NIST-Waterloo Pool Fires

[rmcdermo]

Hi All,

I wanted to start a discussion for the pool fires cases. Please cc anyone in your organization that I may have missed. I don't know everyone's GitHub usernames.

If you are not aware, the plots for everyone's cases are posted here with the zip files in alphabetical order.

@gatw In looking at the BDI results for methanol pool fires, your FDS centerline temperatures are better than ours. I am noticing I have this issue across the board and think it is due to the EXTINCTION 2 model we are using. In fact, I did not realize until too late that Kevin had set SUPRESSION=F on these cases last time (MaCFP-2) to deal with this issue. But it is really just sweeping it under the rug. My question to you is, did you use an extinction model or did you also have SUPPRESSION=F? If you did, please note this in your README.md. Thanks!

Adding relevant users to the discussion:
@ahaminsgithub
@mameehan5
@hong27
@gmaragko82
@NingRen
@atrouve
@bradensouthern
@rlfalken
@tbeji
@ejweckman

[tbeji]

Thanks Randy for initiating the discussion.

Perhaps another point regarding the comparison NIST Vs DBI (prescribed MLR) is the cell size. It seems to me that for the 1 cm cell size, the results are quite comparable for the mean temperature at the centerline of the 30 cm methanol pool fire. DBI used an even finer mesh with a 0.5 cm cell size, which appears to capture better the high temperatures near the fuel surface.

[gatw]

Hi Tarek and Randy,

Sorry for the delay in getting back. I was on vacation for the past two weeks. We used the SIMULATION_MODE = 'LES' for all our simulations (even for the 2cm and 1cm grids). I assume this means that we have used the 'EXTINCTION 2' model in our simulations.

We did not use 'SUPPRSSION = F' in our simulations.

I agree that our README.md file should include more details regarding the simulation settings. I will try to update the file.

Also I agree with the point Tarek has made regarding the finer meshes capturing well the temperatures near the fuel surface in our simulations.

[gatw]

I should also mention that the we had the 'ZMIN' boundary not 'OPEN' and the rim of the pan was located 30cm above 'ZMIN' (= floor) similar to what is reported from the experiments. I see that this was not the case in NIST simulations. We also did not use the new 'GEOM' feature for the pan geometry although it would have been nice to compare the outputs from that setting with the conventional hex-block geometry we have used here.

[rmcdermo]

FYI, I just pushed up NIST results for puffing frequency. The updated plots are here and the README.md file has a table with the computation results for puffing frequency compared to the measured.

[tbeji]

Noted.

[tbeji]

Hi Randy,

I have a few questions about the NIST results.
@other participants: I will address you in separate comments, but I strongly encourage you to intervene here as well.

The questions are mainly about the predictive methanol pool fire simulations.

1. What lip height did you prescribe? What do you do when the cell size is larger than the lip height?
2. Why did you go for an ignition protocol? I guess you did not keep the AIT to its default value of 0K, right? What happens if you do so?
3. The results for total heat flux at the liquid surface look good in terms of convergence and accuracy. Can we explain then why aren't the predicted MLRs not as well predicted? Do you think that the results call upon using the approach where MLR = Heat Flux / Heat of Gasification?
4. The number of radiation angles of 600 is particularly high in comparison to the default in FDS as well as to what has been used by FM, Sandia and UGent (between 64 and 128). What is the reason? Have you tried the default number of radiation angles?

I have one general question about turbulence modelling. Do you think that Dynamic Smagorinsky (DS) can reduce the mesh sensitivity? That may be inferred (without compelling evidence though) from a comparative analysis of the results. I recall your reluctance however to use DS for numerical issues.

Thanks!

[rmcdermo]

Regarding the turbulence model, as you know, I do not think DS solves any of the problems we are having with this case. And I can point to many other problems that it creates. I believe that most of our issues stem from the treatment of the first layers of cells at the surface and also at the edges (maybe mostly because of the edges). DS is not well defined at a surface. Neither is our variant of Deardorff, hence we use WALE for nu_t. The vorticity at edges dominates the puffing behavior of these pools, and this is difficult to get right. As I mentioned, it is something we need to deal with for the cutcell method.

