Recently, I've started a new project that I'm pretty excited about: the Community Starship Loads Model. With all this discussion of materials and structure on the Starship, I was interested in taking a closer look and compiling everything we know. This is a long post. If you're more a pictures-person (me too!), scroll to the bottom to see the model and some results to determine if you may be interested in reading the rest of my novel. :)
So What is a Loads Model?
In traditional aircraft design, a coarse Finite Element Model (FEM) is developed to apply loadcases during the sizing and initial testing phase of the program. I believe there are other methods still used, and I don't really know how SpaceX is developing their craft, but there are significant advantages to this method. For one, the model can be used to determine the load distribution for internal structure during complex flight cases (including gross thermal loads). Additionally, loadcases can be created, applied, and revised to determine feasibility and flight envelope capability (or vice versa). It's usefulness is degraded when you start to consider re-entry effect, but as part of this project I'd like to explore that and see what we can actually simulate.
At this point, it's worth pointing out that there will be a lot of guesswork. SpaceX obviously has design and testing data that we just haven't seen before, but this sub can be quite crafty from the little bits that are actually released. We will have to guess on details from structural layout to material properties and exact loadcases. I plan on laying out the assumptions made in the documentation for this project, so that the community can review and provide feedback on the accuracy.
Well for starters I think it's cool...It's also so I can develop my skills in an area of aerospace that I've been interested in. I have the skillset and tools available so I figured I'd give it a go. We may all learn something during the process!
Nuts and Bolts
Initially, a very very basic fem will be created which reasonably approximates the size and shape of the Starship. The model will strive for "2 and half D" elements (shells wherever possible, combined with bars/beams to represent caps and integrated stiffeners).
Right now I plan on using MSC Nastran, and outputting HDF5 results. A student version of the software can be downloaded for free, but may or may not be able to explore the results due to size restrictions on the license (if I release the full results in the future).
After initial fem creation, loads development will commence. Loadcases will be determined and applied to the model as detailed as feasible, with the objective to cover the full mission spectrum. Initially, the launch profile will be covered, then re-entry, spaceflight, and varying EDL (not necessarily in that order).
After a loads drop release, the fem can be refined and modified to better represent the necessary structure and design changes known by the community. This may be the area where effects of design decisions can be explored, and results here could be very insightful.
Right now, I can't release the full analysis deck, but all other data should be unrestricted (it's a licensing issue which I may sort out in the future). So I should have plenty of pictures and plots to share, just not the raw input or output data.
- Currently, I created the basic lines of the craft from some community drawings. The flight configuration was modeled, and unfortunately I believe a separate model will be needed for each configuration (entry and landing positions of the fin-legs, and any necessary positions in between).
- From there, I created the mesh, and put structural bulkheads where I think they might go, based on intuition.
- After meshing and equivalencing the nodes, we now have a full model ready for loading and properties. For now, a temporary property card was applied to all shell elements, and there are no structural bars/beams. This material just has dummy properties that allow the model to take load without crapping out the solver.
- A test loadcase of uniaxial compression was applied to verify functionality.
- The model was "tied-off" with 3 loose springs at arbitrary points on the airframe, attached to fully-constrained nodes.
- The model needed AUTOSPC to solve, meaning the loadcase is unbalanced. Upon investigation, this can be verified, since the applied nose pressure is distributed over the whole annular surface, and therefore creating a moment about the legs which cannot be balanced without taking tensile load in the tail foot.
- This is why the landing configuration is important. It may not make sense without visualizing it, but think of a 3-legged stool vs a 2-legged stool. The latter can be balanced upon, but is significantly more difficult.
- I didn't realize my load would make the Starship "lean" forward, but that's exactly what it wants to do, and it breaks my grounding constraints. (will upload a displacement plot tomorrow when I have time).
- No matter, this is just a test. It works!
- Community Feedback! What do you think? Does my geometry look correct? Are the canards/leg-fins in the right positions and angles (roughly)? Is the nose too pointy? Think about geometry discrepancies that might significantly change the results. If you like, go ahead and draw up some lines in a format similar to my lines drawing! I can then easily incorporate it into a model revision!
- MaxQ. MaxQ is perhaps the most critical loadcase expected during a nominal mission. This is often referred to as DLL in the aviation world (Design Limit Load). I think it's a great place to start with realistic loads creation, especially since the community has already done some math/data extraction for me!
- Documentation. I have a private github page with all these files, and I plan on writing up the information in this post in pdf format, with appropriate pictures and tables. Additionally, the model groups in the database will be cleaned up for easier use.
- Student Edition and Full Release. As I've mentioned, I cannot release the source code for this project right now. I need to either coarsen up the mesh more, or acquire a personal license to release the full analysis deck. This sucks, so I really want to figure this one out soon!