Forces from FVM to FEM conversion (fluid/structure interaction)

 Register Blogs Members List Search Today's Posts Mark Forums Read

 February 24, 2017, 07:15 Forces from FVM to FEM conversion (fluid/structure interaction) #1 New Member   A.MAROUF Join Date: Dec 2016 Posts: 22 Rep Power: 9 Dear all, i am using an opensource code based on Finite Volume Method, well i want to make an interaction with a FEM software ( Abaqus , hyperworks, Ansys ...) the idea is to have the forces from FVM and go to the FEM do the Modal Analysis, get the mode shapes and come back the the FVM code to run the simulations. The question is that i want to : - convert FVM forces are in cells to forces in nodes for FEM is there any idea how to make it ? Thanks

 February 24, 2017, 08:40 #2 Senior Member   Michael Prinkey Join Date: Mar 2009 Location: Pittsburgh PA Posts: 363 Rep Power: 25 Your FVM code will (likely) have cell pressures and cell pressure gradients available. The most direct way to approach this would be to use those two piece of information to interpolate from the cell data to the node data. Since your FEM code is (likely) continuous, you will have multiple cells providing pressure data to the same node. You can compute them all and average them. Or you can assemble all of the cell data touching each node and build a multidimensional interpolant (say, least-squares). The easy way to do it is to just average the pressures in the cells touching the node. Most wall boundary conditions assume dp/dn = 0, so the cell pressure is likely the same as the pressure at the wall. You just really need to average out the wall tangent pressure variations. I'd do that first. Then come back and see if including the wall normal gradient terms would make a difference.

 February 24, 2017, 09:11 #3 New Member   A.MAROUF Join Date: Dec 2016 Posts: 22 Rep Power: 9 First of all thank you for your aswer, i already had this idea to make an interpolation between nodes and cells. But we may lose precision. What i know is that for example Workbench ansys does it from Mechanical APDL to Fluent ) , maybe there are some tools for it. Don't you think ? Thanks

 February 24, 2017, 09:33 #4 Senior Member   Michael Prinkey Join Date: Mar 2009 Location: Pittsburgh PA Posts: 363 Rep Power: 25 There is no magic here. Pressure (as it is applied in FVM) is done via interpolation to the faces and and single point (piecewise constant) quadrature. That is how the hydrodynamics are computed. Perhaps the "most" accurate form is to interpolate the pressure to face centroid and consider that as the face load applied to the adjointing element. That would be the most consistent with the FVM formulation from a momentum conservation point of view. Then, I'd refer you to the finite element literature to find the best way to distributed forces to an element. For example, the node pressures should (in an intergral sense) preserve the total force applied in both formulations. I think you can use area-weighted averaging to accomplish this. Something like: P_node = sum(A_face P_face) / sum(A_face) But, you need to verify that the integral over the same surface using your finite element and P_node data matches sum(A_face P_face). My guess is that if you are using linear elements, it should work out. But I haven't done this before, so you really need to investigate it carefully.

 February 24, 2017, 09:45 #5 Senior Member   Filippo Maria Denaro Join Date: Jul 2010 Posts: 6,793 Rep Power: 71 I assume that for FV method you mean that you are working with the integral form of the NS equations. Therefore, you need to evaluate the integral of the total stress along each face of the FV. The total stress is defined by the pressure (-pn) and the tangential stress n.(2mu S). Therefore you should match such quantities from the FE solution. I don't know if you have a sort of immersed boundary method to link the stresses. I had some experience with my FV code interfaced with the elastic stresses of a membrane computed in a FE manner. You could see some details in https://www.researchgate.net/publica...ids?ev=prf_pub mprinkey likes this.