[tbeji]

Thanks, Randy. I will do my best to convey the information during the presentation at the workshop.

It would be great indeed if cases without an ignition source can be run, at least for the 4 cm and 2 cm mesh. It is your call of course; I know you are very busy.

I understand your statement about the issue of grid dependence related to the lip. However, please note that there is also no clear convergence in the NIST results for the 37 cm methane fire where there is no lip.

The DS approach will be a good point of discussion, given that it is the retained (perhaps also preferred) option by Georgios (UGent) and Ning (FMGlobal).

[rmcdermo]

I understand your statement about the issue of grid dependence related to the lip. However, please note that there is also no clear convergence in the NIST results for the 37 cm methane fire where there is no lip.

In a "cutcell" the subgrid geometry is not perfectly maintained. The issue of how the vorticity is computed on edges applies to both the pools with an explicit "lip" and the cutcell edges of the 37 cm gas burner. They both require subgrid representations for the momentum equation for the Cartesian cell that owns the cutcell.

The DS approach will be a good point of discussion, given that it is the retained (perhaps also preferred) option by Georgios (UGent) and Ning (FMGlobal).

There is no argument from me that DS can be made to work. Since no other groups attempted all the cases listed in the workshop, I do not see how anyone can claim the approach is clearly more robust and accurate across the board for the general case. Unlike everyone else who focuses on one or two cases, I have run DS against every case in the FDS validation guide. Without exception, when DS worked better it ultimately turned out to be related to uncertainties entering the boundary, not because of better subgrid stress closure or mixing time scales in the bulk flow. At the resolutions we are running, we might as well be using Vreman instead of DS.

[tbeji]

In recent collaborations that I started with the universities of Poitiers and Nancy in France, I always insist on using square burners in order to easily make the computational Cartesian grid perfectly coincide with the geometry. Do you think that this helps mitigate, to some extent at least, some problems and allows to analyze modeling results more easily? Or there is no advantage in doing this?

[rmcdermo]

Well, for us, certainly the Cartesian code is more mature. As I said, my decision to use the new geometry features for this exercise has its drawbacks. But we need to push this method forward.

I think a circular burner should be something we can handle. Others seem to have no problem with it. This is the advantage of an unstructured code.

[tbeji]

@NingRen

Hi Ning,

Some similar questions about the predictive methanol pool fire simulations:

1. What lip height do you prescribe? I guess that with the structural grid refinement in the burner region, you are able to keep the same lip height even when you change the resolution further downstream in the fire and smoke plume regions, right? What is the cell size that you prescribe just above the liquid surface?
2. I guess that you did not use an ignition protocol because you set your AIT to a low value, right?
3. For predictive 30 cm methanol simulations your MLR results are great. However, I noticed that your total heat flux predictions at the liquid surface are underestimated by about 20%. Given that in your approach (if I understand correctly), MLR = Heat Flux / Heat of Gasification, can we say that there is an 'issue'/uncertainty regarding the calculation of the heat of gasification?

Thanks!

[NingRen]

@tbeji I used structured Cartesian grid, and the pool has castellated edge. The computational cell is a cube, so have the same dimension in x,y,z directions. For the 5 mm and 1 cm grid, the lip height is 1 cm. For coarse grid, e.g., 2 and 4 cm, the lip height will be the cell size. The image attached is an example of 5 mm grid.

The AIT is 0. Initially, there is very slow evaporation. As the pool heats up slowly, the evaporation and HRR increases. It roughly reach steady state after about 3 mins.

The predicted heat flux is lower than measurement in the center region. The center has smaller area. If you integrate the heat flux over the pool area, it may not be very bad. On the other hand, the measured value maybe higher than what the liquid received. The liquid evaporation causes the 'blowing' effect, which reduces the convective heat flux. I guess the heat flux gauge don't have this effect. The heat flux data in the submission is the liquid received heat flux, which may differ slightly from the measurement.

