Cone calorimeter test for Glass composite material

[Thevakumar Thevega]

Hi All,

I need to check the combustible behaviour of glass with polyurethane composite material. Then I modeled the cone calorimeter test for the composite material. When I observe the variation of heat release rate with time, it shows that the material started to burn within a few seconds. However, it is not in the real situation as it will take time for the ignition. 

I could not understand the error with my model. I have attached the result and the model. If anyone has any idea about this, could you please share with me?

Thank you for your time.

Best Regards,

Thevega.

[Kevin McGrattan]

1. Update to the latest version of FDS

2. Remove N_THREADS from the MESH lines. This parameter is no longer used.

3. Set TGA_ANALYSIS=T on the SURF line

4. Open gcm_tga.csv and look at mass vs temp and mass loss rate vs temp. These plots indicate that your kinetic constants cause the pyrolysis to occur at ambient temperature.

[Thevakumar Thevega]

Thank you very much for your guidance. 

As you mentioned, I understood from TGA analysis that pyrolysis starts at 20 degrees Celsius (ambient temperature) (Attached are the TGA results below). Then if I know the starting temperature for the pyrolysis reaction, how can I insert it into the model? Or do I miss something in my model?

[Kevin McGrattan]

There is a discussion in the FDS Verification Guide of how you can derive A and E using actual measured TGA data. If you do not have that, you can remove A and E from the input file and just add REFERENCE_TEMPERATURE=XX where XX is the temperature at which the mass loss rate curve is at its maximum value.

[Thevakumar Thevega]

Many thanks.

The analysis works with REFERANCE_TEMPERATURE, HEATING_RATE and PYROLYSIS RANGE. 

I can get these values from the real test data of the TGA analysis. Then how can I identify the pyrolysis range from the mass loss versus temperature curve? 

[Kevin McGrattan]

Eyeball it, and then compare the FDS TGA curve with the measured one.

[Thevakumar Thevega]

Thank you.

I did the TGA analysis but experimental and numerical TGA results vary (Attached below). 

Also, I have a doubt that previously I calculated the kinematic parameters by using the Coats and Redfern method from the TGA experimental result. Is that wrong or will it not predict the real reaction behaviour?

[Kevin]

You say that 5% of the mass is vaporized. What happens to the other 95%?

&MATL ID='GCM',
      SPECIFIC_HEAT=0.8,
      CONDUCTIVITY=0.22,
      DENSITY=1552.0,
      ABSORPTION_COEFFICIENT=10.0,
      HEAT_OF_COMBUSTION=1.62E+4,
      N_REACTIONS=1,
      HEAT_OF_REACTION=455.0,
      SPEC_ID(1,1)='METHANE',
      NU_SPEC(1,1)=0.05,
      A=4.2,
      E=1.75583E+4/

[Thevakumar Thevega]

The glass composite material contains 95% of glass and 5% of polymer. Then I assumed that the 5% of mass is only reacted as 95% of glass is a non-combustible material. 

[Kevin McGrattan]

That has to be made clear. You need to add a MATL_ID and NU_MATL to the MATL line to tell FDS what becomes of the 95% of the mass that is not vaporized.

[Thevakumar Thevega]

Thank you for your guidance.

After adding that, I can able to see the result. However, the pyrolysis reaction starts at ambient temperature in the numerical simulation.

&MATL ID='GCM',
      SPECIFIC_HEAT=0.8,
      CONDUCTIVITY=0.22,
      DENSITY=1552.0,
      ABSORPTION_COEFFICIENT=10.0,
      HEAT_OF_COMBUSTION=1.62E+4,
      N_REACTIONS=1,
      HEAT_OF_REACTION=455.0,

      MATL_ID(1,1)='Glass',
      NU_MATL(1,1)=0.95,
      SPEC_ID(1,1)='METHANE',
      NU_SPEC(1,1)=0.05,
      REFERENCE_TEMPERATURE=275.0,
      HEATING_RATE=20.0,
      PYROLYSIS_RANGE=250.0/
&MATL ID='Glass',
      SPECIFIC_HEAT=0.84,
      CONDUCTIVITY=0.8,
      DENSITY=2500.0,
      ABSORPTION_COEFFICIENT=10.0/

Also, is this data of the remaining material necessary to measure the heat release rate of the composite material? 

