[shailenr]

Hi Everyone,

I am currently facing challenges in matching the in-cylinder pressure trace in diesel combustion simulations. We are using a well-validated multi-component diesel surrogate mechanism with 176 species and 933 reactions. Despite testing several approaches—including varying spray breakup parameters and implementing a two-step cone spray embedding—none of these adjustments have led to a significant improvement in matching the experimental pressure profile. Hydraulic delay is present in my case and we also need to consider, but as of now, I do not have the information yet. However, when I looked at the AHRR plot and you can see the AHRR starts to go negative which was around 722.5 deg CA. So I considered 2.5 deg CA and included to SOI = 0 DEG CA. Also, when I calculated the total cumulative heat release, my results differ by 20% against the experiments. So I also checked the total mass of the fuel (species_mass.out in Converge CFD), it seems to be 5 mg compared to what was injected (7 mg injected). This probably correlates to the emission values as the carbon emissions from my simulations were also low for CO, CO2, and UHC compared to experiments.

I have attached plots for pressure, AHRR, and emissions.

As you can see, the simulation deviates from the experimental results, and I am looking for guidance on which strategies or parameters I could explore to improve the pressure trace agreement.

Any suggestions or insights would be greatly appreciated!

Thank you for your support.

Best regards,
Shailendra

[alenp07]

Originally Posted by shailenr

However, when I looked at the AHRR plot and you can see the AHRR starts to go negative which was around 722.5 deg CA. So I considered 2.5 deg CA and included to SOI = 0 DEG CA.

- AHRR in Converge is a chemical heat release where the wall heat transfer effects are not considered. Since it is chemical it is okay to go negative. Is this diesel surrogate Dodecane or N-heptane ? Can you try the same with another reaction mechanism? Also can you calculate the Apparent heat release as well between experiment and simulation?

- Also can you revist the wall temperatures especially at Piston Liner and Head? What are your assumptions ?

- Since it is diesel your heat release can be a strong function of your injection profile. How confident are you with these ie: Injection Rate Shape profile, injection pressure, injection velocity etc?

Originally Posted by shailenr

Also, when I calculated the total cumulative heat release, my results differ by 20% against the experiments. So I also checked the total mass of the fuel (species_mass.out in Converge CFD), it seems to be 5 mg compared to what was injected (7 mg injected). This probably correlates to the emission values as the carbon emissions from my simulations were also low for CO, CO2, and UHC compared to experiments.

- Which version of Converge do you use? and if your experienceing mass loss during combustion, can you try a case with inliad mesh for injectors if you are not using ?

-7mg injected mass is too low in my opinion are sure with the injected mass ?

There are so many points Sorry! Happy troubleshooting !

[shailenr]

Hi, and thanks for the comments — here are brief replies to each point.

I am running CONVERGE v4.1.2 with a four-component diesel surrogate (decalin, HMN, a2ch3, nC20H42), not dodecane or n-heptane; I have not yet tested a different reaction mechanism but can if you think that’s a useful next step.

Just to clarify, the plot I showed is the apparent heat release derived from the pressure trace, not CONVERGE’s chemical AHRR (which can go negative because it excludes wall heat transfer). My cumulative heat release from the pressure trace differs from experiment by ~20%.

Wall temperatures for head, liner, and piston are set to 400 K based on the experimental data supplied to me; I’m open to revisiting these if you suspect they are too low.

Regarding injection, the pulse is 3° CA at 750 bar with Cd = 0.6071 from the spray-rate preview. I’m testing injection-rate shapes now: a triangular profile produced no ignition, a rectangular profile is running, and I’ve also checked fixed embedding (0.125 mm) and a two-step cone (0.0625→0.125 mm) with no change. I adjusted spray breakup parameters with no meaningful effect on the pressure trace.

On fuel mass: the injected mass per pulse should be 7 mg, but species_mass.out shows ~6.1 mg in cycle 1 and ~5 mg in cycle 2. I summed the surrogate species to get those numbers; can you confirm that this is the correct way to evaluate vaporized/remaining fuel mass or suggest a better check? I have not yet used an inlaid mesh for the injectors — I can run that if you think it will help recover lost mass or improve early jet breakup.

Thank you !

Regards,
Shailendra

[alenp07]

Originally Posted by shailenr

Hi, and thanks for the comments — here are brief replies to each point.

I am running CONVERGE v4.1.2 with a four-component diesel surrogate (decalin, HMN, a2ch3, nC20H42), not dodecane or n-heptane; I have not yet tested a different reaction mechanism but can if you think that’s a useful next step.

Essentially you are using 4 component mechnism to get more real world diesel like behavior. However I suggest you to try a single component mechnism either n-heptane / dodecane so as to check. I've been using these mechanism for ages and until unless if it is a special case it always produced good results. Also under 'Drop Evaportation' what is your evaporation source ? is it a 'specified' , 'all' or 'base' ?

Originally Posted by shailenr

Just to clarify, the plot I showed is the apparent heat release derived from the pressure trace, not CONVERGE’s chemical AHRR (which can go negative because it excludes wall heat transfer). My cumulative heat release from the pressure trace differs from experiment by ~20%.

Wall temperatures for head, liner, and piston are set to 400 K based on the experimental data supplied to me; I’m open to revisiting these if you suspect they are too low.

This is bit worrying I strongly suggest you to do some trials with your boundary temperatures. Also try giving different temperatures (marginally) for piston liner and head. Do a check on the LHV values of the fuel and see if it is different in your simulation

Originally Posted by shailenr

Regarding injection, the pulse is 3° CA at 750 bar with Cd = 0.6071 from the spray-rate preview. I’m testing injection-rate shapes now: a triangular profile produced no ignition, a rectangular profile is running, and I’ve also checked fixed embedding (0.125 mm) and a two-step cone (0.0625→0.125 mm) with no change. I adjusted spray breakup parameters with no meaningful effect on the pressure trace.

