ML20079J188
| ML20079J188 | |
| Person / Time | |
|---|---|
| Site: | 05000142 |
| Issue date: | 01/16/1984 |
| From: | COMMITTEE TO BRIDGE THE GAP |
| To: | |
| Shared Package | |
| ML20079J177 | List: |
| References | |
| NUDOCS 8401240241 | |
| Download: ML20079J188 (71) | |
Text
l F
T o-00CKETED USt:P.0
'84 JAN 23 Pl2:00 CFfiCE OF SECIJ.M COCELTH:!i & 3E UNI'MD STA'IES & AMRICA WF4 NUCIEAR REGULAT 2 Y COMMISSION BEF 2E THE ATOMIC SAFETY AND LICENSING BOARD In the Matter of Docket No. 50-142 THE REGENTS & THE UNIVERSITY (Proposed Renewal of Facility
& CALIFORNIA License Number R-71)
(UCLAResearchReactor)
January 16, 1984 CBG'S REVIEW & UCLA'S " ANALYSIS & SHUTDOWN MECHANISM" Committee to Bridge the Gap 1637 Butler Avenue, Suite 203 Ims Angeles, CA 90025 8401240241 840116 PDR ADOCK 05000142 0
w o
O CBG'S REVIEM & UCIA'S " ANALYSIS & SHUTDOWN MBCHANISM" i
Introduction An independent, in-depth assessment has been-performed of UCLi's " Shutdown Analysis" related to how a major power excursion at the UCIA Argonaut-type nuclear reactor would selfaterminate. The assessment identifies a number of methodological problems with the UCLA Analysis and concludes that the likely termination mechanism is core destruction.
Ahong the methodological problems are:
The UCIA analysis attempts to model a highly complex reactor kinetics probles, occurring on a time scale of milliseconda and highly geometry-dependent, by using assumptions valid only for idealised, uniform, steady-state fluid flow problems occurring over long time periods. Assumptions useful, for example, in estimating steady, frictionless flow between infinite reservoirs are used to describe, millisecond by millisecond, what is i
really a samil explosion, when those assumptions are inapplicable and the true solution considerably more complex.
The mean of a few individual data points from power excursions at one reactor, SPERT I, is taken as a constant, "a common denominator for a spectrum of reactcrs," despite the fact that the " constant" does not apply to even so similar a reactor as BORAX I.
Nonethelssa, it is applied unquestioningly to the unique hybrid design of the UCLA Argonaut, for which no confirmatory l
data exist.
- The transient, dynamic nature of the problems involved are j
ignored. Pressure spikes are inexplicably assumed to be l
sustained pulses, hundreds of milliseconds or longer in
.w.
--p_e,_e-.,
w--w-T'M-**-wMwv-- - - ' - -
-'w--
N'-w'--'-me
-4 aw*
7w-Wm4 time-N-wWsiem'we='ve Mer----
ww*wA~e"*W* * '"m'*"T'-- - -- ---
~
a
' g duration. Surface effects and viscosity functions,. junction losses, compressibility, interphase miring and film boiling effects are not considered. Eighth-of-an-inch cracks are summed as if they were a single 6-inch diameter pipe, unacceptable even for a first-order approximation.
Re crucial input data are unreliable. Se least conservative of numerous irreconcilable values are chosen for key core parameters. S e reliance on 1.1.p. M ucible and conflicting data has serious potential implications for public health and safety.
l Forexample,despitemeasuredvaluesoff(8reportedto vary by 250%, a non-conservative value is selected and three significant figures provided. By so doing, a $3 reactivity insertion is converted into a 14 millisecond period, when it could just as easily from the existing data be a 9 millisecond l
period. Se fundamental first step in determining reactivity l
effects-translating the insertion into an exponential period-1 is impossible because the fundamental parameter of My is unknown for the reactor.
he void coefficient-perhaps the most important value in the entire analysis-is arrived at by discarding all data l
which produce less favorable results ani by relying instead on an extremely questionable value.
=_
u Se answer to the central question of "where the water goes" appears to remain unanswered. Se deflector plate region chosen as the escape route faces a wall of lead bricks.
De premise that copious amounts of water will exit via these brick walls and supposed eighth-of-an-inch gaps simply does not
" hold water."
SUMMARY
& CONCLUSIONS g
4 1.
Se postulated power excursion will not self-terminate as assumed by I
expulsion of water out the top of the fuel box region through the surround-l ing brick walls, s.
Water tossed up, even assuming that it is able to reach the deflector plate by breaking the aluminum plate " membrane" at the top of the fuel box, is merely going to hit a blank wall of lead bricks and bounce off. Very little, if any, water will exit through the supposed eighth-of-an-inch gaps, most of which do not even appear to exist.
b.
Even if the gaps exist, they are r)ot summable. It is not scientifically proper to sua these long, narrow cracks ard treat them as though they were one large pipe opening of circular cross-l section, ignoring the significant surface effects of viscosity, turbulence, steam binding, junction losses, and so on.
c.
Water is not going to rise as a slug and remain suspended, pressurized by steam below, as though there were a clean boundary preventing the denser liquid from falling and the less dense gas from rising. If there is release through narrow cracks, it will l
be primarily release of steam, not water.
1 d.
Even if the sigh %of an inch gaps existed as claimed, and even l
if the water could be forced through those gaps in the time frame 1
c being discussed, most of those pps would lead nowhere: there u
is no void space on the other side of the supposed apertures, only more leed and graphite hricks closely stacked.
i e.
Even if the apertures did exist and did lead to interconnecting void spaces, the void spaces would not sua to the 35 liters or so necessary for permanent shutdown.
Sus, even if the fuel did not nelt on the first pulse, the reactor could be destroyed on su k equent " chugs." Permanent snutdown does not occur l
through expulsion in the manner asserted by UCIA.
(
2.
More appears to be insufficient space for void formation to turn even the first pulse, thus power would continue to rise until shutdown is achieved through rapid disassembly.
3.
Even if sufficient void space is available to turn the first pulse, void forantion will be supressed if high pressures are generated (as assumed by UCIA), thereby increasing peak power and energy release by 50-100%, apin indicating molting on the first pulse.
4 If high pressu:es are not generated as assumed, but only those pressures found at SPERT I-D and BORAX, the rupture disk may not break at all until long after melting.
5 Even if the rupture disk does break, expulsion through that 30-foot pipe occurs after fuel melting would have taken place.
6.
Even if the UCIA Argonaut were an open pool reactor like SPERT I ani l
BORAX I, the Argonaut reactor's fuel would be more likely to melt on an excursion of the same period simply because its neutron lifetime is far longer than that of BORAX I or SPERT I.
,---m,w-,
g.,
y-w w-----r-,
r-
.-----,y,
,,....r,w,
,-1,
,.--r.--
- -.~,..
.-.-,---.-,e-
~5-w 7.
Even were the Ostrander modification of the Forbes model correct-and it is not, failing to accurately predict peak power even in a reactor as similar to SPERT I as BORAX I, let alone the far different Argonaut-- the UGA Argonaut would melt on the first pulse based on 30RAI data.
8 Even were the Ostrander method correct, and even were he justified in discarding the unfavorable kinetics testing data, fuel melting is indicated on the first pulse merely by using more accurate core parameter data for the Argonaut.
9.
water cannot get out of the system through the pathways described for the UG A Argonaut by UG A in the time frames of concern. The shutdown mechanism appears to be via ropid disassenhly of the core.
i l
l
. Difficulties in Estinating Effects of Reactivity Insertions in the Armonaut here are three great _ difficulties in determining accurately and with a sufficient safety margin the possible consequences of a reactivity insertion of a given angnitude in the UCLA Argonaut reactor.
The greatest difficulty is due to the fact that there are no kinetics testing data whatsoever svailable for Argonaut reactors. Whereas TRIGA and pool-type plate reactors have large and detailed bodies of actual testing data from power excursions, there is simply no empirical evidence from which to make determinations as to reactivity effects, shutdown speed and mechanism, and the like for the UCLA Argonaut. his makes an adequate safety determination essentially impossible.
Se second great difficulty is related to the first. Since there are no kinetics testing data available for the Argonaut, extrapola-i tion from kinetics tests on different kinds of reactors is extremely difficult. Se translation equations and correction factors are unknown.
Data txist for SPERT and BGtAX reactors, but the substantial differences in neutron lifetime, geometry, and voidable moderator make extrapelation from the SPERT data useless without the appropriate correction for these differences. Yet there are no empirical data for the Argonaut to use as the basis for such correction.
l l
The third great difficulty is that even if there were reliable methods of correcting for the major differences in the reactor geometry i
and core characterirtics, reliable core parameter and geometry data for the UCIA Argonaut are non-existent. Sus, for example, even were l
l there a reliable method of correcting for UCLA's far longer neutron lifetime, that key core parameter is not reliably known, with asasured values extending over a range of 250%.
similarly, a key parameter in l
l t
~ _..... _ _.., _ __ _ _.. _. _ _ _. _ _. _ _ _ _,. _ _ _ _ _ _ _.. _ - - _... _
h deterai. ling relative effectiveness of shutdown mechanism is the void coefficient. It is known that the smaller void coefficient at UCLA will produce larger energy releases at the same period than the larger void coefficients at SPERT or BGtAX. But even if there were a method of correcting to three significant places for the void coefficient differ-ences, there are no reliable data available as to precisely how small UCIA's void coefficients are, nor the significance of those void effects that are positive at UCLA. Likewise, the information about the core geometry at UCIA is so poor that even knowing how to correct for void coefficient and knowing its value precisely at different locations in the core, one cannot determine accurately how much effective void space there is to permit voiding, whether it will be sufficient to permit shutdown, and whether it, even if enough, will delay shutdown due to system pressure effects.
These difficulties are, we believe, close to insurmountable.
Pool type plate reactors ani TRIGA reactors have been extensively tested, and a large body of etpirical data exist by which to make meaningful judgments about those types of reactors. The Argonaut went into pro-duction without such kinetics testing, and the manufacturers are no longer in the reactor business so testing at this stage is unlikely.
In the absence of reliable empirical data for Argonauts, reliable extrapolation techniques from open pool, short-lifetime, single-moderator reactors to the Argonaut, and reliable core parameter and geometry data for use in auch extrapolation, reliable demonstration of safe self-shutdown from an insertion of, say, $3 00 of excess reactivity is not possible.
What Are the Priumry Differences That Must Somehow Be Accounted For?
UCIA's reported neutron lifetime is approximately three tinos longer than that of the 3GtAX I and SPERT I reactors. We are aware of no empirical kinetics testing of reactors with neutron lifetime in t
l the zange of UCLA's. The reported UCIA neutron lifetime is between I
that of the SPERT II heavy water core and the lightwater SPERT I and 3GtAX I cores.
$lf3 Reactor BGtAX I 8.6 asec SPERT I 7-11 asoc UCLA 20-30 meec SPERT II 100 asec At SPERT II, which was designed to test the effects of longer lifetime cores, limited core damage was observed for tests of initial periods of 70 meec, with estimates of molting commencing at 30 asec. Molting and damage thus occurred at periods roughly ten times l
longer than for the shorter lifetime cores of the light water BORAI and SPERT I ceres.
$3 at UCIA will result in a period estimated by Mr. Ostrander at 14 asec, using an 2//3 of 29.2 asec.
SPERT II data suggest nelting for such a period.
One must also correct for the void coefficients at UCLA.
UCIA's void coefficients are smaller than those of most of the SPERT and BGtAX cores. Furthermore, the Argonaut has positive void coefficients at the edges of and above the fuel elements, due to its unique hybrid design and the higher absorption cross-section for light water than for graphite.
Only a portion of the reactor moderator is light water, the
C.
rest being solid graphite which is, obviously, non-expellable. Further-more, it too has certain positive reactivity effects. And one must correct for the fact ht W BORAI I and SPEtT I reactors were open pool, with plenty of expansion room and opportunity for expelling water from the core, whe) ans W Argonaut core is enclosed in fuel boxes with plugs and a solid core around. In addition, W Argonaut fuel plates are unconstrained, unlike those of SPERT or BORAI, creating the potential for substantial increases in power due to reactivity effects of bowing.
l l
To analyse shutdown of the UCIA reactor, Mr. Ostrander defines four steps. The first is to determine a " common denominator for a spectrum of reactors" to allow extrapolation from data for one type of reactor to another. Step two is to apply this generalized relation to the UCIA Argonaut and to predict temperature rise for a certain period.