The heat of vaporization is from the OpenFOAM thermodynamic database. At boiling temperature, the heat of vaporization is about 1.115 kJ/kg, which is slightly smaller than the value of 1165 kJ/kg (https://www.engineeringtoolbox.com/methanol-methyl-alcohol-properties-CH3OH-d_2031.html)

I have a liquid fuel model, which has not been released to public version yet. This model is able to model liquid spill and in-depth heat transfer. For the 30-cm pool, I have the liquid depth and temperature profile at different time (see attached figure). @rmcdermo It will be nice if NIST provide some liquid temperature data.

I have in-depth radiation in my model. For methanal pool, the in-depth radiation don't change the steady state burning much. For fire spread over high-flashpoint liquid, in-depth radiation can make a big difference.

[rmcdermo]

@ahaminsgithub may be able to provide this data now. I thought we had pushed up these files, but I don't see them at the moment. I think we were trying to keep the repo stable while results were coming in. If it will help the discussion, @ahaminsgithub feel free to send a PR and we can add the in-depth profiles.

[tbeji]

@NingRen Thanks a lot for the information.

1. The integration of the heat flux is a good explanation. If you can do the calculation and provide me with the result, I will mention that during the presentation.
2. There is potentially a good discussion point with the experimentalists about the blowing effect. @ahaminsgithub Any thoughts about this?
3. @NingRen and @rmcdermo During IAFSS, I will present a paper on transient heating of liquid pool fires, including in-depth temperature measurements with a good resolution and fine thermocouples. The experiments are done at the University of Nancy (France).
   Here are some results for 15 cm methanol and heptane square burners with L = 15 cm.
   
   I will be happy to share the data. Just need some time.
   Some FDS results are here for heptane at different heights, z:
   
   The Nusselt number is an INTERNAL Nusselt number linked to an effective conductivity used in the 1D solution of Fourier's equation in FDS. In the paper, we explain how to estimate Nu experimentally. FDS confirms the validity of our experimental estimates.
   In any case, I will be very happy to interact with you on this topic.

[rmcdermo]

@NingRen Anthony now has pushed up the in-depth temperature profiles at r=1 cm and r=13 cm. There are time histories available in the following files:

Methanol_30_cm_Indepth_Fuel_TC_r=1_cm.csv

Methanol_30_cm_Indepth_Fuel_TC_r=13_cm.csv

[NingRen]

@rmcdermo @ahaminsgithub
Thanks Anthony and Randy for sharing the liquid temperature data. The 0.8-cm-deep temperature keeps increasing slowly even after 10 mins. Definitely need to run the simulation longer to reach steady state. The 0.8-cm temperature is close to the boiling point at 2400 s. Do you have any measurement of the thickness of the boiling region? In my current simulation, there is a 3-mm thick of fuel reaches the boiling point, which can be increased by tuning the in-depth radiation. Any idea about the radiation penetration depth for the methanol pool fire?

Best,

Ning

[tbeji]

@mameehan5

Hi Michael,

I have a few questions for you:

1. I am not familiar with the Inagaki model for the SGS viscosity. Can you please provide me with the equation that you use? Why not use more 'common' models like Smagorinsky (static or dynamic)? Is there a specific reason?
2. Can you provide details about the prescribed lip heights?
3. You provide only 2 cm results and mention problems with convergence. Can you be more specific? What happens if you use a coarser mesh with 4 cm cell size or a finer mesh with 1 cm cell size?
4. I noticed that your fuel concentration in the centreline of the 30 cm methanol pool fire is particularly low near the liquid surface. The core of your flame is too fuel lean, as opposed to the experimental data and the numerical data of the other contributors. Can we attempt to explain that? Could there be an 'issue' with the boundary condition for fuel concentration?

Thanks!

[mameehan5]

Hi @tbeji , thanks for reaching out, and sorry for such a long delay in reply! I will be taking a close look at all this information tomorrow and Friday and provide those details then. Thanks!

[tbeji]

Thanks.

