I was recently looking at some of the official vehicles jbeam files, and have noticed that they are now using a jbeam section labelled "torsionbars". Clearly I should have read the 0.14 changelog more closely Code: "torsionbars": [ ["id1:", "id2:", "id3:", "id4:"], //24mm 2543 //25mm 2994 //26mm 3502 //27mm 4073 //28mm 4710 {"spring":4073, "damp":10, "deform":10000, "strength":9999999}, ["arbfr", "fx1r", "fx1l", "arbfl"], ], Does anyone know the specifics of how they work, does it calculate torsion as if it was a pipe following that path, or as if there was a 3d plane between all of the nodes as a solid object? There is also mention of using them for stiffening chassis in the blog post, is there anywhere that they have been found to be especially useful beyond what already existed?
I guess it is now possible to do Ford GT style suspension? In fact, you could even do the conventional springs as hydro's. Then drop the hydro length to the their beamLimit leaving only the the torsion bar to take over the full duties. Granting adjustable ride height and suspension stiffness in real time at the press of a button. Could be pretty neat.
From what I can tell, torsion bars work by defining 4 nodes in the order of Right Attach, Right Pivot point, Left Pivot Point, Left Attach. Then, it links them, and applies a torsional force if they are misaligned with each other (based on their starting positions in the Jbeam file, or if precompression is applied). It can only apply forces in one dimension (clockwise and counterclockwise, relative to the axis defined by nodes 2 and 3) and provides no attaching forces. I'm not sure what unit it is in, presumably N/m^2 though, so if they are given a 1 meter lever arm (as defined by nodes 1-2 and 3-4), it will be equivalent to a spring with the same spring force. Of course, if you have a longer or shorter arm, the actual force applied will be correspondingly larger or smaller.
I hadn't considered that it was limited to 4 nodes, that makes a lot more sense now. I was thinking of it as just a collection of nodes, which left me wondering how they were figuring out what rotates with what etc.
That gives pounds per inch, but you convert that to N/m^2 easily though. Well, that 'should' work, but it can then get bit more complex than that and essentially one might end up adjusting value a lot manually.