[Kevin McGrattan]

PYROLYSIS_RANGE=250.0

Why? This should be roughly the width of the temperature spike.

[Thevakumar Thevega]

I am sorry that I couldn't understand. Could you please explain what you mean by temperature spike?

[Kevin McGrattan]

Look at the mass loss rate curve; that is, the plot of mass loss rate vs temperature. Typically, there a bump or spike corresponding to the temperature range over which the mass loss rate is greatest. What is the approximate "width" of the bump, in degrees C.

[Thevakumar Thevega]

Thank you.

At present, I don't have the experimental mass loss rate vs temperature curve. I think that I can get the pyrolysis range from the experimental mass loss vs temperature curve around 80 degrees C (the range of highest slope of the mass loss vs temperature curve). Is that correct?

&MATL ID='GCM',
      SPECIFIC_HEAT=0.8,
      CONDUCTIVITY=0.22,
      DENSITY=1552.0,

      ABSORPTION_COEFFICIENT=0.0,

      HEAT_OF_COMBUSTION=1.62E+4,
      N_REACTIONS=1,
      HEAT_OF_REACTION=455.0,
      MATL_ID(1,1)='Glass',
      NU_MATL(1,1)=0.95,
      SPEC_ID(1,1)='METHANE',
      NU_SPEC(1,1)=0.05,
      REFERENCE_TEMPERATURE=275.0,

      HEATING_RATE=20.0/

However, when I check the numerical mass loss rate vs the temperature curve, I do not understand why the width of the bump is in a wide range (Attached below).

[Kevin McGrattan]

You derive the mass loss rate curve by numerically taking the slope of the mass loss curve. You can do this yourself. 

The default PYROLYSIS_RANGE in FDS is 80 C. Try that as a start.

[Thevakumar Thevega]

Thank you for your time.

I will try it.

[Thevakumar Thevega]

Hi All,

I have to validate the numerical result with the experimental result. Before that, I tried to find out the suitable mesh size for the simulation. However, I could not see any convergence pattern in the result (Figure is attached). Does anyone have any idea or suggestion for that?

[Kevin McGrattan]

Record the 'SURFACE INTEGRAL' of QUANTITY's like 'GAUGE HEAT FLUX' and 'CONVECTIVE HEAT FLUX' and 'HEAT TRANSFER COEFFICIENT'. The heat flux to the surface is based on empirical correlations, and grid size affects them. At present, there is no way to guarantee that these sub-grid scale models of heat transfer at the surface converge until you enter the DNS (Direct Numerical Simulation) resolution regime, which is on the order of a mm. In DNS mode, the heat transfer is computed directly from the temperature gradient at the surface, but you need a very fine grid to adequately resolve the boundary layer.

[Thevakumar Thevega]

I will check, Thanks a lot.

[Thevakumar Thevega]

Hi All,

If I am measuring the heat release rate while applying the heat flux, how can I use the heat release rate to calculate the characteristic fire diameter for the mesh resolution according to the fds user guide (6.3.6).

Best Regards,

Thevega.

[dr_jfloyd]

From a cone test you know the kW/m2 as a function of heat flux (a cone test measures the heat release rate and the sample surface area is defined by the standard being used for the test). How to translate this to a real fire requires engineering judgement. What is the shape of the real world object that is burning, how is the fire expected to spread over the object (i.e., how much area might be burning at one time), does the object melt into a pool or stay as some 3D shape, etc.?

[Thevakumar Thevega]

Thank you.

[Thevakumar Thevega]

Hi All,

When we check the mesh resolution, should the characteristic fire diameter/grid size be between 4 to 16 to obtain reliable predictions? or May the characteristic fire diameter/grid size be greater than 16 also?