[mameehan5]

Sorry again for the delay. Here is that information you are looking for:

1. The Inagaki model is a wall-resolved LES approach (see reference below), so it requires sufficient resolution at the boundaries to be effective. Without going into too much detail, it is a single equation closure similar to that proposed by Yoshizawa (reference also below) except the dissipation rate adds an additional term 2\mu(d sqrt(k_sgs)/d x_j)^2 . The turbulent viscosity is modified with a blending function of \mu_t = f_\mu \bar{\rho} C_\mu \Delta sqrt(k_sgs) with f_\mu = 1 - exp(-(y_e'/20)^(4/3)) and y_e' = y u_e'/nu and u_e' = (\nu \epsilon_sgs)^{1/4} sqrt(4 y/\Delta). Some recent internal work suggested that this approach may be better for capturing the near-boundary dynamics near the pool surface since this region is very difficult to model due to buoyancy and sharp density gradients.
2. The lip heights are 1 cm for both cases.
3. Given the discussion above in (1), the model is pretty resolution sensitive, especially when the mesh is under-resolved. I didn't include additional results on the various resolutions because there was a lot of complexity that needed to be unpacked before I was comfortable presenting them. I will have a few plots looking at mesh resolution for my poster though!
4. I also noticed that behavior and realized that I had actually been pulling data from the top of the lip rather than the pool surface. This should at least explain why it is so low at first data point (I can re-submit my results with this change if you would like). Otherwise, I will likely need to look at how the SGS model is connected to this behavior; it shouldn't be an issue/mistake with the implementation at the boundary surface.

References
Inagaki, Masahide, Tsuguo Kondoh, and Yasutaka Nagano. "A mixed-time-scale SGS model with fixed model-parameters for practical LES." J. Fluids Eng. 127.1 (2005): 1-13.
Yoshizawa, Akira. "Bridging between eddy-viscosity-type and second-order turbulence models through a two-scale turbulence theory." Physical Review E 48.1 (1993): 273.

[tbeji]

@mameehan5
Thanks for the info.
No need to re-submit data at this stage. Looking forward to discussing with you regarding this aspect.

[tbeji]

@gatw

Hi,

Few questions about the predictive methanol pool fire simulations of DBI:

1. Can you interact with Randy to know why EXTERNAL_HEAT_FLUX did not work for you to have ignition, whereas it did work for NIST? Could it be that the EXTERNAL_HEAT_FLUX was too low; NIST used 20 kW/m2.
2. What is the temperature of the solid block? How far was it from the surface?
3. Same question as to Randy: What would happen if you do not model ignition and keep AIT = 0K?

Thanks!

[rmcdermo]

Yes, I do not like the hardwired limiters.

[tbeji]

Does it make sense though to give the possibility to put a lower bound for the MLRPUA, which would correspond to what is needed to reach flammability and have ignition?

[rmcdermo]

Well, to me that falls into the category of any of these ad hoc ignition models. Whether we are igniting a pool surface or initiating a backdraft, I have yet to see anything that is robust and does not rely on tuning to the specific case. (Not counting the trivial option of SUPPRESSION=F, which obviously does not work if you want to capture extinction/re-ignition in the general case.)

[gatw]

@tbeji

answers to your questions?

1. You indicated that 'the top of the heated block is at the same level as the pool surface'? Do you rather mean 'the bottom'?

Yes, im sorry about the typo.

2. Can you please give details about the liquid phase? Meshing? In-depth radiation (absorption coefficient)?

I will confirm the answer as soon as I have access to the simulation files. We did not change any of the default properties for liquid methanol included in the FDS libaray. So, the adsorption coefficient was set to 1500.

3. The lip height being equal to 1 cm, what do you do when the gas phase cell size is 2 cm?

We set the lip height to 1cm in the 1cm and 5mm simulations. However, on the 2cm simulation, the lip was 2cm (1 cell thick). We thought the lip should exist no matter what the cell size is in our simulations.

I think now that we may have had higher predicted HRR in our simulations when using the heated block might be due to the block also heating the pan. I am just guessing, we dont have boundary files to confirm this.

[tbeji]

@gatw Thanks!

@rmcdermo The minimum MLRPUA would correspond to the concept of 'critical flowrate of volatiles from the surface [...] for sustained ignition to occur', as described for instance by Drysdale who gives an average value of 2.5 g/m2/s for some solids. Perhaps that helps the burning to be sustained in the early stages, even with SUPPRESSION=T. At a more developed stage, the minimum MLRPUA can be turned off.
In any case, we need an engineering approach for the ignition; an ad hoc approach remains of interest.
These are just thoughts, of course...

