ML20211G271
| ML20211G271 | |
| Person / Time | |
|---|---|
| Issue date: | 12/31/1986 |
| From: | Lee T NRC |
| To: | |
| Shared Package | |
| ML20211F389 | List:
|
| References | |
| FOIA-87-8 NUDOCS 8702250328 | |
| Download: ML20211G271 (32) | |
Text
.t..
i UNCERTAINTY PAPER on HIGH PRESSURE MELT EJECTION (DirectContainmentHeating) by Tim M. Lee December 1986 f* "'*%,
E S
i ti s'...../
U.S. Nuclear Regulatory Comission Washington, DC 20555 N
07022gOS2e07020 PDR e&c<o A" gA2 e7-s
HIGH PRESSURE MELT EJECTION (DirectCentainmentHeating)
Tim M. Lee I.
INTRODUCTION t
I.A Definition of Issue _
In certain reactor accidents, such as those initiated by station blackout or a small break LOCA, degradation of the reactor core can take place while the RCS remains pressurized.
Left unmitigated, core melt will slump and collect at the bottom of the reactor vessel.
After boiling off the remaining water in the vessel, molten core materials will start attacking the bottom head cf the reactor. When the bottom head of the reactor vessel is breached in such accidents the core melt will be ejected under pressure. The ejected materials are likely to be dispersed out of the reactor cavity into surrounding containment volumes as fine particles, quickly transferring thermal energy to the containment atmosphere.
In addition, metallic components of the sprayed core debris, mostly zirconium and steel, can react with cxygen and steam in the atmosphere generating a large quantity of chemical energy, heating and pressurizing the containment further.
The term " direct containment heating" (DCH) is used in the present discussion to describe this complicated physical and chemical process.
1 Simple analyses of the cor.tainnent heat balance indicate that even a large dry containment of a Pk'R plant can be pressurized beyond its ultimate strength if a significant fraction of the core materials participate in DCH.
The peal: con-tainment pressure is normally attained within seconds after the melt ejection.
A large amount of aerosols, including refractory fission products, could be generated in a high pressure melt ejection.
If the containment should fail from the DCH loading, a massive release of radioactive materials could result.
12/19/86 1
UNCERTAINTY PAPER
Dispersing core debris could induce other hazards.
If hyorogen existed in the containment atmosphere, dispersing hot debris particles could serve as a cata-lyst to promote recombination of the hydrogen with free oxygen even though the H concentration may be below the conventional flamable limit.
Hydrogen 2
recombination will generate more energy to raise the pressure and the temperature in the containment.
The issue would be further complicated if the reactor cavity is filled with water at the time of the RPV failure.
The pressurized stream of molten core materials is likely to cause a steam explosion that may contribute to debris fragmentation and promote debris dispersion at the same time generating dynamic loading on the containment.
Recently, both Brookhaven National Laboratories (BNL) and Sandia National Labu-ratories (SNL) predicted in their analyses that metallic components in the melt will be completely oxidized by steam in the reactor cavity region during high pressure melt ejection (HPME).
Such reactions would generate a large quantity of hydrogen that can readily mix in the containment atmosphere regardless of debris transport.. Burning of this hydrogen could challenge the containment integrity.
Other potential hazards associated with HPhE/DCH that merit further investiga-tion are containment liner abrasion, effects of high temperature on containment structure and equipment, and possible missile generation.
The risk of DCH is likely to be significantly different for BWR plants because the Automatic Depressurization System may be used to depressurize the RCS. A large body of water in the suppression pool is believed by many to be able to moderate the effect of DCH.
On the other hand the containment volume is much smaller for the BWR plants so that much less corium is needed to prersurize and fail the containment. A BWR generally has a larger core and a higher zirconium content that favor DCH.
The reactor containment of ice condenser plants has a smaller volume and a lower design pressure than a typical large dry containment of PWR plants.
Although it is generally recognized that ice chambers in the ice condenser plants are likely to trap and cuench the bulk of the dispersing debris, hydrogen generated in the metal-steam reactions discussed above should be able e
4 12/17/86 2
TECH UNCERT RPT/ LEE
to migrate through the ice chambers easily.
Burning of such hydrogen in the ice chambers or in the upper co51partment could fail the containment.
I.B Important Subissues and Relatea Uncertainties The ultimate concerns of the DCH issue is whether the containment could fail and, if it could, what is the amount of airborne radioactive materials at the time of the containment failure. The consequences of DCH depends on a host of subissues that are discussed below.
I.B.1 Initial Conditions The severity of DCH is highly dependent on the conditions inside the RPV at the time of the vessel failure.
These conditions are determined by the sequence and progression of the accident and must be provided by sources outside of HPME/DCH programs (such as SASA and MELPROG programs).
a.
Reactor Coolant System (RCS) Pressure The RCS pressure provides the motive force for melt ejection and debris dispersion, and affects the extent of disintegrat' ion and atomization of the melt jet. The RCS at a higher pressure generally stores more mass and energy that will be released to pressurize the containment in the event of l
an RPV failure.
A higher RCS pressure will result in higher DCH.
The RCS pressure is likely to remain near the PORV set point for a station blackout accident in a PWR but should be at a lower level that is determined by the break size in a small break LOCA.
The accumulator set point is the likely lower bound for this parameter. Once the RCS can be depressurized below this point, at which a large volume of water is available for flashing, the re ntor pressure should fall very rapidly.