I found this information in the literature. As I am struggling to do the mesh convergence, I would like to get a clear idea about this.

Best Regards,

Thevega.

[Khalid Moinuddin]

Not all, if you are not prescribing your fire size.

 

You need to compare your HRR vs time profile obtained with different meshes and continue to refine the meshes until HRR vs time profile converges.

 

From: fds...@googlegroups.com <fds...@googlegroups.com> On Behalf Of Thevakumar Thevega
Sent: Monday, 12 September 2022 8:13 PM
To: FDS and Smokeview Discussions <fds...@googlegroups.com>
Subject: Re: [fds-smv] Cone calorimeter test for Glass composite material

 

EXTERNAL EMAIL: Please be cautious before clicking on any links or downloading attachments.

 

--
You received this message because you are subscribed to the Google Groups "FDS and Smokeview Discussions" group.
To unsubscribe from this group and stop receiving emails from it, send an email to fds-smv+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/fds-smv/17738941-ec7e-4b37-a563-1322cc6f250en%40googlegroups.com.

[Kevin McGrattan]

This range of 4 to 16 has become legend. It is not. We did a validation study 15 years ago and reported our range of D*/dx. That is all. In general, the larger the value, the better, but how much better depends on the case and a grid resolution study.

[Thevakumar Thevega]

Thank you all.

When I decreased the mesh size from 20 mm (D*/δ=2.5) to 2 mm (D*/δ=25), there are no grid convergence (peak heat release rate or peak mass loss rate) in the combined reaction (solid and gas phase reaction). 

But the peak mass loss rate converged in the solid phase reaction. Because there are no fire simulations in the solid phase reaction. 

Can I use the converged mesh size in the solid phase reaction for further analysis in the combined reaction? 

The attached images are the grid convergence results for both conditions.

Thank you for your time.

Best Regards,

Thevega.

[Jonathan Hodges]

A few additional thoughts that I didn't see in the earlier discussion.

1. Are you using the default radiation angles? Increasing the radiation angles may increase your convergence without going to a finer grid.

2. Are you using the default number of time steps between recomputing radiation? As you are resolving your grid more, the time step is being reduced per the CFL condition. Since you have significant feedback between the flame and surface which is driving the pyrolysis, any artificial delays in the RTE calculation at the coarser scale will be reduced in the finer scale. But you could get a similar benefit by modifying this time step at the coarser grid without going to the fine grid.

3. Have you defined what your criteria for convergence is? You can keep resolving your mesh until you get to a sub millimeter grid and your answer will still change a bit. At some point you need to establish what is good enough for your application.

4. If the end goal of your analysis is to simulate a large fire, you aren't going to do that with a 2mm grid throughout your whole domain. If you are hoping to establish a D*/dx that works for your situation and then extrapolate that to a larger fire, I would caution that approach. The work we have done with SFPE has shown the D*/dx relationship for convergence is not great when you are looking at heat transfer from a local fire.

Jonathan

--

You received this message because you are subscribed to the Google Groups "FDS and Smokeview Discussions" group.
To unsubscribe from this group and stop receiving emails from it, send an email to fds-smv+u...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/fds-smv/54a3b845-c368-49ac-9ac9-23071657f0f1n%40googlegroups.com.

[Khalid Moinuddin]

Very good points.

In terms of convergence criteria, for the quantitative approach, we use Grid Convergence Index (GCI).

Roache PJ, 1992. Quantification of uncertainty in computational fluid dynamics, Annual Review Fluid Mechanics, vol. 29:123–60

When the HRR is NOT prescribed, GCI can be calculated for the time-series of HRR. If the HRR is prescribed, it can be either time-history of temperature at any specific point or time-averaged temperature of a thermocouple tree.

Lower the GCI, closer/better the convergence. A GCI of 10%-20% will be really good.

D*/dx relationship is only applicable for when the HRR is prescribed. 4 <D*/dx < 16 was based on Baungui et al (2003) where the HRR was prescribed. One of our Masters students conducted simulations of a set of prescribed fires in warehouses (different fire sizes and warehouse sizes). At different locations of the warehouse, they calculated measure of turbulence resolutions. 80% turbulence kinetic energy was generally resolved when D*/dx ~ 12.

