My 6-year old son entered my office when I was working on the video below. He looked at it with amazement and asked me “Oh! Mummy, is that real fire?!” I stopped for a second to think and then answered “Kind of. It is real fire, but in the computer. We calculate how the real fire burns.” After a half-impressed and half-puzzled “wow!?” he cheerily jumped out of my office again and probably forgot all about it within 2 minutes.
However, his question got me thinking. The fidelity of combustion simulation is at a level where one can claim to do “real fire” in the computer. For example, B&B AGEMA is doing amazing studies using simulations to design a gas turbine combustor for minimized NOx emissions and manufacturing costs while maximizing energy density. To facilitate such applications, STAR-CCM+® software takes another leap in accuracy for combustion with the release of STAR-CCM+ version 12.02.
In the previous release (v11.06) we introduced embedded generation of combustion tables for all flamelet combustion. In this release, we take this functionality one step further by significantly increasing the combustion table accuracy, and hence the combustion simulation accuracy, through adaptive table discretization.
Let me start with some background. There are two types of combustion models in STAR-CCM+:
Species transport models
In species transport models, all species are transported in each cell at each time step. In flamelet models, the chemistry is pre-calculated and stored in a combustion table. The adaptive discretization introduced in STAR-CCM+ v12.02 increases the accuracy of the combustion table while keeping the table size at a minimum.
Let me show you how significant this is through an example. To meet emission targets, gas turbine combustors need to be designed for low CO emissions. CO is particularly challenging to calculate as it reaches a concentration of several percent in the flame front but is oxidized to a concentration of parts per million downstream of the flame. This means the combustion table needs to be accurate over four orders of magnitude.
CO in gas turbine combustor. Orange: CO > 1% Blue-yellow: Outlet CO, max 23 ppm
The combustor shown above was modeled in both v11.06 and v12.02 using a range of generated combustion table sizes. The graph below clearly shows the advantage of adaptive discretization: CO reaches a converged value of around 8 ppmvd for a very small table size in v12.02, much smaller (~9x) than the required tables in v11.06. As well as this, combustion table generation time in v12.02 is only 5 minutes, compared to 19 minutes for v11.06. This makes it much easier to find the required table accuracy in v12.02.
Converged value of CO for different combustion table sizes. STAR-CCM+ v12.02 requires a much smaller table size.
Table size required to reach table accuracy in CO emission of 1 ppm.
Why is the required table size so different in v12.02? The graph below shows CO mass fraction as a function of Mixture Fraction Variance for tables created in v11.06 versus v12.02. Both tables use 31 points in the mixture fraction variance, but the distribution of points is radically different. The even distribution in v11.06 gives a lack of points between 0 and 0.1, leading to inaccuracies. In contrast, in v12.02 the point distribution is adapted so there are more points here, ensuring a smooth and accurate representation of CO for low mixture fraction variances.
CO mass fraction as function of Mixture Fraction Variance for Mixture fraction 0.03, scalar dissipation rate 11.4/s
To enable this option in STAR-CCM+ v12.02, you select the desired error tolerance along with Maximum Number of Grid Points for each Table Dimension as in the screenshot below. The error is controlled for all species selected for postprocessing and temperature.
This is just one of our enhancements in STAR-CCM+ v12.02. I hope you find this information useful and that this feature brings you a bit closer to having “real fire” in your computer… at least figuratively.