ML20214X027
| ML20214X027 | |
| Person / Time | |
|---|---|
| Issue date: | 03/05/1985 |
| From: | Ross D NRC OFFICE OF NUCLEAR REGULATORY RESEARCH (RES) |
| To: | Gillespie F, Marino G, Speis T Office of Nuclear Reactor Regulation, NRC OFFICE OF NUCLEAR REGULATORY RESEARCH (RES) |
| Shared Package | |
| ML20213E209 | List:
|
| References | |
| FOIA-87-113, FOIA-87-60 NUDOCS 8706160305 | |
| Download: ML20214X027 (20) | |
Text
-
,/
UNITED STATES NUCLEAR REGULATORY COMMISSION
-h WASHINGTON, D. C. 20555
' MAR 5 1985 MEMORANDUM FOR:
Those on attached list FROM:
D. R. Ross
SUBJECT:
NRC/IDCOR TECHNICAL ISSUES One of the main objectives of the IDCOR/NRC staff interactions over the past year and a half was to define technical areas of agreement and dispute between the methodologies used by both parties in assessing severe accidents in LWR reactors. -Initially, the " issues" were stated generally in a list of about 59 (later reduced to 29) broadly defined areas.
The NRC. staff produced documents outlining their concerns in these areas which were later used as. the basis for discussions during the specific technical exchanges.
Now that the major meetings are over, the NRC staff is attempting to define more clearly those issues which merit.further discussion and resolution.
An initial attempt wa's made to do this via a meeting between RES and NRR staffers and their contractors on January 31, 1985.
A "strawman list" of issues taken from the IDCOR/NRC minutes of the meetings was prepared by G. P. Marino and used as a nucleus for expansion, deletion and discussion.
NRR participants included R. Barrett, J. Rosenthal, J. Mitchell, C. Allen, and Z. Rosztoezy.
RES personnel included G. Marino, M. Silberberg and T. Walker.
Contractors present were T. Pratt (BNL)
T. Kress (ORNL), S. Hodge (ORNL), P. Cybulskis (BCL), R. Denning (BCL), and J. Gieseke (BCL).
R. Barrett presented a suggested process for prioritizing and assessing the importance of the issues relative to their effort on reactor systems and sequences.
Much discussion ensued on the issue presentation of G. Marino l
-which resulted in the deletion of one issue, the combination of three issues, and the addition of five unidentified issues.
Later updates were provided -
i by R. Denning from BCL.
On February 14, 1985 the issues were transmitted to l
IDCOR management at a meeting in Chicago attended by R. Bernero, G. Marino, l
Z. Rosztoczy, and R. Denning.
The final revised set of issues along with a brief-discussion of each is presented in Attachment 1.
Please review the document relative to your research program area and give any comments to G. Marino by March 16, 1985.
I i
e l-
2 R. Barrett will prepare a new paper on a proposed path for issue resolution which will be sent to you under separate cover.
Your input in this effort is needed to assure that nothing " falls through the cracks" in this extremely complicated area of reactor analysis.
O' l
l 'A D.F(.Ross,DeputyDirector
/
Office of Nuclear Regulatory Research
Enclosure:
As stated e
b f
i
~
,.<,------.-m----__-%
4,-
,-e- - -
w-*--
.,.g,,r.--r-.3-
,7-,,-
,--w -r-m -, e -.--,-
--,wt--
-r-,v,
- ~._._ _ - _. _ _ ~_ _
Summary of Major NRC/IDCOR Technical 6r--
Issues for Severe Accidents In the following paragraphs a number of specific technical issues are identi-fied that are either major contributors to the uncertainty in plant risk or represent significant differences in modeling assumptions between the NRC and IDCOR.
It is important to identify another difference between NRC and IDCOR methodologies that is more basic.
In the Severe Accident Risk Rebaselining and Risk Reduction Program (SARRP) uncertainties in phenomenological behavior are treated explicitly.
Not only is a best estimate analysis performed, but pessimistic and optimistic analyses are also performed to provide an under-standing of the range of possible outcomes.
Although some sensitivity studies have been performed in the IDCOR program, their purpose appears to be to support the selection of assumptions leading to point estimate results.
It is the opinion of the NRC and its supporting contractors that the uncertainties in severe accident' processes are quite large and that a meaningful evaluation of plant safety and the possible need for plant modifications must include the explicit consideration of uncertainties.
One significant source of uncertainty is the definition of the accident sequences.
In several cases where IDCOR and NRC have analyzed.the same or similar sequences, large differences in the calculated consequences have resulted because of different assumptions about the sequence of events in the accident.
These differences often reflect a real uncertainty'in the accident sequence definition.
The NRC believes that it is particularly important to reflect these uncertainties in severe accident analyses.
I Core Heatup Stage Issue #1 - Fission Product Release Prior to Vessel Failure t
Uncertainties in fission product release prior to vessel failure can affect both the timing of release and the total quantity of release from fuel.
Whether the fission products are released in-vessel or ex-vessel can be particularly important because of the potential for retention of fission products on reactor coolant system structures for the in-vessel release com-ponent.
Several sub-issues contribute to this issue; namely O
e
e, 2
(a) Assumed temperature for beginning of fuel relocation.
IDCOR models assume and employ a relocation temperature of 2800C (3100K); 1.e.,
the melting point of UO.
This high assumed temperature in conjunction with 2
their fission product release model (see below) allows all volatile fission products such as I, Cs, and Te to be released prior to core slumping and quenching in the lower plenum.
As a result 100% of these F.P.'s are available to be deposited in the upper plenum and other structures of the primary system, thereby resulting in possible non-conservatism in the source term to the con-tainment for certain seque'nces.
The BMI-2104 models assume that melting occurs at an average eutectic tempera-ture of 2550K.