We also tried to use D*/dx relationship when the HRR was NOT prescribed. For different fire sizes (cone calorimeter to forest fires), it ranged between 15 and 40. I don't recommend using D*/dx criteria when the HRR is NOT prescribed. One should not reduce, 20 mm to 2 mm in one jump. The preferred way is to use 20, 10, 5, 3.75, 2 mm (the ratio must be more than 1.3).

From: fds...@googlegroups.com <fds...@googlegroups.com> On Behalf Of Jonathan Hodges
Sent: Friday, 16 September 2022 8:38 PM
To: fds...@googlegroups.com
Subject: Re: [fds-smv] Cone calorimeter test for Glass composite material

EXTERNAL EMAIL: Please be cautious before clicking on any links or downloading attachments.

A few additional thoughts that I didn't see in the earlier discussion.
1. Are you using the default radiation angles? Increasing the radiation angles may increase your convergence without going to a finer grid.

2. Are you using the default number of time steps between recomputing radiation? As you are resolving your grid more, the time step is being reduced per the CFL condition. Since you have significant feedback between the flame and surface which is driving the pyrolysis, any artificial delays in the RTE calculation at the coarser scale will be reduced in the finer scale. But you could get a similar benefit by modifying this time step at the coarser grid without going to the fine grid.

3. Have you defined what your criteria for convergence is? You can keep resolving your mesh until you get to a sub millimeter grid and your answer will still change a bit. At some point you need to establish what is good enough for your application.

4. If the end goal of your analysis is to simulate a large fire, you aren't going to do that with a 2mm grid throughout your whole domain. If you are hoping to establish a D*/dx that works for your situation and then extrapolate that to a larger fire, I would caution that approach. The work we have done with SFPE has shown the D*/dx relationship for convergence is not great when you are looking at heat transfer from a local fire.

Jonathan

On Fri, Sep 16, 2022, 5:34 AM Thevakumar Thevega <mailto:thevakum...@gmail.com> wrote:
Thank you all.

When I decreased the mesh size from 20 mm (D*/δ=2.5) to 2 mm (D*/δ=25), there are no grid convergence (peak heat release rate or peak mass loss rate) in the combined reaction (solid and gas phase reaction). 
But the peak mass loss rate converged in the solid phase reaction. Because there are no fire simulations in the solid phase reaction. 
Can I use the converged mesh size in the solid phase reaction for further analysis in the combined reaction? 
The attached images are the grid convergence results for both conditions.
Thank you for your time.

Best Regards,
Thevega.
On Monday, September 12, 2022 at 4:19:44 PM UTC+5:30 Kevin wrote:
This range of 4 to 16 has become legend. It is not. We did a validation study 15 years ago and reported our range of D*/dx. That is all. In general, the larger the value, the better, but how much better depends on the case and a grid resolution study.
--
You received this message because you are subscribed to the Google Groups "FDS and Smokeview Discussions" group.

To unsubscribe from this group and stop receiving emails from it, send an email to mailto:fds-smv+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/fds-smv/54a3b845-c368-49ac-9ac9-23071657f0f1n%40googlegroups.com?utm_medium=email&utm_source=footer.

--
You received this message because you are subscribed to the Google Groups "FDS and Smokeview Discussions" group.

To unsubscribe from this group and stop receiving emails from it, send an email to mailto:fds-smv+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/fds-smv/CAG5CbNOTGzsingzMABxjXk98BZm5AeJ_W4XEUzQFXgzk9SHa8A%40mail.gmail.com?utm_medium=email&utm_source=footer.

[Thevakumar Thevega]

Thank you for your thoughts.

1,2 I am using the default radiation angles and time step. After your suggestion, I checked by changing those parameters but I don't see that much variation in the results.

