Gas Turbine Combustor Mixing at Cranfield
Natural gas accounts for more than 23 percent of the world's energy production. Although it is not used in airplane engines because of its large storage volume, it has become widespread in industrial gas turbines and is the fuel of choice for many to plug the “energy gap” within the United Kingdom.
The use of natural gas within combustors provides new design challenges. The mixing process preceding combustion in gas turbines is crucial to ensure flame stabilization and recirculation of chemicals that lead to high efficiency and reliable performance. Swirl injectors have demonstrated a unique capacity to produce proper aerodynamic conditions within the combustor, introducing vortical structures that enhance mixing and decrease flow velocity. (See Figure 1 and this video.)
Cranfield University has a strong history in computational fluid dynamics (CFD) methods and combustion modelling, training hundreds of CFD experts as part of the Department of Engineering Physics, Masters of Science in CFD program. The master's students have benefited from the close relationship between Cranfield and Pointwise, which has enabled them to gain experience generating both structured and unstructured grids.
In his thesis, Pablo Aguado Lopez (Masters in Science in CFD 2011-12) developed methods and ran large eddy simulation (LES) calculations of a prototype combustor using Flamenco, a fifth order in space, second order in time in-house code developed specifically for multi-species compressible flows. He implemented non-oscillatory fully compressible species tracking via a set of quasi-conservative volume fraction equations. These simulations examine the cold flow to explore the efficiency of this simplified design.
The fully structured grids required were produced in Pointwise, which provided the necessary control over grid spacing, block size and, importantly, a reliable output to CGNS file format used by Cranfield's code Flamenco. The strategy followed to produce both grids was to use the minimum number of cells possible so that low time and computational effort requirements could be accomplished without missing the essential physics of the problem.
The unsteady computations are quite challenging, even at the current coarse grid resolution. The domain was split into 172 blocks in total to enable efficient load balancing in parallel LES simulations on Cranfield's ‘Astral’ HPC facility. The video Lopez produced shows a visualisation of his results, in which he examined the ability of the algorithms at extremely coarse resolution where the large scales are only just resolved. This is effectively an eXtra Large Eddy Simulation (XLES) where true separation of the scales is not achieved. The rotating structures are isosurfaces of Q-criterion, a measure of the strength of rotation relative to strain which is particularly effective at isolating vortex cores, and the contour flood shows the distribution of methane (red) as it mixes into air (blue). (See Figure 2.)
The approach has shown to be well suited for this type of problem because extremely complex characteristic phenomena, such as the central recirculation zone, precessing vortex core and secondary vortical structures are captured surprisingly well from a qualitative point of view. In the combusting case, these vortical structures are important for recirculating radicals and hot combustion products to ensure a stable flame. (See Figures 3 through 11 and this video.)
Figure 7: Radial Turbulent Fuel fluxes
The next steps are to improve results quantitatively through the use of finer grids using Pointwise, improved initialisation and calibrated boundary conditions.