Oscillating mass flow along tunnel

[Ed]

Hi All

Please see attached plot of the mass flow along tunnel at 50 m intervals.  This is based on a 75 MW fire (3 x 4 m) in a 6.4 m wide and tall tunnel.  The fire is located in the middle of the tunnel.  

I'm puzzled as to why I am seeing this behaviour as for the smaller fire sizes (up to 55 MW with everything else the same), the mass flow remains relatively consistent bar some fluctuations

Many thanks!  

[dr_jfloyd]

It can happen in large enough fires: Pulsation - Full-scale fire test in train tunnel …: http://youtu.be/1feqOVztgtY

[Ed]

Many thanks!

[Ed]

Hi Dr. Floyd

A separate question if I may.  

I understand you are the developer behind the HVAC module from a paper you authored and the FDS developer intro.  

For the HVAC module, I understand the flow should be in one direction only, and reading the solution procedure on FDS math guide for HVAC, if there is flow reversal, the solution will be re-iterated.  Therefore, am I right to think if my HVAC module is located near somewhere with flow reversal, this means any results that I get could be spurious? 

Many thanks!

On Sunday, 24 August 2014 01:44:19 UTC+1, dr_jfloyd wrote:

[dr_jfloyd]

The HVAC model starts off each timestep using the flow solution from the prior timestep as the starting point.  The properties of the fluid in each duct segment are upstream properties.  If the boundary conditions have changed such that flow reverses in duct, then the properties being used in that duct are no longer correct.  Therefore, the HVAC model needs to update the properties and repeat the solution process.  This is what the re-iterate is referring to. If the flow reversal in the duct is the correct response to changing boundary conditions (i.e. FDS having the fire pulsate), then this is not spurious.

[Ed]

Many thanks for clarifying this.  

I'm currently looking at using HVAC for multi-scale modelling in tunnel.  See attached Figure 1.

As you can see, the mass flow over time for a 35 MW in the tunnel is conserved for the full CFD model, but when in the multi-scale model, at about the 200 s mark, the mass flow gradually decreases.  A couple of points:

1) Only portals in tunnel, no other openings for flow. 

2) 35 MW fire around the mid section of 1.5 km tunnel activated at 50 s.  

I'm trying to pin point the cause of decreasing mass flow and so far:

1) Checked mass in and out from the HVAC section.  Mass is conserved to a difference of 0.05%.  

2) Thought about flow reversal, but as per your email above, there shouldn't be a problem here.

3) Checked HRR and burn rate.  Relatively consistent over time and no sudden spikes.  

4) Exactly the same model as the full CFD bar the HVAC model, so shouldn't be a model issue (holes, leakage etc). 

5) Checked the .out file as well, CFL number is fairly steady at > 0.8, pressure iterations at 1-3 and velocity error at 0.x and 0.0x decimal points.  

Next I looked at the pressure along the tunnel.  Referring to the graph, the decrease in mass flow starts around the 200 s mark.  Attached are the screen shots taken of the pressure near the fire.  As you can see, there is a big patch of negative pressure at the 200 s mark, but disappears in the next slice.  Is something amiss here?

I'm trying to identify the cause of this, and I've varied the models as well, e.g. different length of HVAC and distance from fire.  I kept seeing the same issue with the mass flow, it either decreases, or increases over time and don't reach steady state (even when the CFD model has already reached steady state).  

I fully appreciate this is not something being done before other than by Vermesi, so I'm really just looking for any ideas and thoughts to see if I've missed anything.  

Many thanks as always!

[dr_jfloyd]

Some general comments, they may or may not be applicable to your specific inputs

- There is no heat transfer to the tunnel walls in the HVAC portions of the multiscale model.  There is in the full CFD.  Probably a minor effect.

- In the full CFD model you can have flows in both directions at any point in the tunnel.  If you just have a single duct for the multiscale portion, then flow is forced to be uni-directional.  This could be a significant difference.  

- The title of the images suggests you have jet fans.  A single jet fan will move more air than the flowrate of the fan itself, but if you put two or more jet fans in series the flowrates are not fully additive

- There are flow losses along the length of the tunnel that somehow should be accounted for in the HVAC sections.

- Your plot of the various models only goes out in time to 375 s.  The multiscale has not reached steady state over the time period shown; however, at some point it should.  What happens if this is run out in time long enough?  

[Ed]

Many thanks for your thoughts.  

- In the full CFD model you can have flows in both directions at any point in the tunnel.  If you just have a single duct for the multiscale portion, then flow is forced to be uni-directional.  This could be a significant difference.  

>>This might be a good point to consider.  Because I am seeing either overshoot or undershoot in the mass flow, which corresponds to the oscillatory nature of the flow.  So when we are seeing undershoot, it could be the HVAC interface is located at regions where flow are lower?  See attached 5.png figure.  The straight line in the middle is the multi-scale section.  We can see at the HVAC in (500 m) the mass flow going in tend to be at the lower point.  

I will check my plots for models where I am seeing significant mass flow overshoot.  

To isolate this problem, I'm thinking it would be worth plotting out the oscillation at locating the HVAC interface at regions where the mass flow oscillation intersects?  

2) Regarding the pressure slice file, is the blue patch something to be concerned of?  It seems to have a rare shape that starts and end so abruptly, and vanishes in the next slice file.  

In any case, thanks so much again for your comments!

[dr_jfloyd]

The pressure slices are odd looking.  If it the slice looks OK the frame before and after, then maybe this is just a rendering glitch in Smokeview or some write error to the output file?   You could try running again (and maybe add a line of pressure devices through the region).

I have never attempted what you are doing, so I don't have any experience to draw on.  The HVAC model was not designed for this purpose so you are in basically doing research.  Examine closely your full CFD results.  Are you locating the multiscale region in a location where the HVAC model assumptions are appropriate?

[Ed]

Many thanks.  

I've checked the SLCF output via fds2ascii, and the pressure in the same area is positive (see attached figure, note fire is at 1000m and sharp drop in the middle is when the SLCF passes through the HVAC section).  I appreciate the SLCF in fds2ascii needs to be averaged, but given the big drop I thought we should still see something.  In any case, as per your advice I'm re-running this model with additional devices around this region to be doubly sure, plus extended time to see when the mass flow would reach steady state.  

Thanks for your comments, I fully understand this hasn't been done by many people and so I'm just hoping to get different opinions.  

I've tested the HVAC location by switching it upstream and downstream of the fire, and by varying the distance of the HVAC interface to the fire as well.  I'm seeing the same increase or decrease in mass flow.  Hopefully I'll get a better picture if I run the model to steady state as the plots show a distinctive rise and fall of the mass flow at 200 s mark.  

Thanks very much again.