3,4 No, I don't have any idea about the criteria for the convergence to this simulation. As I need to validate the cone calorimeter numerical results with experimental results, I thought to do the domain and mesh convergence before validating the model. And I am not going to use this same model to create a large fire, this is only for the cone calorimeter test. Can we define the criteria for convergence?

[Thevakumar Thevega]

Thank you.

I will check about Grid Convergence Index (GCI).

Previously I didn't consider the  D*/dx and changed the mesh size to 20 mm, 10 mm, 5 mm, 4 mm, and 2 mm. But there is no any convergence pattern then I compared the  D*/dx values for 20 mm and 2 mm mesh sizes.

[Jonathan Hodges]

It is up to you to decide when the model is sufficiently converged for your application. I would suggest a criteria based on comparing the percent difference in the results when you double the grid resolution to the uncertainty in the experimental measurements. The uncertainty in experimental measurements of heat release rate varies from 2.5-13%, with the recommended value in the FDS validation guide of 7.5%. Similarly, the experimental uncertainty in heat flux estimated to be around 10% in the validation guide.

Looking at the figure you provided, I estimate your HRR to be at the following levels:

30k cells - 0.57 kW

60k cells - 0.62 kW

120k cells - 0.69 kW

The percent difference between your cases is around 8-11%. I would say your results are sufficiently converged, but again it will depend on the goal of your analysis.

[Randy McDermott]

Don't forget about the condensed phase resolution.  You need to consider CELL_SIZE_FACTOR, WALL_INCREMENT, SUBSTEP_POWER.

It is also possible that time resolution at the interface is more important than gas phase spatial resolution.  You can be fooled into thinking that spatial resolution is the key because the time resolution decreases with spatial resolution automatically for an explicit code.

To unsubscribe from this group and stop receiving emails from it, send an email to fds-smv+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/fds-smv/bda9df47-7d4d-4ab3-ad48-fa25cb3d999dn%40googlegroups.com.

[Thevakumar Thevega]

Thanks a lot. I will check that.

Also, I would like to understand that how HRR is calculated and what is the relation between grid size and HRR. Could you please explain it?

[Thevakumar Thevega]

Thank you.

When I change the time step (WALL_INCREMENT) for solid phase reaction or combined phase reaction, I observe some fluctuations only in the results there are no changes in the peak value. 

I don't have any idea about  SUBSTEP_POWER. 

[Jonathan Hodges]

I am not 100% clear what your question is on the relationship between HRR and grid size. If it is asking to clarify my comment from a few days ago, the relationship between HRR and grid size I discussed was based on the plots that you provided a few days ago. The message that starts with:

" When I decreased the mesh size from 20 mm (D*/δ=2.5) to 2 mm (D*/δ=25), there are no grid convergence (peak heat release rate or peak mass loss rate) in the combined reaction (solid and gas phase reaction)."

If you are asking for general instruction on what HRR is/how it is calculated in FDS, I would suggest you read chapter 5 of the FDS user guide and chapter 26 of the SFPE handbook (Heat Release Rates).

[Kevin]

Given the exact same boundary conditions, the FDS solid phase solver will converge as solid phase grid is refined. However, the gas phase boundary conditions at the solid surface will not converge in LES mode, only DNS. The reason is that LES mode makes use of empirical convective heat transfer correlations that typically use a "free stream" velocity and temperature. As one refines the gas phase grid, the location of the "free stream" changes. It is taken at the center of the first gas phase cell abutting the surface. 

[Thevakumar Thevega]

Thank you for your reply. 

I asked about the general calculation of HRR in FDS. I don't get a clear idea about the numerical calculation from the FDS user guide or chapter 26 in the SFPE handbook.

[Thevakumar Thevega]

Thank you. I will try in DNS mode and check the convergence. 

I didn't understand the meaning of "center of the first gas phase cell abutting the surface".

[dr_jfloyd]

The equations FDS solves and the details on how they are solved are in the Technical Reference Guide: Volume 1 Mathematics which is in the same folder as the FDS User's Guide.

[Thevakumar Thevega]

Thank you.

