Bavarian Graduate School of Computational Engineering

Honours Project:
OpenFOAM applied in CFD Simulation of Head Pressure Waves of a Train and its Comparisons with CFX

Tao Zhu

Tao Zhu did his honours project at Deutsche Bahn Mobility Networks Logistics.

With several advantages such as free of charge, OpenFOAM is becoming popular in CFD simulations in industrial applications. Yet its capability remains unanswered within the research group of Deutsche Bahn. In this project the capability of OpenFOAM simulating the head pressure distribution around a high speed train is investigated, and the results are compared with those from a previous research using CFX(ANSYS). First, the quality of the mesh of the domain and the train created by snappyHexMesh, a meshing tool packed in OpenFOAM would be tested. Then capability of different solvers such as simpleFOAM, rhoSimpleFOAM and rhoPisoFOAM would be checked and the results would be analyzed.
One example of fully developed pressure and velocity fields around the train

SnappyHexMesh defines several refinement regions of the computational domain around the geometry. In the refinement region closest to the geometry the grid of the domain could be mostly refined; and the domain could become coarser when it is further away from the geometry. This feature allows the simulation to represent the close-to-surface viscous sub layer of turbulence more accurately, and the Law Of the Wall (LOTW) could also hold more precisely. This mesh distribution is similar to a streching mesh, in which the mesh is also fine at the geometry surface, and becomes coarser when it goes further away; yet it is more flexible when defining regions where it should be refined, and avoids the situation that at some locations the finest grid resolution in one direction meets the coarse grid resolution in other directions, which leads to very big aspect ratio. Another advantage of applying snappyHexMesh is that at non-orthogonal surfaces, the mesh becomes tetra-like and edges of the mesh do not cut the surface of the geometry, which aviods the number of grids which has half geometry and half fluid inside. This helps improve the accuracy in calculating quatities such as flux, especially if some methods like the Immersed Boundary Method (IBM) are applied.
Views of the meshed domain and train created with snappyHexMesh
There are three solvers investigated in this project: simpleFOAM, rhoSimpleFoam and rhoPisoFoam. In OpenFOAM, simple means the solver is steady state, which means that many of the cross terms and higher order terms dealing with time are ignored, and the time step size is set to 1. This kind of solvers make the convergence easier to reach but they also make results less accurate. However, in steady flows it has more advantage concerning the faster convergence. Figure 4 shows one example set of experiments with OpenFOAM proving that time step size does not play a rule in steady solver. On the other hand, piso means that the time dependent terms are not ignored, and the simulation is a real-time simulation. Thus, the simulation converges slower but it is more accurate. In this kind of OpenFOAM simulation the courant number must be considered. While discussing the term rho, it simply means that the flow is considered compressible. If the solver does not have this term in its name, it means that the flow is considered incompressible. The figure below compares the head pressure wave from CFX, simpleFoam, rhoSimpleFoam and rhoPisoFoam. The results imply that simpleFoam, rhoSimpleFoam and rhoPisoFoam are all suitable solvers for this case.
Head pressure wave around the train from different solvers