6
NT SEMAR 03/021 Revision 3 EXECUTIVE SUMMARY Introduction Over the last thirty years, the Committee on the Safety of Nuclear Installations (CSNI) of the OECD Nuclear Energy Agency (NEA) has sponsored a considerable number of international activities - in particular, International Standard Problem exercises - to promote the exchange of experience between its Member countries in the use of nuclear safety codes. ISPs are comparative exercises in which predictions or recalculations of a given physical problem with different best- estimate computer codes are compared with each other and above all with the results of a carefully specified experiment. A primary goal is to increase confidence in the validity and accuracy of analytical tools which are needed to assess the safety of nuclear installations, and to demonstrate the competence of the organisations involved. Following an offer made by the French Institute for Radiological Protection and Nuclear Safety (IRSN) in 1998, the CSNI has devoted its 46th International Standard Problem to a code comparison exercise based on the Phebus FPT1 experiment. The ISP-46 exercise is part of the programme of work of the CSNI Working Group on the Analysis and Management of Accidents (GAMA) both in the field of in-vessel behaviour of degraded cores and in the field of fission product release, transport, deposition and retention. Its objective is to assess the capability of computer codes to reproduce an integral simulation of the physical processes taking place during a severe accident in a pressurised water reactor, i.e. including fuel degradation and associated hydrogen production and fission product release, fission product and structural material transport in the primary circuit, aerosol behaviour in the containment and iodine radiochemistry. The ISP was conducted as an open exercise, with all the relevant experimental results being available to the participants. It was divided into four phases: Fuel degradation, hydrogen production, fission product and structural material release (‘bundle’, phase 1); Fission product and aerosol transport in RCS (‘circuit’, phase 2); Thermal-hydraulics and aerosol physics in containment (‘containment’, phase 3); Iodine chemistry in containment (‘chemistry’, phase 4). The emphasis was on integral calculations (all phases). The aim was not to carry out interpretation work, but to use the codes as in plant studies, i.e. with standard models/options as far as possible, representing the facility in a similar level of detail; this constituted the mandatory ‘base case’ calculation. A more detailed ‘best-estimate’ sensitivity study could also optionally be performed. Schedule and Participation The ISP started in November 2001, with a timescale of two years. An initial Workshop was followed by issue of the main Specification Report, that took account of comments made at the opening meeting. The period for calculations lasted six months, during this period a supplementary Workshop was held to clarify points arising. The Comparison Workshop and Final Workshop were held about one year and eighteen months respectively after the start of the project, with participants having had the opportunity to submit revised calculations in the meantime. The ISP was well supported, with participation from 33 institutes, companies etc. in 23 countries and international organisations. The latter comprised EC-JRC, Austria, Belgium, Bulgaria, Canada, Croatia, Czech I

EXECUTIVE SUMMARY - oecd-nea.org

  • Upload
    others

  • View
    2

  • Download
    0

Embed Size (px)

Citation preview

NT SEMAR 03/021 Revision 3

EXECUTIVE SUMMARY

Introduction Over the last thirty years, the Committee on the Safety of Nuclear Installations (CSNI) of the OECD Nuclear Energy Agency (NEA) has sponsored a considerable number of international activities - in particular, International Standard Problem exercises - to promote the exchange of experience between its Member countries in the use of nuclear safety codes. ISPs are comparative exercises in which predictions or recalculations of a given physical problem with different best-estimate computer codes are compared with each other and above all with the results of a carefully specified experiment. A primary goal is to increase confidence in the validity and accuracy of analytical tools which are needed to assess the safety of nuclear installations, and to demonstrate the competence of the organisations involved.

Following an offer made by the French Institute for Radiological Protection and Nuclear Safety (IRSN) in 1998, the CSNI has devoted its 46th International Standard Problem to a code comparison exercise based on the Phebus FPT1 experiment. The ISP-46 exercise is part of the programme of work of the CSNI Working Group on the Analysis and Management of Accidents (GAMA) both in the field of in-vessel behaviour of degraded cores and in the field of fission product release, transport, deposition and retention. Its objective is to assess the capability of computer codes to reproduce an integral simulation of the physical processes taking place during a severe accident in a pressurised water reactor, i.e. including fuel degradation and associated hydrogen production and fission product release, fission product and structural material transport in the primary circuit, aerosol behaviour in the containment and iodine radiochemistry.

