Olivier,
thanks a lot for the compliment. ;-)
The numerical methods I use for this project are rather intransparent. Catia is a quite powerful tool, but as far as FE is concerned, it is in my opinion really questionable. I use it for designing and nothing else.
At first I started with deriving a load collective ("acting" on the car). I took these loads and converted them (its just an inverse matrix operation) onto loads acting on the suspension pickup points. Then the major masses in the car have to be discretized in terms of C/G position and mass, which gives additional loads that stress the structure. The laod collective is very critical, because they affect all the results.
After this step I used a topology optimization (-> Hyperworks is quite good for this, also Nastran, but I find it more inconvenient, as it has no preprocessor) to find sensitive loading paths (in other words the optimal stiffness distribution).
And now comes the really complicated step: To redesign the topology iso-surface into sheet metal parts. This sounds easy, but you always have to keep in mind the stiffnessloss because of rivets, bends,.. and even more important the manufacturability (I don't wanna use 20 different bending presses for example). After this step I have surfaces, but no good wallthicknesses. So i mesh them (-> Hypermesh, my absolute favourite) and set up a parametric optimization (as it is just to get stiffness right, which means linear elastic solutions, Nastran SOL200 proved to be really nice for this). This optimization is maybe going to run for 5 or 6 days, and the result will be the optimum wall thicknesses. Slight redesign can be made after this step, if eg some walls show to "disappear", last thing you have to do before fabrication is to wind all the sheets down into flat pieces, which are going to be nested for laser cutting.
It all sounds a little strange I think, but there is no tool in the world (and it will be never invented, I'm sure), that can do all that tasks for you in once. The computer makes some more sophisticated things possible, but in the end it is always the engineer in behind those things.
I don't understand what you mean with "specific radiuses abaqus". If you mean if I use Abaqus, then the answer is yes. It is perfect for nonlinear problems, like elastc-plastic material behaviour.
Money: I'm sure you will understand that I don't want to give numbers here. But I can tell you some other interesting facts: Our largest expenses until now were the engine and the gearbox. Mouldmaking wasn't that bad because I did 85% of it by myself, so there were the machine-hours and the material left to pay for. I had a discussion with quite some mouldmakers in the mean time, and if you would give them cad files and wanted to get moulds back, you surely would have to pay around $50G for a toolset for a carbody.
But let me tell you that the most important thing here is NOT the money.
I just want to mention the nearly 3000 engineering hours I've put in so far.
I hope this info answers your questions.