The range of this parameter is therefore 600-2400 psig.
1 Recently, the possibility that natural circulation inside the RCS may l
induce a failure elsewhere on the primary boundary is being investigated.
If the failure size is sufficient to depressurize the RCS before the core melt breaches the RPV, HPME, and consequently DCH, will not take place.
1 12/17/86 3
TECH UNCERT RPT/ LEE
s
+
p Preliminary findings of the investigation indicate that such failure is possible. These works, however, have not been subjected to an intensive peer review.
It may be of interest to note that natural circulation was not in evidence during the core melt accident at TMI-2.
5 b.
Melt Temperature
~
~
-*4 The melt temperature will determine the amount of thermal energy carried '
into the containment by core debris.
It also affects transport properties" of the melt, such as surface tension, viscosity, etc., that control the size of debris particles.
Finer particles favor heat transfer and chemi-cal reactions because of a higher surface-to-volume ratio.
The melt temperature could range from the melting point of steel (1800*K) to that
~~
of uranium oxide (3100*K).
A higher melt temperature will result in higher DCH.
c.
Fraction of Core Melted and Ejected The larger the amount of ejected core materials, the more severe DCH will be. A large uncertainty exists regarding the fraction of the core th'at became molten at the time of the melt through.
It is believed that this fraction is in the range of 20-80% of the reactor core.
d.
Metallic Components in the Melt Portions of metals (Zr and Fe) in the melt are likely to have reacted with steam inside the vessel prior to the ejection.
But a substantial amount of them could remain in metallic forms and be oxidized in the containment atmosphere.
The higher the metal contents in the melt the more chemical energy is available for DCH. At this moment,<it is believed that anywhere between 20-80% of Zr and Fe in the melt could remain unoxidized at the time of the vessel melt-through.
Results of recent NRU tests seem to suggest that in-vessel metal oxidation may be very extensive for some accident conditions.
12/17/86 4
TECH UNCERT RPT/ LEE
.=
e.
Dissolved Gas and Steam Hydrogen and steam can be dissolved in the core melt as the accident pro-gresses. When the melt is ejected, the dissolved gas and steam will boil out, contributing to disintegration of the liquid jet and atomization of the melt.
A higher content of dissolved gas and steam tends to break up the melt into finer particles that favor DCH and contribute to aerosol generation.
Test data are scarce in the range of the reactor accident conditions.
f.
Mode of Vessel Failure It is assumed in most DCH studies that an instrument tube will fail at its weld to initiate the high pressure melt ejection.
There is, however, a controversy as to the possibility of multiple instrument tube failure.
Even a circumferential break of RPV is not ruled out.
Results of High Pressure Streaming (HIPS) tests at SNL also confirmed the Zion Pro-babilistic Safety Study's prediction that the aperture will ablate and grow in size during the melt ejection.
The rate of hole ablation can be predicted fairly accurately, but the effect of such ablation on debris dispersal and DCH is difficult to quantify.
Generally, a larger flow area will increase melt and gas discharge rates, and that tends to favor core debris dispersion and DCH.
I.B.2 Effects of Water It is generally believed that core melt ejected from the RPV can be quenched to moderate or mitigate DCH if water is available.
Results of High Pressure Streaming (HIPS) tests at SNL seem to indicate that water in the reactor cavity may not be as effective in quenching the melt as previously believed.
In the two HIPS tests (HIPS-4W,' -6W) conducted with a water-filled test cavity, a sharp pressure spike was produced in both tests immediately following the. melt ejection.
High speed movies clearly showed that water was pushed out of the cavity by the pressure spike ahead of the dispersing core debris.
The level of water in the reactor cavity can be determined by the analysis of a specific accident sequence for a given plant so that uncertainties regarding the 12/17/86 5
TECH UNCERT RPT/ LEE
e 8
availability of water is small; but the effect of the water on DCH is difficult to determine with the present st. ate of understanding. Another effect of the water that merits further investigation is the possibility that ex-vessel steam explosions may promote fragmentation and dispersion of the core debris thereby enhancing DCH.
I.B.3 Debris Transport e
a.
Mechanism of Debris Dispersal Results of High Pressure Streaming tests (HIPS) at SNL suggest that entrainment of melt particles in a high velocity gas flow may be the dominant mechanism to disperse core debris out of the reactor cavity region.
The situation, however, is very much different from the film entrainment in which liquid particles are' stripped from the surface of a liquid pool or film.
Disintegration and splashing on the cavity boundary of the melt jet could have created airborne melt particles that are ready to be carried forward in a gas stream, b.
Melt-Structure Interaction Trapping of core debris particles by structures in and around the reactor cavity is generally considered to have mitigating effect on DCH.
Recent tests at SNL (HIPS-7C and -8C), however, raised doubts about the effec-tiveness of shielding against debris dispersion provided by the structure.
In HIPS-7C, the addition of a semi-enclosure at the exit of the cavity keyway that simulates the instrument shaft at the Zion plant did not seem to appreciably reduce dispersal of core debris out of the test cavity.
In HIPS-8C, 25-30% of ejected materials was found dispersed up through a gap that simulates the annular gap around the RPV of the Zion plant.
It is believed that melt particles impacted on a concrete surface will splash and bounce right back into the stream and be carried away if the local flow velocity is sufficiently high.
When high temperature, high speed melt particles impact on a thin steel surface the steel surface is likely to be ablated or even penetrated.