The third step is to estimate the voiding of the water in the fuel box. Step four considers b question of expulsion of water from the system. We will avamine each of these steps below.
S'IEP OFE K = 73: "The Common Denominator for a Spectrum of Reactors"?
Mr. Ostrander describes the first step in his analysis as finding the " common denominator for a spectrum of reactors." In his attempt to find such a common denominator, he puts forward a " constant of - @ t.ionality," which he defines as k = 73. This " constant," he y
asserts, makes possible the accurate'prsliction of peak temperature in the UCIA Argonaut for a $3 insertion, based on SPERT and BORAI data and his method of correcting for the various differences between the Argonaut and the SPERT I and BGtAX I reactors.
The SPERT and BGtAI series of experiaects, performed under l
ABC auspicos for a period of roughly fifteoa years, attempted to understand the extremely ccmplex nature of reactivity accidents. Even after many, many years of intensive research, the prediction of system damage in different kinds of reactors remained extremely difficult.
The limits of confidence in analytic methods and the accuracy of values for fundamental parameters used in such analyses were quite.
large. In particular, the use of empirically derived parameters from t
one reactar system in predicting and analysing potential consequerces of reactivity incidents in a system of another type remained highly uncertain. To attempt to explain major differences in reactor behavior l
l due to substantial differences in design and core parameters on the basis of a single " common denominator" or a numerical " constant of proportionality" belies the extreme complexity of reactor kinetics and excursion behavior. If there actually were such a constant (such as Mr. Ostrander's k = 73), fifteen years of detailed research l
e,--,
--m..
..,n
, van,s
-n
-,,-a ene -,_
,,,----,------..,..,--,-..,--,---,~.---,--~.-.-.------n-..-.---,n--
a trying to understand the complexities of power excursion behavior would not have been necessary.
Much work was done during those years of research attempting to understand qualitatively the andanisms of reactor shutdown. Ora such analysis was performad by Forbes, gt d (" Analysis of Self Shv.t-down Behavior in the Sport I Reactor," IDo-16528), upon which both Mr.
Hawley and now Mr. Ostrander have relied. However, in the case of Mr.
Ostrander, the Forbes methodology has been dismissed and his own revisionist version put in its place. We shall discuss this below.
i l
What was the Purpose of the Forbes Methodolomy?
Ss Forbes paper was a review of data from several plate spacing arrangements in the same open-pool SPERT I core. By changing i
plate spacing, void coefficients were variei by a substantial factor, and the effect of varying void coefficients while holding essentially all elsa constant could be examined. It was known that void coefficir,nts were important in determining the effootiveness of reactivity compen-sation mechanisms and thus the speed with which an exponential power rise would be turned and terminated. Se Forbes paper attempted to determine whether previous models suggesting a strong dependence of total energy release on void coefficient were correct.
I Forbes argued that the linkage between void coefficient and I
total energy release was non-linear, and estimated it to be something i
between an inverse of the reactivity coefficient b (void coefficient over the neutron lifstine) raised to the i power and the 1/15 power.
Using the square root (i.e., the i power), he was able to get reasonably j
good agreesent with the data from the SPERT I cores (i.e., an estimate
. =
of a seven-fold mas:iana difference in power when the data indicate a 5 5 difference).
The important thing to remember is that the Forbes i
method was applicable for cores in the same reactor vessel. Astdo from one steel core, nothing changed except the plate spacing and number of plates,, and Forbes found that power and energy were dependent upon the void coefficient that resulted from that spacing arrangement.
Importantly, he found no need to correct for the mass of the core beyond the correction already done by considering the effect on void coefficient.
It is also important to keep in mind that the variable that was being changed was the void coefficient-not the neutron lifetime.
And that although some of the void coefficients considered were in the range of those reported by UCLA, none of the neutron lifetimes were.
UCIA's neutron lifetime is anny times longer than that of the cores Forbes considered, and the Forbes cores had extremely little variation in neutron lifetime.
geactor Core
/O void coefficient SPERT I B-12/64 11 asec
-0.116 $/% void **
SPERTIB-16/40 10 mooc
-0.206 SPERTIA-17/28 7 asec
-0.224 SPERTIB-24/32 7 asec
-0 325 SPERTID-12/25 8 asec
-0 36 UCLA Argonaut 29 asec*
-0.277* to -0.076 (measured values up to 50 maec)
It is obvious that there was very little variance in the neutron lifetime in the Forbes analysis, ard that the primary variable at work
- value used Iqr Ostrander.
the void coefficient used by Forbes was in units of g/ml, and varied even more widely than in the units chosen by Ostrander.
-%)-
1 was the void coefficient. Peduced neutron lifetimes ranged from 7 to n asec, while the void coefficients varied by a factor of three. None of the lifetimes were anywhere near UCLA's, being a factor of roughly three lower than the value Mr. Ostrander uses.
S us, Forbes found that in varying void coefficients within thesansreactorvessel,whileessentiallyallelse(particularly neutronlifetime)remainedmoreorlessconstant,arelationship somewhere between an inverse square root effect ani an inverse 1/1 5 Power could account for auch of W variation in peak power and energy release, irrespective of the mass of the core. Se Forbee I
model is not intended for, nor based upon, reactors of widely different design or neutron lifetimes yet that is the use to which Mr. Ostrander l
puts it, after first substantially altering W model itself.
l Se Ostrander Revisions to the orbes Model he Forbes model provides a rule of thumb for predicting, within certain error margins, peak power and total energy (not power or energy density) in a relatively narrow class of reactors open pool, moderated solely by light water, and having short neutron lifetimes.
Forbes found that he could predict fairly accurately the behavior of the various cores on the basis of a shutdown coefficient b, consisting of their void coefficients and neutron lifetime, and irresnective of the wide variation in the size of the cores.
Mr. Ostrander, in addition to certain other modifications *,
has substantively altered the Forbes model by dividing the value resulting from the model by the ratio of the mass of the cores.
t i
i i
- For example, Ostrander divides the void coefficient by $/4 rather than f as in the Forbes model.
l
_.______.._.,...,_m__..,_,_,,
___,r,__
-.,_.m.__.____
This, of course, reduces Mr. Ostranfer's predicted energy releases for the Argonaut by very substantial margine, as the mass of the Argonaut is smaller than any of the SPERT cores, by as auch as a factor of three. In other words, if one were to apply the Forbes model as Forbes originated it to the Argonaut, the predicted energy releases and temperatures could be far higher than indicated by Ostrander, and well into the region of melting.
l Forbes and colleagues found they could predict peak powers and energies in SPERT cores based primarily on differences in void coefficients, despite cores varying substantially in size. Core size l
was not considered a factor except in so far as it implicitly affected void coefficients and neutron lifetime, which were found to be primary i
l determinants of peas power and total energy. Mr. Ostrander, by essentially discarding the Forbes method and inserting once again
[
his mass-of-the-core factor, reduces by a substantial factor the predicted consequences of a power excursion that result when the Forbes model is used as designed by Forbes, n al.
j Se Enerar Density Argument l
Mr. Ostrander again raises his energy density argument, and we need not belabor again the problems with it.
In short, he has the concept upside dcun. Se assa of ths core will determine the l
energy density, given a certain energy release, but it will not determine the energy release. In other words, a small core will have a higher temperature (energy density) than a large one in which the l
same total energy was released, but the size of the core tells one nothing about how much energy will be released in the first place.
L
-15.
Mr. Ostrander himself proves this once and for all with his,
admission that his mass-of-the-core theory is unable to explain the 1
BORAI data, which will be discussed in more detail below.
l "k = 73" There is no magical numerical constant which will permit extrapolation from a set e.,f reactor kinetics data for one reactor type to an entirely different type of reactor for which kinetics data do not exist. Mr. Ostrander askes the fuMamantal error of taking the mean of five individual data points, which have been assessed to reduce data spread, as a censtant.
We have stressed over and over again the complexity of power excursion phenomena and the multitude of effects that differences in core design and core parameters can have on excursion behavior.'
The scientific method-particularly in a matter involving public health and safety-does not permit imposirve a theory on data',
particularly when the data are extremely limited, and data which l
contradict the theory are thrown out, and when the theory is then not even tested.
In this case, Mr. Ostrander began with just five data points.
These data points were not, aside from SPERT I-D, obtained from actual numerical values, but from eye-balling a smoothed curve on a log-log chart that cannot be read with accuracy to much more than one significant figure. Since the data from which the curves were derived were not checked, the actual data points might be above or below the curve relied upon. The spread in data was narroup by
choosing a different unit than.that used 4 Forbes, and an average then taken. Incredibly, even the average is in error-- Mr. Ostrander takes the average of his five values of k to be 73, when the actual average is 74.2.
Claiming that this massaged average is a constant, he then applies it to the Argonaut for which no kinetics data exist, and for which the applicability of the " constant" is undemonstrated.
More importantly, the applicability of the " constant" to even the SPEtT cores from which it was derived 4 Mr. Ostrander is undemonstrated. After obtaining this average from the five data points, he does not attempt to test the mothed at different periods, and with different available data. Furthermore, even at the same period, the method does not apply for example, it underpredicts 4 several factors the power rise in the SPEtT P core, which the original Forbes method was able to predict with considerable accuracy. Likewise, as will be discussed below, it underpredicts severely the behavior of the BORAI reactor. If the " constant of proportionality" cannot predict behavior in reactors so similar, how can it be expected to have any validity for the Argonaut, which is so different?
The scientific method does not permit eye-halling a graph, taking a few numbers off of it and multiplyi v and dividing the few data points by various factors until one has narrowed the spread of the resulting values, throwing out the data points that do not fit.
l and taking an average of the remaining ones and calling it a constant on which the safety of thousands must rest. Kypotheses must be tested, large enough bodies of data must be acquired and must support the hypothesis, standards of statistical significance ard controlled experimentation must be met.
K = 73 (74.2) is nothing more than hand-waving at data.
D e BORAX I Data Disprove "k = 73" In 4 21, Mr. Ostrander was asked why he did not use BGtAX I data in his k'= 73 method. He answers that although that reactor was very sin $1ar to SPERT reactors, "its performance was quite different."
He goes on to explain that he does not know how to explain why the
(
energy density was "significantly higher" in BORAX I than "in the i
apparently similar SPERT I-A core." K = 73 does not work for BORAX.
It severely underpredicts the actual energy density rises at that s
reactor. But instead of discarding the k = 73 theory, Mr. Ostrander discarded the data that did not fit the theory.
We had in previous testimony pointed to the fact that BORAI I produced considerably more energy at the same period than the l
SPRT I-D data upon which the Battelle study was based, and that this might therefore be nonconservative, severely underpredicting resultant energy at UCIA.
Mr. Ostrander at that time attempted to explain away the difference based on the larger mass of the BORAI core. Yet with the new k = 73 method, which compensates for the differences in cora i
masses, Mr. Ostrander still underpredicts actual power rise at BORAI by a factor of 2.
1 Using Mr. Ostrander's theory, k=P*(C *O/l)k y
Khere k = 73, P = maximum power density in kw/cm3, C is the void y
coefficient in $/5 void, and I// is the reduced prompt neutron lifetime.
'the input data for BORAX I are as follows (taken from Mr. Ostrander's direct testimony and the Dietrich article):
l l
- -.-~
-,n
,---.--.,,,...--..,-----.~,-,.-n
active volume of core 29230 cm3 void coefficient
-032$/% void f/
8.61 asec t
1herefore, if Mr. Ostrander's theory were correct and his "cor.atant of proi.tianality" were indeed 73, power density for BORAX I at a 14 asec Period should be P = 73 (032/8.61x10~3)Y
-12kw/cm3 Peak power, as vyr =1 to power density, at B0"JLX should thus be c
12 km/cm3 x 29230 cm3 - 350,760 kw 1
I or about 350 nu If Mr. Ostrander's 2T rule is applicable (and he indicates that it is for BGtAX), then the BGtAX I should have yielded for a 14 aseo period 351 Mu x 2 x 0.014 see 9.8 Mu-see However, Dietrich reports that the actual enersty release in the 14 mec range at BGtAX I was approximately 20 Mw-sec. rouschly twice what Mr.
Ostrander's model would predict.
And the BaltAX, of course, was far l
nore similar to SPERT than '_4 the hybrid, closed system of the UCIA l
Argonaut.