The ISP was conducted as an open exercise, with all the relevant experimental results being available to the participants. It was divided into four phases:

• Fuel degradation, hydrogen production, fission product and structural material release (‘bundle’, phase 1);

• Fission product and aerosol transport in RCS (‘circuit’, phase 2); • Thermal-hydraulics and aerosol physics in containment (‘containment’, phase 3); • Iodine chemistry in containment (‘chemistry’, phase 4).

The emphasis was on integral calculations (all phases). The aim was not to carry out interpretation work, but to use the codes as in plant studies, i.e. with standard models/options as far as possible, representing the facility in a similar level of detail; this constituted the mandatory ‘base case’ calculation. A more detailed ‘best-estimate’ sensitivity study could also optionally be performed.

Schedule and Participation The ISP started in November 2001, with a timescale of two years. An initial Workshop was followed by issue of the main Specification Report, that took account of comments made at the opening meeting. The period for calculations lasted six months, during this period a supplementary Workshop was held to clarify points arising. The Comparison Workshop and Final Workshop were held about one year and eighteen months respectively after the start of the project, with participants having had the opportunity to submit revised calculations in the meantime. The ISP was well supported, with participation from 33 institutes, companies etc. in 23 countries and international organisations. The latter comprised EC-JRC, Austria, Belgium, Bulgaria, Canada, Croatia, Czech

I

NT SEMAR 03/021 Revision 3 Republic. France, Germany, Greece, Hungary, Italy, Japan, Korea, Mexico, Russia, Slovenia, Spain, Sweden, Switzerland, Turkey, UK and USA. The participating organisations included utilities, regulators and their technical support organisations, research institutes and private engineering consultancy companies, thus providing a good range of backgrounds to the technical work. Fifteen different codes were used: ASTEC, ATHLET-CD, COCOSYS, CONTAIN, ECART, FEAST, IMPACT/SAMPSON, ICARE/CATHARE, IMPAIR, INSPECT, MAAP4, MELCOR, SCDAP/RELAP5, SCDAPSIM and SOPHAEROS, of these 4 are integral codes (ASTEC, IMPACT/SAMPSON, MAAP4 and MELCOR). For the base case, 47 calculations were received, with 21 for the optional best-estimate version. Of the base case calculations, 14 were integral (at least 3 phases calculated).

Representation of the Facility For the base case, a noding scheme was recommended in the specification report The bundle is divided into 11 axial nodes and typically 3-5 radial rings, with normally 1 or 2 thermal hydraulic flow channels. The circuit is divided into 11 nodes, this being the minimum considered necessary for an adequate calculation of deposition. The containment model is simple, with 1 node for the main volume and 1 for the sump, taking advantage of the well-mixed conditions. For best-estimate calculations, noding density was increased by typically a factor 2 or more, at the choice of the user.

Analysis of the Results The results were analysed in detail, comparing the results amongst each other and with the FPT1 data. There was considerable scatter amongst the results obtained from each code by different users, the ‘user effect’. To minimise this effect, representative cases were selected where necessary, taking into account the quality of key output variables, completeness and accuracy of the technical reports, and including code developers where possible. This analysis led to an assessment of the main models in each of the four areas considered. These are grouped below, in order of their perceived adequacy. There was on the whole little significant difference between the base and best-estimate cases, with at most a small improvement only in the results of the latter cases, so conclusions could be drawn on the basis of the former.

Assessment of Codes and Models The following phenomena/parameters are in general well simulated by the codes:

• Bundle – thermal response (given adjustment of input nuclear power and shroud thermal properties within experimental uncertainties), hydrogen production (including oxidation of relocated melt), bundle final state material distribution (given suitable reduction of the bulk fuel relocation temperature from the ceramic value, in the longer term a more mechanistic model is desirable), total release of volatile fission products;

• Circuit – total retention of fission products and structural materials (but after cancellation of errors);

• Containment – thermal hydraulic behaviour (as exemplified by average gas temperature, pressure, relative humidity and condensation rate), depletion rates;

• Chemistry – models of the Ag/I reaction in the liquid phase are adequate for FPT1 (this cannot be extended to other cases where the Ag is not so much in excess with respect to I; due to the large excess of silver, in the experiment, radiolytic production of gaseous iodine

II

NT SEMAR 03/021 Revision 3

and dissociation of silver iodide did not play an important role in the overall iodine behaviour).