Melt particles that are intercepted by heavy steel structure could freeze on the surface of the steel.
Thickness 12/17/86 6
TECH UNCERT RPT/ LEE A
of the frozen layer is likely to be limited by the heat transfer at the interface.
Frozen debris particles may behave differently.
c.
Mixing of Debris Particles in the Containment Atmosphere The extent of containment pressurization by DCH depends on how well the ejected core materials mix with the containment atmosphere.
The better the mixing, the more complete themal and chemical interactions will be.
The chemical reaction and heat transfer are most active when the debris particles are suspended in the air.
Interaction with structure and equip-ment inside containment may affect the airborne time of the debris particles thereby affecting the extent of DCH.
d.
Ex-Vessel Metal-Steam Reactions Hydrogen generated in the ex-vessel metal-steam reactions discussed earlier can migrate relatively freely in the containment. Analyses showed that burning of such hydrogen could fail the containment under certain conditions.
If confirmed by experiments, this consideration could change the perception regarding the mitigating effect of the structure.
I.B.4 Containment Atmosphere Composition The atmospheric composition in the containment dictates chemical reactions the debris particles will be undergoing, l
If the atmosphere is inerted by nitrogen (such as BWR, Mark I containment),
oxidation of metals will not take place.
Only the themal energy of the melt will contribute to DCH.
In the presence of steam, metal-water (steam) reactions will take place.
i Metal-water reactions take place at higher temperatures than oxidation in air and generate less energy, although hydrogen generated in metal-water reactions can later recombine to generate additional heat if oxygen is available.
Steam concentration could range from that in the containment atmosphere during normal operation to 100% steam which may be expected in the lower compartment of an l
12/17/86 7
TECH UNCERT RPT/ LEE
o
\\
ice condenser when it is completely purged by blowdown steam in certain acci-dent conditions.
l Possible recombination of pre-existing hydrogen by hot core debris sprayed into the containment atmosphere was discussed earlier.
Hydrogen accumulated in the RCS will be added to the containment atmosphere when the RPV fails.
Heat gen-erated in burning such hydrogen will further pressurize the containment.
I.B.5 Aerosol Generation In SNL's System Pressure Injection Tests (SPIT) and HIPS tests, an intense cloud of aerosols was observed following each high pressure melt ejection.
Data collected from SPIT-18 and -19 suggest that aerosols generated in a pressurized melt ejection could amount to 1-5% of the ejected mass.
SPIT test samples indicate that aerosol sizes are bimodal:
a fraction of a micron and several microns.
It is believed that these aerosols are formed by two different mechanisms; condensation of vaporized materials and mechanical breakup of melt particles.
Consequently, radioactive materials contained in these aerosols could include refractory as well as volatile fission products.
I.C Review of Current Modeling The Source Term Code Package does not include a containment code that calcu-lates DCH.
IDCOR has been taking a position that DCH will not take place because they think structures around the reactor cavity would confine dispersal of core debris.
Any debris dispersed out of the reactor cavity were assumed to be quenched by water.
It is not known whether IDCOR has a containment code that can calculate DCH. They were invited, but did not participate in DCH test calculations.
In the containment Loads Working Group's standard problem exercises, all l
participants used a single node assumption and calculated mass and energy balances.
Chemical reactions were assumed to be complete, thermal equilibrium l
attained, and heat losses were neglected.
When quenching of the core debris by water was considered, it was simply assumed that thermal energy of a portion of 12/17/86 8
TECH UNCERT RPT/ LEE
the core melt was expended in generating steam.
Chemical energy in the quenched portion of the core meit was ignored.
Recently, several containment codes have adopted rate equations in otherwise a lumped-parameter approach to calculate chemical reaction and heat transfer more realistically.
There, however, is no existing code that can calculate core debris transport. The prospect for these codes to acquire capabilities to calculate rates of debris entrainment and dispersal, particle sizes, airborne time, etc., does not appear promising.
However, the metal-steam reactions discussed above may moderate the need for such modeling.
6 l
I 12/17/86 9
TECH UNCERT RPT/ LEE
O b.
E II. DESCRIPTION OF PAST, PRESENT AND FUTURE RESEARCH II.A Past Research The possibility that core debris could be swept out of the reactor cavity in a high pressure ejection of core melt from the RPV was first recognized in the Zion Probabilistic Safety Study (ZPSS). While conducting System Pressure Injection Tests (SPIT,1:20 linear scale) to confirm the debris dispersal, SNL realized the possibility that thermal and chemical energy of the dispersed core debris may quickly heat and pressurize the containment atmosphere.
SNL continued with larger scaled High Pressure Streaming tests (HIPS, 1:10 linear scale) to characterize core debris dispersal.
SPIT-18 and -19 were conducted with an interaction chamber while all other tests were conducted outdoors. ANL has conducted 1:30 scale tests under EPRI's contract independent of SNL's activities. ANL test results were included in IDCOR Technical Report 85.2.
They form the basis for IDCOR's position.
II.A.1 System Pressure Injection Tests (SPIT)
Nineteen tests were conducted in this series; but only the last two provided useful quantitative data.
In SPIT-18 and -19, iron-aluminum thermite was used to simulate the core melt.
A concrete cavity simulating 1:20 linear scale of the Zion reactor cavity was used in SPIT-19 while an alumira cavity was used as the test cavity in SPIT-18.
The test apparatus was placed in a makeshift interaction chamber that was of sheet metal construction with an estimated pressure rating of 3 psig.
