|
[Sponsors] |
June 29, 2016, 05:19 |
Combustion Model
|
#1 |
Senior Member
Tobias
Join Date: May 2016
Location: Germany
Posts: 264
Rep Power: 10 |
Hi,
i am using CONVERGE for Gasoline PFI and DI engines. I have always worked with SAGE, because it is pretty straight forward. All what is needed is the fuel within the engine and a suiting reaction mechanism and a Ignition source. However, i wanted to know if someone has used other combustion models for specific cases? I noticed that especially with several cycles the WallTimes increase, even though i added the skip species. Thanks for respones. Tobias |
|
July 1, 2016, 12:10 |
|
#2 |
Member
Dan Probst
Join Date: Jan 2016
Location: Convergent Science, Madison WI
Posts: 38
Rep Power: 10 |
Yes, you can use simplified combustion models for gas PFI and Diesel. For gas PFI, the G equation model could be used, while for Diesel the shell/CTC could be used.
While these models might require tuning to match data, they can offer shorter runtimes. Detailed chemistry, however, is a good choice when predictive results are required. |
|
July 6, 2016, 04:07 |
|
#3 |
Senior Member
Tobias
Join Date: May 2016
Location: Germany
Posts: 264
Rep Power: 10 |
Yes, at the moment i am completly on the predictive side, as no experimental data is available yet.
But i think i am running into another issue with the validity range of the mechanism. The SAE 2013-01-1098 paper is mentioning the range for the jia mechanism, which is 10-50 bar. However, i might have cases with even lower pressures at spark timing (3-4 bar). What can i do to compensate this? Adding spark energy? In that paper mentioned above, a reduced mechanism especially for low pressure range has been used. But i dont want to try reducing a mechanism, because it would be a completly new field. What do you think about the idea of implementing a mechanism database in the converge download area of your homepage? |
|
July 6, 2016, 16:30 |
|
#4 |
Member
Dan Probst
Join Date: Jan 2016
Location: Convergent Science, Madison WI
Posts: 38
Rep Power: 10 |
If the operating conditions are outside of the applicable range for the reduced mechanism, the results may not be satisfactory. Adding spark energy would not be a solution for a poorly performing mechanism.
I would suggest first running the baseline case using an appropriate mechanism and then assess the results against experimental data. We do not currently have a mechanism library on the downloads page, but we may have something like this in the future. We do recommend mechanisms with our example cases, which are available for download. Thus, you can look at the example case that most closely matches your situation (SI, DI, etc...) for a recommended mechanism. |
|
July 8, 2016, 04:40 |
|
#5 |
Senior Member
Tobias
Join Date: May 2016
Location: Germany
Posts: 264
Rep Power: 10 |
I am using the Jia PRF / TRF mechanism.
When i compare the 0D Ignition delay times with the detailed mechanism from LLNL within a range of 1-11 bar, they seem not that far off. How can i further check the performance of the mechanism? edit: Ive set up the same simulation with G-Eq. and default inputs. G-Eq. doesnt need a mech.dat and Temperature AMR, right? Last edited by MFGT; July 8, 2016 at 08:34. |
|
July 8, 2016, 11:31 |
|
#6 |
Member
Dan Probst
Join Date: Jan 2016
Location: Convergent Science, Madison WI
Posts: 38
Rep Power: 10 |
The performance of the mechanism can be assessed by running simulations and comparing to experimental data.
Geqn does not need Temp AMR. The combustion is determined by correlations for flame speed. You can look at our example case for SI8 Geqn to see recommendations. |
|
July 11, 2016, 03:25 |
|
#7 | ||
Senior Member
Tobias
Join Date: May 2016
Location: Germany
Posts: 264
Rep Power: 10 |
Hey Dan,
the G-Equation model performed a lot better for my current case (2000rpm - 2bar bmep). I took all default inputs from the Example case. edit: Sorry again, but i am running into problems with the G-Equation model. During gas Exchange, i suddently run into the following issue (here is the log): Quote:
Quote:
Or what else is causing this issue? I am checking if the simulation can continue from 380°CA without combustion model. Last edited by MFGT; July 11, 2016 at 09:05. |
|||
July 11, 2016, 10:08 |
|
#8 |
Member
Dan Probst
Join Date: Jan 2016
Location: Convergent Science, Madison WI
Posts: 38
Rep Power: 10 |
It looks like there is a problem with your case setup. G eqn uses passives to track the flame.
Please email your case setup to support@convergecfd.com for us to help resolve this issue. |
|
July 11, 2016, 10:48 |
|
#9 |
Senior Member
Tobias
Join Date: May 2016
Location: Germany
Posts: 264
Rep Power: 10 |
Thanks.
I´ve done that. Incuding a furher description and coressponding map files. |
|
July 12, 2016, 09:04 |
|
#10 | |
Senior Member
Tobias
Join Date: May 2016
Location: Germany
Posts: 264
Rep Power: 10 |
Quote:
What is the "best" CA to reinitialize G? Can i limit the G-Equation model to region0? |
||
July 12, 2016, 12:00 |
|
#11 |
Member
Dan Probst
Join Date: Jan 2016
Location: Convergent Science, Madison WI
Posts: 38
Rep Power: 10 |
You need to reinitialize G for multiple cycles. The SI8 Geqn example case shows how to do this. This example case is available for download.
