In the CFD area, it is common sense to set interfaces between two bodies that have different meshes. My questions are:
1) Which set of variables are mapped between two different meshes on the interface? For example, Velocity? Or Massflow?
2) How the solver solves the fields between the interface? Are the fields in the two bodies (on the two sides of the interface) solved together or separately?
Thanks in advance.
I don't think, that your question can be answered independently from the concrete implementation of a CFD code.
1) usually a conservative formulation in terms of transferred massflow across the interface (i.e. mass flux, momentum flux, energy flux, conservation of species concentrations).
2) For sure dependent on the CFD code implementation.
For parallelized codes, each body usually represents at least one partition in terms of the applied domain decomposition for the purpose of parallelization (can be split in several partitions). And then a solution matrix is computed for the partition with appropriate coupling on the partition interfaces throughout the iterative solution process, which usually involves (MPI) communication between different processor cores of the parallel machine with data exchange on the partition interfaces. During that communication and data exchange it has to be handled, whether this data exchange is related to a conformal or non-conformal partition interface.
Regards,
Dr. Th. Frank.
It depends heavily on the software, numerical methods, and requirement of precision.
The conditions to be satisfied at the interface depend upon the physics of the problem. These conditions are different depending upon the type of the body.
(i) if the two bodies are the same fluid, then the velocity, pressure and all other variables should be the same at the interface.
(ii) if the two bodies are two different fluids with different properties, then only the velocity is the same; pressure and density may be different.
(iii) if one body is fluid and the other body is solid, then the velocity at the interface should be the same.
The way these conditions are imposed depends upon the numerical method under consideration.
This is a very interesting topic. The problem of interface is always accompanied by the large-scale calculation of CFD, but it is also a very difficult problem to solve thoroughly.
In single-phase flow, interface may appear in cross-scale calculation, which is used to deal with the trouble caused by the grid size difference between the small-scale flow domain and the large-scale flow domain, or the rotor dynamic static interface often uses interface boundary connection. The main solution is to use MMRF such as MRF, MPM, or SMM.
From the code level, it is mainly to ensure the conservation of some flow parameters and the limitation of interpolation error, but this is usually difficult to achieve, because the grid nodes on both sides of the interface boundary are generally not corresponding, and the difference in spatial position will be closely related to the flow distribution, so it is difficult to guarantee these two points at the same time.
In the gas turbine and other rotor machinery with supersonic flow leading to shock wave and other problems, the dynamic and static interference is a problem that can not be ignored, but it just exists in the interface position. In the presence of multiphase flow such as combustion, the problem will be more complicated.
Therefore, interface processing is not a separate problem, but a subsidiary problem related to the characteristics of the flow itself. How to process the interpolation depends not only on the interface solution itself, such as MRF, but also on the algorithm used in the flow calculation and the flow parameters to be guaranteed.
Thanks a lot for the professional answers from Thomas Frank Chennakesava Kadapa Yuan Tang Jiarong Hong . According to the answers, I hereby post several further questions for discussion. Let us extend the problem constraints to:
1) Water, liquid single-phase.
2) The mesh consists of two flow domains. One is structured mesh, the other is unstructured mesh.
3) The interface is to join the two flow domains that have different mesh types. Mass, energy, and solute transfer happen through the interface.
4) No MPI implementation, the two flow domains are solved by one core or one thread.
The questions are:
1) If the two domains are set to form one solution matrix and are solved fully-implicit by the solver, I guess the two domains are bound by constructing the relations of the basis vars e.g. velocity, density, temperature to form the entire solution matrix. This is the same idea with Chennakesava Kadapa .
2) According to the answer from Thomas Frank , some conservative formulation is transferred across the interface e.g. mass flux and energy flux. If this is the case, how can we project them to the basic solution items e.g. velocity, density?
3) Which kind of error sources does the interpolation have? Yuan Tang
Since I have CFD experience as you experts, some of the questions may seem logical loosely and even wrong. I hareby appreciate all your efforts and time on this question.
- Badges
- Science topic