The melt generator was pressurized to 1500 psig by nitrogen gas in both tests.
95% of the ejected melt was dispersed out of the test cavity in SPIT-19 while only 58% was dispersed in SPIT-18.
The lower dispersal in SPIT-18 was attrib-utable to higher heat loss, and consequently higher freeze-up of the debris, in the alumina cavity.
Measured pressure rises of 3.5 and 2 psi were not consid-ered meaningful because the interaction chamber suffered substantial damages-in 1
12/17/86 10 TECH UNCERT RPT/ LEE
[.'
both tests that resulted in large leakage.
In SPIT-18, several anchor bolts were " pulled" through a 5 milliineter thick steel plate causing gaps of 4 to 8 centimeters between the shell of the interaction chamber and its foundation.
It was estimated that I to 5% of the ejected melt was aerosolized.
II.A.2 High Pressure Melt Streaming (HIPS) Test The HIPS apparatus consists of a melt generator, a concre'te test article that is a 1:10 linear scale model of the Zion reactor cavity.
The experiment used a mixture of iron oxide and aluminum pofer to produce a melt by thermitic reaction to simulate molten corium in a core melt accident.
The melt generator was pressurized to a level ranging from 3.3 MPa (480 psia) to 11 MPa (1600 psia) by either nitrogen or carbon dioxide.
The HIPS tests concentrated on investigation of factors contributing to the mechanism of debris dispersal over a range of conditions.
A total of 8 tests were conducted in the HIPS series.
HIPS-4W and -6W were conducted with a water-filled test cavity while in HIPS-7C, structure was added at the exit of the test cavity to simulate the instrument shaft and the seal table areas of the Zion plant.
In HIPS-8C, the annual gap around the reactor vessel of Zion 1 was simulated to investigate the potential for core debris dispersal directly into the upper containment dome via this gap.
In all tests, practically all molten materials ejected from the melt generator were dispersed out of the test cavity even at a pressure as low as 3.3 Mpa (480 psia),
The presence of water in the test cavity did not appreciably affect the fraction of dispersal.
High speed movies shcwed that a slug of water was expelled ahead of dispersing debris in both HIPS-4W and -6W.
A pres-sure spike in the order of a thousand psi was measured in the cavity following the melt ejection.
The reinforced concrete test cavity was destroyed in both tests and in HIPS-6W the whole test facility was lifted 6 feet in the air.
No appreciable retention of debris was observed in HIPS-7C where the additional structure was expected to trap dispersing debris.
In HIPS-8C, approximately one-third of dispersed debris escaped through the annual gap.
This fraction is 12/17/86 11_
TECH UNCERT RPT/ LEE
1 o
roughly ecual to the ratio of the flow area thrcugh the gap to the total cut-flow area, supporting the theory that the predominant mechanism for debris dis-persion is entrainment by high velocity gas flows.
A large cloud of aerosols was observed in each HIPS tests.
II.A.3 ANL Experiments The Ahl apparatus includes a thermite vessel representing the RPV, an interaction chamber to represent a reactor cavity, a pipeway simulating an instrument tunnel and an expansion vessel simulating the containment. The interaction chamber, the pipeway and the expansion chamber are all of steel construction.
The experiment used materials composed of uranium dioxide, zirconium dioxide and stainless steel to simulate the core melt.
The molten materials were ejected into the interaction chamber by gas that was pressurized to a level ranging from 0.21 MPa (30.5 psia) to 5.7 MPa (826.5 psia).
The scale of this facility is approximately 1:30 linear scale of the Zion' plant configuration, but no attempt was made to maintain geometric similarity.
ANL also conducted a separate series of tests using Wood's metal as the melt simulant, to investigate the influence of the containment configuration cutside the reactor cavity on the core debris dispersion.
Wood's metal has a melting point of the 73 C.
The injection pressure used in this series of tests ranged from 0.25 MPa (36 psia) to 1.4 MPa (200 psia).
The corium tests showed sweepout fraction of 1 to 60% and a peak pressure rise from0.12MPa(17.5 psi)to0.38MPa(55 psi).
The temperature rise ranged from -4 to 50*C.
These results do not include a contribution from oxidation of metallic components as the tests were conducted in an inerted expansion chamber atmosphere. The Wood's metal tests showed that the fraction dispersed out of I
the test cavity ranged from 10 to 90%.
II.A.4 Discussion of Experimental Results Several considerations pertinent to the application of the test results are discussed below:
12/17/86 12 TECH UNCERT RPT/ LEE
7 a.
Scaling i
Generally speaking, any test results are valid only for the conditions under which the test is conducted.
Extrapolation of test results is pos-sible only to cases eith physical similarities.
It is imperative to con-duct a scaling study if such extrapolation is contemplated.
SNL has provided a scaling study in their HIPS Program Plan supplemented by calculations matching the Kutateladze number with postulated accident conditions in the Zion plant to preserve similarity in entrainment and dispersion of core debris.
IDCOR Technical Report 85.2 provided no scaling study that could provide guidance for application of ANL test results to a reactor accident test conditions.
Recently, BNL has conducted a scaling study and grouped variables that affect core debris dispersal into 6 dimensionless parameters.
Theoreti-cally, if a test can be designed that matches values of all six parameters to the reactor accident conditions, results of'the test should be directly applicable to the reactor accident evaluation.
In practice, matching more than one dimensionless parameter in a test is very difficult, if possible at all.
