Effect on Turbulent Combustion Simulations of Various Solver Options
Turbulent combustion is a complex subject, and getting CFD of turbulent combustion to consistently yield the right answers can be very tricky. Using a supersonic hydrogen-air case as a testbed, the effect of many different options has been studied. On other pages, we have delved into the effect of using
different turbulence models
on this case and also the effect of
different chemistry options
(see these pages for more discussion of the case being simulated). This page deals with a collection of options that doesn't exactly fit either category.
These simulations are based on, but do not exactly reproduce the Burrows and Kurkov experimental conditions. That, however, does not detract from their ability to shed light on the sensitivity of the solution to various code options. As such, the important thing is the difference between the various turbulent combustion simulations and the baseline case. The temperature contours for the baseline are plotted below.
Effect of TVD Limiter
One option that has been found to be important in other situations is the
choice of TVD limiter.
Therefore, instead of the default “Superbee” limiter, a simulation was run with the Koren limiter. Temperature contours for this case are shown below. Based on this plot, it is clear that the changing the TVD limiter is not making a large difference (although running without a limiter would likely not work well).
Effect of Frozen Chemistry
Another issue which should always be considered when putting together a CFD case is whether or not the chemical reactions are significant. Sometimes options are used just because they are there, but that is definitely not the best practice. This case is a bit more obvious than some, in that we have a reaction zone in the flow domain that is significantly changing the temperature and chemical mixture within it.
But even in a reacting flow, sometimes the effect on the quantities of interest doesn't really require modeling the reactions. Below are the temperature profiles from a frozen chemistry simulation. Note that, if what you really needed to know was the pressure distribution on the upper wall in this domain, a non-reacting version would likely do about as well as a reacting flow simulation.
Effect of Wall Temperature
Another issue which is critically important for simulating reacting flows is getting the boundary conditions right. Look at the effect on the solution of running with hot walls and a slightly elevated inflow temperature (shown below). From the notes on the plot, you can see that there are several differences from the baseline in this case.
From the study of the
effects of turbulence model on turbulent combustion simulations,
we know, however, that the incompressible form of the SST model gives much the same results are the compressible form used in the baseline. Also, in the page examining the
effects of different chemistry options,
it was seen that using the SPARK curve fits for thermodynamic data made no difference in the results. Ignoring the third body effects tends to move the reaction zone downstream. Thus, it appears that simply by changing the boundary conditions we have moved the reaction zone almost 15 cm upstream.
The Burrows and Kurkov case has been studied extensively in the past, and more than one study has concluded that that an accurate simulation of this case requires that the incoming freestream boundary layer profile closely match the experimental measurements. The best results, in terms of comparison with experiment, were found when no upstream duct was simulated at all. Instead, the experimentally measured profiles were applied directly at the entrance to the test section as inflow conditions. Note that all the cases here used an upstream duct to generate a plausible boundary layer at the inflow of the test section.
From the above results, and the experience of other investigations, it seems obvious that one of the most important factors in getting good results from a simulation of turbulent combustion is getting the boundary conditions correct. Also, while a frozen chemistry approach may be justified in many cases (or ideal gas, for that matter), if there is a significant region of combustion in your domain, it will probably be important to model the chemical reactions.
When you are finished learning about the sensitivity of
turbulent combustion simulations to various options, head back to the CFD Applications
Or, head back to the
Innovative CFD home page
and browse some of the other topics.