From manual intuitive selection to full numerical optimization of an airfoil for a range of design points

Since my graduation project I have been fascinated by the potential to design an airfoil perfectly matching some specified design conditions. In my graduation project, I used the thin airfoil theory, the NACA 4-digit airfoil shape function, my little notebook, and Excel to numerically determine the airfoil shape ensuring a given lift coefficient for my first boxwing aircraft. It was a rather weak and quite imperfect means of figuring out how the "aerodynamic soul" of this aircraft might look like. But - it was just a starting point.

The next step was to dive deep into the relationship between airfoil geometry and characteristics. Before numerically, more or less blindly, optimizing an airfoil, I needed to make sure I was able to understand, anticipate and predict what was going to happen to airfoil performance once specific geometric changes are made. This was an exciting journey of exploration. The result is summarized in many of my airfoil-related PPTs, one of which I will upload to this site. Since than airfoil design has been my favorite topic for discussion with my students.

My first step towards automatization of airfoil design and analysis was to establish a Matlab - XFoil interconnection. To do this, I first used a freely available script. I also used it to learn how to code such Matlab functions, and now I write the interconnection functions for other executable programs such as XRotor.

Each time I noticed the software PROFIL mentioned in academic papers, I tried to find it and learn to use it to solve inverse design problems. My anticipation was that with this software, my design power would be unlimited. There are only Fortran-based source files of the code available on the Internet, and I had to ask a friend to generate an executable file for me. Eventually I learned to use it, but the result was painfully far from my engineering dreams, in part because of my unreasonable expectations. The code is limited to the potential model, so there is no way to design an airfoil taking into account viscous effects. Furthermore, the input data are a set of angles of attack at with positive pressure gradient changes to the negative one, which for me is extremely inconvenient and contra-intuitive.

In the meantime, I learned to write Matlab functions reading data from files. That is, I came to know how to use Matlab to automatically upload some input data to various executable files, launch software, and then read and plot the output data.

And finally the airfoil design dream became reality when I started using XOptfoil in our LEMFEV project. This is the XFoil-based global and local optimization software, flexible and agile, with a range of measures of merit, and allowing multiple design points to be specified. A couple of additional Matlab functions made the interaction with this program even more pleasant and rapid. Just love it.

So, this journey took several years, and I feel so happy observing the wide design opportunities which are now available to me.