[Thevakumar Thevega]

When I search about LES and DNS mode, the gas phase reaction was also performed in LES mode and the grid size was converged. I have attached the referred literature. Could you please explain it?

[Khalid Moinuddin]

The FDS code is substantially changed since we did this study 15 years ago. We used FDS version 4. Now we have FDS version 6.7.9.

 

From: fds...@googlegroups.com <fds...@googlegroups.com> On Behalf Of Thevakumar Thevega
Sent: Monday, 26 September 2022 9:02 PM
To: FDS and Smokeview Discussions <fds...@googlegroups.com>
Subject: Re: [fds-smv] Cone calorimeter test for Glass composite material

 

EXTERNAL EMAIL: Please be cautious before clicking on any links or downloading attachments.

 

When I search about LES and DNS mode, the gas phase reaction was also performed in LES mode and the grid size was converged. I have attached the referred literature. Could you please explain it?

--

You received this message because you are subscribed to the Google Groups "FDS and Smokeview Discussions" group.
To unsubscribe from this group and stop receiving emails from it, send an email to fds-smv+u...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/fds-smv/f4d1ce2e-eba0-44ef-bf33-48f7c45c0f6fn%40googlegroups.com.

[Thevakumar Thevega]

Thank you for your response.

[Thevakumar Thevega]

Hi All,
I performed the grid convergence while using DNS mode for the gas phase reaction. However, it does not converge to the smaller grid sizes (attached below).
When I reduce the grid size in LES mode, the measurement uncertainty of HRR is reduced. However, in DNS mode, it is high.
Also, I show a few recent studies in that researchers used LES mode for gas phase reaction and validated their results.

I am concerned that I may have overlooked anything in the fds numerical model. Could you please assist me in identifying that?

Thank you for your time.

[Kevin]

Even if SIMULATION_MODE='DNS', if you are still using the simple chemistry model, there are grid dependent mixing time scales involved. In other words, it is difficult to perform a simulation without some empirical sub-model that is grid dependent. You probably have to do a DNS simulation with finite-rate combustion kinetics in the gas phase. 

[Thevakumar Thevega]

Thank you for your response.

I cited the fds technical reference book (Page 56), where the Arrhenius reaction is referred to as finite-rate chemistry. So does it imply that the model should utilize Arrhenius parameters while in DNS mode? Or does Pyrosim have any explicitly defined sub-models?

Also, if we use the predefined fuel type instead of a simple chemistry model, don't we think about empirical sub-models?

Thank you.

[Kevin]

DNS mode does not imply finite-rate reaction kinetics. For that, you must specify these parameters on the REAC line. PyroSim does not have any explicitly defined sub-models that are not included in the FDS code. Pre-defined fuel type does not imply anything other than that its physical properties are known. It implies nothing about the combustion reaction.

[Thevakumar Thevega]

Thank you for your clarification.

Then, do I need to assign kinetic parameters in the REAC line and also the MATL line?

[Kevin McGrattan]

The REAC line is for gas phase reaction parameters. The MATL line is for solid phase reaction parameters.

[Thevakumar Thevega]

When I assign the Heat of Combustion (Gas phase reaction parameter) in the REAC line, the variation of HRR is zero (Attached fds code). 

I researched from different models that the pyrolysis (Solid phase reaction parameter), combustion (Gas phase reaction parameter) and thermo-physical properties are assigned in the MATL line. Are there any specific conditions for various models?

Thank you!

[dr_jfloyd]

There are two problems:

1.

Your mesh size is 5 cm (this is too large for DNS), and you have defined your burning surface with:

&OBST ID='Obstruction', XB=-0.05,0.05,-0.05,0.05,0.0,0.01, RGB=28,240,240, SURF_ID='INERT'/

&VENT ID='Vent surf', SURF_ID='Surface01', XB=-0.05,0.05,-0.05,0.05,0.01,0.01, RGB=230,37,255/

The color for your OBST is a light blue color. The color for your VENT is a light purple color. If I run your case and look at it in Smokeview, I only see light blue. Your VENT was not applied. It is good practice to use Smokeview to see if FDS has interpreted your inputs correctly.

