Coupled Level Set Volume of Fluid Simulations of Prefilming Airblast Atomization With Adaptive Meshing

The fuel atomisation process and the resultant spray drive nearly all aspects of gas turbine combustion system performance, including emissions. A greater understanding of the fuel atomisation process is therefore required if future requirements are to be met. The majority of investigations into fuel atomisation mechanisms and performance to date have been performed at atmospheric conditions and use simplified geometry. This is done to reduce cost and reduce complexity for both experimental and numerical studies. Real combustion systems operate at temperatures, pressures and velocities very different to these and, as such, high-fidelity simulations are attractive to understand the atomisation process in real systems. However, due to the higher Weber numbers encountered, the range of length scales that must be resolved in a simulation of the atomisation process is higher for these operating conditions. This, together with the added geometrical complexity of swirl vanes and other features, leads to increased computational costs when attempting to perform high-fidelity simulations of practical combustors. Hence the computational efficiency of any such simulation method is very important.

This paper presents an efficient numerical method for simulation of gas turbine fuel atomisation. A fully Coupled Level Set Volume of Fluid (CLSVOF) solver has been developed using OpenFOAM, building upon the native VOF solver interFoam. The inclusion of the level set field allows for a defined reconstruction of the interface within a cell and allows a reduced resolution for a given level of accuracy and therefore reduces the necessary cell count. An algorithm based on tetrahedral decomposition greatly increases the efficiency of the interface reconstruction step meaning that the reduction in cell count substantially outweighs the time overhead of the reconstruction step. The CLSVOF method is used with Adaptive Mesh Refinement (AMR), which moves the mesh with the interface, allowing much greater resolution for a given cost than would be possible with a static mesh.

A simple 2D planar prefilming airblast atomiser has been used as a validation case for the method, and in order to assess the cost benefits of using this method. The geometry and boundary conditions are taken from published literature, with experimental and previous computational results available for comparison. An initial mesh of 78,400 cells is used, to which three levels of automatic refinement are added to give a resolution of 12.5 microns at the interface. This results in a cell count that fluctuates between 2 and 3 million. If a static mesh was used the cell count to give the same resolution at the interface would be approximately 40.14 million. The behaviour of the film and subsequent breakup is consistent with what has been reported previously. As the film reaches the trailing edge the surface tension causes the fuel to roll over and gather into a reservoir. From this, sheets, ligaments and bags are drawn out by the airstreams above and below which are then further broken down into a spray of droplets. To simulate 20ms, 100,000 CPU hours using 112 cores were required, an 87% reduction compared to published computational work which used the native OpenFOAM VOF solver. The majority of this is due to the AMR, however the inclusion of the level set field allows a reduced resolution for a given level of accuracy, further reducing cost. This cost reduction will allow a larger range of spatial scales to be resolved and more complex geometry to be used, giving the capability to perform future simulations on more realistic geometry.