Moreover, BNL's scaling concerns only hydrodynamics of the debris dispersal; it did not consider chemical reactions and heat transfer.
It did not even include the thermal effect on a compressible fluid flow, b.
Scale of the Test Facility l
HIPS attempted to simulate 1:10 linear scale of the Zion Plant configur-ation, and SPIT 1:20 linear scale.
ANL's facility is approximately 1:30 linear scale of Zion.
It is generally recognized that the surface-to-volume ratio increases as the scale of the test facility decreases (in 1:10 scale, ten times higher; 1-30 scale, 30 times, etc.).
This increases l
the heat loss that promotes freeze-up of melt particles.
Debris retention in a small scale test facility will, therefore, be increased and the fraction of dispersal decreased.
The effect of disproportionately higher heat loss and debris retention should have been intensified by all steel construction of ANL's test facility that is not prototypical of commercial l
12/17/86 13 TECH UNCERT RPT/ LEE
'i*.i reactor plants.
The increased surface-to-volume ratio in the small scale facility also increases the flow resistance and decreases the duration of blowdown from the RPV; both contr-ating to further reduce the debris dispersal.
For the same reason, test results are also believed to under-predict, albeit to a ' 3ser degree, the core debris dispersion.
It should be noted that, in a small scale facility, the dispersing debris will have a shorter "mean flight path." Airborne time, during which the debris undergoes vigorous chemical and thermal interactions with the atmo-sphere, is reduced accordingly.
c.
Test Pressure Entrainment of debris particles in a high velocity gas stream appears to be the predominant mechanism for the dispersal of core debris out of the reactor cavity. The extent of core debris dispersion, therefore, depends on the momentum flux of the gas flow which, in turn, depends on the pres-sure in the melt vessel.
From the above discussion on the. scale of the test facility, it is apparent that the test pressure should be slightly more than that in the RPV the test is simulating to provide an equivalent condition for entrainment and debris dispersion. Test pressures for SPIT and earlier HIPS at SNL are in the range of the RCS pressure discussed in Section I.B.a.1.
Later HIPS test pressures skirted the lower end of this range. All of ANL's tests, except two corium tests, were conducted with injection pressures substantially below the lower bound (600 psig) of the above mentioned RCS pressure range; some at a pressure as low as I bar above atmospheric pressure, d.
Effect of Structure It is recognized that structures around the reactor cavity could affect the core debris dispersal; the question is "how?"
In addition to the smaller scale and lower test pressures discussed above, ANL's test configuration also lacks geometric similarity with any operating commercial reactor plant. The flow field in and around the Interaction Chamber and Pipeway in ANL's tests, therefore, is not similar to that 12/17/86 14 TECH UNCERT RPT/ LEE
o' expected around the reactor cavity of a nuclear plant during the j
postulated reactor accident.
Consequently, it is difficult to relate these test results to a reactor accident.
SNL has conducted HIPS-7C with a scaled concrete structure simulating the instrument shaft and the seal table at the exit from the Zion reactor cavity.
HIPS-7C yielded no measurable difference in the fraction of debris dispersal.
The situation could be significantly different in the presence of heavy metal pieces such as structural steel or equipment.
Substantial freeze-up of the melt particles may be possible on the surfaces of these pieces, especially if they are located in areas of considerable flow deceleration.
It must be cautioned that the problem associated with the higher surface-to-volume ratio that is inherent in a small scale facility will be worsened with additional structures. Some provision to compensate for this effect is needed in the design of tests to investigate the trapping effect of the structure.
e.
Effect of Water In both SNL's tests and ANL's corium tests that were conducted with a water-filled test cavity, the water was observed to be ejected out of the cavity as a slug ahead of the dispersing debris.
Complete mixing of debris with water, therefore, was not observed in the test cavity and in flight. Quenching of debris by the water in the test cavity appeared much less than that assumed in many analyses.
SNL's HIPS tests were conducted outdoors so that no observation was made on the possible interaction with structure'of the water slug and debris swept out of the reactor cavity.
ANL's corium test results showed that a pool of water on the floor of the Expansion Chamber (simulating reactor containment) can mitigate direct heating of the EC atmosphere.
The non-prototypical arrangement in the ANL test configuration where the stream of debris is directed horizontally and downward toward the water pool has not yet satisfactorily explained.
Wood's metal tests are of little value in the assessment of the effect of water.
12/17/86 15 TECH UNCERT RPT/ LEE
.v f.
Chemical Reactions The atmosphere of the Expansion Chamber was inerted in ANL's corium tests.
This excluded chemical energy of metal oxidation from heating the containment.
II.B Present Research Two experimental programs are on-going; the Surtsey/DCH test program at SNL and the separate-effect test program at BNL.
II.B.1 Surtsey/DCH Tests Surtsey is a steel vessel that is approximately 33 feet high and 12 feet in diameter.
Its designed maximum working pressure is 150 psig.
The vessel has an internal volume of approximately 3,600 ft3 that is adequate to simulate the volume of a 1/10 linear scaled large dry PWR containment.
Eleven tests are currently scheduled for the Surtsey/DCH series (Table 1).
The first two of the eleven tests were completed; DCH-1 in June 1986 and DCH-2 in October 1986.
Preliminary observation of these test results confirmed substantial pressuriza-tion and aerosol generation in the vessel. Analyses of the test results are continuing.
