Simulation of mechanical systems - FLOPS required?

[Borek]

I suppose some of you know the game called "Bridge Builder" (or Pontifex, or clones).

I need some help determining how many FLOPS are required during the simulation and how do they scale with the bridge size (number of elements). Some of these bridges contain hundreds of elements and tens of nodes, this is a video with many examples:



I guess model behind is just some simplified application of FEM to solve the equations describing stresses on each element, but while I do have some very vague idea about these things I don't know enough to even guesstimate required processing power. Let's say we need to run the simulation at 30 frames per second. Any help would be great.

(This is - hopefully - my final post in the quest for FLOPS, games and physics engines. Sadly, so far I have almost nothing. Just an interesting anecdote - I have learned that Spindizzy, game some of you can remember from the 8-bit eighties, had a very simple physics engine based on 8.8 fixed point math.)

 

[Bandit127]

I have no idea of the complete answer to your problem. However, I found a good page as an introduction.
http://www.datam.de/en/products-solutions/simulation-with-fea/copraR-fea-rf/fea-calculation-time/
Although the information is somewhat specific to the software, it does have some good data on compute time vs mesh increment and CPU processor.

Interesting that mesh increment vs compute time seems to be roughly linear. My guess was that it would be squared.