Mr. Ostrander's model thus underpredicts hr a factor of two the release from an excursion in the very similme BORAX reactor. For his model to work for BGtAX, k would have to be about 150 Even assuming that k for UCLA-which is unknown-is not greater than ths.t at BGtAX, all of Mr. Ostrander's estimates of peak energy density at
UCLA must be increased by a factor of two, even if one accepted his " correction" for the mass of the core. Using his method, the increase in fuel temperature at UCLA, regardless of the geometry issues later discussed as to whether room exists for shutdown or whether backpressure win increase energy release, win be twice his 382*C value, yielding a temperature rise in the 800'C range, well l
over molting. If one eliminates his mass of the core revision of l
the original Farbes model, even higher temperatures result.
_Further Problems with the Ostrander Model Ostrander indicates uncertainties in the input data from the SPERT testa as to the void coefficients-whether they are identicany defined for all reactors. He also identifies uncertainties in reading the power values off of the Forbes graphs, which adds significant possibility of error to the comparisons. Furthermore, as is indicated in, for example, tha Dietrich graphs of energy release versus period which include the data pointa as well as the smoothed curve, there is considerable scatter of data about the smoothed curve.
Thus, in addition to the errors involved in taking data off such a crude graph, there are additional errors associated with the degree of scatter around the curve to begin with. These errors alone, even accepting the k = 73 method and value, might be sufficient to push the estimated 380'C temperature rise to over the molting point.
. -. ~. -
J t
.go_
?
Se2T*M(anx)" Rule" i
I Not having obtained actual energy release data for most of the SPERT curves, Mr. Ostrander takes peak power data off the generalised curves from Forbes identified above, and then attempts to translate l
peak power into energy by a rule, of thumb that multiplies peak power times twice the exponential period. Mis introduces further potential
-l l
for error, as it is a rathar gross approximation in an area where public safety is at stake. More importantly, however, the approximation j
is of questionable validity for a reactor such as UCLA's with a long neutron lifetime.
l Forthe2t*/(max)" rule"toholdtrue,therateof shutdown at UCLA aust be identical with that of the BORAI I and Sport I-D reactors for which Mr. Ostrar.ler indicates the relationship is valid.
i l
However, the entire premise of the Forbes model which Mr. Ostrander has adopted with modifications is that the total energy released l
l at the same period varies with the shutdown coefficient, which depends upon the effectiveness of both the void coefficient and the neutron lifetime. In particular, a long prompt neutron lifetime will " broaden" the burst, because it reduces the effectiveness of the shutdown mech-anism. If the shape of the curve broadens, simply multiplying the peak power times two exponential periods any considerably underestimate the actual energy release.
i l
l Assuming that an Average is Conservative Mr. Ostrander, in addition to throwing out data such as l
. _ _ - - _ _ _ _. _ _ _ _ _ _ _ _ _ _. _ _ _. _ _ _.,. _ _ _. _ _ ~,. _
s l
1 l
the P core data used by Forbes and the BORAI I data which would indicate "k" values much higher than the 73 he calculates from the L
more favorable data, uses the mean value of his five favorable data points and claims this is conservative. As indicaud previously, by an arithmetic error won the mean (which equals 74.2) is miscalculated.
But since the five favorable data points include values as high as 86, not to mention less favorable data points that have values in the hundreds, use of k = 73 (even if the methodology were valid) is clearly non-conservative. UsingtheSPERTI-Dkof84.3,ortheB-24/32kof 85 8, Pushes Mr. Ostrander's temperature rise up about 20% on that basis alone. 'Ihe probability that k values for the UCLA reactor may occur outside the bounds of the five favorable data points, much less be exactly the mean of these data points, invalidates the whole process.
What About Mr. Ostrander's Method of Convertina Estimated Power Density to Temperature?
i It is axiomatic that there is some relationship between l
power density and temperature, but the specific relationship is far more complicated than Mr. Ostrander makes out in his analysis. For examplis, at SPERT it was found that very small rises in power density led to very substantial rises in fuel plate temperatures, due to onset of steam blanketing effects. Furthermore, as indicated previously, l
comparing recorded temperature rises from thermocouples at SPERT for the purpose of estimating actual temperature rises that would occur at UCIA is made very complicated by the fact that at SPERT the thermocouple data tended to under-record actual fuel plate temperatures. Use of such data would thus tend to underestimate actual conditions: fuel
could be molting where thermocouple data indicated it was below the molting temperature.
Assuming that one can simply come up with a power density, transform it into energy density by a rule of thumb approximation, and determine temperature rise through use of specific heat values is an l
l attempt to use principles of physics and engineering that may be i
applicable to problems involving long time framos but lose their validity in the millisecond time frame in which power excursions occur.
Likewise when Mr. Ostrander assumes that he should divide the total energy release he has estimated for the UCLA Argonaut-a value we have shown above may underestimate true values by a large factor-by the total mass of metal in the Argonaut core in order to get an average power density and thus temperature rise in the fuel material. He includes as mass of the fuel not just the fuel meat but the cladding as well. The heat, however, is not being generated in the cladding, of course, but in the meat. While it is true that eventually the heat will ha distributed relatively evenly throughout the fuel plate, both most and clad, that is not true in the time scales of concern. Eventually, of course, the heat would be distributed also to anything outside the fuel, so why, if Mr. Ostrander's method of calculating temperature were correct, include only the cladding as a heat sink for the fuel meat?
l The fuel meat, where the power and energy are being generated, represents approximately half the volume assumed by Mr. Ostrander by his including the cladding. The purpose of " adiabatic" analysis, which he says he is performing, is to exclude heat loss from the system or material in question. That is a useful methodological assumption when one in talking of events occurring in t$w frames comparable to heat
transfer times, as is the case here. The whole problem with power excursions in reactors is that power-- and therefore tamperature -
can rise in the fuel mest faster than it can be transferred to the cladding and from the cladding to the coolant, so that the fuel melts before that heat transfer has taken place. It is therefore inappropriate, particularly in a supposed adiabatic analysis, to have included the clad as part of the active mass and to assume even distribution (correctedonlyforpeak-to-averageflur.inthecoreregiongenerally) through meat and clad. Heat will be generated in the meat and it will take time to be transmitted. If it didn't take time, boiling would be instantaneous, shutdown automatic without time delay, and water-cooled reactors would have the prompt shutdown coefficient of TRIGA reactors.
Thus, the assumption that the " adiabatic" calculation performed by Mr. Ostrander will significantly overestimate actual temperatures is doubtful. He has included twice the active ansa in the calculation, and does not assume adiabatic conditions. It should be noted that calculated adiaktic rises at SPERT were close to and sometimes less than the measured temperatures, and, as indicated before, those measured values often underestimated actual temperatures.
SMP WO Application of the Ostrander " Constant of Proportionality" to the UCLA Argonaut We discussed at the outset the major difficulties involved in assessing the course of a power excursion at the UCLA Argonaut in the absence of kinetics testing data on Argonaut-type cores of the sort available for TRIGA reactors and plate-type reactors of the open-pool, single moderator, short lifetime variety.
Mr. Ostrander has attempted to overcome that threshold difficulty by proposing a " constant of proportionality" or " common denominator" he claims is applicable to a spectrum of reactor types.
He proposes a numerical value of 73 for this supposed constant "k" and a methodology for translating kinetics testing data from one type of reactor to another. We have discussed in the previous sections the problems with that methodolog.
f In Step Two, Mr. Ostrander attempts to apply his constant of proportionality to the UCLA Argonaut. 'Ihis Step involves one of the other major difficulties in assessing the UCIA Argonaut, besides the lack of kinetics testing data and an accurate method of extrapolating from SPERT and BORAX: the lack of accurate input data for the basic core parameters ofr the UCLA Argonaut. A discussion of Mr. Ostrander's
~
attempt to apply his k = 73 methodology to the UCIA Argonaut in the face of a dearth of reliable core parameter data for the Argonaut follows.
What Period is Produced by a $3 Insertion at UCIA? No One Knows Mr. Ostrander's constant of proportionality of k = 73 is
l said to be valid only for periods of 14 milliseconds (and, as I
demonstrated above, is low by a factor of two for the BCRAI I reactor atthatperiod).
Mr. Ostrander's first problem in applying his method i
to the UCLA reactor is in demonstrating that the insertion of the presumedamountofexcessreactivity($3)willyielda14aillisecond period.
The relationship between reactivity and period is, of course, governed by the inhour equation, which in turn is governed by the value of / /# valid for the particular reactor in question.
Mr. Ostrander obtains a 14 millisecond period from a
$3 insertion at UCLA by use of a simplified version of the inhour equation, and the assumption that f /4 at UCLA is precisely 29 2 asec.
It is obvious that a different 1/4 will produce a considerably different period, and thus energy release, for the same reactivity insertion, and thus the value of //4 used must be known with considerable confidence for any of the rest of the calculation to have validity. But //4 at UCLA is not reliably known, with measured values extending over an extraordinary 250% range. The choice of 29.2 asoc (a value given to three significant places!) is totally arbitrary, and the true value could be auch, auch different, as would the results of any calculations dependent on it.
Mr. Ostrander has argued that energy release is largely dependent upon exponential period. We have argued that of equal importance is the speed and effectiveness of the shutdown mechanism-how quickly the exponential rise is turned. Both factors, however, are deeply tied up with ///3
e ~
l What is Y-//3 at the UCIA Arnonaut? No One Knows The reduced prompt neutron lifetime is one of the most important core parameters, upon which so many key aspects of reactor i
safety depend, that it is among the very first reactor parameters measured. It must be accurately known, and there are excallent l
methods fee accurately determining it.
l However, in the UCIA case, Mr. Ostrander reports that measurements range from 20 to 50 milliseconds for 2 /4. That is i
an extraordinary degree of uncertainty the true value, obviously could even be outside that very large range. Nonetheless, Mr.
i Ostrander chooses a value of //8 to three significant figures-l 29 2 usec-in part because it corresponds to a " reported" value of 4
1.9 x 10 seconds for neutron lifetime and " plausible" values of g.
4 However, reported values of neutron lifetime range as low as 135 x 10 seconds and greater than 1 93 reported values for 8 for Argonaut reactors, including UCIA's, likewise have a large range.
Mr. Ostrander relies in part on the calculatsd value for l founi in the Neogy memorandum, yet rejects the calculated value found there for 8, which contradicts the values he has chosen.
Se calculated value for l
$//3 found in the Neogy memorandum is contradicted by calculated values found in UCIA's Application. All the calculated values are contradicted by actual measurements:
the actual measurements are each contradictory, one to the other, over an extremely large range.
Rus, no one knows what the actual reduced prompt neutron lifetime is for the UCIA Argonaut-error margins for any estimate are greater than 100%. It is impossible to calculate, therefore, even the exponential period on which a $3 insertion would send the reactor,
l 1
let alone the far more complex matter of how the different $//3 values would affect shutdown.
Will $3 Insertion Result in a 14 asec Period?
The answer to that question, given measured values and l
calculated values for / /4 so widely in disagressent, is almost i
certainly no. But precisely what period would result is impossible to say without professional, reproducible, reliable measurements having been corducted. What can be anid ire that Mr. Ostrander's choice of 29 2 asec is neither technically defensible nor conservative.
For example, Mr. Ostrander reports one value of 2//3 for the UCLA l
Argonaut as 18.9 asec. If that were W correct value - and there are measurements at least close to that calculated valus-- the resultant period would not be 14 asec, an Ostrander has assumed, but rather 9 asec.
(Recall tnat the original lissards Analysis concluded that the onset of molting is in the range of 9 asec, and numerous non-conservatisas would push that far higher). Thus, Mr. Ostrander's use of an ///5 of 29 2 meec, chosen from an extraordinarily wide range of measured and calculated values, any result in a significantly non-conservative exponential period, which could seriously underestimate energy release.