The following phenomena/parameters were reasonably well simulated, but some modelling improvement is desirable:

• Bundle – outlet coolant temperatures (overprediction), time dependence of volatile FP release (generally too fast a release at low temperatures, e.g. for CORSOR-type approaches);

• Circuit – distribution of deposition in the circuit (underestimation in the upper plenum where vapour condensation and thermophoresis are the dominant mechanisms, overestimation in the steam generator hot leg where the mechanisms are thermophoresis for all elements + vapour condensation for I and Cd), noting that too coarse a noding leads to underestimation of deposition;

• Containment – relative importance of the two main depletion processes (diffusiophoresis and gravitational settling), but it is hard to make firm conclusions owing to the variability in the results;

• Chemistry – no items identified.

The following phenomena/parameters were not well simulated and substantial model development is necessary:

• Bundle – release of medium and low volatiles (e.g. tendency to calculate low for Mo, very high for Ba, reasonable order of magnitude for Ru and U but considerable scatter), and of structural materials (Ag/In/Cd from the control rod where the basic process of evaporation from a molten AIC pool is not captured, tin from the Zircaloy cladding);

• Circuit – iodine speciation and physical form;

• Containment – no items identified;

• Chemistry – gas phase reactions, organic iodine reactions, including production and destruction through radiolytic processes (definition of optimum parameters for the modelling codes such as adsorption velocity and desorption rate on/from painted surfaces, and the facility to input the gaseous iodine fraction at containment entrance, are recommended).

The good prediction of hydrogen production, generally near the upper bound of the experimental uncertainty range, (+10%), is an important safety-relevant conclusion. The good prediction of the bundle material distribution in the final states requires a suitable reduction of the bulk fuel relocation temperature from the ceramic value. Implications, in the short term and, in the longer term, the need for a more mechanistic model will be discussed later on as integral aspects. Although the structural materials do not themselves have radiological significance, they potentially react with fission products, their source terms are therefore needed for accurate calculation of chemistry and transport in the circuit. A particular need is to saturate the iodine reaction. The semi-volatile fission products are also of importance, either because of their radio-toxicity and influence on the residual power, or by their propensity to react with other fission products. The structural materials also form the bulk of the aerosol mass, affecting the aerosol concentration and the agglomeration processes.

III

NT SEMAR 03/021 Revision 3 Concerning the circuit, the overestimation of bundle outlet temperature cannot fully explain the upper plenum results; its main effect is to displace the zone where vapours nucleate. For some elements, part of the discrepancy in the deposition pattern is due to the wrong prediction of the chemical form, and thus of its volatility; Cs is generally calculated as a vapour at 700°C, whereas it was condensed in the experiment. However, this is also not enough to explain the underestimation in the upper plenum and overestimation in the steam generator rising line. Finding explanations is presently part of the work performed in the frame of the interpretation of Phebus-FP tests.

Care is needed in extrapolating the rather good results for the containment directly to the reactor case, as the Phebus containment thermal hydraulics are relatively simple, and the role of gravitational settling is overscaled, with a shorter residence time of aerosols in the atmosphere and probably less effect of agglomeration than for real plant.

Concerning the chemistry, the reaction of iodine with silver, forming non-soluble silver iodide, dominates the phenomenology in the liquid phase. In the FPT1 conditions with a large excess of silver, the models behave sufficiently well, provided enough silver is injected into the sump water. Gas phase chemistry is dominated by early injection of gaseous iodine from the primary circuit, that will be discussed later on as an integral aspect, and by inorganic iodine adsorption on the atmospheric paints followed by organic desorption, together with destruction mechanisms. The results were particularly contrasted, with a large scatter on the total gaseous iodine concentration, and a fraction of organic iodine ranging from less than 10% to nearly 100%. Overall, they range from unreliable to very good (after tuning).

Computing Assessment Key output variables for code assessment, such as those requested in the ISP, were not always accessible to the user; these should be available as standard code output. Graphics dumps to enable post-processing of results should be a standard feature to aid in code assessment, to aid in detailed analysis. Computer (CPU) time and timestep information should be available for plotting, to help optimise code use, while temporal (timestep) and spatial (noding) convergence studies should always be performed. Platform dependence (both concerning computer hardware, and sensitivity to compiler options) should be eliminated as far as possible.

