Bouchair Solar Chimney case: flow-rate normalization and heat transfer

[Koranten]

Hi,

I have a question about the NIST FDS validation case “Bouchair Solar Chimney”.
For the specific case, the plot is titled “Solar Chimney (0.1 m inlet; 0.2 m cavity; 60 °C)”.

In Bouchair’s thesis, the normalization appears to use the cavity length. On printed p. 71, the author says the flow rates are calculated “for a unit length of the cavity width” and that the measured flow is divided by “the length of the cavity.” On p. 68, the fixed dimensions are listed separately: cavity length 1.50 m, inlet length 1.40 m. Table 4.5 (p. 76) gives an inlet flow rate of 0.113 kg/(s·m) for a 0.1 m inlet, 0.2 m cavity and 40 K surface-to-ambient temperature difference.

Reading the thesis literally, I would multiply by 1.5 m, giving about 0.165 kg/s. However, Section 12.5.1 of the FDS Validation Guide says the thesis data are per unit length of the 1.4 m inlet slot, which gives about 0.154 kg/s using the rounded value. Could you clarify which length should be used to recover the total experimental flow?

I also reran this FDS case. I obtain about 0.127 kg/s and roughly 1 kW of convective heat transfer (enthalpy flux z). Bouchair’s Table 4.33 (p. 107) reports a 12 K air-temperature rise for the same 0.1 m inlet and 0.2 m cavity when the wall-to-ambient temperature difference is 40 K (60 °C walls and 20 °C ambient). Using 0.165 kg/s, that corresponds to approximately 1.99 kW carried by the airflow.

Does the NIST validation case assess only the mass flow, or is the outlet-air temperature / heat carried by the flow also expected to agree with Bouchair’s measurements? I would also appreciate clarification on whether my approximately 1 kW result is comparable to the ventilation enthalpy gain.

[Koranten]

I would also like to mention a supplementary diagnostic case. Instead of prescribing the heated-wall temperature at 60°C, I prescribed a convective heat flux corresponding to an estimated convective heat input of about 2 kW, inferred from the thesis data using the air mass flow rate and temperature rise. In this case, FDS predicts a mass flow rate of 0.155 kg/s.

This is close to the approximately 0.165 kg/s obtained by multiplying the thesis’s reported 0.11 kg/s/m by the 1.5 m cavity width, while the official FDS validation case gives 0.127 kg/s.

This flux-controlled case does not reproduce the experiment’s 60°C wall-temperature condition, so I see it as a complementary diagnostic rather than a direct validation case. It may help assess FDS’s prediction of the buoyancy-driven flow and pressure losses when the convective heat input is prescribed. The higher predicted flow suggests that the heat-transfer model may contribute to the lower flow rate in the 60°C case, although this comparison alone does not establish the cause.

What do you think of this interpretation?

[rmcdermo]

Hi @Koranten ; If you set HEAT_TRANSFER_MODEL='RAYLEIGH' on a SURF line, you should get the experimental convection model. This is really the research I would love to be doing. But I have been pulled away seemingly never to return. I would be interested in your experience with it. My results indicate it works well for natural convection. And where the research stalled was in extension to forced convection. There is a path forward, but I have not had time to work on it.

I think it would be interesting to look at the La Pica experiment. Can you give a full reference---my search was ambiguous.

[Koranten]

Hi Randy :)
I'll try the Rayleigh mode on Bouchair, La Pica and VTT and will tell you how it went.
The La Pica experiments are here :
A. La Pica, An experimental investigation on natural convection of air in a vertical channel, International Journal of Heat and Mass Transfer, 1993 (sorry I said 1991 but this was 1993), https://doi.org/10.1016/0017-9310(93)80036-T
I enclose one of the FDS model I did for La Pica, for the 75 mm gap. Like in the Bouchair experiment, there are many gaps.
lapica_075.txt

[Koranten]

Randy, may I ask more details about this Rayleigh mode? It's not in the documentation it seems. What's the theory behind it?

[Koranten]

So, I re-ran Bouchair with the normal law, the Rayleigh law and the log-law.
Normal : ~0.125 kg/s, ~4.5 W/m/K, ~1.05 kW.
Rayleigh : ~0.12 kg/s, ~3.8 W/m2/K, ~1 kW
Log-law : ~0.07 kg/s, 0 W/m2/K (error I guess), ~0.16 kW
Experimental : ~0.165 kg/s, ~10 W/m2/K (inferred from temperature rise and calculated with mean temperature), ~2 kW
So, Rayleigh and log-law do have an effect, but not really in the good way.
Maybe the Bouchair experiment is flawed, I don't know. But the La Pica experiment also indicate a too low h_conv.
Also, not here but in many cases and relevant to a broader discussion about convective coefficient, FDS has a tendency to activate the DNS-resolved mode way too "easily", even when y+ is extremely far from resolved, we have seen the DNS triggered with y+ still at 30 or more, resulting in divergence in mesh sensitivity studies for convective heat flux (and condensation heat flux as well).

[rmcdermo]

I have some notes on Rayleigh, but they are not a complete paper yet. Send me an email and I will send you a draft.

I have not studied these cases at all, so it is hard for me to speculate at what is going on. From my experience with Rayleigh, it does a good job if the flow is natural convection. Grid independence for forced convection and stagnation flows (like ceiling jet impingement) are where all these models fall apart.

Can you send your set up for the La Pica cases?

[Koranten]

You can send the draft at corentin.macqueron@gmail.com

The La Pica model is given in a previous message. Don't you see it on your side?

[rmcdermo]

Yes, sorry, I see it now. OK, will send you the notes today. Let's figure this out...

[mcgratta]

There was a mistake in the experimental data reported in the FDS Validation Guide. This PR corrects it. In short, the mass flow rates were reported by Bouchair in units on kg/m/s. I multiplied these values by 1.4 m instead the correct value of 1.5 m.

[Koranten]

Thanks!
I also noticed that if I switch to LOG LAW for heat transfer, the reported heat transfer convective coefficient is 0, while the convective heat flux is not zero. I guess there's just a problem with the post-processing. With default and Rayleigh modes it works fine. I attach the model and the results.
loglaw.txt
loglaw_devc.csv