What Would Be the Effect of Usirut Another l/e3 ?
i l
If the value from the Hazards Analysis were used, for example, or the measured value of 20 were used, a pariod of roughly 9 asec would result. At SPERT I-D, such a period resulted in about
a 7 5 m-sec energy release. Using Mr. Ostrander's method of taking the ratios of the square roots of the shutdown coefficients, and using the lower ///9 value, one gota a shutdown coefficient of 0.277+18.9 x 10-3 or 14.7 for UGA, compared to W value Mr. Ostrander reports for SPERT I-D of 44.1.
h cores are alaoat identical in mass, so h energy release should be, if Mr. Ostrander's method is correct, about 173 times greater at h Argonaut than at SPERT I-D, using his value for void coefficient and 18.9 maec for
/
his would indicate a release of about 13 Mu-sec. Se Hawley report estimates temperatures of 586'c from a 12 Mu-sec burst 13 would be in h molting range.
l i
f
l What is the Void Coefficient at the UCIA Arnonaut? - No-One Knows To use his " constant" of k = 73 in extrapolating from the selected 1
Sport data to the UCLA Argonaut situation, Mr. Ostander needs only two major items of input datas an accurate value for the reduced prompt L
neutron lifetime, ani an accurate value for the void coefficient at UCIA.
Aswehaveindicatedabove,nosuchaccuratevaluesexistforA[8, with measured ulues extending over a range of 250%. The values for void coefficients at UCLA extend over an even larger range cf uncertainty.
In his previous testimony, Mr. Ostrander cited three values for the void coefficients
.164%k/% void h e most recently-measured value, l
asasured after changes to the core l
to increase power to 100 kw.
.18%k/% void valte in the original Hazards Analysis
.2%k/% void a value assertedly measured during start-up in 1960, prior to the 1963 modifications to increase power Roserangethusfrom22W/%voidto27#. In his new analysis, )tr. Ostrarkier has selected the least conservative of these values, making it even slightly larger than he had previously used (by increasing it to 277d/% void).
Se significance of the void coefficient, and the need for a very high degree of accuracy in identification, is due to two factors:
(1)the void coefficient will determine the relative effectiveness of the shutdown parameters for the reactor (in other words, how fast the exponential rise will turn and fall, and thus the total energy releassi), and (2) it will determine whether sufficient expansion room exists within the closed syste's of the Argonaut to permit sufficient voiding to occur to bring about shutdown.
It is thus a factor absolutely crucial to a determination of safety at the facility.
-3%
i And yet, there is considerable uncertainty as to what precisely the void coefficient of reactivity is at-the UCIA Argonaut.
As we shall see, that uncertainty extends over a larger range than even the 22d to 27 7d identified above.
We have noted reference in operating logs and other documents to several void coefficient measurements, but have been informed I
that records of those measurements no longer exist. We frankly find this very disanying, because the void coefficient is a permaster of the utmost importance, and records of several asasurements have been lost entirely, whereas the available documentation of the two primary available measurements (1960, 1963) is unacceptably poor. Se accuracy I
of the first void coefficient measurement is questioned in the document which reports the measurement itself, and the more recent measurement, required by the ABC as a condition of permitting core modifications to be done to increase power sai reported to the AEC as the accurate value, is now called into question and discarded by UCLA.
Bere are no accurate measurements, no acceptable records, and no verifiable values for the most important of core parameters, one essential for an analysis of whether this reactor can safely shut itself down. It is, of course, through the void coefficient of reactivity that shutdown occuras and no one knows what the coefficient is for UCIA.
It could be anything.
Use of the more recent of the two coefficients produces a 50 C tenPerature increase over that estimated by Mr. Ostrander using the less conservative values using the SPERT I-D k value and the more recent void coefficient produces a 33% higher energy release than that
~-.r--
,,._-n.-,
,v+----mn,-,r,-~n-----,-.-
.,.~~..~.-----,,-..-~r-
... - - - - - - - - - - ~ ~ - - ~ - - - -, - - -. -
0 calculated by Mr. Ostrander, or about 115 0 higher. Unfortunately, it does not appear that merely using the more recent value solves l
the problem, because there is substantial evidence that the true void. coefficient may be several times lower than the value used by Mr. Ostiander. Ironically, that evidence comes from the very measurements he cites as the basis for his far higher value.
The Start-Up Report l
Mr. Ostrander, citing the Start-Up Report of 1960, indicates that the only documented value for the UCIA reactor is 0.2%k/% void or 23 inhours per void of 110 al.
It is through converting inhours to l
cents (d) that Mr. Ostrander arrives at his 27 7d figure.
Mr. Ostrander reports the value as an average or uniform coefficient for the core based on sampling in a number of locations. Bis overstates the validity of the measurements, as documented, and perpetuates a rather extraordinary error in averaging, as well as reliance on izzqducible data.
%e measurements on which Mr. Ostrander relies for his figure of 23 inhours per void or 27 7d/% void ~are recorded in Table 8.1 (attached) of the Start-Up Report. As is seen therein, 9 voids were inserted into the core and the reactivity effect noted. Thereafter, 3 additional voids were inserted into different locations, and the additional reactivity effect recorded. Then voids were removed by stages, four at first, then one, then two, then two more, and then three. The worth l
of the voids as measured on their removal was several times that of l
their measured worth when inserted. his was explained by the difficulty in reproducing bundle spacing exactly with each void insertion or removal, l
l t
~
TABLE 8.I VOID COEFFICIENT i
i 1
Reg. Rod Reactivity Run Void s*
Position (th)
Worth No.
Added Location Before After Before After A
ih/ void 104xf.k/ml 76 9
6 NE 1/4 NE t
6 SE 1/4 NE
,u 6 NW 1/4 NE E
e 6 SW 1/4 NE j
6 NW 1/4 N l
6 NW 1/4 S 5 SW 1/4 S 4,8, SW 1/4 N 20.0 28.0 219 162
-57 6.3 1.46 l
77 3'
4,5,6, NE 1/4 S 28.0 43.0 162 79
-83 27.7 6.4 79
-4 NE Box 59.5 40.0 29 93 64 16.0 3.7 l
80
-1 6 NW 1/4 S 40.0 33.0 93 132 39 39.0 9.0 81
-2 4,8,SW 1/4 N 33.0 27.0 132 168 36 18.0 4.2 82
-2 6 NW 1/4 N 5 SW 1/4 S 27.0 20.0 168 219 51 25.5 5.9 84
-3 4,5,6, NE 1/4 S 34.0 22.0 126 202 76 25.3 5.8
..J_____...
N
- A void is a piece of polystyr ene foam 24 1/4" x 2..//8" x 1/ 8. The volume of one of these strips was measured by immersion in water to be 110 ml/ strip.
due to the reactivity effects of slight changes in bundle spacing.
( mis is a matter about which we have commented previously, as having the potential for reactivity addition beyond that licsased or for additional reactivity insertion in the midst of an excursion.)
It poses a markad effect on the void measurements, making it impossible to reproduce the effect of a void at the same location. A void at one location is worth 6.3 inhours on Vlacement and 18 upon removal, for example. Se experiment produces such contradictory results for precisely the same effect that the resalts are of questionable utility.
A void must have the same worth when removed as when inserted. Even so, the 23 inhours per void reporter 1 by Mr. Ostrander is not even a l
l correct average value of the resulting data.
Se column marked " worth" in the attached Table 8.1 gives average reactivity worths for the voids inserted on each of seven runs.
Sus,63ih/voidistheaverageworthofninevoids: 27.7 ih/ void the average of three voids and so on. Notably, the value based on the largestnumberofdataisalsothesmallest,i.e.,63ih/ void.
(
Improper Averaging Inflates the Asserted Void Coefficient l
To derive the asserted 23 ih/ void figure, UCIA has simply I
takest the mean of the seven average values listed in Table 8.1.
But Applicant makes the fundamental error of failing to weight each value for the number of voids it is tased on.
Thus,UCLAtreats63ih/ void and 39 0 ih/ void as equally significant, although the former is tased on nine separate voids and the latter on only one. By se doing, the resulting figure is improperly inflated.
A more correct procedure (given the limitations of the data) would give the following results:
a) Se insartion of voids-9 voids at 6 3 inhours each, and 3 voids at 27 7 inhours each-yields an average of 11.65 ih/ void.
b) Removal of the same voida yields an average of 19.2 ih/ void.
Note that when properly averaged, even the removal of all voids, though l
wceth nearly twice their valua on insertion, is still less than the valueof23ih/coidMr.Ostrandercitesastheaveragepervoidover the ca e for the whole experiment.
One might be tempted to combine the average void worth at insertion together with that for removal, but that Modd be an error.
As indicated above, the fact that insertion of the voids was worth so much less than removal of the same voids indicates error due to bundle shifting. Assuming that either value is accurate, conMning them would average good measurements with bad, providing only a bad average measurement.
Bus we have at least two different values for void coefficients from this experiment, both of them subject to doubt because they contr-l dict one another. Oneaveragevalueis11.65ih,orabout14e/5 void.
Seotheris192ih,orabout23d/% void. Rather than picking the smaller of the two-- which we would view as not sufficiently conservative given the data spread-UCLA asserts a value higher than either average.
Because of the in yoducibility of the data and the inter-forence from bundle shifting, use of any but very conservatively chosen values from this experiment is not responsible. But if one had to rely on thsae data-because, as Mr. Ostrander says, virtually no other documented data exist-one.certainly cannot improperly average and endupwith277d/% void. Atbestonewouldhavetouse14W/% void
l^
l -
(
(or0.10%k/% void): and it wouM be more reasonable to use the 6.3inhourspervoidvalue,whichis7.6d/% void (or0.05k/.% void).
l Se former wouM raise energy release and temperature rise estimates
(
by 40%
use of the latter value would result in apprad==tely a doubling of estimated release.
Se h11er Void Coefficients Are 5=u M by Other Measurements l
Sere is a context of several other void coefficient l
l asasurements which lend support to the smaller values measured at i
UCIA.
I h e University of Washington Argonaut value is reported by UCIA (Application, p. III/6-3) as 0.0002 %k/cm3, which corresponds to 0.07 5 k/ % void, or about log / % void.
Se Argonaut at Iowa State has performed a multi-group diffusion calculation of the void coefficient using modified versions of the IEPARD, FOG, and PERT programs, obtaining a value of 6 3 x 10-2
%k/ % void (SAR for the Iowa State Argonaut, August 1981), or about 9 7 e/ % void, in very close screement with the value reported for the University of Washington.
Furthermore, UCIA has performed sons water level variation experiments which provide relevant information.
Mr. Ostrander, in using his 27.7d value, estimates that he must remove all of the water above the fuel plus 15 of the water within the fuel region to compensate
$4.00 of reactivity. 15 is about 3) inches from the top of the fuel.
i Thus, if Mr. Ostrander is correct about the void coefficient he has used, l
draining water down to 3) inches below the top of the fuel should be worth $4
l l
l
-39 An experiment was performed at UCIA to measure this effect, l
and the results do not confirm Mr. Ostrander's void coefficient value, but are in fact closer to those obtained from the Univ. of Wa=Mneton and Iowa State Argonauts, and the insertion of voids experiment at UCLA.
In the early 1960s, it was determined that the level of water in the core that permitted criticality with the control blades fully withdrawn was just under N inches from the top of the fuel, and the excess
(
reactivity limit at the time was about 0.6%. Thus, the effect of dropping the water from the ton of the fuel down 3i inches is arnroximately l
l
$1. and not the $4 assumed by Mr. Ostrander. One must expel considerably mera water than assumed in the analysis. and the void coefficient assumed har Mr. Ostrander is indicated empirically to be quite high.
These data have been confirmed by other experiments at the Univ. of Washington.
CBG Exhibit C-I-1, a 1962 inspection report, indicates dropping the water from the top of the fuel at the U of W down four inches results in a reactivity loss of only 0 53 %k.
Mr.
Ostrander's analysis would require that those four inches of water loss reduce reactivity by at least 2.6 %k, four times more than empirically determined. It is obvious that more water than assumed in the analysis must be expelled to produce the reactivity effect needed, l
In sum, it would appear that, if one were to use the Start-Up Report void coefficient data at all, the data from the insertion of the voids would appear to be more reliable due to their agreement with so much additional data. A summary of these void coefficient values follows:
UCIA Start-Up Benort Voil Insertion Data 9 voids 7.6g/% void 12 voids
,14g/% void Univ of Washington Argonaut Void Coefficient 10g/% void
]
s Iowa State Argonaut _ Void Coefficient 97g/% void i
UCIA Water Invel Variation MeasurementA I
as found 7 M % void l
adjusted for normal fl u profile 12g/% void l
(blades partially inserted)
Univ of Washington Vater Imvel Data 5W/5 void (not adjusted far flux profile)
Sese values all range from one half to one fifth that used hr l
l Mr. Ostrander in his analysis: they can be described as roughly 1015t/% void,ratherthanthe277eheused. no Start-Up Report does not.wyyod his use of that latter value, and it and numerous other data suggest that void coefficient values far lower must be used. Se effect of this would be to vastly increasu the estimate of power and temperature on the first pulse, even accepting his methodology and all his other assumptions and inputs. Avalueof5W/5voidwould yield temperature rises 2 35 times higher than assumed by Mr. Ostrander the 10 g figure temperature rises 1.66 times higher, and the highest i
figure of 14 W temperature rises 1.41 times higher.
