Over predicted combustion from RANS+SAGE
 

Hi all,

How are you doing?

I met some issues or similar problem again while using RANS + SAGE. Please check out the attached photos. I plotted the output results with Matlab.

Engine running conditions:
5000 rpm, WOT, no EGR, lambda=1
SA=1.8 CA ATDC (this is coincident with the experimental data)


Simulation setup:
SA=1.8 CA ATDC
RANS + SAGE (Prof. Ming Jia 48 species & 152 reactions)
AMR level 3 on temperature
Base grid 2mm
Fixed Embedding on both flame kernels and intake/exhaust valves (refered to Converge examples)
Fixed embedding level 2 on entire cylinder permenantly (refered to Converge examples)
Combustion modeling started 1 degree before Source/Sink Modeling kicked in
Reaction multiplier=1



As observed from the figures, pressure and temperature aligned well with the 1D data (since this 1D data has been calibrated based on experimental data already) before the combustion. However, the combustion rate from 3D started to go off and over-predict the combustion. I have tried retarded SA timing but the over-predicted still existed.


My questions are summarized as below:
1. Has anyone else meet some similar situation like this? Please share your thought as much as you could. This could be very useful.

2. Does it relate to the mechanism itself? obviously the ignition delay was shorten through 3D CFD and that is not how the experimental (or 1D data) looks like.

3. Does it relate to the turbulence modeling? RANS, LES, etc,....

4. Is that the case that if I want to match the pressure/temperature curve, the only way I can do is to manually shift SA?


I have been successfully simulating a 3000rpm case (SA= -18.8 CA ATDC) with 23.77% external EGR before, and I followed the same philosophy to go through this 5000rpm case. Obviously it is not showing the same trend.

Thank you and Auf Wiedersehen

The SI Wiebe is 1D matching the EXP?


thats a strange 5000 WOT pressure profile, altough its CR16.
How does it look like if you adjust your spark timing to lets say 10° aTDC?


I usually dont take the EXP and simulated IGN timings too seriously, a couple of degrees difference is not uncommon. Btw, is that the 1D IGN timing or EXP ignition timing or are these two the same?

Hallo Tobias

My friend, it is good to see you again. Wie geht's?

Yes, that is a strange 5000rpm WOT profile, since it comes to knocking limit.
Check out the attached figure. This one is the SA retard to 6.8 CA ATDC (the experimental SA is 1.8 CA ATDC, as same as 1D simulation), and everything is starting to line up with experiment.


About the 1D simulation: Since 1D simulation is usually based on Wiebe function or SI-turbulence model, it can manually adjust to match up with experiment pretty well. So yes, the SA in 1D is the same as experiment.


The reason why I am really diving into these questions is that, it is not the first time for me to see papers or dissertations talking about RANS+SAGE over-predicted combustion rate. In fact, folks only discuss this kind of issue in their thesis or dissertation BUT NEVER IN THEIR PUBLICATION. In order to do some deeeply precise calculation before any experimental data available, firstly we must figure out how much bias does the simulation generate.


Unfortunately, the last time I got reply from Converge application engineer (and it is pretty much the same question, Over-predict combustion by RANS+SAGE), Converge staff only said that there are many publications over there and that's it.



Danke, mein Freund