Gd burnup in VVER-1000 fuel assemblies (benchmark with HELIOS)

[Antonio Dambrosio]

Hello all,
I'm currently facing what seems to be an issue with the depletion of Gd isotopes, specifically the 155 and 157, in VVER-1000 assemblies. 

As part of a master thesis research work, we are conducting a benchmark between SCALE/TRITON and HELIOS on new fuel assemblies for Temelin, a VVER-1000 reactor.
The initial conditions, geometry and material compositions of the models are identical. I have directly copied the number densities I got from SCALE into HELIOS.

The modelled assembly has 4 different pins, with enrichments of 4.95%, 4.7%, 4.2% and 3.6%. The 18 pins at 3.6% contain 5 (wt)% of Gd2O3.
Each fuel cell uses an 8x8 mesh grid.
The burnup steps are the following [in MWd/tU]:
0, 10,
100 to 1000 in steps of 100
1000 to 20000 in steps of 1000
20000 to 70000 in steps of 2000

I'm currently using SCALE 6.3.1, with the ENDF-VIII 252g library.

The first result we compared is the k-inf: at time 0 and after 22000 MWd/tU of burnup, the results differ of maximum 150 pcm, whereas between 1000 and 20000 MWd/tU the discrepancy is much higher. 
I have also tried increasing the burnup steps, namely using nlib=2 and nlib=4, without improving the results much.

The main difference seems to be a much faster gd-155 and gd-157 depletion predicted by SCALE, which causes an initial stronger dampening of the k-inf and a greater increase of the k-inf as soon as both isotopes are (almost) fully depleted.
After a long discussion with Studsvik, I have been made aware that this is a relatively old issue, where the calculations tend to generate results that don't match experimental results (what they have defined as the "bathtub" k-inf trend). The way Studsvik has addressed this issue is using a quadratic Gd depletion in CASMO and a projected predictor-corrector in HELIOS, which is something I don't think I can implement myself in a TRITON model. 
My questions are:

- has this problem been highlighted or discussed before?
- is there anything that could help improving my results? Possibly still using TRITON: in the future I may consider moving to POLARIS


Thanks in advance for your answers,
Antonio Dambrosio

[Pavlo Ivanusa]

Hi Antonio,

I would recommend moving to Polaris if you can, but if you want to stick with using TRITON NEWT, there are probably a few things you can try. First, you'll have to make sure that your gad pin is radially split into different materials that can deplete at different rates. If your gad containing pin is one material, then you will get incorrect answers. CASMO, for example, does split the gad pins automatically into different radial sections. 

Second, you check on the Dancoff factors if there are significant differences among the pins in the assembly. However, this can be a significant amount of work and it's not something I would do unless you really want to stick with NEWT.

Third, you can run a TRITON case with KENO (e.g., 1 cm axial height with periodic or mirror boundary conditions) instead of NEWT to get a baseline that removes any assumptions made in the transport solution in the presence of strong absorbers.

Are you able to provide one of your input files for this question? 

[Antonio Dambrosio]

Hello Pavlo,

sorry for the late reply.

For the sake of the master thesis itself, also considering the limited time we have, it is impossible to do any attempt woth Polaris or KENO, mainly because i have never used them, so it would take some extra time to adapt my TRITON input. About the first idea, I was erroneously convinced that a very dense mesh, like the 8x8 per cell I'm using, was sufficient for the Gd cases. I can definnitely try to split the pin into 6 radial rings (the same we use in HELIOS) and see if something changes. I agree that manually searching for each Dancoff factor would be painful.

Nonetheless, this is also a work that we are doing for the Czech TSO, to prepare a detailed cross section libraries to use in PARCS. So, as soon as the thesis is concluded, moving to POLARIS will be my priority. 

Side question for the first idea: once i split the pin into multiple rings, how would I define the cell for the latticecell definition? Should I leave it as is, using a full pin with only one of the materials?

Extra note: I'm unable to share the input since it contains proprietary data of the Czech utility

Thanks,

Antonio

[Pavlo Ivanusa]

You will have to use a "multiregion" lattice treatment. https://scale-manual.ornl.gov/XSProc.html#unit-cell-specification-for-multiregion-cells

You can then define several radii for the fuel pin. Just make sure that you convert your pin pitch to the correct cylindrical radius to have the same moderator to fuel ratio. 