It is recognized that voids at the upper portion of the active core any not have the same woeth as voids elsewhere. However, even correcting for the flux profile in the box (which any over-compensate, because the l
flux variation was measured with control blades partially inserted at the i
top of the box, depressing the flux, whereas the water level experiments, and presumably any power excursion, would occur with control blades withdrawn),
the void coefficient indicrted by these experiments is in the range of the other values indicated above. Assuming a 1:17 ratio of flux at the top of the box to the average flux in the box, the water level variation experiments indicate a void coefficient of about 0.085 %k/ % void or about 12 e/ % void.
l.
l l
l Positive Void Coefficients.
i Se Start-Up Report measurements did not measure the worth of voids at all locations throughout the core, kt only between fuel I
plates within fuel elements. The value, and sign, of voiding at the edge of bundles was not determined. However, other experiments indicate that these peripheral voids are positive in reactivity effect, further complicating comparisons with SPERT and BCRAX where the void coefficients j
were generally negative throughout the core. Furthermore, consideration of a void coefficient based solely on voiding effects between platee--
and not at the outer-side of bundles or above bundles-wouLi considerably underpredict energy release, because at the sans time that some voids were helping shut the reactor down, others would be adding reactivity to the system.
A unique result occurred in one experiment at UCIA for which records do exist.* Se experiment consisted of inserting a duany fuel bundle consisting of eleven dummy fuel plates in an empty corner of a l
fuel box. Se duany bundle displaced water, and the reactivity effect of that displacement was found to be positive, not ne6ative. Bus removing water from the periphery of an active fuel bundle by adding a duany bundle increased reactivity.
He same effect was experimentally determined by Vitti.
He found that as he increased plate spacing on his experimental bundles, in some of his experiments a second reactivity effect be6an to add to the effect of the increased water moderator between the plates. With
- 1xperiment Report by Eugene Weinstine, June 13, 1967
sons of the bundles, the effect of increasing plate spacing had the effect of displacing water at the edges of the htudles, which had a noticeable positive reactivity offect. Thus, he found two reactivity values for increasing plate spacing: one that was due purely to the increased plate spacing, the other that was due to both the additional moderation between the plates and the reduced absorption outside the bundles due to water displacement at the periphery. The effect of i
reduced absorption on the periphery roughly doubled the effect of increased spacing merely due to compensating for the undermoders(ed condition between the plates.
In the experiment where effects of reduced water on the periphery were significant, Vitti found increasing plate spacing (and consequent water displacement at the edges of the bundle) resulted in a reactivity increase of roughly 0.05 * / water channel when the channel was increased to about twice its normal size.
He found in the experiments where the periphery effects were minimized that the value was approximately 0.0225 # ehannel for doubling the channel size.
He concluded that the latter figure was more reliable for estimating the effects of increased plate spacing alone, due to the severely undermoderated nature of the core, and that the larger figure was due to the dual effect of increased plate spacing and the displacement of water at the periphery.
(In estimating, in our previous testimony, $2 additional reactivity available from modest bowing, resulting in the equivalent of three channels per bundle increasing by a tenth of an inch or so, we used the lower value found by Vitti, ignoring the potential additional positive effects of displacement of water at the periphery i.e., 3 channels per bundle x 0.0225 # channel x 24 bundles - 1.62 9,
which would be shout $2 50 consideration of the effect of such bowirs
~
t on peripheral voiding could increase that effect significantly, as evidencedbytheVittimeasurements.)
Se reason for the effects noted by Weinstine and Vitti is explained in experiments on similar positive effects above the fuel, noted both at UCLA and the Univ of Washington and described previously.
Light water, although very effective at slowing down neutrons, also has a very high absorption cross-section for neutrons. Thus, at the anae time it is moderating and reflecting neutrons, it also absorbs them, taking them out of the system and asking them no longer available for fis.,ioning of uranium. Graphite, however, has a very low absorption cross section. He effectiveness of a moderator is best measured by the Moderating Hatio, the ratio of slowing down power to macroscopic absorption cross section.
_ Moderator Moderating Ratio Light Water 70 j
Graphite 170 (fromGlasstone,1955)
As seen from the table above, the Moderating Hatio for graphite is 2i times better than for water, when the absorption is taken into account.
Thus, for fuel bundles surrounded by a layer of water in the fuel box and a large quantity of graphite outside and above the fuel box, removal l.
of water at the periphery has a positive effect, reducing absorption while not reducing to any significant degree moderation or reflection.
I Displacement or voiding of water at the periphery merely permits the more effective, less absorbing graphite to act, increasing the neutrons available for fission.
l
h Argonaut core, with its current plate spacing, is
~
essentially undermoderated between the fuel plates and overmoderated at the periphery-creating positive void coefficients at the periphery and potential for positive effects due to bowing due both to displacement at the periphery and increased moderation in the center. These effects are not taken into account by the void coefficient measurements taken at start-up in 1960, and thus the use of void coefficients.,evet if valid for voids between fuel plates, any not be valid when the potential for voiding at the periphery is taken into account.
In another of h documents recently provided, the confirmation of these positive effects above the care led to speculation that there could be fuel savings if W fuel bundles were raised in the box, so as to lower the effective water above the fuel and the absorption there.
It turned out, because of control blade placement, not to be worth doing, 4
but the existence of h oe positive void effects at the periphery and I
above the bundles indicates how a single value for void c6 efficient between plates vastly oversimplifies what would actually go on during a power excursion. Expansion of water between the plates could cause, in addition to positive effects due to bowing between plates, displacement l
of water at the peripherys boiling would also produce bubble formation at h periphery, all of which would have a positive, rather than a negative reactivity effect. Bubbleriseand/orexpulsionofthewater above the fuel would also have a positive effect.
To effect parasnent shutdown, Mr. Ostrander says the ten inches of water everburden above the fuel and a certain amount of water between the fuel plates must be expelled from the system. But the reactivity effect i
l of h loss of that first ten inches or thMy liters or so of overburden i
has a positive reactivity effects only when water level drops in the fuel i
region does a negative effect take place. fne matter is thus far more complex than a simple single value for a void coefficient, plugged into a " constant of proportionality" would suggest.
Positive Graphite Effects In addition to the problem of positive reactivity effects in the water moderator, there is the problem of positive effects in the other moderntar in this hybrid reactor, i.e. the graphite.
l l
Mr. Ostrander, in his neu analysis, discusses this positive t
l effect in light of the Robles measurements of it and Mr. Ostrander's own l
calculations of what fraction of the neutrons reach the graphite. Since the analysis was filed, Mr. Ostrander has found certain errors in his j
neutron calculations, invalidating this section of the analysis. Other errors in those calculations will be identified in daalW with the Wigner energy issue for which they were primarily performed. Ist it simply be said at this stage that most of the neutrons do indeed reach the graphite.
As to Mr. Ostrander's dismissal of certain of the Robles data i
regarding the magnitude of the positive graphite coefficient, he errs here.
In particular, Mr. Ostrander asserts it is improper to rely upon a reactivity value based on the average temperature change at two thermo-couples in the center graphite island. He chooses to rely instead on only one of the thermocouples, one in close proximity to the heaters placed in the irradiation holes.
Mr. Robles asasured the graphite coefficient by placing heaters in the irradiation holes in the center graphite island and monitoring
42-O the reactivity effects against tosperature readings from two thermocouples in the central graphite island. One thermocouple was quite close to the heaters and thus recorded a relatively rapid rise in temperature, whereas the other thermocouple was closer to the fuel bases (not at power) but further from the heaters and thus its temperature rise was considerably less rapid. During normal operation, l
these tuo thermocouples have very similar readings, as the entire central graphite island (and auch of the surrounding reflector as well, for that matter) is heated. Reactivity effects from a temperature rise in the graphite during normal operations or a transient thus involve an i
average rise in temperature throughout the central graphite island and a portion of the reflector.
l Mr. Robles was quite correct in averaging the two therno-couples. To not do so, as Mr. Ostrander has done, would be to skew the results.
Measuring the reactivity effects of heating the graphite island by only monitoring the thermocouple very close to the heat source, while the rest of the island remains relatively cold, would be to explain the I
reactivity effect of heating the graphite right around the heaters as though the entiro graphite were heated to the same level, which was not j
the case.
The increase in reactivity for an average temperature rise in the central graphite island is given in the attached Figure 2.10 from the Robles thesis. Note that the effect is not linear, and that the initial effect does indeed approach 0.006%k/'F,ingoodagreement i
with the AEC inspection i. Mt for UCIA. The fact that the effect is not r
linear any help to explain the reason why Mr. Ostrander's measurement
--mr rw w.~
g-<--
.,n-, - -,,,,,, - - -
-,,-,wa g-w..
-,,,,een
.,----=-,---.------se
.,n--n-w--
---m-
--v--
qa l
from the Robles t.h. esis 0.005 -
I' p.:..
s.
n
.m o
.g e
..s j.
s..-
g.
3 i
o 4
.c M H-S f..
.....-a...
o s
e J
.g.
n, nn3.
e. --..
.u.~.~.
.j.
l l
a l
I I
0.002 l
I d-f l
0 l'0 20 3'O 4'O l
I 8T av Figure 2.10: O(.gav vs.
?,*Ev for the first g raphite expe ri.me.nt.
t.
s... a.
s. n s.+-
D
- e 80 p
m.
8 i.
so severely underpredicts the actual value. Mr. Robles' starting tospezature was approxiantely 684, whereas Mr. Ostrander's starting temperature was about 85'F. Furthermore, the positive effect of the graphite in Mr. Ostrander's experiment is masked 4 the negative water 0
effect until the graphite reaches about 130 F.
For both reasons, the far larger coefficient found by Bobles to occur in the early sta(res of l
temperature rise had begun to saturate out by the time the graphite reached the temperatures at which Mr. Ostrander ande his measurements.
l The Bobles measurements, in fact, any not be sufficiently l
conservative, since he measured the reactivity effect of raising the I
l average temperature of the central graphite island alone, while during normal operation and during an excursion both the central island and the outer reflector will heat up, increasing the reactivity effect. The 0.006 % coefficient may thus be low hy a factor of 50 to 100%.
Using the Robles data, and correcting for Mr. Ostrander's error about most of the neutrons supposedly not getting out of the fuel boxes, the excursion assumed in the new analysis would produce an 8.4'c prompt temperature rise (15'F), resulting in about an inanMate 10 g rise in reactivity, 15 to 20e if the effects of the reflector are considered as well. Effects subsequent to the first pulse would add to that, as i
conduction to the graphite from the heated water would raise the graphite temperature further. Ten, twenty, thirty or more cents any seen inconsequential, but in the course of an excursion where one is trying to remove reactivity rapidly, not add it, they can be significant.
i l
l
- ~ _ - - _ _ _ _, _ _ _ _ _., _, _ _ _ _ _ _ _ _. _ _ _ _ _ _ _ _ _
-M-Conclusion of Sten Two Review Even assuming there is such a thing as a " constant of proportionality" by which one can extrapolate to the UCLA case from SPERT data, correcting for differences in core parameters, one must reliably know those parameters.
As demonstrated in the previous discussion, the most fuM=mantal of core parameters f// and void coefficients-are not reliably known for the UCLA Argonaut.
Measured and calculated values all vary by wide margins.
The values used by Mr. Ostrarder are seriously non-cor_servative.
Furthermore, positive effects are improperly taken into account, which makes impossible an attempt to determine the dynamic behavior of the Argonaut with a single, static, "k" constant correction.