On fuel mass: the injected mass per pulse should be 7 mg, but species_mass.out shows ~6.1 mg in cycle 1 and ~5 mg in cycle 2. I summed the surrogate species to get those numbers; can you confirm that this is the correct way to evaluate vaporized/remaining fuel mass or suggest a better check? I have not yet used an inlaid mesh for the injectors — I can run that if you think it will help recover lost mass or improve early jet breakup.

To calculate the injected mass, for liquid mass use 'spray_rate_inj.out' and use Inject_Mass. To calculate the evaporated mass, use 'thermo_region0.out 'and calculate the difference in trapped mass before injection and some crank angles after injection. The difference between evaporated mass and Inject_mass should be as low is possible.

Is this a sector case? if not in a post processing tool do a check if the liquid parcels coming from each injector is symmetric or not. Also give a try with inlaid mesh with the same refinement settings as that your embedding.

All the best!

[shailenr]

Thank you for the suggestions. The evaporation source is currently set to “all,” which I believe is appropriate for this mechanism. Regarding the LHV, I am not entirely sure where to check it in CONVERGE; currently, I load the parcel species from the library and specify their mass fractions. For reference, the experimental LHV is 43.16 MJ/kg, while the mechanism targets 42.061 MJ/kg. I think this difference is not significant.

I appreciate the guidance on assessing vapor mass. I am running full engine simulations, not sector cases. For now, I am planning to test cases with a different mechanism and others with slightly varied wall temperatures for the piston, liner, and head. I will consider implementing an inlaid mesh later, after analyzing these results.

Thank you again.


Regards,
Shailendra

[shailenr]

Hi again,

I have a few questions regarding the calculation of total cumulative heat release. I suspect that my current heat release calculations might be inaccurate.

In particular, I am unsure how to properly select the range of pressure data for the analysis. For my calculations, I used in-cylinder pressure data from IVC to EVO. I then set all AHRR values before the start of injection (SOI) to zero and considered the data from SOI up to EVO for the cumulative heat release calculation.

Do you think this is a reasonable approach for determining the total cumulative heat release? Additionally, what is the proper way to determine the mass of fuel burned and the leftover fuel? It is possible that the fuel burn may not be efficient, so the heat release may not match the experiments. Do you have any comments on this?

By the way, I checked the liquid parcel injection in Paraview, and there is a clear indication that the spray is symmetric, with an injection velocity of 390 m/s (after considering contraction effects).

Regards,
Shailendra

[alenp07]

Originally Posted by shailenr

Hi again,

I then set all AHRR values before the start of injection (SOI) to zero and considered the data from SOI up to EVO for the cumulative heat release calculation.

Do you think this is a reasonable approach for determining the total cumulative heat release?

Very interesting question! I will give my perspective may be it can be different from others. If you are doing a comparison between experiment and simulation the key is to ensure that the 'method' you use in simulation and in experiment is consistent. For example in experiment historically test engineers tend to use a fixed 'gamma' or some empirical correlation based on temperature . Similar consideration is applicable if you are using GT -Suite data calibrated from experiments. Then there is a question on how do you calculate the volume in experiment and often it is done using the slider crank mechanism.

While using the Converge CFD AHRR utility, make sure it is consistent with your experiment calculations. If you have your pressure tracesfrom both experiment (with a decent Crank Angle resolution) and simulation do your calculations on your own with Converge Tool or a custom python code in such a way that you are consistent.

With this background you can select IVC and EVO or couple of crank angle before SOI till EVO or SOI till 150 CA after SOI all should work. And if you are getting your AHRR correct, your Cumulative HRR also should match. I think modifying AHRR values before SOI before performing Cumulative HRR is not a neat way to do it.

Originally Posted by shailenr

Additionally, what is the proper way to determine the mass of fuel burned and the leftover fuel?

For Liquid Fuel -->liquid_parcel_region<cylinder>.out -->liq_spray_mass: It must be zero at EVO

For Gaseous Fuel -->species_mass_region<cylinder>.out -->all your fuel species component (s) should be zero at EVO

If it is not then there is a problem!


Regards,
Shailendra

[shailenr]

Hi again,

I've written a Python code to calculate the apparent heat release rate (AHRR) and the total cumulative heat release (TCHR). I verified my code by comparing the AHRR results with those obtained from the CONVERGE AHRR utility, and they match closely, so I believe the code is correct.

The issue arises with the TCHR calculations. When I use the dataset from IVC to EVO and simplify the data by zeroing out all values before the start of injection (SOI)—which represent intake and compression work rather than heat release—and then use all the data until EVO, I obtain a TCHR of 202 J compared to the experimental value of 193 J.

Interestingly, when I checked the unburned hydrocarbon (UHC) amount at EVO, it is around 0.63 mg compared to the total fuel amount of 7 mg. Even when accounting for heat losses and combustion efficiency, the TCHR seems too low for the numerical results. If the burned mass is high but the calculated heat release is low, this suggests there could be an issue with the lower heating value (LHV) used.

In addition, as another scenario for obtaining TCHR, the experimental AHRR contains some degree of noise throughout the dataset. I filtered the AHRR data again, used it up to EVO, and then zeroed out the negative values from the entire dataset and the values before SOI. After this procedure, I obtained a TCHR of 251 J for the experiments and 212 J for the numerical results.

Not sure how to proceed here, but there is definitely something wrong in my simulation setup and I do not know what.

Thank you for the cooperation.

Regards,
Shailendra