Hi Guys,

I'm trying to solve a structure/fluid interatcion problem. There are two parts to my code. The first is an aerodynamic solver that generates aerodynamic forces; the second is a structural component that determines the response of a structure to a given load.

My problem is that, for several reasons I won't bore you all with, the aerodynamic equations are cast with one timescale and the structural equations use another. I know that the standard procedure would be to write both equations in non-dimensional form using the same timescale but for this particular application it isn't an option.

I label physical time 't' and non-dimensional times 'T1' and 'T2'. The aerodynamic code uses a scaling based on the chord-length and the upstream velocity:

T1=(c/U)*t => dT1=(U/c)*dt

The structural component uses a characteristic frequency of the structure 'w' to non-dimensionalise time:

T2=w*t => dT2=w*dt

In my code I select a timestep dT1, evaluate the current aerodynamic conditions and then apply the loads to the structural code. The corresponding 'physical' timestep is dt=(c/U)*dT1; the appropriate timestep for use in the structural component is then given by w*dt=(w*c/U)*dT1.

To recap: I select a value of the non-dimensional timestep T1 and this corresponds to a physical time of dt. The timestep in the strucutural code is scaled in a different way and so I have to work out the value of dT2 that corresponds to the same physical timestep as that used in the aeroelastic solver.

Does anyone see a flaw in what I'm doing here? I'd really appreciate any advice. I have solved unsteady systems before but am worried that I'm perhaps adding in non-physical effects due to the different timescales.

I understand that the best way to deal with a compound problem like this is to recast the equations so that they both have the same time-scaling. One timestep would therefore give consistent results and would make my life a lot easier. In this particular case, however this isn't an option and I'm trying to do the best I can with the information I have.

Many thanks, Frank

I would like to learn how to write the FSI code in order to better understand the theory behind it. I know the concept.Can you teach me how to write the code.

Sorry but no I can't.

The code we're working on is a combination of several people's work. I'm not really confident enough in all the theory to try and teach someone else I'm afraid. There are a lot of people in the forum whom are much better qualified than me and who may be willing to help you though (indeed that's why I posted my message here in the first place).

Hope you find someone more useful than me! Frank

As long as your physical timestep works out to be the same for both domains you will be ok. The only possible, but not likely, issue would be differences in round-off errors for the different domains.

Thanks Peter,

I had hoped that my logic was sound - it gets a bit confusing working everything out on paper!

Thanks very much for your input, Frank