[Rike Bostelmann]

Hi Antonio,

Another reason why your isotopics show differences may be found in differences in the applied recoverable energy from capture/fission, the "kappa" values. For the SCALE 6.3 release, some Gd kappa values were updated, as described in this report. Changes in kappas can cause changes in power normalization which impacts the depletion calculation. 

TRITON is executing a simple predictor-corrector procedure, with one predictor step to the middle of a depletion step, followed by a corrector step to the end of the depletion step that uses 1-group XS from the middle of the depletion step.

Polaris is certainly recommended for LWRs. However, hexagonal geometries are not yet supported, so it may not be an option for your VVER model.

I'm not sure if this helped. But maybe it can at least support the discussion of your results in your thesis.

Thanks,

Rike

SCALE Team

[Willem van Rooijen]

Please note (perhaps a bit late) that sample case TRITON8 seems to provide an example of what you are trying to do, i.e. depletion of a Gd-bearing fuel pin in a multi-pin model (in this case, a BWR pin). Sample case TRITON8 also shows the application of aliases and the MULTIREGION treatment for the self-shielding.

Namizono

2025年8月1日金曜日 1:00:32 UTC+9 Rike Bostelmann:

[Ian Gauld]

The OECD/NEA International Benchmark on Gd depletion (Burnup Credit Criticality Benchmark Phase IIID: Burnup Calculations of Gadolinium- Bearing Fuel Rods in Boiling Water Reactor Assemblies for Storage and Transportation) is an excellent resource for understanding Gd modeling approximations and nuclear data. Although it is for a BWR configuration, the benchmark includes a range of voids including 0%, so it is potentially relevant to the VVER configuration. See

https://www.oecd-nea.org/jcms/pl_76188/burnup-credit-criticality-benchmark-phase-iiid-burnup-calculations-of-gadolinium-bearing-fuel-rods-in-boiling-water-reactor-assemblies-for-storage-and-transportation?details=true

Also note that there have recently been recent experimental measurements of irradiated PWR Gd rods made as part of the REGAL program in Belgium (SCK CEN) for code benchmarking (including SCALE). See https://doi.org/10.1016/j.anucene.2022.109057 . There may also be valuable information on Gd modeling challenges in this and related papers. 

Best regards

Ian

[Antonio Dambrosio]

Thanks to everyone for answering and adding useful info!

I have acknowledged that POLARIS is not supporting hexagonal geometries yet, so I will insist on TRITON for now.
I will be reading all the documents that have been linked, since they look very interesting, and not only relevant for the thesis.

As for the multiregion solution, I have used the same approach as in HELIOS, i.e. 6 concentric rings with the same volumes, and modified the radius of the water channel so it keeps the same moderator-to-fuel ratio:
"
multiregion cylindrical right_bdy=white end
 8 0.159217
18 0.225167
28 0.275772
38 0.318434
48 0.356020
58 0.39
39 0.3965
42 0.455
74 0.6694 end zone
"
The solution was sufficient to fix the Gd depletion, which is now in line with HELIOS, and the k-inf trend as well. Now, around the burnups where the Gd 155 and 157 are completely depleted, the difference between k-inf is, at max, around 300-500 pcm: the higher the Gd content, the higher the difference. To try to improve the results further, I will try to play with mcdancoff to calculate the dancoff factors for the gd pins and apply the corrections in TRITON.
For the time being, I'm more than satisfied with the results. As I have said in the previous messages, I would have sworn that a very detailed mesh grid was sufficient to address the self shielding of the gd pins, so I'm happy to have learned something new.

[Willem van Rooijen]

Well, in the scope of "scandalous self-promotion":

W.F.G. Van Rooijen; Mohammad Jahirul Haque Khan; Alain Hébert; Vivian Salino, " Analysis of a VVER-1000 in-core fuel management benchmark with DRAGON and DONJON", International Journal of Nuclear Energy Science and Technology, 2017 Vol.11 No.4, pp.345 - 376

DOI: 10.1504/IJNEST.2017.090658

Regards,

Van Rooijen

2025年8月1日金曜日 9:43:02 UTC+9 Ian Gauld: