A more precise analysis of the aerodynamic flows around the unmanned air vehicle (UAV) can be used to obtain better estimates of the lift, drag, moments, or stall behavior than can be gained by scaling from the fixed coefficients and angles used in the spreadsheet analysis. It may also be desirable to estimate local pressure forces on the lifting surfaces for full stability analysis or subsequent structural analysis. There is generally no point attempting to further size and optimize wings and powerplant without more realistic aerodynamic data. Ideally it would be possible to swiftly take the entire air vehicle geometry, as created in Rhino, select an angle of attack (AoA) and flight speed, and run a series of calculations to gain the information required. It is, however, still not easy to accurately predict the aerodynamic performance of airframes by calculation, so a range of approaches should be adopted to suit the project in hand.

At the most basic level, the many tabulated sets of experimental results in Abbott and von Doenhoff [20] or Hoerner [9, 10] can be used to estimate forces on individual airframe elements, but it is difficult to allow for the interference effects between the multiple elements that make up a complete air vehicle. To proceed further, computational fluid dynamics (CFD) methods are generally used. The main task in CFD-based aerodynamic analysis, given a suitable outer wetted surface model, is the generation of an appropriate panel or mesh representation that can be used to drive the chosen CFD solver. The choice of solver to use will depend on the type of results required. The simplest reliable physics-based computational tools are generally panel codes with coupled empirical boundary layer models. The most well known of these that are freely available are probably XFoil,^{1} which can analyze two-dimensional airfoil sections, and XFLR5,^{2} which combines results from XFoil to allow wings and combinations of wings to be studied (sometimes with simple fuselages). Neither code will predict stall with complete accuracy, but they are good at understanding lift at reasonable angles of attack and drag at such angles for lifting surfaces. When used properly, it should be possible to allow

1 http://web.mit.edu/drela/Public/web/xfoil/.

2 http://www.xflr5.com/.

Small Unmanned Fixed-wing Aircraft Design: A Practical Approach, First Edition. Andrew J. Keane, Andras Sobester and James P. Scanlan.

for the interference of the flows between the main wings and the tail and to estimate local pressures on the lifting surfaces for use in stability and structural analysis. To go beyond this, some form of mesh-based approach is currently the aerospace industry standard, typically in the form of a Reynold’s averaged Navier-Stokes (RANS) solver such as Fluent®,^{[1]} Star-CD®,^{[2]} or OpenFoam®.^{[3]}

During the earliest stages of design for a small UAV, the costs of very detailed RANS-based CFD analysis can rarely be justified, so panel code results should definitely be used to first check that the primary lifting surface lifts and drags at moderate angles of attack are broadly as expected. If sensible lift and drag coefficients have been used in the spreadsheet analysis, this should always be the case. If the use of panel codes reveals significant differences in the low-AoA lift and drag, something will be wrong and this should be resolved before proceeding. If consistent results are being generated, these may be all that is required before carrying out a full stability analysis and then moving on to carry out the structural analysis. Sometimes, however, it is desirable to deal with more detailed geometrical descriptions or more complex vortex and boundary layer flows, such as those occurring near wing-tip devices and in slotted flaps. If such information is required, work can begin on RANS-based analysis, which can deal with arbitrarily complex geometries. Common to such solvers is the need for an appropriate mesh to describe the surface of the airframe and extending into the fluid domain. Given such a mesh, and provided suitable solver choices are made, it is possible to attempt predictions of the flow around complete air vehicles at quite high angles of attack.

Unfortunately, getting accurate results from RANS codes, particularly for separated flow, is by no means straightforward; any designer new to such methods is well advised to treat their results with great caution unless and until they have been able to produce results for a few known configurations that agree with experiments. When carrying out validation tests, the NASA technical memorandum server^{[4]} is a goldmine of valuable data as is the NASA Turbulence Modeling Resource.^{[5]} The turbulence modeling resource gives examples of using CFD to establish the behavior of several well-studied airfoils showing what may be expected from the very best quality meshes. Even their studies are unable to accurately predict everything that one might wish for. RANS-based CFD meshing and solution is, of course, a huge topic of research and there are major conferences devoted to the subject of drag prediction of aerodynamic surfaces, for example. Clearly, in a text such as this, only a very simple overview of good practice can be given. Consequently, the approaches offered here are aimed at small teams with limited resources; in no way are they meant to represent the best practice of large aerospace companies.