In conjunction with the empirical CORSOR release coefficients, this temperature is high enough to release essentially all of the volatiles (except tellurium) in-vessel, but may underpredict the release of involatile materials.
A more mechanistic fuel melt progression model is being developed for the MELPROG code based on experimental evidence from KfK, PBF, and ACRR.
(b) Modeling of in-vessel release of fission products The IDCOR fission product release from fuel model is based upon the oxidation of UO by steam.
The model assumes that sufficient steam and contact area 2
(with UO ) is present at al.1 times during the heatup to oxidize the UO to a 2
2 higher state, thereby significantly enhancing the release of fission products.
Each of the species considered is released from the fuel at the same rate.
Partial pressures of the vapor species are used to determine the amount of released material that can be transported as a vapor.
The condensed component is apportioned between aerosol and fuel surfaces according to an input factor.
The BMI-2104 model, CORSOR, is entirely empirical based on a variety of in-pile and simulant experiments.
The BMI-2104 model does not attempt to account for differences in mass transfer limiting processes that may exist between the experiments and the real system.
O a
3 Recent PBF experiments indicate that the CORSOR and IOCOR models both probably overestimate in-vessel release of fission products.
The VICTORIA code which is m
being developed as an element of the MELPROG package will provide a more mechanistic model for fission product release which can be compared with integral experiments in P8F and ACRR.
(c) Te retention in-vessel A major difference exists in the treatment of tellurium release between the 10COR model and the BMI-21'04 model.
Experiments at ORNL and at the PBF strongly indicate that the tellurium release is reduced during this stage of the accident if approximately 25% or more of the Zircaloy remains unoxidized.
Because of the rapid heat-up of a core during boiloff, most sequences result in unoxidized Zircaloy contents greater than 25%.
Therefore, in most sequences, the CORSOR release model allows most of the Te to be retained in the core while it is in the reactor vessel.
The signi,ficance of this effect is that if the Te is released early - as IOCOR advocates - it will be deposited in the upper plenum and not be available in the source term estimate; however, if it is retained in the molten fuel at this stage - as experimental evidence indicates
- it will be released when the core exits the R.V. and interacts with the concrete.
Thus, the Te would be released to the containment volume without the potential for being deposited in the reactor coolant system and, perhaps, at a time closer to the time of containment failure.
Although the CORSOR model accounts for the interaction of tellurium with the Zircaloy cladding, the model is crude and the supporting data are sparse.
The overall significance of this issue is that, for certain sequences in both PWR's and BWR's, significantly higher amounts of volatile fission products (I, Cs, and Te) may be available for release to the containment during the core /
concrete interaction stage than computed in the 10COR models.
Depending on the containment failure time, this effect can and does lead to much larger source terms than computed by 10COR.
The PIE of the P8F tests should help clarify this issue in early FY 1986.
r 4
Issue #2 - Recirculation of Coolant in the Reactor Vessel
- m Neither IDCOR nor current NRC (BMI-2104) models are able to calculate recir-culation patterns of steam in the R.V. after core uncovery.
Based on simpli-fied analyses at SAI (EPRI) and Purdue, it appears that recirculating ' flow could have a significant effect on the core heatup bahavior of PWR's at least for high pressure sequences.
It is speculated that recirculating flow can affect the core heatup rate, in-vessel release of fission products, quantity of hydrogen produced in-vessel, structure temperatures, and deposition of fission products in the reactor coblant system.
Analytical studies have been initiated with the COBRA, COMMIX, and TRAC codes to resolve this issue.
These studies will be completed by December, 1985.
Experimental results are being obtained by Westinghouse which should offer an opportunity for model validation.
Issue #3 - Release Model for Control Rod Materials A major discrepancy between the IDCOR and BMI-2104 models is the mode of be-havior of silver and cadmium during core meltdown.
The IDCOR model allows the control rod material to rapidly melt and runoff to cooler regions of the core where it freezes and takes no further part as a potential source of inert aero-sol material.
In contrast, the BMI-2104 models ignore the potential for runoff of liquified control rod material prior to regional slumping and, as a result, may tend to overpredict the release of control rod silver and cadmium.
Neither the IDCOR or BMI-2104 analyses consider the possibility of chemical reactions with B C control material that could result in increased hydrogen 4
production or changes in the chemical form of fission product species.
The SFD1-4 experiment in PBF is designed to study questions related to the behavior of silver-indium-cadmium control rods.
Investigations at ORNL'are addressing 8 C control material behavior.
4 e
S Issue #4
' Fission Product and Aerosol Deposition E
in the Primary System
-(
c The uncertainties in the existing predictive capability (TRAP-MELT 2 is the mostytailed of the methods available) for deposition in the reactorloolant system are very large.
The vapor and aerosol transport processes are complex t
and interacting.
The upper plenum geometries and flow regimes are treated in a L
i '-
crude. approximation of the prototype.
In initial discussions with IDCOR, the focus of NRC concern was the use of a log normal aerosol size distribution in the RETAIN code.
This forinulation has now been replaced with an empirical aerosol model, the validity of which is of equal concern in the RCS geometry.
4
^
Results of verification experiments at ORNL and the integral Marviken tests will provide a means to evaluate the existing models.
j II Melt Progression and Fuel Relocation Stage gs 5
l Issue #5 - Modeling of In-Vessel H, Generation
,, w,
j Substantial agreerrent exists in the modeling of the steam /Zircaloy reaction in 4
the reactor core as long as the original geometry is maintained.
However, when P
cladding melts and slumps and flow channels begin to block, issues arise as to when blockage will occur, how effective it will be, whether cladding material l
will run out of the hot zone, how much oxidation will occur as it moves, and l
whether the relocated cladding will subsequently be reheated and exposed to i
steam.