Useofmorereasonable/// and void coefficient values l
l indicates melting on the first pulse, irrespective of the geometry concerns discussed in the next section as to whether there is even sufficient room for voiding to turn the first pulse at all without core disassembly.
l l
l 1
l
49 STEP THREE Is %ere Room for Shutdown by Void Formation?
After having attempted to find a " common denominator" or
" constant of proportionality" by which one could estimate the reactivity effects of reactors with different shutdown coefficients, and after attempting to determine what values of those coefficients to use for the UCLA Argonaut, Mr. Ostrander's third step was to assess how the different geometry of the Argonaut wouM affect the shutdown mechanism itself.
Vhestems the SPERT I and BGtAI I reactors were open pool reactors, l
the UCIA Argonaut is a closed system. Step one and two were to deternita, if possible, what the peak power would be at a SPERT reactor with UCIA's void coefficient and [//.
Step three is to determine what effects, if 3
any, being in a closed system would have. In particular, the fundamantal question for Step Three is whether there is sufficient space in the system to permit turning the first pulse, and whether any constraint would inhibit shutdown to any degree.
Is here Enough Space to Permit Sufficient Void Formation?
Mis question is far more complex than the UCIA analysis indicates.
It depends on the dynamic nature of shutdown and void formation, the correct value for void coefficients, whether and to what extent positive void effects exist, precise determination of pressure buildup and pressure release effects which are highly dynamic and geometry dependent, as well as intricate questions about the physical geometry of a core for which no accurate drawings appear to exist.
.. ~
No Accurate Core Structural Data Exist _
The UCLA analysis makes assertions about the approximate size and locations of various apertures and water levels and coolant system elements, but no accurate information on the actual structure of the The drawings and other information we reviewed core appears to exist.
each provide contradictory data on these matters, and we note that no For example, the drawings exist for the system as it currently exists.
[
fuel boxes were refabricated in the early 1970s, but the only drawings Whether there are two inches, one and a half, are for the original boxes.
or some other amount between the mid-line of the overflow pipes and the top of the fuel boxes simply cannot be determined from the various We note that failure to keep and update technical contradictory records.
drawings is very poor practice, particularly when matters of safety analysis are dependent upon them.
The UCIA analysis apparently assumes that there is, at the start of the excursion, two inches of void space at the top of the fuel box due to the water level at the commencement of the event being at the UCLA midpoint in the exit line and two inches being available above.
=
further assumes an additional inch-equivalent is available above the box
+
by breaking an aluminum " membrane" making water entry into the deflector The total available room assumed by UCLA is thus plate region possible.
29 square inches interior dimensions by two inches in height within the i
box, plus an additional equivalent inch in the plug above.
Fuel box Plug Total 0.951/ box 0.48 1 1,431/ box 5 71/whole core 2.85 1 8551/wholecore To cancel $4 of reactivity requires - even using Mr. Ostrander's s
extremely non-conservative void coefficient of 27 7d per % void- 0.89 liters per box of voids formed between fuel plates, and at least that auch room for displacement elsewhere in the system. Ae UCLA estiantsa only 0 95 liters per box, with an additional 0.48 in the plug if the aluminum barrier t
sealing it is timely broken, it is clear that at best it is a very tight fit as to whether there is sufficient room for voiding.
Assume for the moment that the aluminum barrier does not break and the expansion room must be found in the fuel box itself. If Mr. Ostrander's 27 7p void coefficient is correct, one needs 0.89 liters per box and one has only 0.95 liters. We are told there are "approximately" two inches I
of space above the midpoint in the exit pipe. If those "approximately" twoinchesareonlyliinches,only0.71litersof" bounce"spaceexistin the box, which would be insufficient. EvenifIand3/4inchesexist,that I
means only 0.83 liters, again insufficient.
Or if the event occurs when unter is not, as assumed, at the midpoint of the exit pipe, but higher, again the amount of. space available becomes insufficient. It is thus apparent why very accurate specifications of the precise box goonstry are necessary and the precise location of maximum water level in the box l
prior to commencing of the event. Neither point of information is reliably known.
Vill the Aluminua "Montrano" Itreak (In Time-or At All)?
he UCIA analysis indicates that a plug sits atop the fuel boxes, and that a " thin aluminum acabrane" seals off the deflector plate l
l region in the plug from the fuel box below during normal operations to prevent water from entering that region. De issue is whether it will break and permit water to enter that extra inch of space, and if so, 1
1
-.----- _, - n-
,. -..,,,. -.. _ - - - - - - -, - -. - - _ - - ~ ~ - - - - - - _ _ -
P -
l 1
whether there win be any time delay or back pressure that any elongate the excursion, delay shutdown for even a few milliseconds, and thus incrosse energy release.
The phrase " thin aluminum mentrane" any create the impression that the tarrier is aluminum foil. However, heavy duty household aluminum foil is approximately 3 mils thick. The UCIA analysis iniicates that this material is 20 mile thick. That is aluminum sheeting, not foil.
l Itisthickerthanthecladdingonthefuelplatesandabout1/3the thickness of the plate itself. While it any be true that it win break or tear at 15 Psig uniform loading, that does not ten us that it will indeed 1reak in time (prior to peak of power, in time to provide the extra voiding room), or for that matter, treak at all. If there is any delay, or backpressure, that may' increase the energy release.
j Millisecond by millisecond, what win happen? Power will I
rise in the fuel meat, be transferred to the cladding and from there l
to the coolant, bubbles will form, water win be displaced to permit the voiding in the form of irregular surface rise or splash. This water rise will splash upwards-in the case of SPERT and BORAX I, unobstructed.
In the case of UCIA, it can only rise unobstructed two inches before the vanguard of the splashing water hits the aluminum barrier and splashes off, returning back into the core. This creates backprosaure, collapsing l
bubbles and delaying shutdown. As the first of the water bounces off l
the aluminum sheet and fans back into the core, more water is splashing upwards, creating turbulence and hitting the barrier too, creating more backpressure. The turbulence caused by this splashing makes a water-i l
air mixture, reducing the effective void space at the top of the box.
When sufficient pressure finally develops, something finally gives, and i
the pressure is released.
l l
._._m_._.
-49 It is likely, however, that the pressure release will occur not by bursting the aluminua plate, which would take approximately 45 Pounds over the 30 square inch surface, but by tilting theplug over so slightly.
)
h plug weighs only 100 pounds: the 45 Pounds necessary to break the bottom plate would be more than sufficient to tilt the plug, with most of its weight sti n supportec, merely having shifted the conter of gravity slightly. As with the cover on a boiling pot, the cover win
)
tilt slightly and pressure will be released in the form of stama (which is arders of angnitude less dense than liquid water, and thus scores of times more readily released through saan apertures). It is through such a nochanism that the air at the top of the box is likely to be released. 'thus, pressure any not result in treaking the aluminua plate j
but merely a slight tilting of the plug, releasing steam but almost no water, which because of its much greater density win have great difficulty (pexticularly at pressures of about 1 psig). getting through the tiny cracks.
h effective void space thus would be-- even assuming the event did i
begin with water at the half-way mark in the exit pipe, armi even assuming that the mid-way point is a fun two inches, not something less, from the top-- less than two inches. 'Ihe space in the deflector region would remain sealed off and the splashing water, bouncing off the aluminum plate as its upward action is arrested, ankes the region at the top of the box j
an aireter-ateam mixture.
Assume for the moment that the aluminua plate does treak.
There is still the first part of the scenario - water's upward travel arrested, f.my back into the core after only two inches upward splash, l
backpressure and bubble collapse, time delay until pressure builds up l
i
-9 to reak the plate, the time necessary for tearing it. These factars alone can dehy shutdown-- we are +=
M about events in the millisecond zange. If the plate does tear, an am6e will tear open, permitting a pathumy for the pressure to be released, although through an aperture of relatively===11 aroes sootional area. The tear any be an inch or so in diameter, and the water will be delayed-or greater system pressures may have to be developed to permit exit at the same rate, again producing i
delay and inhibition of shutdown-in fully entering the deflector region through the tear. Entry through the tear will increase turbulence and l
aske ilmpossible instantaneous fimM of the full space behind the tear.
l There would thus be additional time delays, and backpressures, inhibiting l
further voiding below. The shutdown mechanism-- whether the aluminua plate tears eventually or not-is constrained compared to the open systems at
$PitRT I and BCRAI I, and any constraint can delay shutdown while power continues to rise.
(VerylittleifanyreliefcanbeexpectedthrouEh the very small exit pipes in the boxes water will be thrown upmrd in a kind of small explosion, the pipe is at least half-full at the onset l
anyway, and its cross section is so small that it would take extraordinary pressures to provide any voiding at the rates necessary.) In short, it is likely that the additional inch above the aluminua plate will not be available for additional void space, at least not early in the excursion.
If it is at some point ando available, sono delay in shutdown compared to the non-constrained situation et the open pool SPERT and BORAX reactors must be expected.
l Even if the Available " Bounce" Space were Somewhat Greater than the Nece--n Void. That would Not be Enough for Rapid Shutdown Even if Mr. Ostrander's 27 7d void coefficient were correct, l
o 1 -
and even were there two inches d space at the top of the box, and even were an additional inch available in the plug above the aluminum plate, would that be enough?
Sere is an assumption at work in the UCLA analysis that if X milliliters of k bbles in the active core region are required for shutdown, I milliitters of air space at the top of the box is sufficient to permit unimpeded shutdown, as though the reactor were an open pool.
Behind this assumption is the apparent conception that voids will occur s
precisely where they are most needed and nowhere elser and that somehow the two inches of void space at the top of the box will simply trade places with the equivalent volume of water between the fuel pla h s.
The phenomenon of power excursions involving rapid boiling does not quite work in such a static fashion. Boiling is a very complex mechanism, and the apparent image of hbbles only inside the kndles and a square slug l
of water rising uniformly to fill every available nook and cranny in the fuel box does not jibe with the dynamic reality.
Se UCIA Analysis counts on 886 al of voids per box occuring within the fuel bundles. It assumes nose of the hbbles form between l
bundles or at the ed6es of the kndles, and that none of the h bbles which formed at the very onset, for example near the top of the fuel, will rise. All voids ths.t displace precious void space in the boxes, where the fit is so tight, are assumed to form only where needed, none where they have reduced usefulness in turning the excursion or even an opposite effect in terms of positive reactivity contriktion. It is also assumed that no additional cents here and thea are added during the excursion (due to graphite heating, bowing, displacements or bubble formation at the periphery, and so on), that would require additional voiding to counter.
l 1
w
---e--
m
w-
-,w
,, m
.r-,.- - - - -mwe
--w e.,--.,,e,-----e
,---,me--*-r----w-%
w--,
ee--+.
e.--u----
-52 And it is assumed that the w.ter rise will be perfectly even, not a splash or plume, and that the effect of water rise being halted after two to three inches will not result in bachmessure or bubble collapse or water return to the care, alongating the excursion or double-Ws l
it as opposed to the open core power traces with which it is being compared.
A power excursion is a dynamic, complicated phenomenon. Even if, say, 886 al of voids were all that were needed to form between the fuel plates, and even were 950 al availam in the box, or even 1430 with the full region of the deflector plate added in with no time delay for breaking the aluminum barrier, that would likely not be sufficient.
One needs considerably more room than that-otherwise the system is essentially constrained, and thereby delayed in shutdown.
I It is our opinion that sufficient shu W wn space does not exist. 'the aluminum diaphragm either will not break, reducing the space available, or it will delay shutdown by the time necessary for tack-splash and breaking, and squirting through the tear. Water cannot be presumed to be exactly at the midline of the exit pipe, and the records do not provide confidence that that midline is indeed two inches below the top.
And even if there were 10% more room than needed as bubbles in the plates, or even 50%, that would be insufficient, because a slug of water in the shape of a two inch high block is not going to instantaneously trade places with the two inches of al.r space at the tops bubbles will occur at places besides just the interior of.the fuel bundles. 'there will be an air-water-steam mixture at the top and an air-water-steam mixture below, thus requiring considerably more space than is available.
t 1
.=
- 53 4
Since the Void Coefficient is Wrong. Far More Snace is Needed than is Assumed ~ in the UCIA Analysis he entire issue of whether there is sufficient " bounce space" to turn the excursion revolves around two issuesa precisely how much space there is (not well known) and how much void space is needed (likewisenotwellknown). Se latter point depenis largely on the void coefficient, discussed at length in the previous section. If Mr.
