[KarenRei]

I've been working on this problem for the past several weeks. I have a simulation modeling the combustion of oxygen, paraffin and aluminum with several intermediary species. It's built atop a stable rhoSonicFoam run, dealing with it without combustion before switching over to reactingFoam.

It runs for a fair while, but almost inevitably, eventually "something happens" that leads to a blowup. In general, it appears to usually be along the lines of turbulence somewhere causing a localized patch of heating, which increases the reaction rate, increasing heating faster, etc such that temperatures skyrocket or go negative and things blow up. Yet it also happens (faster, actually!) if I set the turbulence model to laminar and/or the combustion turbulence to laminar.

I can generally prevent these blowups by dropping the timestep way down. But then it runs at a crawl. It's like 99.999999% of all calculations are fine, but that last 0.000001% blows things up, and to prevent that I have to tremendously slow things down for the 99.999999%. It would be lovely if chemistry could limit the timestep the same way as the courant number does... but I see nothing in either the docs nor the code to allow that.

* I make use of a temperature limit in fvOptions, but it doesn't seem to do anything useful.
* All of my fvSchemes are first-order that can be, most commonly Gauss linearUpwind grad(U).
* I use no relaxation parameter in fvSolution because they don't seem to help much, only delaying the inevitable
* My experimentation is mixed as to whether varying tolerances in fvSolution helps any or make things worse. And if they're high enough then strangeness crops up in the run, so for now I'm just keeping them low.

What do you do in such a situation to try to prevent these blowups? I've even considered a cheap hack, running reactingFoam via a looping external script that automatically lowers the timestep (then raises it after a period) if things blow up. But often when I restart reactingFoam the existing cell contents change for some bizarre reason - for example, all of the fuel might disappear and have to be repopulated, which of course screws up the burn, introduces more turbulence and just makes the problem worse.