Got Blob?

I recently published a short post on the new aerospace geometry tool we've developed at Aerion. I'd like to showcase a few of its capabilities.

The tools we bought or built before this were pretty good at creating or modifying simple shapes: airliner wings are rather like long flattened sheets with a little thickness, fuselages are like tapered tubes, and engine nacelles are like hollowed-out tapered tubes. But many airplane parts are quite... blob-like. How about this typical fillet on a low-speed wing?

It may be a challenge for a homegrown geometry tool, but it's a cinch with any reasonable CAD program. But you don't want just one fillet, the whole point is to try a thousand different ones; you want to automate that. Your CAD program probably won't run on your Linux cluster, it won't run headless, it'll be slow, and it'll probably cost a gazillion bucks in licenses. You probably have to write, I don't know, a macro or something, in Visual Basic. And your parametric CAD model had better have been built by someone who knew what they were doing, or you won't even get a geometry for 50% of those designs. Now, would you like to finite-difference that beast for gradients and connect it to an optimizer?

With our geometry tool, I wrote a hundred lines of Python over the course of one evening to make a parametric CAD model driven by JSON data. The triangulation above was generated by that Python script in a few seconds, is watertight, and is ready for CFD analysis. With an

import numpy

I can generate a million random fillets in a few hours. With an

import multiprocessing

I can probably cut that down to an hour. Sure, everything is easy in Python, but even so, that's pretty good.

Finally, a few more pictures of the transonic airliner fuselage/wing fairing in the title picture.