I am looking for a free CFD program to analyze loss of cooling to a spent nuclear fuel pool. The pool is used to store and cool used nuclear fuel. The pool is 40 ft deep by 80' x 30'. The fuel bundles are 12' tall and 9" square. Each bundle is stored in a square container open at the top and bottom. Upwards of 2000 may be contained in the pool.

Current analysis by the NRC assumes that that heat load is uniformly distributed in the pool and that on a loss of cooling the time to boil is a simple relationship between the heat load and thermal capacity of the total volume of water. The reality is that the total heat load of the pool is roughly 60E6 BTU/hr of which 40E06 btu/hr is clustered in a corner of the pool about 15' square. I question whether convective currents can transfer heat to the far side of the pool before local boiling occurs in the hot corner.

I have available a dual PIII machine running NT 4.0. I also have several Linux machines. I have done some prelim searches and PHOENICS keeps turning up. Please note that I am trying to deal with a nuclear safety issue, any help would benefit many.

(1). We don't know how the hot fuel bundles are placed in the pool, and we don't know how each fuel bundles are designed, so, it is hard to answer your question. (2). I am sure that each fuel bundle contains many fuel rods, and each is surrounded by space in which the water can flow from the bottom of the 12' bundle to the top.(since you have mentioned that both the top and the bottom of the square container are open (to cooling water I guess). (3). So, you have relatively long and thin chimney surrounding each fuel rod, which will effectively pump the cooling water from the bottom to the top. Based on the number you quoted, the volume ratio of the cooling water to the fuel bundle volume is about 7. If we assume that half of each bundle is occupied by the fuel, then, the volume ratio will be 14 when the cooling pool is at its maximum capacity. (4). As the hot water is being pumped by the convection upward, cooler water will be sucked in from the bottom. By continuity, the hot water at the top of the pool will have to come down from the side of the pool, because it is not heated on the wall. So, the general picture will be that above the hot fuel bundle, the flow will move upward, and then spread toward the sides of the pool , and then sink down to the bottom, where it is recycled through the bottom of the fuel bundles. (5). This motion is established by the continuity equation and the temperature gradient. So, my feeling is that this circlar motion will be a global one. That is the fluid will always move downward on the wall, then, recycle through the fuel bundle from the bottom of each container. (6). If you place the fuel bundles in a corner, then the upward moving water will still move away from that corner towar the rest of the three corners, and then follow the walls and move downward to the bottom. It will behave like the outer boundary and send the water toward one corner on the bottom of the pool. So, the doughnut shape flow motion will be established across the whole pool, unless there are some obstacles on the bottom of the pool to deflect the flow moving toward the loaded corner. (7). So, my feeling is that in such case, the flow motion will be global across the whole pool, and thus, the conduction and the mixing of hot and cool water will be on the global scale. (8). Naturally, it still depends on the arrangement of the fuel bundles in the pool. (9). You can put a heating element into a glass fish tank (with fish removed from the tank first) and observe the the motion of the water by mixing with some milk. (10). This is just my own feeling, without knowing your exact configuration and the question. And I don't know whether you will be able to model each fuel rod in the pool with a CFD code. As you said, there could be 2000 fuel bundles, and I think, there will be many more fuel rods in it.(I could be wrong here, because I don't have the design information about the fuel bundles)

Thank you for a prompt and detailed response. I like the idea of a fish tank to try it out on. I am certain that eventually the pool would get circulation, but establishing full natural circ can take a long time. if boiling in the hot corner occurs before full circulation takes place, undesirable results could occur. The large density difference once boiling does occur will accelerate the cooling flow and may quench boiling.

You are correct in surmising the bundle properties - it is a 17*17 grid with .3088" fuel rods. The chimneys are a pre-existing structure into which the fuel assemblies are loaded. Since not all cells are filled, there may be 'local' downcomers.

The current belief by the regulator is that all will be well on a loss of cooling due to the large thermal sink. I am trying to confirm that local bulk boiling will not occur.

Is phoenics suitable for this application? I have seen applications where the fuel is smoothed into a porous heat producing media with known hydraulic resistances.

(1). You can get in touch with the code vendor to see if they have similar experience in pool cooling of fuel bundles. (2). You can also try to do the website search of DOE laboratories, to see whether they have done similar cfd simulation or not. (3). In your case, the issue is whether the simulation can be done accurately or not. (you are talking about the prediction of the heat transfer coefficient, and the mass flow rate through the fuel bundle (and also the pressure difference between the bottom inlet and the top outlet). (4). I think, it is an interesting problem, and you should be able to find some cfd simulations in the related areas.(purely my guess)

AEA Technology's CFX package CFX4 was originally written for solving complex heat transfer and multi-phase flows in the Nuclear industry.

Therefore, I imagine that many of the applications you might wish to simulate will have been done by CFX and many routines and models will already be in the code to help you model the phenomina.

Barry,

Wow, 80x30 is a pretty big SFP. I'll have to look up to see which plant that is. We've been analyzing these problems using FLUENT for 5 or 6 years now.

At the time we selected FLUENT we also looked at PHOENICS and just about everything else avaiable. We could never get PHOENICS or any other FEM-based solver to even come close to obtaining a converged solution. We had much better luck with finite-volume codes.

IMPORTANT: We evaluated all the available codes years ago and I don't know anything about the current versions of PHOENICS or any other code besides FLUENT. I don't want to say FEM-based won't work now, just that they didn't work very well half a decade ago.

Several codes are available from the DOE to perform SFP analyses, most notably COBRA-SFS from ORNL. COBRA is a subchannel code, not a true CFD solver, but the NRC and others use it successfully. The NRC recently had a code developed, I think, the name of which currently escapes me. I'll post the name if I can remember it.

If you can tell me who you are working for and/or affiliated with, I may be able to provide you with much more information. E-mail me directly if you want.

PHOENICS is a Finite-Volume Code and has never had any "FEM-based solvers".

Its huge record of industrial applications includes many Nuclear Safety issues. One might like to have a go through numerous proceedings, reports, papers and thesises to pick up (and quickly reproduce even with PHOENICS shareware) something very similar , if not exactly as required, to the problem in question.

Regards

Sergei Zhubrin

Barry,

It appears that several of the commercial firms have staff monitoring this web site.

And at least 3 commercial codes have been mentioned, two by folks using the CFD firm's e-mail.

You might just ask them directly if the problem is of sufficient marketing potential for them to give you some free help.

After all, you did state up front that you were looking for a free code!

Apologies to all at Cham. As I said, I'm working from 5-year old memory.