In the previous post I wrote about one of my projects working at the GME and the new tools for me I used. This post relates to the other project I work on it.
The GME is partner of the MAAXIMUS Project (More Affordable Aircraft through eXtended, Integrated and Mature nUmerical Sizing). We put our knowledge about structural optimization to produce better materials and elements to be used in the aircraft industry–mainly in the AIRBUS Consortium.
Fuselage (source: eurocopter.com)
Well, one of our tasks is to design and test some optimization methods. Because the complexity of the real elements to analize these are a collection of numerical methods working together. In fact the process implies using some computer programs: design, meshing, structural calculus, postprocessing tools, and, at the center of the system, the optimization tool.
The optimization tool generates via different engines series of designs that must be tested by the other tools. Each engine implements a method or family methods. And this is my project: I must develop an engine to be integrated in a Dakota-style framework. I would use the old Fortran to do it. But I needed to program a text file parser, so I decided to make some experiments with the Python programming language.
I hadn’t programmed in Python before. I avoided it in part because I thought Python would have a lot of hype as some years ago it happened with Perl. Some proofs of concept after, I was really happy with the results. It’s true the basic Python was a bit limited to develop the numerical part of the engine. But then I discovered NumPy (now part of ScyPy libraries).
NumPy allowed me quickly replace some code developed in Matlab language by my fellows and write my own, more clear and simpler code. And, after the experience of implementing some numerical methods, I seriously think it can be a good idea for me to leave Fortran (with the exception of educational purposes) and code Python+NumPy.
Python++