[amjadfirst]

In solution controls, when changing the under relaxation factors, does anyone know what terms and in which equations these factors take effect? so for example I change the Energy under relaxation value from 1 to 0.9, what equation is affected by this and what term is under relaxed? is it temperature, or heat flux, or enthalpy? same thing for momentum, what changes and in what equation?

Thanks in advance,

Amjad

[LuckyTran]

It's algorithm dependent! But I'm assuming you're asking about the pressure-based solver using either SIMPLE or PISO.

Most of the under-relaxation factors are applied to equations (momentum, "enthalpy", etc). I.e. implicit under-relaxation. Material properties, etc are explicitly under-relaxed. For a temperature-based solver, the enthalpy (~CpT) is implicitly under-relaxed by the urf. If there is a urf for the material property, Cp in this example, then it would be explicitly applied to Cp later. For momentum, the urf would be applied to (rho*U, rho*V,rho*W) and not u,v,w, by themselves. Whether or not that makes any difference is a question for another day!

The COUPLED solver is a mess, it uses both explicit and implicit under-relaxation. The density-based solver is also a mess for similar reasons.

[amjadfirst]

Thanks for the reply. Yea for the coupled solver, this is what I thought they meant, but for properties (density and others) it's more understandable than equations. How do you apply an under relaxation factor for an equation? say momentum for example. I tried to find out in the ANSYS documentation but couldn't find anything.

[LuckyTran]

I believe it is well described in 21.4.4.2 Under-Relaxation of Equations. I don't claim to know exactly what Fluent does under the hood but the way it is written in the manual is how I would do it.

Forget the word "equation." Recall that the discretization schemes reduces the (non-linear) partial differential equation into a system of linear equations. Once you know the appropriate weight factors for each term you are you essentially solving for the, I'll say, "property" at each cell. For x-momentum equation you are solving for the x-momentum at each cell (rho*u) which is dependent of course on neighboring cells. The urf is applied to the "equation" or "property", which in this case is the momentum (rho*u).

The implementation looks a little weird but essentially, you inject the existing "old" solution into the system of equations. This is equivalent to solving the new system by itself and then calculating a delta, multiplying the delta by the urf, and then adjusting the old solution by the delta*urf instead of delta. Delta in this sense is the property at each cell.

[amjadfirst]

Thank you, that's clearer. I'll look up that section again in the documentation.