Variables
The variables of our problem describe the aerodynamic shape of the aircraft. Over the last two decades, many engineers have tried to describe the shape of a wing accurately with only a few parameters while maintaining flexibility. There arc three general classes of aerodynamic shape functions:
• Linear combination of existing wing sections. This geometric approach to the airfoil design problem is a derivative of the original NACA method {330}. Although this method is computationally efficient, at DA this method has often given only trivial solutions (c. g. the old airfoil) due to the low flexibility inherent to this geometrical representation. In addition, there is little physical foundation for the implicit assumption that the drag (objective! of a linear combination of airfoil geometries is linearly or at all related to the drag of the individual airfoils.
• Analytical shape functions arc linearly superimposed to define a wing geometry. Examples Of these include Hieks-Hcnnc functions. Wagner functions and the patched polynomials discussed in reference 1344j. as well as the Legendre polynomials used by Reneaux (346J and
the splines used by Coscntino (336]. Unfortunately, the number of parameters required for these shape functions arc usually too high to allow multipoint three-dimensional design. Every lime, a minimum of 30-40 points arc required to define each airfoil. Even if only five airfoils are sufficient for defining a wing, about 200 parameters will be necessary. This number of variables is simply too great for practical optimization at present. Coscntino optimizes a 3D wing by allowing variations over only a small portion of the wing. Lee has proposed patched polynomials as a way of flcxiblcly modeling an airfoil section with only sixteen parameters. Although this approach does produce a flexible geometry, the performance of the sections designed with this method cannot compete with those produced by experienced engineers using inverse methods.
• Special aero-functions were proposed by Rcncaux [346] for designing airfoils with a minimum number of parameters. This approach automates the steps that an expenenced designer follows when using an inverse method. ‘Good* pressure distribution types are defined, along with the shape or special aero-functions that produce these pressure distributions. Unfortunately. for a given pressure distribution type, these aero-functions vary with Mach number, and this method is therefore impractical for multipoint designs. This method has all the disadvantages of inverse methods discussed in the Introduction.
At DA (352) the author has introduced another type of function, a highly nonlinear special aero-function, which can optimize the shape of a 3D wing with adequate flexibility using current computer technology. It is now being used for the industrial design of transonic and supersonic wings. Its application will be shown in the next section. These shape functions produce the aircraft geometry in any resolution.