Which solver would be better for this case : Implicit or Explicit ?

[missil3]

I am trying to simulate the inflation of an inflatable structure for one of our products. Currently I am using an explicit solver for it (explicit FEA+CFD) . The problem is the inflation time is 4 seconds. The analysis is highly non-linear with surface to surface contacts with outside bodies and also self contact. The material model is also non-linear with large deformations. It involves fluid structure interactions (employing both Eulerian and langrangian mesh). All these point towards explicit methods but given the time for inflation is 4 seconds means it takes a lot of time to solve (about 40 hrs currently) to solve for 4 seconds. Moreover the meshing is to be done keeping in mind the timestep needs to be as big as possible for a reasonable solving time.

I have been considering trying out an implicit solver (dynamic, coupled fluid structure) since it will allow me to use smaller elements without worrying about the timestep. However, I am worried if it will actually improve or increase the solving time. Cause issues with convergence due to the high degree of non-linearity. This would mean changing the solver we use as well, so license costs are another factor.

Which do you think is better suited for this problem ?

 

[burningDiesel]

If you have a folded structure, it will be very hard (if not impossible) to calculate it using implicit methods. In abaqus, you can use fluid inflators to model inflation (typically used for airbags), to avoid using the FSI interactions. Other programs probably have similar options.

 

[missil3]

Thanks, I think the methods you are talking about are uniform pressure methods ? Those don't capture the physics correctly for our problem though.

 

[burningDiesel]

Yes they are.

 

[sunnymarks]

Hi missil3, I agree with sdebock. For the inflatable structure using uniform pressure methods using fluid inflators to model inflastion for CFD simulation would be a good idea than using implicit methods.

 

[missil3]

Uniform pressure methods won't work for our problem. It would have been so much easier if I could use that as it is very fast to solve.

 

[burningDiesel]

You can try other techniques to make your simulation faster: mass/time scaling, selective subcycling (when using parallel solving), submodeling or substructures and coarsening your mesh.
Most things influence your solution accuracy though.