A 1 cm tall OBST is not resolvable on your grid. FDS is treating it as a thin obstruction placed on the domain boundary.  It appears the combination of a thin obstruction on the boundary with a vent being applied to it is resulting in the vent not being successfully applied. You could either delete the OBST or you could delete the VENT and use SURF_IDS on the OBST.

2.

You have defined your fuel as METHANE with a HEAT_OF_COMBUSTION of 16200.  This will result in a non-physical EPUMO2, and your fuel will not pass the extinction check made by the default extinction model (see the Technical Reference and the User's Guide).

[Thevakumar Thevega]

Thank you for your comments.

1. I did the simulation with the mesh size of 2 mm and now running with 1 mm. But it is still taking a long time even with 32 MPI processors, is it possible? (attached figure)
     With a 1 mm mesh size, I can able to see the obstruction and vent in SMV view (attached figure). However, I will try with your suggestion and check the results.

2. While using DNS mode, I considered putting HoC in the REAC line as a gas phase reaction property. Should I assign the kinetic parameters for fuel and add them to the REAC line to produce a fine rate reaction if that is incorrect? 

In LES mode, I am able to do grid convergence for pure polymers. I am finding it difficult to complete the material with less polymer (PU). Does it have anything to do with estimating a lower HRR and less combustible behavior?

[Thevakumar Thevega]

[dr_jfloyd]

DNS takes a very long time (2 mm is still probably too big for DNS); small grid cells means very small timesteps. 

Read the Technical Reference discussion on the default extinction model. Defining METHANE as having an HoC of 16000 is non-physical and will not work with the  default extinction model.  If the solid material you are burning has an HoC of 16000, then the fuel gas it gives off is not a hydrocarbon. 

[Thevakumar Thevega]

Thank you for your guidance.

Since I was unable to run the much finer mesh in DNS, I chose to use LES. I then made an effort to comprehend the grid convergence from earlier research. I ran a number of simulations (Pyrosim-LES) with various materials, and I have included the results.

1. The 10 mm grid size for PMMA polymer is converged and validated by experimental results. However, the outcome starts to deviate from the result of the 10 mm grid size when the grid size is progressively reduced.

2. For PBTGF material, grid sizes of 50 mm to 5 mm produce nearly identical HRR values.

3. The findings of my grid convergence for glass composite material started to diverge after 10 mm (the pHRR increases rapidly and the time for PHRR changes).

On the basis of these observations and grid convergence studies in other CFD software, might the divergence be caused by numerical instabilities of eddies?

Thank you!

[Thevakumar Thevega]

Hi All,

I want to know the accuracy of the Large Eddy Simulation (LES) simulation which I used. Therefore, I need to calculate the ratio between the resolved and total turbulent kinetic energy (k resolved / (k resolved + k sgs)) in order to determine whether I am resolving 80% of the turbulent kinetic energy with my grid.

Is it possible to calculate these parameters using the Pyrosim analysis's output?

Please let me know if you have any suggestions.

Thank you.

[dr_jfloyd]

This is a question for Pyrosim technical support; however, I think the answer is you would need to do the calculation using another tool for post-processing (e.g. Excel, Python, Matlab) 

[Kevin McGrattan]

21.10.27 A Posteriori Mesh Quality Metrics

Is this section from the User's Guide relevant here?

[Thevakumar Thevega]

Thank you for your response.

Yes, It is explained via 21.10.24 A Posteriori Mesh Quality Metrics. However, I was unable to locate "SUBGRID KINETIC ENERGY" in Pyrosim's output. How do I obtain it?

[Thevakumar Thevega]

Hi All,

I calculated and compared the turbulent resolution for each grid size. From the results, I observed that initial energy resolution is quite low when I reduce the grid sizes too much (4 mm, 2 mm). Will it have an impact on the end result or can I ignore that part?

I have attached the energy curves and the hrr outputs for different grid sizes. Could you please give your thoughts?

Thank you.

Best Regards,

Thevega.