Optionally, you can turn this
symmetry plane off, or add a ground symmetry plane to compute ground effect. Note that a single wing can
be made up from multiple lifting surfaces. This allows you to simulate a wing with winglets, or with a partial
span flap, for example.

You find out about other available airfoils and how you can add your own airfoil ClCdCm data on this page.

Note that after running with the web form, a link to a plain text file which contains the complete input data
will be provided at the bottom of the results page. This can be modified and re-used with the "Run from File"
option at the bottom of the input page.

Interpreting the results

Successful Runs:

Run output data will be returned in the following order:

Geometry
First a plot of and then values for the leading and trailing edge line coordinates are returned.
Before looking at anything else, verify that this is what you had in mind.
 Residual History
Next is the residual history, or the variation in root-mean-square error in time. Inspect the
residual history for convergence. If several angles of attack were specified, the residual history
will jump vertically when the angle is changed. Generally small angles of attack should have
almost monotonic convergence. At higher (stalling) angles of attack, the residual will initially
oscillate, but should still converge smoothly when the residual is small.

Total Force and Moment Coefficients

Next the forces and moments for the complete configuration (i.e. the sum of all surfaces) are
given. These are non-dimensionalized using the input reference area and length. Both the
induced and profile components of the drag are shown (CDi and CDp), along with the total drag
(CD). Inspect the variation of coefficients with angle of attack. If you have placed your moment
reference point at the centre of gravity, and your CM curve has a negative slope, congratulations!

Final Spanwise Gamma and Cl Distributions

The spanwise distributions of force and residual are then plotted and tabulated for each surface.
In the first plot, you can compare the magnitude of the local residual with Gamma, the local
lifting-line circulation per unit length. If the residual is an appreciable fraction Gamma, you will
need to re-run with a lower tolerance. In the second plot, the distributions of Cl, Cd, and Cm are
shown. These coefficients express force per unit span and are non-dimensionalized using the
local chord (surface dimension in the x-direction). Oscillations in the circulation and Cl
distributions at angles above stall can be controlled by adding dissipation (k2). Oscillations in
the distributions at low angles of attack might be an indication of one surface lying directly in
the wake of another.

Final Load Distribution

Finally, the lifting surfaces are plotted with colours determined by the local value of Gamma.
This is an interactive plot that allows you to rotate, scale and zoom the image. If you are unable
to see anything, check to make sure that Java is enabled in your browser. If the visualisation still
does not appear, it may be necessary to upgrade your browser, or your browser's Java Virtual
Machine driver.

Ensuring convergence:

Due to the non-linearity of the problem, its inherent instability with the negative lift curve slopes encountered
in post-stall regions, and the normal difficulties associated with possibly intersecting singularity distributions,
it can require effort to obtain converged solutions for complex geometries. Review this page before
attempting to run a complex case.

First version 06/2000, last updated 05/2007.