The g_eqn_init file is referenced in combust.in: "g_eqn_init.in" g_eqn_init_value Initial G-value. G should be reinitialized at a appropriate crank angle after combustion is expected to be complete - such as 120 dATDC. |
|
July 13, 2016, 10:00 |
|
#12 | |
Senior Member
Tobias
Join Date: May 2016
Location: Germany
Posts: 264
Rep Power: 10 |
During ignition, i ran into time-step limit =dt_min forced by dt_src_passive.
dt_min is 1e-8 Luckily, it can recover an get back to values in the range of 2e-6. Is this supposed to happen? Whats the reason? I couldnt find any information about dt_src_passive. edit: Quote:
Or is CONVERGE receiving a value from the g_eqn_init.file? |
||
August 1, 2016, 13:54 |
|
#13 |
Senior Member
Tobias
Join Date: May 2016
Location: Germany
Posts: 264
Rep Power: 10 |
Dan helped me out here as well.
As help for other users, i had forgotten to add the G passive at the IN- and OUTFLOW boundaries. Of course this leads to a problem, if i flush my G passive out of the domain with multi-cycle Btw, the initialization with the g_eqn_init.file worked fine. |
|
August 4, 2016, 10:56 |
|
#14 |
Senior Member
Tobias
Join Date: May 2016
Location: Germany
Posts: 264
Rep Power: 10 |
Another question regarding the combustion model, this time about SAGE.
What is the best practice to start and end times of SAGE? In the engine sector cases, SAGE is activated 1°CA before SOI. In the SI8 PFI case, SAGE is also activated 2°CA before SOI. However, in the Premixed SI8 case, SAGE is activated 2°CA before IGN. In my PFI and DI gasoline simulations, I have seen heat release even before my Spark Timing, especially in multi cycle simulations. There is no real temperature rise, so i dont think its auto-ignition. |
|
August 4, 2016, 12:58 |
|
#15 |
Member
Dan Probst
Join Date: Jan 2016
Location: Convergent Science, Madison WI
Posts: 38
Rep Power: 10 |
You should be sure SAGE is turned on 1 or 2 CA before you expect the earliest SOC. If you might expect auto ignition before spark, the SAGE start time should be advanced accordingly.
The end time is typically EVO but can be longer or shorter as necessary. The primary condition for keeping SAGE on after EOC is emissions. |
|
August 16, 2016, 08:08 |
|
#16 |
Senior Member
Tobias
Join Date: May 2016
Location: Germany
Posts: 264
Rep Power: 10 |
Yep, that makes sense. For now, i switch off SAGE before EVO and my skip_species timing.
|
|
September 28, 2016, 10:52 |
|
#17 |
Senior Member
Tobias
Join Date: May 2016
Location: Germany
Posts: 264
Rep Power: 10 |
In a simulation with SAGE, i have a slower burn rate (and pressure rise) compared to measurement data. I guess its reasonable to tune SAGE with the help of the reaction multiplier then?
Or is this something you would rather not do? |
|
September 28, 2016, 17:19 |
|
#18 | |
Member
Shengbai Xie
Join Date: Aug 2016
Location: Convergent Science, Madison WI
Posts: 60
Rep Power: 9 |
Quote:
The lower burn rate can be resulted by many reasons. You may have underestimated turbulence, or the mechanism you used is not good, etc. I think you'd better first check your case setup and make sure everything is correct. You can also compare other properties such as TKE with the measurement data to figure out where the problem is. Tuning the multiplier may help to get a correct burn rate, but it may also mask some other problems in your case. It is purely numerical. So you'd better at first make sure everything else is OK. |
||
September 29, 2016, 03:01 |
|
#19 |
Senior Member
Tobias
Join Date: May 2016
Location: Germany
Posts: 264
Rep Power: 10 |
Thanks for your feedback. The mechanism should be fine. I am simulating a 1500 WOT case and the mechanism has been tested for similar conditions in SAE 2013-01-1098.
I am trying to rerun it with new inputs, as i also have troubles to get the desired cylinder charge. Therefore i am using GT-Power pressure and temperature profile inputs. |
|
November 28, 2016, 09:15 |
|
#20 |
Senior Member
Tobias
Join Date: May 2016
Location: Germany
Posts: 264
Rep Power: 10 |
I figured out to not touch the SAGE multiplier. With optimized other inputs i was able to achieve better suiting conditions at IGN timing and the combustion was matched better.
Now, new question regarding the G-Eq. Combustion Model: i use G-Eq. for idle/low load operating conditions. However, i came across a few situations where i had to retard the G source timing by up to 20CA to match pressure rise. Otherwise i had a way to high peak pressure and early pressure rise. Also, i had to lower b1 down to 1.5 to match the combustion intensity. Now today i came across the dillution species mass fraction, which i had never adjusted and i obviously had lots of EGR in the above mentioned cases of low loads. So would it be right to set it to e.g. 0.222 if my GT-Power model has an EGR ratio of 22.2%? Last edited by MFGT; November 29, 2016 at 03:21. |
|
|
|
Similar Threads | ||||
Thread | Thread Starter | Forum | Replies | Last Post |
Combustion Model - Laminar Flamelet Model | spg1g14 | CFX | 0 | August 5, 2015 18:13 |
Combustion model vs. chemistry model (thermo type) | wenxu | OpenFOAM | 2 | July 19, 2015 21:15 |
Liquid rocket engine non-premixed combustion model | Erik | FLUENT | 2 | November 28, 2013 14:09 |
eddy dissipation model: combustion doesn't occur | roukaia | FLUENT | 0 | December 24, 2011 09:10 |
combustion model | Hennie van der Westhuizen | Siemens | 7 | February 27, 2002 02:10 |