Learning Forum

[abtharpe42]

I trying to run a simulation of a combustor with air flowing in annularly and then swirling around a fuel inlet. The 8 blades were angled such that the swirl number is 3, and the overall cylinder is 15 cm in diameter and over 150 cm in length. A polyhedral mesh consisting of over 700,000 cells and a 15-layer smooth transition boundary layer was used, as it's the only way I've found to consistently stay within the confines of the student license limit.

The combustion model used is Non-Premixed Steady Diffusion Flamelet with Inlet Diffusion, Compressibility Effects, and P1 Radiation enabled. The Coupled Psuedo-Transient solver is used and everything is set to second-order. The under-relaxation factors are left at default. I've read in papers and forum posts that the Reynold's Stress Model (RSM) is the best turbulence model to use for a high swirl situation such as this, but I've yet to find, even in the User and Theory Guides, if Linear Pressure-Strain (LPS), Quadratic Pressure-Strain (QPS), Stress-Omega (SO), or Stress-BSL (SBSL) is the best variation to use for high swirl or even simply how to best set up the solver to handle a high swirl combustion simulation. QPS always ends in floating point error in anything I've tried to use it in, and the other three immediately start with extremely high residiuals no matter the intialization type I use as shown below with a run using RSM-SO:

I've hit a roadblock on this one, and any guidance I get on more appropriate solver settings to reach baseline convergence will be greatly appreciated.

[Essence]

Hello,

RSM tends to be transient state model is generally difficult to converge. First use two-equation models without combustion (non-reacting flow) and converge it. Then switch to RSM and again converge it. Then turn ON the combustion (reacting flow), converge it and afterwards, turn ON radiaion model and converge it. Hope you got the point here.