Ostrander's void coefficient-which he asserts is an average of voids from different parts of the core measured reliably during the startup tests in 1960- is correct, one needs 886 al of voids between the fuel plates.' However, if the actual average value found for inserting voids in that 1960 test is used, one needs in reality twice the void space assumed by the UCIA Analysis, or apprwhtely 1750 ml per box, whereas there are only 1430 ml available if every nook and cranny of the box ani deflectcr plate region are filled.
(If the nine-void average value is used, 3224 al are needed per box.) If one employs either the Univ of Washington or the Iowa State Argonauts' void ccefficients, 2450 al of void space is needed per box, 70% more than is possibly available. If l
one uses any of the water level variation measurements, even corrected for l
flux profile, one cannot shut the reactor down one the first pulse by void formations there simply is not enough space.*
It is recognised that one may not need to compensate $4 as assumed by UCIA in its aWysis. However, if $3 were inserted, it is not at all unlikely that the combined bowing effects, graphite heating effects, and positive coefficients at the periphery would result in needing to compensate $4 or l
oven considerably more. While it is true that the burst starts to turn when the prompt reactivity has been compensated, it is also true that eventually all l
the reactivity must be compensate. Sere must be room to compensate both the prompt ard delayed neutrons. In an open pool reactor, compensating the prompt reactivity is the turning point, because there is sufficient expansion room that hr the time the delayed neutrons come along, so much additional boiling l
has taken place that any contribution they may make has already been compensated l
for. However, that implies there is room for that void formation to occur, l
which appears not to be the case at the UCIA Argonaut.
e Conclusian of Step ihree Review The long and the short of it is that there does not appear to be sufficient void space in the UCLA reactor to safely shut the reactor down in a power excursion hy void formation and water expulsion. Thus, unlike the open pool reactors where there was a n the expansion room needed, power win continue to rise and terminrtion win be through core disassembly. The excursion will initiate bubble formation, the power rise will start to slow, but the available void space will be l
fined before sufficient reactivity to funy turn the rise has been 1
i compensated. Power win then continue to rise, but at a slightly slower rate, evsn perhaps largely on the delayed neutrons which must l
also eventually be compensated. Shutdown win be delayed, energy release win increase, and power will continue to rise until sufficient pressures have been generated to relieve the pressures this win take muy e-folding times, and the excursion win terminate by core disassembly.
l l
1 f
SEP FOUR What is the Mechanism for Permanent Self-Shutdown at the UCLA ArJtonaut?
or Where Does the Water Go?
We have demonstrated so far that Mr. Ostrander's methodology of translating SPERT and BORAI data to the UCIA case is incorrect that i
even were one to accept the methodology, it indicates fuel molting on the first pulse, irrespective of whether sufficient void space exists:
l and that sufficient void space does not exist to turn the first pulse, l
no matter what its peak if in an open core reactor, and that the termination l
of the first pulse would appear to be core disassembly.
l The fourth step in the UCLA analysis is an ettempt to determine how water would exit the reactor-- assuming that the first pulse did not l
bring the fuel to the molting point-to induce final self-shutdown and prevent the possibility of " chugging." The analysis posits an imaginative release route through a network of eighth-of-an-inch gaps asserted to be separating the fuel box plugs from the surrounding lead bricks, as well as the generation of rather substantial pressures as a driving force to propel water at a rapid rate through those tiny cracks. Even if the reactor somehow managed to avoid fuel molting on the first pulse, it is not going to shut itself down in; the manner posited by UCIA in its analysis, except in the sense that what UCIA has really described is a millisecond by millisecond picture of a steam explosion and core disassembly, with shield plugs and ten-ton concrete blocks being blown into the air and the fuel blown apart.
M o Do the Eighth of an Inch Gans Exist?
When we iMicated in our direct testimony on the potential for fire that some of the graphite logs had gaps between them of roughly 1/8th of an inch, UCIA disputed this, asserting that the gaps were two j
orders of magnitde smaller. Now, in its analysis, it asserts that uniform eighth of an inch gaps exist around the fuel box plugs, on all I
sides. We are not convinced of this from our review of the photogra? s,h and we note that if this assertion is based upon personal recollection by NEL staff members, that recollection must be nearly ten years old, as we understand the last core entry occurred nearly a decade ago. On such an important matter as the existence, unifora dimensions, and actual width and location of cracks relied upon for safe self-shutdown of a reactor, ten year <@i recollections seem thoroughly unreliable. If the gaps exist, are they not perhaps 1/16th of an inch, or 3/32nds? What are the chances that they are indeed uniform, or are truly interconnecting?
l l
S e existence, location, and dimensions of these gaps seems poorly founded in scientific fact. However, this doesn't really matter, because even if they exist, shutdown does not occur through them.
Sinnificant Qaantities of Water Will Not Exit 3 rough Eighth-Inch Cracks Assuming that the aluminum barrier at the top of the fuel box is indeed ruptured, water thrown up against the deflector plate and from there against the lead brick wall will simply splash off with very little water penetration. To force significant quantities of water through those cracks, UCIA postulates large pressure rises. It does so through the use t
of a series of elementary steady-state assumptions for what is a dynamic, short-time-frame event. The Analysis takes the energy level of two open l
pool reactors after their first pulse on a certain period, and assumes
-$7-the entire energy produced is transferred to surrounding coolant, yielding a steam production rate and from there a constant pressure rate. Yet the empirical data from SPERT and 3GIAX testa show that pressure traces were highly oscillatory, certainly not constants furthermore, one cannot assume 100% of the energy produced is immediately 4
transferred to coolant. Sis ignores stes,a blanketing effects and other dynamics of heat transfer in complex turbulent systems on millisecond time scales.
Se analysis states the truisa that pressure will build to that level which permits its release (which manna merely that if the system must explode to relieve the pressure, it will). The analysis back-figures the pressure it believes necessary to expel sufficient water for shutdown, and then e,ttempts to demonstrate that such pressures will be created and will be sufficient to expel 35 liters of water in the time frames of concern. In so doing, the ana3ysis askes a serien of fundamental errors.
l Me Eighth Inch Cracks Are Not Sunamble he Analysis asserts that the available expulsion pathways consist primarily of approximately 20 feet of eighth inch cracks around the deflector plate aperture and the bottom of the fuel box plug (which isassumedtobeliftedandsuspendedindefinitelyinmid-air),aswell as a==all cross sectional area from the overflow lines in the fuel boxes.
Se Analysis suas these, treating twenty feet of cracks as though they were a single circular cross section pipe opening of about 200 cm2 (a pipe of 16 cm diameter!). Assuming that the pressure losses associated with these cracks-which would be immense-are really just the minimal r.,--.,-,
,as,
loss associated with entry and exit from a 16 cm diameter pipe, UCIA estimates that 47 psi would be necessary to remove 35 liters at the necessary velocity.
However, in so doing, the Analysis totally ignores an array of surface effects involving viscosity, junction losses, steam binding, phase mixing, addy currents, and the like. Water will not go through a network of eighth inch cracks as readily as it will a single 6 inch l
pipe. Se circular aperture of a pipe minimises the surface to volume ratio,,thus minimising surface effects that result in energy losses and turbulence. Cracks, on the other hand, almost by definition are the anximisation of surface to volume ratio and thus of surface effects.
Whereas 47 psig may be sufficient to expel water at the rates desired through a single 6 inch pipe opening it would be nowhere near sufficient to do so through the actual non-idealised geometry of cracks. The velocity through those cracks would be very, very high, creating extraordinary turbulence. The elementary pressure formulas used in the Analysis do not apply in situations of high Reynolds number involved with the kind
~
apertures. He problem l
of turbulence that rould accospany these narrow is far more complex than the simple, steady-state, long-time frame formulas used in the Analysis are intended for. Roy any be appropriate for elementary problems like fluid flow through an idealized garden hose, but not for complex geometry-dependent, short-time scale, two-phase problems like we have here. Bere is going to be very little liquid flowing through eighth inch cracks at the suggested pressure, even if the cracks do exist.
Se Cracks Can Not be Treated As a large Aperture Leading from a Tank td a Reservoir The Analysis considers only exit and entrance losses, ignoring
-59 In particular, the Analysis treats the problem everything in between.
an idealized circular as though it occurs in only two dinensions:
cross section aperture is imagined in the reactor fuel box or plus Even if there is a small gap region, with void on the other side.
between the plug and the brick wall, that gap is still facing a brick For water exiting through those sv y sed gaps and ending up at wall.
void spaces several feet away, the water must travel through a three dimensional maze of cracks between solid graphite and lead bricks, making numerous right-hand turns, going through forced restrictions and the like.
The junction losses and so-called " minor" losses will be enormous.
Whereas UCLA assumes one velocity head each for entrance and exit losses, as though this were a pipe without junctions or valves or narrowings or widenings, the actual situation will involve many substantial losses in At 47 psig, between, all of which are ignored in the UCLA Analysis.
even assuming all the cracks exist and all lead somewhere, expulsion would occur over time periods many times longer than those postulated by UCIA.
Most of the Cracks Lead Nowhere_
The failure to do a realistic three-dinensional analysis, doing instead an idealized analysis ignoring what happens to the water when it Having enters the supposed cracks, is symptomatic of a larger problem.
identified the cracks in a solid core, UCIA treated the cracks as apertures But the reactor core is solid logs ani bricks, and the into a reservoir.
Most of the cracks lead cracks are not an aperture into a reservoir.
l nowhere.
1
.=-_
1 i -,.
i In response to the Board's inquiry, UCIA supplemented the analysis, ettempting to identify where the water would go if it did indeed exit through these cracks. Se primary area identified is in and around the overflow lines from the fuel boxes, and the manifolds a l
couple of feet away. What little space there is would be from square pipeways aroutd curved pipes, voids of extremely small magnitude.
l (Se drawing on page 29 of the supplement is somewhat misleading in this regard, as it suggests large spaces on the sides of the pipes whereas the photographs indicate clearly that graphite wedges have been jaaned in i
against those sides. What space there is, is where the square box cut doesn't perfectly fit the curved pipe.).
Se void space outside all six overflow lines only suas to half a liter six liters are said to exist around the manifolds, they are a i
foot and a half away. To even get to the half litar space said to be around the overflow line, water must be tossed at some considerable pressure via the deflector plate against the brick wall, make a ::ight-angle downward turn through a tiny gap that any or may not exist, and find a connecting l
crack that likewise may or any not exist that will permit it to aske another rignt hand turn to a very small space around the pipe. It is quite un-clear that there are cracks that interconnects but if there are, the pressure j
drops will be tremendous and at best a very small amount of water can go there.
Sus, even accepting that there are the cracks the Analysis claims exist, and even accepting that the spaces around and in the pipe exist, the vast anjarity of the cracks assumed to be apertures do not lead to the void spaces assumed to be reservoirs. For example, the fifty inches of perimeter of fuel box plug per fuel slab assumed to be available because the hundred pound plugs are assumed to be lifted evenly in the air and remain suspended there are almost entirely fifty inches facing the wgg_ng direction (away from the idsntified voids in the areas around the pipes). Se only
3 part of the perimeter ihat faces the right direction is facing a crack.
"already taken" by the supposed aperture from the deflector plate opening a fraction of an inch above. One can't get twice the amount of water through the same aperture at the sans pressure assumed for half the unter.
(For that untter, the plug is likely to tilt, not lift, and the whole perimeter wouldn't be available anyway.) Sus none of the aperture supposedly created by the plug lifting is actually an exit to a void space.
Likewise, most of the gap assumed to be created by the deflector plate opening faces no reservoir or void area and is therefore useless. Only the bottom lip faces in the direction of an area-the pipeway around the overflow pipes-identified as a potential void, and it is uncertain whether the cracks actually connect so as to permit passage thereto. Dat leaves the 2
small apertures in the fuel box overflow lines (a paltry 30.4 cm, ignoring the fact that they are already half-filled at the time of the excursion) and the bottom lip of the deflector plate apertures (6 5" x 6xl/8"- 4.875 square inches or 31 cm ), for a total possible aperture of approximately 60 cm, 25%
the amount considered in the Analysis. Brough those cracks, if they inter-connect, and with those pressure losses connected with the high turbulence, l
surface effects, right-angle turns and the like, must flow 35 liters of water (not the easier-to-release steam) into voids that can hold at best a few l
liters, Pressures auch greater than those assumed by UCIA would be necessary.
~
Re obvious observation is that even if the cracks existed, the void spaces in the core of solid blocks don't, certainly in nowhere near the quantity needed for shutdown.