II.B.2 Separate-Effect Tests The program at BNL will investigate subissues that are important to DCH, but are either difficult or too expensive to be studied in the Surtsey facility.
An example of such a subissue is the core debris dispersal.
The high tempera-ture (2500'K) melt simulant used in Surtsey tests makes it difficult to observe trajectories of dispersing debris.
Many tests wil,1 be needed to study sensi-tivities of the debris dispersal to various conditions predicted for different accident scenarios.
The relatively high cost of Surtsey tests makes it imprac-tical for such studies.
The initial phase of BNL's program includes construc-tion of 1/40 scale transparent models, using Plexiglass, of Zion, Surry and Watts Bar reactor cavities and major structures around the cavities.
Water and 12/17/86 16 TECH UNCERT RPT/ LEE
y P
Table 1.
DCH TEST MATRIX FOR THE SURTSEY DIRECT HEATING FACILITY Test Characteristic 1
Small mass (20 kg) 2 Large mass (80 kg) 3 Surry cavity 4
In-containment structures 5
Defined flow paths 6
Inert atmosphere 7
Air, stem, & H 2
8 Water sprays 9
Corium melt 10 Water-filled cavity 11 Shallow water pool i
i l
l 12/17/86 17 TECH UNCERT RPT/ LEE t
k'ood's metal will be used as melt simulants to study debris transport.
It is expected that BNL's tests can identify physics important to analytical modelir.g of the debris transport and provide a data base for such modeling.
II.C Program Strategy In view of the complexity of the DCH process, it is highly unlikely that a small scale experiment can be designed that preserves geometrical, physical and chemical similarities with postulated commercial reactor accident conditions.
Small scale integrated tests, therefore, are not very meaningful because the results cannot be related to the reactor accidents.
Full scale mock-up tests, of course, are not feasible.
The plan to resolve the issue of DCH is to build up, from experimental programs at SNL and at BNL, data bases that are needed for development of analytical models simulating physical and chemical phenomena important to DCH.
These mod-els will be incorporated in existing containment codes to analyze the conse-quences of HPME/DCH in operating reactor plants.
These models will be validated against new test data from time to time to reduce uncertainties.
Refinement of the models will be brought about as necessary.
II.D Future Research II.D.1 Phenomenological Research l
As our understanding of the process improves, additional subissues that could have major impact on the outcome of DCH start to surface. Additional research will be needed to investigate these subissues.
There are a few such subissues that have already been identified:
a.
Ex-vessel Metal-steam Reactions Both BNL and SNL calculated that metal contents in dispersing core debris l
could be completely oxidized by steam blown down from the RPV before the debris leaves the reactor cavity region.
Hydrogen generated in such reactions can readily be transported, regardless of the debris dispersal, 12/17/86 18 TECH UNCERT RPT/ LEE
7 and burn in other parts of the containment.
Because this finding could change the belief that DCH can be moderated or mitigated by containment structures that intercept and trap the melt particles,'it is recommended that tests be conducted to confirm and quantify these reactions.
Related subissues are burning of hydrogen in a high temperature atmosphere and the catalytic effect of hot debris particles in promoting hydrogen recombination, b.
Ex-vessel Steam Explosion In both HIPS-4W and -6W tests, where the melt was ejected into a water-filled cavity, violent steam explosions took place that destroyed the test cavities.
Steam explosions associated with high pressure melt ejection are believed to be more intense than when the melt is ejected under gra-vity because, in the HPME, ejected melt is likely to be atomized increasing the surface area available for heat transfer.
In addition to generating dynamic loading and possibly missiles, the steam explosion could further fragment the debris and enhance their dispersal. We plan to add measurements for steam explosions in those Surtsey tests with a water-filled cavity.
Additional tests will be proposed if the measurements clearly indicate the need for additional and more comprehensive tests.
c.
Melt Simulants Neither iron-aluminum thermite used for SNL's tests nor corium thermite l
used in ANL tests are representative of molten core materials expected in a severe accident.
The use of different melt simulants has impact: on(1) the melt temperature that affects aerosol generations and transport, (2) thermalandchemicalenergycontents,and(3)transportpropertiesofthe melt that affect mixing and interactions of debris in the containment atmosphere.
17/17/86 19 TECH UNCERT RPT/ LEE
U II.D.2 Analyses The CONTAIN code, together with DHEAT and IDHM, has been ussd to conduct exten-sive sensitivity studies of DCH in Surry and in Sequoyah.
Results of these studies were referenced in formulating staff positions in NUREG-1150.
DHEAT is an abbreviated version of CONTAIN while IDHM is an addition that enables CON-TAIN, a lumped-parameter code, to analyze the rate-sensitive DCH process.
The University of Wisconsin is using the HMC code to analyze DCH.
HMC is a union of three existing codes; HECTR, M1 and CORCON.
Other participants in DCH test calculations include DHCVIM by BNL and a still unnamed code by ANL.
All of the above discussed codes are lumped-parameter in nature.
Each code includes rate equations for chemical reactions and heat transfer, but had to make assumptions regarding the core debris transport to analyze DCH.
Among others, each code requires user input for the mean flight path, the debris air-borne time, or some other equivalent parameter that is highly dependent on plant geometries and configurations.
It is worth noting that the value of this quantity calibrated on one facility is not likely to be applicable to another facility.
Debris transport is one area that lumped-parameter codes are not likely to be able to solve. Additional research is needed in this area.
