Topics relate to Fluent, CFX, Turbogrid and more

Unphysical self-ignition in partially premixed combustion model – simulation

    • Johann Emhofer

      Hello all,

      I am currently trying to simulate an explosion in a vessel (grey in image below) filled with an initially quiescent stoichiometric propane/air mixture connected to ambient air (red in image below) via a circular port.

      I use the partially premixed model with chemical equilibrium – state relation and non-adiabatic energy treatment. Furthermore, I consider compressibility effects, use the C-equation, the Zimont flame speed model and calculate the pdf in advance.

      The initial mixture fraction inside the vessel is about 0.06 (stoichiometric) and 0 at the outside (boundary condition at port). The temperature in the vessel is initialized with 300 K and is 300 K at the pressure outlet. For ignition, I patch a small sphere in the center of the region with premixc = 1.

      Everything looks fine after the initialization and the first 2-3 timesteps:

      but then the mixture ignites close to the pressure outlet, where I have also a back flow with pure air into the vessel.  After 5 ms the combustion progress variable in the symmetry plane looks like this:

      In the described case, the initial mesh was set-up with inflation layers at the walls and the mesh was automatically refined during the simulations using the predefined criteria “Combustion – Flame Indicator” from Fluent.

      However, if I use a mesh without inflation layers, the situation is even worse, and I end up after 5 ms with something like this:

      Does anyone know, why my mixture ignites at the walls or close to the outlet?

      Thanks in advance, Johann

    • Judy Cooper
      Ansys Employee

      Hi Johann:

      The results of the partially premixed model are sensitive to turbulence, so an area that is borderline for ignition may be quenched/reignited if turbulence is low/high.  When there is no boundary layer mesh at walls, articifial turbulence may be generated by the solution of the turbulence equations, because bad values are produced. 

      For the issue at the outlet, pay attention to the turbulence and mixture fraction backflow values.  I believe the turbulence intensity defaults to unrealistically high values on backflow.  If the outlet is experiencing any backflow or recirculation, this can also generate artificial turbulence. It may be better to extend it so that backflow is avoided entirely, however.

Viewing 1 reply thread
  • You must be logged in to reply to this topic.