The IDCOR models employ two parameters:
a metal-water reaction cut-off j
temperature and a flow blockage parameter that effectively limit the extent of l'
hydrogen" production that occurs in-vessel.
Similarly the BMI-2104 code, MARCH 2, has externally controlled parameters that can limit the extent of metal-f water reaction.
In practice, however, the "best-estimate" assumptions made by the IDCOR and BMI-2104 analysts have lead to substantially different results.
0 2
i I
c
'6
- l t
y For a PWR SBLOCA sequence.without ECCS, IDCOR models calculate 200 Kg of H 2
[
produced. whereas, the.the BMI-2104 calculation yields 450 Kg of H I"'
' ~
2 significant dt(ference of a factor of 2.25.
Experimental data of H2 production after fuel relocation is difficult to obtain, but careful analysis of the PBF h
Phase I program and planned experiments in the NRU facility wherein full-length ti elements will be tested to slu'mping and held for long times will help to resolve this issue.
The MELPROG code will employ more mechanistic fuel slumping models.
and a two-dimensional fluid flow model which will also provide a better under-i standing of the issue.
Issue #6 - Core Slump, Core Collapse, and Reactor N
Vessel Failure Models The IDCOR and BMI-2104 models-~of core slumping are greatly simplified repre-l 4
sentations of very complex processes.
Some key features of the models are:
(a)
IDCOR assumes that molten core material becomes isolated from the rest of the system until a slumping criterion is satisfied and then it instan-taneously slumps-to the lower core support plate (LCSP).
The NRC model treat's approximately the effect of in-core fuel' relocation 'and growth of the molten region.
l 1
(b) iDCOR models assu:ne' failure of the LCSP after a user-input fraction of the core is molten..NRC models c'alculate failure of lower core support structures due to heating by slumped core material.
(c)
IDCOR assumes immediate vessal failure after failure of LCSP.
The NRC model permits the user to select early local head failure or later gross l'
head overheating.
The condition of the core material (mass, composition, and temperature) at the time of' vessel failure can have a major influence I
on the subsequent loads on containment (steam spike, direct heating) and the extent of dispersal of the core debris.
e e
L n-,
-e
-n--
-y
--,,,m,--.-
,,n,
.e,,-.,--,.n-
-,v,,
na--,-----.-,nn,,---
,..,,w,
-,a,m
7-
.The best hope in resolving this issue is by detailed best-estimate calculations of melt-progression using the MELPROG code.
Since integral experiments in'this area are clearly impractical, the code models will have to be validated with data from the ACRR debris relocation experiments scheduled to be completed in I
FY 1985.
Issue #7 - Alpha Mode Containment Failure by In-Vessel Steam Explosions The major significance of this issue is whether sufficiently energetic molten
]
fuel / coolant interactions can occur in the lower plenum to produce a missile of sufficient energy to breach the containment (commonly referred to as the Alpha-mode failure).
IDCOR contends that this is not possible.
There have been many discussions by NRC staff, NRC consultants, and IDCOR staff on the validity of i
IDCOR's contention.
To date, however, the question has not been resolved to everyone's satisfaction except for a general feeling amongst all parties that
" explosions in excess of 2000 MJ would be required to fail containment.
Explosions of such high energy are deemed unlikely, but have not been demon-strated to be impossible."
4 Work is continuing to resolve this issue by experiments at SNL, in the UK, and by better, more-detailed calculations of the masses involved using the MELPROG code (see issue #6).
A report on this subject will be issued in March, 1985 by an NRC-sponsored expert review group (NUREG-1116).
III 'Ex-Vessel State Issue #8 - Direct Heatino of Containment by Ejected Core Material This issue is critical to containment failure timing for high-pressure sequences, such as TMLB',
in large-dry PWR containments.
In the IDCOR calculation of the TMLB' sequence for the Zion Plant, it is assumed that half e
b e
v y
,r~
m-
-,,---w, vn-m.,g-
~~-ow
,..w--,.--w-
.-.--.--------,~--.r---n-
-,,.v,,,
.rn,
,---r
-m,.
8 of the molten material ejected from the vessel under pressure is swept out of the reactor c'avity onto the containment floor.. The analysis does not account for potential rapid heating of the containment atmosphere by the core detiris or m
the oxidation of the core debris during transit and further heating of the atmosphere.
The potential for " direct heating" of the containment atmosphere is being investigated by the NRC in an experimental program which is underway at SNL.
Results will be available in FY 85.
The issue of direct heating is related to' Issue #2.
If recirculating flow patterns within the reactor coolant system result in alternative failure locations such as the hot ' leg, then the RCS will depressurize prior to melt-through of the lower head and the core debris will not be dispersed from the reactor cavity.
As stated before, COBRA, TRAC, and COMMIX calculations are being performed to study this possibility under the supervision of the CLWG.
Issue #9 - Ex-Vessel Fission Product Release A potentially significant modeling difference between the IDCOR model and the NRC models' (CORCON and VANESA) is related to the release of refractory fission products during the core / concrete interaction process.
The NRC models allow for the production of volatile oxides and hydroxides of these fission products by reaction with steam and carbon dioxides sparging through the melt.
Uncer-tainties,in the prediction of the ex-vessel release of fission products involve the composition (masses of materials and oxidation state) at the time of vessel failure, initial temperature of the melt, extent of core dispersal, modeling of core-concrete attack as well as the complex chemical behavior within the melt.
Experiments are planned at SNL in which simulant fission products will be included in core-concrete attack tests.
These data should provide a basis for testing CORCON and VANESA modeling assumptions by late 1985.
Issue #10 - Ex-vessel Heat Transfer Models from Molten Core to Concrete / Containment This issue is associated with the magnitudes and mechanisms of energy transfer from the molten core debris to the concrete, to the containment atmosphere, and
~
l
~
--.m
__m m. m.m..
9 to overlying water pools (if present).
The issue can potentially impact the mode and timing of containment failure and the chemical forms of fission products.
This is an area in which major differences exist in the modeling
~~77 assumptions used in the IDCOR and BMI-2104 analyses.
In general, the BMI-2104 analyses involve more heat going into concrete attack, more rapid production of
~
non-condensible gases, more rapid pressurization of the containment, but lower atmosphere temperatures, particularly in the BWR analyses.
These differences are partly the result of assumptions made regarding debris dispersal or spreading and, partly, differences in core-concrete attack models.
When water is present in the cavity, it is assumed in the IDCOR analyses that a coolable debris bed will form.
In the BMI-2104 analyses this possibility was treated parametrically.
Core-Concrete tests in the BETA facility and at Sandia (involving sustained urania/ concrete heating tests) in FY 85 will' provide an expanded data base for improving and validating models.
Issue #11 - Revaporization of Fission Products in the Upper Plenum Both the IDCOR and NRC models for in-vessel transport and deposition of fission products predict that a large fraction of fission products released from the core can be deposited on. structures in the reactor coolant system.
Associated with these fission products is a significant portion of the decay heat with the potential to result in direct heating of structures and the revaporization of deposited fission products.
In the most recent IDCOR analyses revaporization is taken into account resulting in some enhancement of the environmental source term for some reactor types and accident sequences.
A number of major uncer-tainties remain, however, including:
a very sparse data base related to the mechanisms of reaction between fission product species and surfaces (-and the potential for revaporization), the modeling of natural convection driven flow patterns in the reactor coolant system prior to, and subsequent to, lower head
-failure and the behavior of RCS insulation material in the accident environ-ment.
Coupling of the TRAP / MELT and MERGE codes has been achieved to permit analysis of the effects of revaporization during the period prior to vessel e
i
10 meltthrough and in a parametric manner after meltthrough.
A more rigorous treatment will be provided by the coupled RELAP5 and TRAP / MELT codes which is being undertaken for the SCDAP code package.
Results of analysis should~be available in FY 85.
~
Issue #12 - Deposition Model for Fission Products in Containment
.Because of the limitations in the applicability of the log-normal size distri-bution assumption for aerosols, which was incorporated in the RETAIN code, IDCOR has developed an empirical correlation for aerosol settling which has
-been included in the MAAP ' code.
This new model has also been criticized by NRC contractors for not including a dependence on particle size and the lack of wide-range comparisons to experimental data.
This issue will probably not be settled until direct comparison calculations are made for specific plant /
sequence events using both the NRC and IDCOR models.
Issue #13 - Amount and Timing of Suppression Pool Bypass Analyses of suppression pool scrubbing in the BMI-2104 study indicate that the effectiveness of suppression pools is quite good even when the temperature of the pool is at saturation.
In general, the extent of pool bypass was found to be a more important source of uncertainty than the details of suppression pool modeling.
In the IDCOR analyses suppression pool decontamination has been treated simplistically using constant decontamination factors and ignoring bypass under the assumption that bypass leakage would become plugged by aerosols.
The potential for bypass is quite plant-design dependent.
- However, since pool bypass will govern the magnitude of environmental releases in Mark III BWR designs, it should be explicitly considered in the analyses for this type of containment.
Issue #14 - Modeling of Emergency Response The predicted consequences of severe accidents can be very sensitive to the modeling of emergency response.
This is particularly true for early fatalities
11 because of their threshold nature and the high dependence of dose to proximity to the release point.
In the base case IDCOR analyses, the population in the evacuation zone is removed at a given rate without recognition ~of a straggler population that is slow to evacuate or refuses to evacuate.
In combination with long assumed warning times and the assumed mode of containment failure,
~
this assumption results in the prediction of no early fatalities, even for sequences in which on the order of ten percent of the volatile fission products are assumed to be released from the plant.
The modeling of emergency response is complicated by a number of factors including:
the human element (reluctance to leave home, panic, etc.), site dependence, and dependence on the weather and time of day.
Key sub-issues:
Evacuation Model; Definition of Warning Time; Containment Failure Mode.
Issue #15 - Containment Performance The potential magnitude of the source term to the environment is largely controlled by the mode and timing of containment failure.
If the containment remains intact for a number of hours following melting of the core and the release of fission products, the potent.ial consequences will be substantially reduced either through natural deposition processes or by the action of con-tainment safety features such as sprays, coolers, suppression pools, and ice condensors.
Over the past few years, the analyses of containment performance and model experiments at Sandia National Laboratories have verified the expec-tation that substantial margins exist between the design conditions and condi-tions at which major leakage of containment structures can be expected to occur.
Considerable uncertainty still exists, however, as to how rapidly leakage will grow a,s a function of pressure and temperature for different containment designs, the mode by which a containment will fail, and the location at which failure will be initiated.
In the IDCOR analyses, contain-ment failure is typically characterized by a leak rate sufficient to prevent e
l 12
+
1 further pressurization.
This assumption has the effect of increasing the residence time of fission. products in the containment and enhancing the effec-tiveness of evacuation.
In the BMI-2104 analyses,~ failure is typically charac-
'~
terized as a large hole that leads to depressurization of the containment, but does not involve total destruction of the building.
Sensitivity studies' have '
also been performed for leak rates that vary as a function of internal pressure.
The importance of better determining the conditions leading to containment failure is' plant-and sequ,ence-dependent.
Predicted consequences for a large, dry containment (such as Zion) can be insensitive to the pressure at which failure occurs, if the potential for early failure can be precluded.
In con-trast, the Mark I source term results are found to be very sensitive to the location of failure and the conditions leading to failure.
Experimental programs at Sandia will provide important data on the behavior of gasketed penetrations, electrical penetration assemblies, and the structural response of concrete containments. -These, combined with further plant-specific analyses, should serve to resolve the outstanding issues.
Issue #16 - Secondary Containment Performance The IDCOR analyses tend to give greater credit for secondary containment per-formance than the BMI-2104 analyses.
The differences are particularly evident in the IDCOR V sequences in which considerable deposition is predicted to occur in the auxiliary building even in the absence of water pool scrubbing.
It is also evident from ORNL SASA analyses that the reactor building surrounding the Mark I primary containment could significantly mitigate the consequences of severe accidents under a given set of conditions or assumptions.
The principal uncertainties regarding the effectiveness of secondary containment buildings relate to:
the mode of primary containment failure and its impact on the sur-vival of the secondary building; potential for hydrogen deflagrations to occur in the secondary building; and the modes of leakage and failure of the e
. _. ~.. - _, _
13 secondary building.
Secondary building performance is of particular interest in accident sequences involving early failure or bypass of the primary contain-ment envelope and, thus, has the potential to mitigate consequences.
The'
~.
uncertaintie's in secondary containment performance could be reduced by addi-tional analyses since the processes of interest are essentially the same as those in the primary containment for which methods have been developed.
To the extent that the mode of failure of the primary containment influences the per-formance of the secondary buildings, the uncertainty will always remain large.
Issue #17 - Hydrogen Ignition and Burning There are substantial differences between the IDCOR and NRC treatment of igni-tion, burning, and flame propagation in air-hydrogen-steam mixtures.
IDCOR analyses base ignition on a calculated flame temperature criterion which is a function of the composition of the atmosphere within the compartment.
NRC analyses, in addition to considering hydrogen and oxygen concentrations within a compartment, give explicit considerat. ion to steam inerting and the availa-bility of ignition sources.
The IDCOR models appear to predict continuous burning in essentially all cases, whereas the NRC treatment tends to predict a number of discrete burns.
The NRC's approach tends to allow the buildup of higher hydrogen concentrations and hence can lead to the prediction of higher containment pressures.
The differences between the IDCOR and NRC treatment of hydrogen ignition and burning are particularly pronounced in multi-compartment systems, such as the ice condenser containment, and in the absence of deli-berate ignition.
A number of related sub-issues are noted below, i
Effect of Natural Convection IDCOR models include consideration of natural convection driven flow between compartments which appears to lead to enhanced hydrogen burning at low concen-trations.
NRC models do not include consideration of natural convection flows.
Flame Propagation Due to a combination of differences noted elsewhere. NRC treatment of hydrogen combustion indicates greater likelihood of flame propagation into the upper b
e n.
,--n
--=r
--v-
,---r v.
14 compartment of the ice condenser containment where impact on containment pressurization is the greatest. Potentially Detonable Concentrations NRC's treatment of hydrogen combustion indicates the possibility of developing potentially detonable compositions in local areas, e.g., the upper plenum of the ice condenser.
IDCOR's treatment appears to preclude such localized hydrogen buildups.
Effect on Chemical Form of Fission Products There is experimental evidence that hydrogen combustion can alter the chemical form of airborne fission products, e.g., release of molecular iodine from cesium iodide aerosols due to hydrogen flames.
It is not clear whether such changes in chemical form increase or decrease the consequences of severe acci-dents.
Neither IDCOR nor the NRC analyses at present consider such changes in fission produ'ct chemistry.
Resolution of the outstanding issues will come, in part, from continued com-parisons between experimerits and analyses.
It must be recognized, however, that experimental data may not be available to address issues related to burning in complex geometries.
Issue #18 - Essential Equipment Performance Neither the BMI-2104 nor IDCOR analyses have provided a detailed assessment of the ability of essential equipment to survive the conditions associated with a severe accident environment (high temperature, high humidity, pressure differentials, flames, high radiation, and high aerosol loadings).
This equipment has been qualified to survive the LOCA environment which is in many ways similar.
However, experiments at Sandia have indicated that the more severe environment of core meltdown accidents could lead to the degradation of equipment, particularly cables.
Once substantial core degradation has occurred, protection of containment integrity becomes the key safety function.
e
15 Whether or not containment safety equipment is vulnerable to the severe acci-dent environment depends on the design of the plant.
Not only does the issue involve questions of the performance of the equipment, but also of the ability
~#~7 of the operator to monitor conditions in the containment and to control essential equipment.
Thus, the degradation of monitoring and control systems could also potentially degrade the effectiveness of the containment or accident management strategies to protect the public.
e e
m 9
_- -__~. _
l
(
Addressees - Memorandum dated G. P. Marino, RES i
D. F. Ross, RES F. Gillespie, RES T. Speis, NRR Z. Rosztoczy, NRR R. Vollmer, NRR M. Ernst, RES R. Bernero, NRR i
G. Arlotto,' RES
- 0. E. Bassett, RES R. Denning, BCL x
R. Barrett, NRR
\\
Cardis Allen, NRR
\\
[
J. T. Han, RES
\\
M. Silberberg, RES
//b I
-s J. Rosenthal, NRR T. Walker, RES u
- b C. Tinkler, NRR
\\.,b
\\
B. Burson, RES x
V. Noonan, NRR J. Cos'tello, RES W. Farmer, RES
\\
L. Soffer, RES
\\
J. Martin, RES
\\
P. Baranowsky, RES B. Agrawal,'RES C. Overbey, RES J. Murphy, RES M
G. Burdick, RES L. Chan, RES J. Telford, RES R. Barrett, NRR G. Bagchi, NRR R. O. Meyer, RES B. Sheron, NRR J. Glynn, RES R. Feit, RES J. Mitchell, ASTP0 P. Nyogi, RES J. Malaro, RES P. Worthington, RES T. Lee, RES W. Beach, RES L
an~
OBaHelle naumno t.au,r.a....,
- CMt he.mc
.., n,w,..... m,....
- i. a.......,. m : : :... :.
u 1. ::..:
May 7, 1985 Mr. Mel Silberberg U.S. Nuclear Regulatory Commission Willste Building 7915 Eastern Avenue Mail Stop 11305S Silver Spring, Maryland'20910
Dear Mel:
This is to convey my thoughts on the April 30, 1985, NRC/IDCOR meeting on technical issue resolution.
Upon some reflection on the technical issues that were discussed, it appears to me that a number of aspects may have been glossed over too quickly.
I will try to touch on what appear to me to be the more significant outstanding differences.
Corium-Concrete Interactions and Ex-Vessel Fission Product Release The NRC and IDCOR calculations of source terms to the environment differ appreciably in many cases, particularly with regard to the significance of the ex-vessel releases.
Whereas the NRC source terms are at times dominated by the l
ex-vessel release component, IDCOR generally predicts little or no contribution to the source terms from corium-concrete interactions.
The point was made at the subject meeting that the reasons for the differences were not due r.o the I
fission product release methodology, but due to differences in the treatment of l
corium-concrete interactions which drive the fission product releases.
This point is at least partially correct.
The results published by IDCOR were based on a methodology that differed considerably from that used by the NRC; it was pointed out at the meeting that the methodology is being further developed so that it will be more like the approach used by the NRC.
The treatment of corium-concrete interactions does vary substantially between the two groups.
Not only are there differences in the treatment of the interactions themselves, but there are also substantial differences in the boundary conditions that are typically assumed.
There may also be some questions of self-consistency in the individual analyses.
As you are well aware, the thermal hydraulic analyses in BMI-2104 are based on the use of MARCH 2 which incorporates the INTER subroutine for describing corium-concrete interactions; the ex-vessel fission product releases were evaluated by VANESA, with CORCa" M001 used as a driver for it.
The initial and boundary conditions for CORCON were provided by MARCH.
The use of one model of corium-concrete interactions in the thermal hydraulics and another for fission product release is recognized as undesirable, but at the time was unavoidable.
)
~~
E7 w -
i se
Mr. Mel Silberberg 2
May 7, 1985 i
IDCOR uses the DECG W subroutine in MAAP to describe corium-concrete interactions.
, t IDCOR went to great lengths to point out that in their analyses the heat transfer off the top of the debris is closely coupled to the rest of the containment through the treatment of natural convection among compartments.
The NRC analyses do not consider natural convection effects.
If, however, heat losses from the top of the debris are controlled by conduction through the t
crust, as is generally acknowledged, it is not at all clear that the inclusion of natural convection should have any appreciable effect.on the debris temperature. For example, where appropriate, the NRC analyses include heat transfer from the top of the debris to overlaying water without seeing any cverwhelming effect on the debris temperature.
It is difficult to see how natural convection to the containment atmosphere could provide greater heat losses than the boiling of a water pool.
It could be that such natural convection cooling manifests itself imediately after release of the debris before the crust has formed.
The analyses conducted by IDCOR generally assumed greater dispersal of the debris on the containment floor than did the NRC's BMI-2104. analyses; the latter assumed that the debris remained in the reactor cavity and/or the reactor pedestal.
Both of these are assumptions, and while plausible arguments can be made to support one or the other under'particular circumstances, these arguments are not necessarily compelling nor universally applicable.
Whatever assumptions-are made in a particular analysis regarding debris dispersal, it is important that subsequent events are treated consistently; it is not clear that this has been the case.
High primary systen pressure at the time of reactor vessel failure is probably the most compelling basis for assuming dispersal of the core debris out of the reactor cavity. This mechanism should, however, be invoked only for that portion of the core that is molten and leaves the vessel at high pressure.
In the IDCOR analyses the initial slumping.of the core into the vessel head leads to imediate failure of the vessel head; thus only the initially molten and slumped portion of the core would be subject to this high pressure dispersal.
The remainder of the core melts and is released from the vessel by gravity over a much longer period of time and thus would be much less likely to be dispersed.
Interestingly, in the BWR analyses melting of only twenty percent of the core is required to lead to initial fuel slumping with attendant vessel failure, with the majority of the core being released later over a much longer period of time.
Core dispersal could also take place in the absence of high primary system pressure just by gravity flow of the molten core debris.
For the debris to be very fluid high temperatures would be required.
In the MAAP analyses there is no interaction of the core debris with structural materials and little or no interaction of the slumped core material with the water in the bottom head; thus the temperture of the debris leaving the vessel could be essentially the same as the assumed melting point.
In MARCH analyses we generally have considerable interaction between the slumped core material and supporting structures as well cuenching of the debris by the water in the vessel head; at head failure the I
y.
-.-a ime-Nw"
' " ^ - " * - - - ' ~
l Mr. Mel Silb2rberg
~
3 May 7, 1985 debris temperature is generally between the melting point of the bottom head and i
the assumed liquidus temperature of the core materials. Thus typically the MARCH calculated debris temperatures are lower than those calculated by MAAP, though the sequences of events associated with head failure and debris release to the cavity are substantially different.
Another mechanism that could lead to debris dispersal out of the reactor cavity is debris-water interation. The likelihood of dispersal would depend on cavity configuration and the availability of suitable quantities of water; thus it would apply to only certain designs and specific accident sequences.
Since ex-vessel release of non-volatile fission products is sensitive to all models that affect the debris temperature, it is important that the debris temperature be determined and applied consistently.