I 12/17/86 20 TECH UNCERT RPT/ LEE
?3 III. TECHNICAL UNCERTAINTY EVALUATION The RES staff's best estimate of uncertainties regarding this issue is provided in Appendix A.
This estimate represents our present state of understanding.
On-going research is expected to provide additional insights that could narrow these uncertainties. We understand that the NRR staff has a different view on the range and the degree of belief of some parameters.
Their view will be presentedinAppendixJ.5ofNUREG-ll50(inpreparation)'.
III.A Uncertainties Expected to be_ Reduc,e,d,bL the Current Procram Uncertainties associated with initial conditions of the high pressure melt ejection (HPME) will be addressed by other research programs such as Severe Accident Sequence Analysis, Core Melt Progression, Natural Circulation Inside RCS, etc. Technical uncertainties this program will seek to reduce include those associate'd with containment heating and pressurization, aerosol genera-tion and airborne aerosol concentration as a function of time.
Experimental prcgrams at SNL and at BNL are expected to provide data bases needed for such exercises.
III.A.1 DCH tests in the Surtsey Facility The mainstay of the DCH research program is the experiments in the Surtsey facility at SNL.
By the end of FY88, all tests in the attached DCH test Matrix (Table 1)areexpectedtobecompleted.
Results of these tests will contribute to substantially reduce uncertainties regarding:
1.
Rates of chemical reactions and heat transfer 2.
Effect of water; both from the pool water and from suspended water droplets in the containment atmosphere.
3.
Effect of structures on core debris dispersal and DCH and 12/19/86 21 TECH UNCERT RPT/ LEE
,J u a
4.
Aerosol generation and transport.
III.A.2 Separate Effect Tests at BNL i
Currently, BNL's activity is concentrated in the study of the effect of three different plant configurations on the debris dispersal:
Zion, Surry and Watts Bar. Results of BNL's tests are expected to reduce uncertainties in the fol-lowing areas:
1.
Influence of different reactor cavity designs on core debris dispersal.
2.
Effects of structures outside the reactor cavity on core debris dispersal.
1 Sensitivities of debris dispersal to the ejection pressure.
3.
i 4.
Flow fields of the gas-debris mixture in various scaled containment
{
configurations.
l S.
Scaling to extrapolate results of small scale tests for assessment of j
risk associated with debris dispersal in comercial plants.
i III.B Programs Needed to Further Reduce Uncertainties Areas in which significant uncertainties may still remain at the end of the current program were discussed in Section II.D.
Following programs are recom-mended to reduce these uncertainties.
III.B.1 Ex-vessel Metal-steam Reactions An experimental program is needed to confirm oxidation of zirconium and iron in j
a steam-inerted containment atmosphere and to quantify the rate of hydrogen j
generation.
An additional program is needed to investigate burning of hydrogen l
at elevated temperatures and in the presence of hot debris.
It is believed i
{
12/18/86 22 TECH UNCERT RPT/ LEE
n 1
that hydrogen could behave very differently under these extreme conditions.
Chemical reactions are likely to be more vigorous under these conditions.
III.B.2 Ex-vessel Steam Explosions Programs are needed to quantify the intensity of steam explosions initiated by HPME and their effects on fragmentation and dispersal of core debris.
If the results of the above programs indicate that the effects of ex-vessel steam explosions could be substantial, additional programs will be needed to quantify these effects--especially the dynamic effect on the containment structure and the possible impact on DCH.
III.B.3 Analysis No computer code currently available for DCH analyses can calculate debris transport, and the prospect for a lumped-parameter code to acquire such capability is not very promising. This is because the debris transport is basically a three-dimensional, two-phase flow problem while the lumped-parameter code is essentially one dimensional.
A program is needed to develop additional sophistication in the analytical tool to handle this subissue.
III.B.4 Melt Simulants A couple of tests will be needed to resolve this issue using a melt, generated by induction heating, of the composition predicted by the best available core melt progression analysis.
This is a state-of-the-art undertaking that will require a certain lead time to develop the technology.
Either relocation of the existing Large Melt Facility (LMF) or installation of a new melt furnace will be needed.
12/18/86 23 TECH UNCERT RPT/ LEE
, -e l
IV.
IMPLEMENTATION OF RESEARCH RESULTS It appears premature at this moment to discuss implementation of research results because the issue is highly controversial and uncertainties are large.
Many people are still debating whether the high pressure melt ejection could take place at all.
Even granted that HPME could take place, considerable uncertainties still exist regarding potential mitigating effects of the containment structures and of the water on the floor or suspended in the containment atmosphere.
Until such time when the technical community agrees on these considerations, it may not be appropriate to consider implementation of the research results in the regulatory sense.
Regardless of the regulatory decision, results of the research will be incorporated into analytical tools, such as CONTAIN, as new knowledge becomes available.
Such tools can then be used to calculate subsequent tests, and when needed, to analyze the hazard of DCH at operating plants.
12/18/86 24 TECH UNCERT RPT/ LEE
V.
SUP91ARY 4
The issue of HPME/DCH is highly controversial.
It has a potential of signifi-cantly changing the risk profile of operating reactor plants. At the extreme conditions, the containment could fail at the time a large concentration of radionuclides is airborne. The issue is relatively new and has large uncertainties regarding the possible consequence mainly because the lack of' data bases needed to quantify the effect of complicated, interacting variables.
The biggest uncertainty concerns the initial conditions of HPME.