[ahaminsgithub]

Our measurements suggest that about 55% to 65% of the heat feedback to the 30cm methanol fuel pool is due to radiative heat transfer (which is not uniform) with about 35% to 45% due to convective heat transfer. Thin film theory applied to the 30 cm methanol pool fire indicates that convective heat transfer would increase by a factor of 2 if the mass burning rate were cut in half, and decrease by a factor of 2 if the mass burning rate were increased by a factor of 1.4 - so evaporation is in a range for which convective heat transfer appears fairly sensitive to blowing. In a somewhat related matter, Orloff and deRis report experiments that show that the presence of the burner lip increases the convective heat transfer coefficient by about 30% (as compared to zero lip).

…

On Fri, Sep 15, 2023 at 12:03 PM tbeji ***@***.***> wrote: @NingRen <https://github.com/NingRen> Thanks a lot for the information. 1. The integration of the heat flux is a good explanation. If you can do the calculation and provide me with the result, I will mention that during the presentation. 2. There is potentially a good discussion point with the experimentalists about the blowing effect. @ahaminsgithub <https://github.com/ahaminsgithub> Any thoughts about this? 3. @NingRen <https://github.com/NingRen> and @rmcdermo <https://github.com/rmcdermo> During IAFSS, I will present a paper on transient heating of liquid pool fires, including in-depth temperature measurements with a good resolution and fine thermocouples. The experiments are done at the University of Nancy (France). Here are some results for 15 cm methanol and heptane square burners with L = 15 cm. [image: image] <https://user-images.githubusercontent.com/144003045/268332789-f8e42907-0755-4976-9fc1-532af83314dd.png> I will be happy to share the data. Just need some time. Some FDS results are here for heptane at different heights, z: [image: image] <https://user-images.githubusercontent.com/144003045/268333558-1160f4b5-db52-4da9-a707-45c9b2c44984.png> The Nusselt number is an INTERNAL Nusselt number linked to an effective conductivity used in the 1D solution of Fourier's equation in FDS. In the paper, we explain how to estimate Nu experimentally. FDS confirms the validity of our experimental estimates. In any case, I will be very happy to interact with you on this topic. — Reply to this email directly, view it on GitHub <#438 (reply in thread)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ADYO4OBTHGZAUTGPHXFV5F3X2R343ANCNFSM6AAAAAA4PG7O5U> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[tbeji]

@ahaminsgithub Thanks!

[tbeji]

@hong27

I would like to get some additional information on the 37 cm diameter methane that you simulated.

1. Just like for the FM-Burner case, can you please share some images of the case/mesh setup along with the rendered flame structure?
2. Are you modelling soot for this case, like for the FM_burner?
3. Is the radiative fraction predicted?

Thanks.

[tbeji]

Additional question:
4. How many chemical reactions are you using? Infinitely fast? Or, as I presume, slow with Arrhenius-type equations?

[hong27]

Hi @tbeji
Some information is provided as follows.

1. the mesh:
   
   

the flame structure: (mean temperature - instantaneous temperature - rendered instantaneous temperature)

2. Soot is neglected in this simulation. Our soot module had not been readily available when running this simulation.

3. Radiative heat loss is predicted rather than prescribed. As mentioned in the README.md file, we used WSGG model to account for the spectral radiative properties. TRI is accounted for by the presumed filtered density function method. Details can be found in this paper. The global radiant power is not available now because I did not average the relevant fields during the simulation.

4. We applied a flamelet-like model (radiative flamelet/progress variable), which allowed the incorporation of finite-rate detailed chemistry with affordable computational cost. Specifically, we applied the GRI-3.0 chemistry mechanism for methane combustion, which includes 53 species and 325 reactions. However, in the CFD solver, we only need to track the mixture fraction, subgrid mixture fraction variance, progress variable and total enthalpy. The species and temperature are derived from a flamelet lookup table by these four tracked variables, and therefore, the computation is cost-effective. Chemistry is computed prior to the CFD flow solver, with the GRI-3.0 mechanism, to generate the flamelet lookup table. Details can be found here.

Please let me know if you need any further details regarding the simulation.

