Soft shell structure under hydrodynamical load
by Yi Zhang
In the middle of constructing a reduced-order model (ROM) of a highly flexible structure interacting with surround flow, I built a structure model and loaded it with oscillating pressure at the lower part. The animation from the simulation is as follows.
There are several challenges to this type of FEA problems, mostly are related to the highly flexibility nature of the material. The structure is first subject to a quasi-static hydrostatic load, which makes it buckle to a new equilibrium position. After that, ambient oscillating pressure with a much smaller magnitude than the initial buckling load is applied to make the fabric vibrate.
This is a perfect example for using explicit-implicit switching for time integration. Since it’s usually difficult for an implicit solver to find the bulking configuration, as a result of the fact that the Newton-Ralphson method does not gaarantee convergence, one needs to issue an explicit analysis for the initial loading stage. Some solvers provide specific “bulking simulation” controls for this type of the problem. After the initial loading the structure to a new equilibrium (passing a bifurcation point, as one clearly see in the animation), we proceed to impose the oscillating load. This type of load is supposed to be of a much smaller magnitude than the initial load, so that the structural vibration would be a perturbation near the new equilibrium.
Since explicit method suffers from a general stability problem (notorious forward Euler blow-up in every toy example in every textbook), it is only applicable to short-duration problems such as impact/contact, and it is very easy for the blow-up to happen during a cyclic loading problem, due to change of sign of the derivatives. Thus we need to turn off the explicit simulation and replace it with an implicit one. The benefit? (Unconditional) stability and hence greater allowable time step. The cost, on the other hand, is much higher since now we are actually inverting the stiffness matrix and perform in-time iterations.