Hello!

In FEM there is a term known as submodelling, wherein the flow solution for an isolated region is sought. Which means for the purpose of better resolution and easy solution a part of the whole computaional is separated, the boundary condiotions are applied accordingly and solution for that specific region is obtained separtely. Once a converged solution for the region is available the computations are further proceeded using this as the recent guess values.

The Question is whether such a term exists in CFD. Is submodelling used in Industry to solve complicated CFD problems?

Information is needed regarding this....

(1). There is a zonal approach, which divides the flow field into several distinct zones based on the characteristics of the flow. Thus, the appropriate equations and models (simplified in most cases)can be derived and solved for each zone. (2). For example, the flow over an airfoil can be divided into inviscid outer zone and the boundary layer region. The name for this approach is "zonal approach". (3). Since the coupling between zones requires additional work, the modern brute force approach is to solve the whole flow field by one method and code using say, the Navier-Stokes solutions.

Hello Mohan,

There are several approaches to this problem, depending on what you really want to do.

In some situations eg. heat exchanger simulations, you typically do not care about the microscopic flow features (eg. fins or turbulators) but they do affect your main flow by additional pressure drop or heat transfer. In these cases, a submodel can be solved, and a linear/quadratic equation can be fitted relating the pressure drop to the velocity. Then in the main flow model, this subregion can be represented as "porous cells" where the momentum equation is reduced to a simple Grad P= K.V where K is the porosity tensor.

In other situations, the details of the flow are important in both the submodel and the full model. For example, if you want to model the underhood region of a car along with its external aerodynamics, you might consider splitting into two models. However, in this case, you may have to iterate between the two solutions, mapping the B.C's from one to the other.

I think the difference between FEM Solid solutions that you are referring to and the fluid model CFD solution is that the CFD equations are nonlinear, and, except in simple situations like my first example above, the solution is not linearly dependant on the B.Cs So a given region cannot be replaced without loss of accuracy. (I'm not an expert on FEM, so I may be wrong on this).