Thanks!
Regards,
Jianhong

[tbeji]

@hong27
Thanks, Jianhong.

[ahaminsgithub]

@NingRen <https://github.com/NingRen> There is one data point regarding in-depth radiative penetration for methanol from an experimental study (Burgess and Hertzberg, Radiation from Pool Flames, Chapter 27, page 425, Table 2 in Heat Transfer in Flames, Eds: Afgan and Beers, 1974, John WIley & Sons, NY). They show that 3mm of liquid methanol absorbs 100% of the radiative emission from a methanol flame. We also have measurements at various liquid depths in the 100 cm methanol fire. In this experiment, the TC (about 0.5 mm diameter type K thermocouple) was moved every 30 sec through the liquid pool. There is an indication of a hot layer near the pool surface (at r =35cm), which is close to the boiling point (see data below from Sung et al, NIST report# TN 2083, page 38). These results seem to be consistent with the radiation penetration value from Burgess and Hertzberg (mentioned above). We also observe standing waves in the liquid pool with an amplitude on the order of 1 mm beating in sync at the puffing frequency, which one might guess ought to enhance mixing near the fuel surface.... *Test 2 at 45 min into the 100 cm methanol pool fire experiment* *Z(mm) T(deg_C)* *-0.1 64* *-1.2 66* *-3.3 65* *-5.2 58 * *-10.0 46* *-20.3 34* *-40.4 27* By the way, if radiative penetration is of interest, there is liquid methanol absorption data available from NIST and others as a function of wavelength in the IR (for example: https://www.hindawi.com/journals/dpis/2013/329406/). One could use RADCAL2 (which includes methanol) to calculate an approximate fire emission spectrum to the fuel surface and then apply the liquid absorption spectrum. Ryan and I are planning measurements of radiation penetration for various fuels sometime in the near future.

…

On Wed, Sep 20, 2023 at 9:42 AM NingRen ***@***.***> wrote: @rmcdermo <https://github.com/rmcdermo> @ahaminsgithub <https://github.com/ahaminsgithub> Thanks Anthony and Randy for sharing the liquid temperature data. The 0.8-cm-deep temperature keeps increasing slowly even after 10 mins. Definitely need to run the simulation longer to reach steady state. The 0.8-cm temperature is close to the boiling point at 2400 s. Do you have any measurement of the thickness of the boiling region? In my current simulation, there is a 3-mm thick of fuel reaches the boiling point, which can be increased by tuning the in-depth radiation. Any idea about the radiation penetration depth for the methanol pool fire? Best, Ning — Reply to this email directly, view it on GitHub <#438 (reply in thread)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ADYO4OFOONOJMDCSNHKLIT3X3LXDRANCNFSM6AAAAAA4PG7O5U> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[NingRen]

@ahaminsgithub
Hi Anthony,
Thanks for these extra information. If a 3 mm of fuel absorbs 100% flame emission, then the penetration depth should be smaller than 3 mm. I am using 1.5 mm now. The air-liquid interaction, e.g., the standing wave you mentioned, should enhance the mixing, which is totally neglected in my simulation. So some discrepancies between measurements and predictions are expected.

For methanol fires, the radiation penetration doesn't change the predicted burning rate. I have some simulations of fire spread in a liquid pool of high-flashpoint fuels, such as corn oil, propylene glycol. In those cases, the radiation penetration do change the fire spread velocity. The main mechanism is that it changes the transient surface temperature, which then changes the thermo capillary convection velocity and eventually change fire spread velocity. Radiation penetration for a variety of fuels will be very helpful. Does NIST have any plans to do some measurements of fire spread over some medium to high flashpoint fuels?
Thanks,
Ning

[tbeji]

@NingRen
Hi Ning,
2 additional questions:

1. Do you use WSGG for the absorption coefficients?
2. How do you get ignition for your predictive simulations of methanol?

[tbeji]

In relation with 2. Do you specify an initial external heat flux, given that at the very start you do not have a flame exerting a heat flux on the liquid?