Integral Aspects This part considered the results with respect to the ‘key signatures’ of plant sequence calculations, namely the core final state (relevant to in-vessel retention) and the fission product source term. Good agreement for the bundle final state could be obtained with suitable reduction of bulk fuel relocation temperature, but this is unlikely to be representative for similar tests such as Phebus FPT2 and PBF SFD1.4 which show evidence for a higher temperature. Therefore, default values should not be reduced on the evidence of FPT1 alone, and similar studies on other experiments such as these are encouraged. In the longer term, a more mechanistic treatment of bulk fuel relocation is desirable, and it seems unlikely that a simple temperature criterion will suffice. More detailed model development, possibly needing new separate-effects data, is therefore indicated, as the mechanisms involved are not clear (effect of irradiated fuel, presence of Fe in the melt …… ). In the meantime, plant studies need sensitivity calculations on relocation temperature with the modelling in its current state. Further studies are recommended on control rod failure (influence of control rod materials) and the fuel rod oxide shell breach criterion (first movement of U/Zr/O melt). A general integral point for the bundle is the need to take into more account the interaction between bundle state and fission product/structural material release, especially for the latter.

Concerning the source term, the accuracy of containment calculations in integral treatments is sensitive, often highly, to results of previous stages (propagation of uncertainties). Key features are

IV

NT SEMAR 03/021 Revision 3 the calculation of FP release from the bundle, and of the structural materials Ag, In, Cd and Sn (the kinetics of release of these and of FPs are as important as the final amounts); the temperatures at the entrance to the circuit, which strongly influence the deposition pattern; while for those codes which calculate the chemistry, the speciation is influenced by the calculated release. The release of structural materials was often undercalculated or not calculated at all, leading to undercalculation of total mass of aerosols, but this had only a weak impact on overall retention in the reactor coolant system (RCS) and depletion in the containment. Iodine speciation and physical form in the circuit was poorly predicted - no code reproduced the observed gaseous iodine fraction in the RCS.

Given these limitations, it is hard for an integral calculation to predict well the containment chemistry, however detailed the modelling for its phenomena – the uncertainty on iodine release from fuel, aerosol transport in the RCS and behaviour in containment is overwhelmed by uncertainties in chemistry. This has implications on conduct of plant assessments, for example it may be better for the chemistry calculations to be carried out in a stand-alone manner, using a range of sensitivity studies, rather than as part of an integral calculation.

Finally, in determining the priorities for code improvement, attention should be paid to finding the weakest link(s) in the chain of calculation which contribute most to uncertainty in the assessment of risk - a ‘cost-benefit’ approach - is it a model itself or the input to it?

Implications for Plant Studies A strong user effect is visible in ISP-46, as in previous ones, therefore the user effect in plant studies cannot be ruled out. A major objective must be to limit its consequences on the quality of the study. It is recommended that this could be achieved by: checking that previous training has been efficient; using adequate procedures are used for controlling the results and peer reviewing, involving experienced specialists in the field; and by checking that enough support is provided by developers when necessary.

The quality of the models must also be taken into account. A number of necessary improvements in codes and models have been identified above, the main ones being: a better estimation of structural material release, especially for control rod elements and tin from Zircaloy cladding, and of semi/low-volatile release; the possibility to take into account the presence of gaseous iodine in the RCS; and the definition of optimum parameters for iodine chemistry codes. As not all the necessary improvements can be achieved in a short term, users have to be well aware of the validation status of codes and must take into account their limitations when performing plant studies.

Severe accident codes are difficult to handle, and their validation is not complete. They should not be used as “black boxes”, i.e. their results have to be interpreted, according to the goal of the study for which they are used. Extensive training of new users should be mandatory, and efficient quality assurance procedures for reactor studies have to be used, involving review of the results by experienced experts not directly involved in the work.

Finally, users should not trust automatically the results of their calculation, but make a critical analysis! Do the results seem consistent and reasonable (“reality check”)?

Concluding Remarks This first integral ISP has enjoyed a wide and varied participation, with almost fifty submissions. There was strong support for the bundle and circuit phases, moderate for the containment aerosol phase, and least for the containment chemistry phase. Follow-on studies may be proposed to continue the exercise, focussing on areas where the greatest uncertainties remain.

V

NT SEMAR 03/021 Revision 3 Recommendations have been made on model development needs, which have been agreed after discussions with the development teams of the major integral codes, and progress is already being made in addressing some of the issues raised. Various implications for plant studies have been identified, and the need for user experience and training is emphasised. Effective review procedures for reports are seen as essential in ensuring the quality of such applications.

Acknowledgements The authors gratefully acknowledge the code developers, for their involvement, and for having provided their own views on the results of the exercise; the OECD/CSNI/NEA and its Secretariat, for having offered the opportunity to conduct such an exercise, all the participants to the ISP, for their intensive and fruitful cooperation, and finally the European Commission DG Research, for the support provided through the thematic network ‘THENPHEBISP’ in the 5th Framework Nuclear Fission Safety programme, contract number FIKS-CT-2001-20151.

VI