There is a belief, supported by some code calculations, that natural circulation inside the RCS could induce a failure elsewhere in the RCS boundary to depressurize the system before the core melt breaches the bottom head of the RPV.
- HPME, therefore, would not take place in this scenario.
There is considerable uncer-tainty in this concept also, and evidence is not conclusive at this moment to obviate the concern for DCH.
In any event, it may be a questionable proposi-tion to depend on an uncontrollable failure of a safety-related system to miti-gate an accident of this magnitude.
Other initial conditions, such as the melt temperature, fraction of the core melted, and unoxidized Zr content in the melt are also highly uncertain.
It is hoped 'that further investigation in the area of Core Melt Progression can help narrow these uncertainties.
Trapping of the melt particles by containment structures and quenching of the core debris by water are believed to be two most promising means of mitigating DCH, given HPME.
They remain to be confirmed by experiments.
Major uncertainties are debris transport, ex-vessel metal-steam reactions, and ex-vessel steam explosions.
A broad data base is needed to develop analytical tools for best-estimate analyses.
12/18/86 25 TECH UNCERT RPT/ LEE
l Appendix A Issue Uncertainties Figure A.1 is the result of DHEAT code analyses for the Containment Load Working Group (CLWG) Standard Problem #2.
It will be used as the basis for our estimation of uncertainties in the pressure rise in Surry containment due to DCH. Sensitivities of the pressure rise to changes in the value of various parameters suggested in Table A.1 are an approximation from results of later DHEAT analyses.
Ranges, degrees-of-belief, and sensitivities provided in Table A.1 were then sampled by a Latin Hypercube procedure to compile a composite total pressure vs. probability curve that is shown on Figure A.2.
4 12/18/86 26 Appendix A
Fi c.A 1.
Surry DHEAT2 Calculations
~
Without Blowdown Steam and no Steam Spike 16.0 2400.0 n
i
- Pressure m ~~~. ~~'
14.0
= Temperature
... **.. ~,.
2000.0 j
9 Q
L 12.0 v
i d
e 1
00 L
v g
10.0 i
1600.0 a
eu M
y 8.0 a)
O.,
i 1200.0 g
,_ ~.
6.0
~
d
.~
e C
W C
E-.
4.0 E
i 800'.0
.=
2.0 l
i i
0.0 d i
i i
i 400.0 0.0 20.0 40.0 60.0 80.0 100.0 Percent Core Melt-Ejection
.-o Table A.1 - Surry DCH-Input for Statistical Analysis of Pressure Rise versus, Probability 1.
RCS Pressure (0-2400psig)
Range
_ Degree-of-Belief (DOB) 1000 psig and up 0.1 600 - 1000 psig 0.2 below 600 psig
0.7 Sensitivity
Considered here is only the effect of the RCS steam inventory on containnent pressurization.
Linear inter-polation has been used for intermediate values.
Add
(+)18-25 psi
=Fj for the range 0 - 2400 psig 2.
Melt Temperature (1800-3100*K)
Range DOB 2500 - 3100 *K 0.2 2300 - 2500 *K 0.65 1800 - 2300 *K 0.15 Sensitivity Multiply (X) 0.85 - 1.10
= F, for the range 1800 - 3100*K 3.
Fraction of Core Melted and Ejected (20 - 80%)
Range DOB 60 - 80 %
0.3 50 - 60 %
0.5 20 - 50%
0.2 l
Sensitivity No correction from Fig. A.1. Value read from, Fig. A.1
=P3 l
4
[
12/19/86 27 Appendix A
.o e,
4.
Unoxidized Metal Contents in the Melt (20 - 70%)
Range DOB 55 - 70 %
0.25 40 - 55%
0.5 20 - 40%
0.25 Sensitivity Multiply (X) 0.8 - 1.0
=F4 for the range 20 - 70%
5.
Effect of Water (0 - 50% quenched)
Range DOB 0 - 15%
0.3 15 - 30%
0.4 30 - 50%
0.3 Sensitivity Multiply (X) 1.0.- 0.85
=F S
for the range 0 - 50%
6.
Effect of Structure (75-100% dispersed)
Range DOB 90% and up 0.85 75 - 90%
0.1 below 75%
0.05 Sensitivity Multiply (X) 0.85 - 1.00
=F6 for the range 75 - 100%
7.
Completeness of Thermal and Chemical Interactions (50-95%)
Range DOB 85 - 95%
0.25 70 - 85%
0.6 50 - 70%
0.15 12/19/86 28 Appendix A
=
a s
l Sensitivity Multiply (X) 0.63 - 0.97
=F 7 50 - 95%
8.
Hydrcgen Recombination (0 - 6% H, by volume)
Range g
5 - 6%
0.2 2 - 5%
0.6 0 - 2%
0.2 Sensitivity Pultiply (X) 1.0 - 1.23
=F8 for the range 0 - 6% H, Estimated Peak Pressure P
=P3xF2 4
5 6
xF xF xF xF7+Fj + P' (F8-1) max P. = P at 0% core Fraction = 2.67 bars (Fig A.1) 3 12/19/86 29 Appendix A
FIG. A 2
.?
I ES-I M A-~ ED D EA < - P R ESS U RE i
N JCH FOR SU R RY (WF RES DO3)
.2018--
.1792--
j
.,s....
.nu--
l }.mO-
.089.--
s
.0672--
I 4
.0448--
.0224--
O P
25 35 45 55 65 75 85 95 105 115 125 135 145 L
Peak Premmurce
-