[NingRen]

@tbeji
For the radiation, I used a simple radiant fraction model.
Because the methanol flashpoint is lower than the ambient temperature, no external heat flux is provided to ignition. At room temperature, fuel has very small evaporation rate, but enough for auto ignition. The fuel evaporation uses h_mrho(Y_s-Y_infinity)/(1-Y_s). Y_s is the surface methanol mass fraction, which is a function of saturation pressure, Pst. Pst is correlated with surface temperature. As fuel heats up, the evaporation rate increases until reaches steady state.
Ning

[tbeji]

@NingRen

So, if I understand correctly, to calculate MLR, you are not using HEAT FLUX / HEAT OF GASIFICATION, but rather(like FDS) a mass transfer number approach, right?

1. Which correlation do you use to calculate the Sherwood number, Sh, in h_m? Or do you do something else?
2. Why not using LN (1+B), meaning that you take into account Stefan's flow ?
3. For Ys, do you use Clausius-Clapeyron or Antoine equation?
4. For Y_infinity: Is it the concentration in the first gas cell above the surface or something else?

Sorry for all these questions. Liquid evaporation is a topic that I particularly like and I think that a lot of seemingly small details matter.

Tarek.

[NingRen]

@tbeji

1. I just simply used a constant h_m. In the steady state of the simulation, fuel surface is in boiling mode (although some measurements show the surface T is slightly lower than boiling point). In this mode, the evaporation rate is generally determined by the heat flux / heat of gasification. h_m seems only affecting the transition stage.
2. I believe the Stefan's flow is resolved in the gas-phase. No need to model it. For the convective heat transfer, I do have a blowing effect included in a wall-model, similar to the B number theory. For higher evaporation rates (e.g., 100 cm pool compared to 30 cm pool), the convective component is smaller due to the blowing effect.
3. I used the Clausius-Clapeyron equation to calculate Y_s.
4. For Y_inf, I used the near-wall cell center value.
   Best,
   Ning

[tbeji]

Thanks.

[ahaminsgithub]

@NingRen <https://github.com/NingRen> there's uncertainty in all the liquid temperature measurements on the order of +/- 0.5 deg K (as well as spatially, so the temperature is really about the same in the top liquid layer near the fuel surface (top 3 mm). And we observe the surface temperature to slowly increase in the liquid fuels with time for acetone, ethanol and methanol (on the order of 1500 sec), likely due to back diffusion of gaseous water vapor from the fire, condensing on the fuel surface.

…

On Fri, Sep 22, 2023 at 10:42 PM NingRen ***@***.***> wrote: @tbeji <https://github.com/tbeji> 1. I just simply used a constant h_m. In the steady state of the simulation, fuel surface is in boiling mode (although some measurements show the surface T is slightly lower than boiling point). In this mode, the evaporation rate is generally determined by the heat flux / heat of gasification. h_m seems only affecting the transition stage. 2. I believe the Stefan's flow is resolved in the gas-phase. No need to model it. For the convective heat transfer, I do have a blowing effect included in a wall-model, similar to the B number theory. For higher evaporation rates (e.g., 100 cm pool compared to 30 cm pool), the convective component is smaller due to the blowing effect. 3. I used the Clausius-Clapeyron equation to calculate Y_s. 4. For Y_inf, I used the near-wall cell center value. Best, Ning — Reply to this email directly, view it on GitHub <#438 (reply in thread)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ADYO4OBJEL7S4O4U7AQEDVTX3ZEAFANCNFSM6AAAAAA4PG7O5U> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[rmcdermo]

@tbeji Please note the addition of U Waterloo results pushed up today.

Waterloo_Methanol_UWaterloo_Plots.zip

[tbeji]

Thanks.

[rmcdermo]

For those who have submitted computational results for the NIST-Waterloo Pool Fires case, we have a favor to ask. Ryan Falkenstein-Smith (@rlfalken) has recently pushed up uncertainty data for some of the experimental plots. I have added the experimental uncertainty to these plots in my plot configuration file here. Look for the column titled Exp_y_Error_Col_Name. Please copy this from my config file and add to yours for the appropriate rows and send the updated file in a PR.

If you do not have time to add this, we understand, but we would appreciate your help in adding the uncertainties if you can. Thanks!

[rlfalken]

Hi All, I was able to complete these plots, so no need for the extra work.