Publikationsrepositorium - Helmholtz-Zentrum Dresden-Rossendorf

Reactivity-initiated accident (RIA) fuel rod codes have been developed for a significant period of time and they all have shown their ability to reproduce some experimental results with a certain degree of adequacy. However, they sometimes rely on different specific modeling assumptions the influence of which on the final results of the calculations is difficult to evaluate.

A conclusion from the 2009 CSNI workshop on RIA was that RIA fuel rod codes are now heavily used, within the industry as well as the technical safety organizations (TSOs), in the process of setting up and assessing revised safety criteria for the RIA design basis accident.

It is then very important to master the use of such codes for reactor accident studies, particularly those involving safety analyses. It is essential to identify and understand real accident conditions that deviate from those of experiments. As a conclusion of the workshop, it was recommended that a benchmark between these codes be organized in order to give a sound basis for their comparison and assessment.

In order to maximize the benefits from this exercise, it was decided to use a consistent set of four experiments on very similar highly irradiated fuel rods tested under different experimental conditions:

• low temperature, low pressure, stagnant water coolant, very short power pulse (NSRR VA-1),
• high temperature, medium pressure, stagnant water coolant, very short power pulse (NSRR VA-3),
• high temperature, low pressure, flowing sodium coolant, larger power pulse (CABRI CIP0-1),
• high temperature, high pressure, flowing water coolant, medium width power pulse (CABRI CIP3-1).

Each of these four sibling rods are of identical fuel design and cladding material. Each experienced essentially the same pre-test irradiation history. For most practical purposes, the experiments are identical except for test conditions. The intent was to examine the so-called “temperature” effect (i.e. how to transpose results from experiments at low temperature to reactor conditions at high temperature?) in various RIA test programs.

The participation to the benchmark has been very important: 17 organizations representing 14 countries provided solutions for some or all the cases that were defined. In terms of computer codes used, the spectrum was also large as solutions were provided with FALCON, FEMAXI coupled to TRACE, FRAPTRAN, RANNS, RAPTA, SCANAIR, TESPAROD and TRANSURANUS.

The first noticeable fact is that, nearly all the participants used code that rely on simplified geometrical representation usually referred to as 1.5D codes. Although some 3D calculations may be done (one example was shown by one participant), it appears that given the conclusions below, the detailed geometrical description is not a priority. Rather, it looks more important at this stage to put the efforts and continue working on physical modeling.

During the benchmark, one source of differences between the results of the participants was identified to be due to the way input data, in particular the power pulse, are interpreted within the different codes. It is recommended that the code developers carefully examine the way the input data are used because this source of difference, that appeared to be significant, should be completely removed.

It was not possible during this benchmark to assess the influence of the initial state (resulting from base irradiation) of the fuel on the behavior during RIA. Nevertheless, this would be an important thing to do in the future in order to evaluate how much it accounts for on the scatter of the results.

With respect to the thermal behavior, the general conclusion is that the differences in the evaluation of fuel temperatures remain limited, although significant in some cases. The situation is very different for the cladding temperatures that exhibited considerable scatter, in particular for the cases when water boiling occurs. The film boiling heat transfer model was responsible for large differences between the calculations.

With respect to mechanical behavior, when compared to the (known) results of an experiment that involved only PCMI, the predictions of the cladding hoop strain from the different participants appeared acceptable, even though there was a factor of 2 between the highest and the lowest calculations. The conclusion is not as favorable for a case for which both the experimental results are unknown and water boiling is predicted to appear because then a factor of 10 on the hoop strain between the calculations was exhibited. This is due for a large part to the differences on the cladding temperatures.

In this benchmark, the fission gas release evaluations were also compared. The ratio of the maximum to the minimum values appears to be roughly 2, which is estimated to be relatively moderate given the complexity of fission gas release processes.

Finally, failure predictions that may appear as the ultimate goal of fuel code dedicated to the behavior under RIA conditions were compared. When looking at predictions in terms of enthalpy at failure, which is of interest in practical reactor applications, typical variations between calculations were found to be within a +/- 50 % range. Although major causes of the differences were identified, it is recommended to perform more systematic sensitivity and uncertainty analyses in a new phase of the benchmark to further assess the significance of the results produced.

It has been possible to evaluate the so-called user effect for the FRAPTRAN and SCANAIR codes. For both of them, it was found to be very limited on the cases of this benchmark, nearly negligible if compared to the differences between the results of the different codes. To generalize this conclusion would require more case to be studied.

The broader objective of the benchmark was to assess the possibility of evaluating the “temperature effect” that can be stated as: is it realistic to use the RIA fuel codes to transpose results, in particular enthalpy at failure, from experiments performed at low temperature to typical reactor conditions? Based on the conclusions formulated above, it appears obvious that it should be done with caution given the scatter that exists between the predictions of the different codes mainly due to the different approaches used to assess the rod failure level.