Some observations regarding the IDCOR approach are noted below.
In the IDCOR analyses the core debris are delivered to the reactor cavity (or containment) piecemeal over a protracted period of time.
For the assumed delivery of the core debris over a protracted period of time, it is not at all j
clear that all the debris would be subject to the same dispersal mechanisms.
For example, the initial debris leaving the vessel may be subject to high pressure dispersal but later-melting portions of the core would not.
The initial mass of debris, which may be only twenty percent of the core in the case of BWR's, would be at a very high temperature, but would quench rapidly due to the assumed dispersal.
Subsequent portions of the core would leave the reactor vessel at a high temperature, but are apparently assumed to disperse and homogenize instantly with the debris already there.
Thus the assumed dispersal of the debris together with the assumed mixing of the " fresh" debris with that already there tends to minimize the time at temperature for the debris.
For purposes of the evaluation of fission product release during corium concrete interaction apparently the temperature of the debris crust rather than that of 4
the bulk debris is utilized.
In the NRC analyses all the debris are assumed to be released to the reactor cavity at the time of reactor vessel failure.
At the present time neither MARCH-INTER nor CORCON has the capability to accept time-dependent debris release from the reactor vessel.
While it would be desirable to have the
(
capability to treat time-dependent debris release from the r? actor vessel, it again should be recognized that the IDCOR approach is not necessarily any more correct than that used by the NRC. As indicated by some of the observations with regard to the IDCOR treatment of this process, time dependent debris release requires a rather complex treatment which may not be adequately addressed in MAAP.
There was the implication that release of all the debris to the cavity at once tended to maximize the debris temperature.
This implication is not necessarily true for several reasons.
First, it will be recalled that in the BMI-2104 analyses an effective liquidus temperature is util; zed for the core materials wnich is lower than the melting point of the uranium dioxide fuel.
Second, in the MARCH calculations structural materials in the core, the lower core support structures, as well as the molten portions of the reactor vessel heat are typically added to the core debris leaving the vessel; these structural
Mr. Mel Silberberg 4
May 7, 1985 materials tend to lower the temperature of the debris.
In the IDCOR analyses only the uranium dioxide fuel and the cladding are included in the debris. The additional structural materials could have their own effect on the subsequent behavior. The latter point was not brought out during the discussions at the subject meeting.
There is an aspect of CORCON that may tend to keep the debris temperature elevated, but it applies primarily to the long-term behavior of the debris.
CORCON does not model conduction heat losses from the melt to the concrete; it considers only the ablation of concrete at an input ablation temperature.
For high debris temperatures and resultant high concrete ablation rates the neglect of conduction ahead of the ablation front is a good approximation.
As the debris cools and its temperature approaches the input ablation temperature, however, there is no way to transfer heat from the debris to the concrete except by keeping the debris temperature above the ablation temperature.
Thus, in the long ters, the current CORCON model may be keeping the debris temperatures higher than would be realistically expected; this, in turn, could be contributing to the sustained high release rates of certain species as calculated by VANESA.
As was noted previously, in the BMI-2104 analyses the core debris leaving the, reactor vessel were assumed to be confined to the reactor cavity and/or reactor i
pedestal.
Dispersal of the core debris'over a larger area can be easily modeled by code input, if it should be desired to do so.
i Hydrogen Ignition and Burning As was duly noted in the discussions at the meeting, there are rather substantial differences in the perceived significance of hydrogen ignition and burning as predicted by the NRC and IDCOR analyses.
This, of course, was not a new revelation but was clear from earlier interactions.
Clearly the differences in the predicted extent of in-vessel hydrogen generation have an obvious influence on the prediction of subsequent burning in the containment.
My views on the treatments of in-vessel hydrogen generation by the two groups were communicated to you earlier.
In what follows I will try to address what I consider to be the key points related to hydrogen behavior in the containment.
The BMI-2104 analyses of hydrogen ignition and burning are based on the MARCH treatment of these phenomena.
While certain aspects of the MARCH treatment can l
be questioned, particularly its application to complex multicompartment geometries, the general features of the MARCH predictions have been corroborated by similar HECTR analyses.
Such corroboration has been shown in independent analyses as well as in the context of the Containment Loads Working Group (CLWG) where both codes were applied to the same " standard problem".
The ignition and i
burning models in MARCH are very similar to those in HECTR and, in fact, have the same origin; the differences come in the treatment of intercompartment transfers and in the modeling of the ice condenser.
Both MARCH and HECTR are lumped parameter models utilizing well-mixed assumptions within each compartment of the containment; the predictions can be sensitive to the containment
,-,a-.
,,--,-a
__-,n-n..n-,,wn,-,-...-<,,,_,,-..,
Mr. Mel Silberberg 5
May 7, 1985 compartmentalization utilized.
Ignition of hydrogen is typically subject ~to the 7
achievement of an input hydrogen concentration, the availability of oxygen, and a consideration of inerting by diluents.
Burn velocities, completion of burning, as well as flame propagation between compartments are typically calculated by the code, but can also be user specified by input.
The presence of igniters can be explicitly modeled, with the hydrogen concentration for ignition user specified.
In the absence of igniters hydrogen burning can be assumed to take place on time, hydrogen concentration, or certain stage in the accident sequence.
Continuous hydrogen burning can also be modeled but is normally subject to the usual considerations of oxygen availability and inerting by diluents. Combustion of carbon monoxide as well as hydrogen can be considered.
Default parameters are provided in the codes that are based on current understanding of ttfe phenomena, but these can generally be overridden by user input if desired.
IDCOR uses the so-called flame temperature criterion to assess flamability of i
hydrogen-air-steam mixtures.