Se argument that the cracks should not be viewed as reservoirs but rather as avenues of transitory passage on the way to the pr<scess pit ani sump 30 feet away fails to understani the time scales involved. It will take
r minutes to reach those distant reservoirs the event will be over long before water could reach there.
It is Not Clear the Reauisite Pressures Would Develon In the Borax tests, pressures greater than 5 Psig did not develop prior to the destruct test. Similarly, pressures greater than 10 psig did not develop in the SPERT ID tests prior to excursions that resulted in molting. If pressure does develop, it occurs as oscilatory pressure spikes and traces, not a steady pressure uatil the water is exponed.
Water Will Not Be Suspended As a Floati u Sl w Pressurized br Steam Below Me analysis somehow assumes that, if pressure does develop, the pressure will be in the form of a block of steam forcing water out of the cracks. However, stama will rise, and water fan, there win not be a clean boundary preventing the denser liquid from faning and the less dense gas from rising. At the ap h a there will be a steam-water mixture, and the primary release will be that of the less dense steam.
Even Assumim Production of 400 Liters Per Second of Steam. it Will Be Primarily Steam Erpelled. Not 140 Liters /Second of Liouid It is evident from basic physical principles, that the less dense a anterial is, the faster it will be able to exit through an aperture given the same pressure. Merefore, whatever pressure arises in the Argonaut, it will be primarily steam, not water, expelled through the cracks. At the pressures considered, steam is approximately 400 times less dense than liquid water, therefore it will exit the apertures (even ignoring other surface effects) about 20 times faster than water.
l t i While a steam-water mixture will exit, it will be primarily steam.
Therefore M 0 liters per second of steam produced will force out not M0 liters per second of water, as assumed in the analysis, but primarily steam.
What About the Incoming Vater?
Se Analysis assumes slightly less than a litar of liquid water is converted to #5 liters of steam per seconi. If, as indicated above, it I
l will be prianrily those M 5 liters of steam per second released through the tiny apertures, then the incoming flow rate of approximately 1 liter per t
second of cold water will simply make up for the water being turned into l
steam. Se system thus has a nochanism for adding water to compensate for l
water lost, and if the water is not lost through some other means, the reaction could go on indefinitely. Furthermore, the incoming water any slightly broaden the initial and subsequent bursts and increase their amplitudes.
l Will Se Ruoture Disk Provide the Avenue of Escane for the Water?
It is possible that the rupture disk any not koak at all, at least not prior to fuel molting having occurred. S e rupture disk is
{
supposedly set to' burst at up to 5 psig above what is defined as normal l
operating pressures. Because of line transients during normal operations, l
that normal pressure is defined as 2-3 psi above atmospheric. Thus, even if properly manufactured and installed, the disk may not be set to koak until 8 psig.
(It should be noted that the rupture disk,like the control blade scram nochanism, is an engineered as opposed to inherent. safety feature, and depends upon proper function, ins +=11= tion, and annufacture.)
Se Borax II reactor, operated with a closed reactor tank, never experienced pressure rises in the steam space above the reactor water of more than about 5 Psi as a result of an excursion. S e SPERT ID pressures
(until the final destruct) were likewise relatively mild. Se locations of the M=hact pressures were near the fuel and pressures thirty feet away through a pipe with numerous turns could be expected to be substantially loss. Since the fuel box plug will tilt at about 1 psig and completely lift at about 3 5 Psig, it is quite possim the rupture disk won't iroak at all, at least not until after the fuel has melted.
Even If It Ruptures. M ll Water Drain in Time?
Even if a pressure spike bursts the disk, the rupture disk is, as the Analysis rightly points out, not a rapid enough route for water expulsion to be of use in the time frames of concern. It is certainly of no use whatsoever in turning the first pulse if there is insufficient void j
space in the fuel boxes. Even at 47 psig, voiding sufficient water for shutdown would take approximately a second (the Analysis makes an error in assuming 47 prig on a 3 inch pipes the available drawings indicate that the openings into the bottom of the fuel boxes are 1" pipe, connecting thereafter toa3"systen.) But as indicated above, it is quite possible that pressures no greater than a few psig will be generated, and then not steadily. Even l
ignoring incoming water from the pump, it is likely to take on the order of l
I 15 seconis for sufficient water to drain once--and if-the rupture disk were to burst. Se first 10 seconis or so of that time would produce no reduction i
in excess reactivity, due to the positive effects of draining water from above the fuel. Vere chugging to occur during that period,.with water t
l tossed up and falling back to be tossed up again, that 15 seconds coula be extended somewhat.
l Can Chugging Occur?
Yes. If the water is not expelled from the system within a vety short time, it is possible that chugging night result.
1
]
.~
49 L
The UCIA Analysis seems to misconstrue chugging as gentle recriticality after the first pulse. Chugging was a form of instability noticed in certain Borax and SPERT runs in which subsequent pulses of relatively high amplitude followed the initial pulse. In some circumstances j
those chugs became divergent, with each pulse larger than the previous one.
In one case, had a control bisde not been broken loose ty the violence of i
the chugging, ending the excursion by accident, the reactor might have been i
severely damaged by the violence of the chugging. It should be noted that g
I i
l it is quite likely that, given the design of the control bisdes and shrouds at the Argonaut and the solid lead and grapite tricks around them, as well as the history of sticking blades due to minor core displacements in the l
past, that an excursion will result in pinning of control blades ant preven-l tion of their eventual insertion in the core rather than fortuitous inter-ventica. It is for that romson, of course, that research reactors are to be inherently safe, so that safety is not dependent upon proper operation of P
engineered features, particularly those likely to be prevented from proper functioning by the very incident for which they any be needed.
Se time scales for chugging at UCIA could be expected to be far l
shorter than the second or so found at the open tanked BORAI reactor, where l
water thrown up had an unobstructed path upwards: water could rise quite high at Borax, whereas at UCIA it can only go two to three inches before splashing tack down into the core.
Thus cycle times of several per second could occur, = Mas several chugging cycles possible before auch water had drained from the core, even assuming functioning of the rupture disk.
Se concept of a slug of water suspended in aid air by a solid block of steam miscomprehends the nature of the excursion phenomenon.
Water would not somehow be suspended above the core, unable to reenter, until the fuel had cooled down. As indicated before, there is not a clean
4&
boundary between steam aM water, but rather steam-mater mixtures. In an excursion, bubbles form, then collapse, then form agains if done coherently, chugging can result. Likewise, water can be expelled out of the fuel plata region, fall back, and be tossel out again.
Chugging provides a means for rapid insertion of reactivity, as water is thrown up and fans baek down, and as bubbles grow and then collapse. @W=r is a kind of coherent pulsing and compensation and pulsing and compensation. Tse energy released can clearly be additive, and temperatures can rise as % chugging progresses. %us, even if the fuel didn't melt on the first pulse, it is quite possible that melting and severe mechanical damage could occur from subsequent pulses.
i l
What About the Assertion %st Fubl Plusts and Sheild Blocks Vould Lift?
As indicated bofore, it quite possible that the fuel box plugs win tilt, although it is highly unlikely thst they win fully lift and remain suspended in air. Wey are tall, heavy plugs in narrow channels lifting pres-sure would be uneven, particularly if via the deflector plater it is es-sentiany impossible for them to lift perfectly straight up, and therefore l
j they win catch on the walls of the lead bricks surrounding them, like.'y pinning the control blades or distorting the fragile 1/8th-inch-thick control hlade shrouds, making them useless even on the long time scale.
Although there may be a fraction of an inch between the top of the fuel box plus and the concrete shield block above, most if not an of that gap is taken up ty the thick lifting handle and bolts at the top of the plug. Even if the plug did lift a fraction of an inch or so, it would do no good for the course of the excursion because, as indicated above, I
the plug is surrounded by trick wall and lifting provides no new aperture leading to a void space. Furthermore, to suggest that void space is created by suspending the plug implies that it rompina suspended forever it will,
l
=,
of course, drop back down, probably in an oscillatory fashion with pressure buildup and release. Any water that finds its way to the extra fraction of an inch space that might be created by lifting the plug win be jammed down a fraction of a second later when the plug fans back lar gravity.
S e Analysis suggests that if pressure near the fuel box top rises to 10 psi or so, the 6 feet of concrete blocks above the core can be lifted "to form new and larger void spaces". Sis is, first of all, wrong numerically, by orders of magnitude. The concrete shield blocks (two of ha) each weigh about 10 tons. Ani of course, even if they could be lifted, this would not create new void spaces, as what Coos up must come down.
While it is true that, with an area of 25 square feet (3600 square inches), pressures of approxiantely 10 psi or so would be sufficient to r
l lift those two ten-ton blocks, it would roeuire 10 nei along the entire 3600 souare inches, not just the 300 square inches of the fuel box tops.
10 psi near the fuel box tops would be insufficient to do more than nudge those ten ton blocks.
Se argument that " water pressure win rise to whatever level is necessary" to fill existing voids, no matter how tiny, or to rearrange h core to make new voids, is simply a way of saying that water will exit via explosion. If there is a significant heat source, as would be the case in a power excursion, and water must get out and pressures are rising, indeed pressures will continue to rise until they create a pathway of release.
But that means core disassembly, and tremendous damage to h reactor.
If it is assumed that ten ton shield blocks win be lifted in W sir, we are talking not about an inherent form of safe self-shutdown, but rather termination of W excursion by rapid disassembly.
The UCLA l
Analysis of how the water exits and W reaction terminates is a description of an explosion.
1 I
b r CONCLUSION 2e model proposed by UC2A for estinating the effects of a major reactivity insertion at the UCIA Argonaut has a number of serious flaws. It attempts to establish a " constant of proportionality" for estimating reactivity effects at one reactor from kinetics data obtained from a significantly different type. Se UCLA model takes an average of five data points as some kind of universal constant, even though it is clearly inapplicable to a reactor so similar to SPERT as BORAX. Application of the model to the unique hybrid design of the Argonaut, in the absence of any kinetics testing data for Argonauts, is unsound.
However, even if one were to use the Ostrander model, i
varying any one of nearly a dozen factors results in fuel molting for the Argonaut. Use of the Borax data rather than SPERT, or eliminating the " mass of the core" factor, or adding error bars, or using other measured values for prompt neutron lifetime or void coefficient, or consideration of positive reactivity effects, or additional energy release that would result from system pressure or time delay in voiding-each of these alone result in core destruction.
Even if fuel melting were not predicted, based on the Ostrander methodology of extrapolation from open-pool, short-lifetime, single moderator reactors, the particular geometry of the closed UCLA system indicates l
nolting. It appears, in particular, that there is insufficient void l'
space available to turn even the first pulse.
Even if molting did not occur on the first pulse, the major difficulties in removing water in a timely fashion from the core mean molting could result from subsequent " chugs."
l l
.... - - -.. -. -.... -. - -. -,.... - -. - - ~... -., - -.
b e 'the shutdown mechanism for the UCLA Argonaut appears to be core disassembly.
i 1
1 i
i I
l
>Q q
VARIANCE CONCLUSIONS Assumption / Variable UCLA Fuel Melting Indicated Yes
. N, o effect of UCLA's long neutron lifetime (SPERT II experience) xx use of Forbes model w/o Ostrander " mass of the core" revision xx use of Forbes model with Ostrander revision, using. Borax I data xx use of Forbes model w/ Ostrander revision, using SPERT I data xx with error bara use of Forbes model w/ Ostrander revision, using SPERT I data, no error bars, Ostrander 27 7d void coefficient,but L{6 of 20 xx -
useofPorbesmodelw/Ostranderrevision,usingSPERTIdata, e
no error bars, OstranderL(6 of 29 2, void coefficient of 10g xx j
useofForbesmodelw/Ostranderrevision,usingSPERTIdata, no error bars, Ostrandert/&and void coefficient, but adding positive reactivity effects (bowing, positive voids, graphitecoefficient) xx use of Forbes mMel w/ Ostrarder revision, SPERT I data, Outrander Lt6and void coefficients, but increase in power due to time delay in shutdown associated with backpressure.
- ime for breaking of aluminum " membrane", etc.
xx use of full Ostrander assumptions, but insufficient void space to turn first pulse xx use cf full Ostrander assumptions, but no mechanism for expelling xx sufficient water in timely fashion to produce fihal shutdown Note: The above are inriependent variables, taken singly.