If the calculated flame temperature of the mixture in question exceeds the flame temperature criterion, the mixture is said to be flammable and ignition is apparently assumed. The use of the flame temperature criterion may be important for cases in which an explicit ignition source is not present; it may not be particularly important in the presence of igniters.
The IDCOR presentation stressed the following differences between their and the NRC's treatnient of hydrogen burning: consideration of natural circulation within the containment, consideration of incomplete combustion, and consideration of hydrogen-oxygen recombination.
The NRC approaches also include consideration of incomplete combustion where appropriate; frequently, however, the hydrogen concentrations are predicted to rise to levels where complete combustion is reasonably assured.
Typically the required concentration for complete combustion is taken as eight volume percent hydrogen.
Thus use of incomplete combustion modeling does not appear to be a real issue, although the specific assumptions regarding completeness of combustion as a function of hydrogen concentration may be different.
The consideration of natural convection flows between containment compartments is a real difference between the IDCOR and NRC analyses. At present none of the NRC's severe accident analysis tools can conveniently accomodate natural i
convection effects.
Such considerations may be amenable to analysis by detailed stand-alone codes, but such an approach would not necessarily be adequate since it would not couple natural convection to all the other accident processes.
Without corresponding tools and in the absence of hands-on experience with the MAAP code, it is difficult to comment on the validity of the IDCOR conclusions i
regarding the effectiveness of natural convection in promoting continuous burning of hydrogen.
This is an area that certainly merits further consideration.
IDCOR's treatment of hydrogen behavior was stated to include hydrogen-oxygen recombination under conditions where normal combustion might be precluded; i.e.,
1 in the presence of high steam partial pressures, but with a high temperature heat source such as the core debris.
The latter type of recombination is quite plausible and should indeed be considered in the analysis.
The NRC analyses to ny knowledge have not explicitly considered such recombination, but its
,m-~
.o Mr. Mel Silberbarg 6
May 7, 1985 possibility has been recognized, e.g., in the context of the direct heating-issue.
It may be noted that such recombination should also be considered if the
~
hot core debris are assumed to be dispersed out of the reactor cavity; the latter could lead to increased containment loadings in such sequences as large dry containment TMLB that are not subject to normal combustion.
Additionally, IDCOR apparently considers hydrogen burning if the hydrogen is released to the containment at temperatures above the spontaneous ignition limit.
This again may be quite plausible under the right circumstances, but it is not altogether clear how some of the details of the combustion should be treated.
For example, is the recombination confined only to the very hot cases, will the combustion propagate to cooler regions, how does the oxygen get to the point of release of the hot hydrogen if a sustained flame is assumed? MARCH 2 is presently not set up to explicitly address hydrogen-oxygen recombination due to the presence of hot core debris or spontaneous combustion of hot hydrogen.
These effects can, however, be at least partially explored by appropriate input modifications to the default hydrogen burn parameters in MARCH.
There are a few additional questions that may be quite significant but which were not explicitly addressed at the meeting.
It is not at all clear what burn velocities or correlations are assumed by IDCOR.
The NRC treatments incorporate a correlation of burn velocity as a function of hydrogen concentration that yields burn velocities that are considerably higher than published laminar flame speeds.
Also, it was not clear if the IDCOR analyses explicitly considered burn propagation between containment compartments, and if so, what criteria were utilized.
The NRC's approach bases burn propagation on the relative orientations of the compartments, hydrogen concentrations, local burn velocities, and characteristic dimensions of the compartments.
l Revaporization of Fission Products in'the Upper Plenum A few observations on the issue of revaporization of fission products in the upper plenum may be appropriate to ensure that we all understand the issue in the same way.
The primary system fission product transport and deposition analyses in BMI-2104 did consider revaporization of fission products as the surfaces heated up.
What BMI-2104 did not consider was the effect on surface temperatures, and hence on revaporization, of the decay heating of the deposited fission products.
Some very early scoping analyses performed in conjunction with Volume I of BMI-2104 indicated that.the effect of such fission product heating should be minimal in the time frame up to vessel head failure, but could be quite significant in the longer term.
Extrapolation of calculated temperature histories indicated that at least some structures would reach very high (melting) temperatures. These early impressions have been basically confirmed by the limited analyses that we have been able to perform so far with the coupled MERGE and TRAP / MELT codes.
The latter explicitly take into account the effect of the heating of structures by the deposited fission products.
The combined MERGE and TRAP / MELT codes are still limited to the time frame up to vessel failure.
The IDCOR analyses generally predict substantial fission product deposition
[
uithin the primary system. They have explicitly modeled the effects of fission
r Mr. Mel Silberberg 7
May 7, 1985 product heating on structures and apparently predict significant redistribution e of deposited fission products within the primary systems as a result of this
~-
effect.
They also predict some rerelease of these deposited fission products late in the accident sequences, though the magnitude of these rereleases is apparently not large.
Apparently one of the reasons why these rereleases are not large is because they calculate large heat losses from the primary system and an eventual balance between such. heat losses and the heating by the deposited fission products.
The NRC analyses have not considered the effects of such possible heat losses and at this time it is not clear whether the IDCOR conclusions are realistic or not. As you may recall from the discussions, IDCOR placed substantial emphasis on these heat losses and alluded to data that indicated that the losses may even be greater than they had initially assessed.
One key aspect of the assessment of these losses is the degradation of the primary system insulation during the course of the accident sequences. Another important aspect is the effectiveness of the entire primary system as a heat sink for the deposited decay heating.
In sumary, upon reflection on the discussions at the subject meeting, it appeared to me that some of the differences between the NRC and IDCOR analyses were not adequately recognized.
I hope that the above discussion will be useful to the resolution of the outstanding issues.
sinc
.f.,
" 'O f' '
-Pe er Cybulskis PC/sm 9
j.,
- -. - -