ML19250B406
| ML19250B406 | |
| Person / Time | |
|---|---|
| Issue date: | 10/19/1979 |
| From: | Goldberg F, Jerrica Johnson, Oh C Office of Nuclear Reactor Regulation |
| To: | |
| Shared Package | |
| ML19250B405 | List: |
| References | |
| NUDOCS 7910300474 | |
| Download: ML19250B406 (4) | |
Text
'
TB?A
SUMMARY
RISKSOFPLUT0NIUMFUELFACILITYQAMAGEFROMHIGHWINDS J. Johnson, C. Oh, F. Goldberg Probabilistic Analysis Staff Office of Nuclear Regulatory Research U.S. Nuclear Regulatory Commission Washington, DC 20555 Backaround and Statement of the Problem _
Wind damage to plutonium fuel processing facilities could release and disperse As plutonium oxide (Pu0 ) causing a radiological hazard to the population.
2 a result of this potential hazard, NRC rules r.equire that existing. plutonium fabrication facilities be examined to determine their ability to withstand adverse natural phenomenon such as high winds.
In connection with this I
requirement, the Probabilistic Analysis Staff conducted risk analyses to aid in characterizing health hazards associated with postulated wind damage to
. s.
a specific fuel fabrication facility.
This paper describes the techniques utilized in the risk analyses to help supply The risk analysis utilized information to the licensing decision process.
Extreme value theory provided extreme value theory and isotonic regression.
a technique to derive point and interval estimates of high wind probabilities while isotonic regression provided a nonparametric method to smooth the risk curves derived from extreme value theory.
886 059 Approach The risk measure defined for this analysis was the probability per year of exceedin dose resulting from postulated structural damage caused by high a population Pu02 Lungs and bones were the organs of concern-because they dominate the 5 winds.
7910300 I
,_wmm
i
_2 year committed dose equivalents from inhalation.
Peak-gust wind speed data for a
~'
26 year period, 1950-1975, were' acquired from the closest meterological station to the facility.
The extreme value distribution (type 1) was used to model the variability in the yearly maximum wind speeds.
For a large class of sampling distributions, if the observations are independent,2 the distribution of the, maximum value of a random sample tends toward an extreme value dis,tri-bution as the sample size increases.
In addition to the theoretical bases, various goodnesses of fit tests were conducted which did not reject the hypothesis that the peak-gust wind speed data was representable by the extreme value distribution.
Utilizing the extreme value distribution, the probability F(x) that the maximum annual wind speed is less than some value x is:
F(x) = EXP(-EXP(- x(x - u),
where w, u are distribution parameters to be estimated.
.s The probability that the maximum yearly wind speed eyceeds x is given by the complimentary cumulative distribution (1 - F(x)).
The extreme value model provided estimates of upper confidence bounds, point estimates, and lower confidence bounds for the probabilities that maximum wind speeds would be greater than various values of intere.st.
Probabilities for wind speeds of 38 m/s, 42 m/s, 49 m/s, and 58 m/s were specifically selected because a structural and release analysis estimated release and dispersion magnitudes and probabilities as a function of these speeds.
886 060
As a part of the risk analysis, a separate consequence calculatiun estimated
~
the structural damage, releases and subsequent, radiological. doses which would result from the wind speed occurrences.3 The best estimate dose and an upper Each confidence bound dose were specifically obtained for each wind speed.
best estimate, lower bound, and upper bound wind probabilities were multiplied by the two release probability values for each wind speed to obtain a represen-tation of the possible probability and consequence points.
From past ex'perience and WASH-1400,4 plots of the probability density function v?rsus consequence indicated that as consequence severity increases the associated probability decreases.
Isotonic regression was therefore selected..to smooth the derived set of points of probabilities and consequences to obtain an empirical probability density function which continually decreases as consequence increases.
Isotonic regression is a least squares smoothing technique which minimizes the 5
mean square loss when a monotonically decreasing density function can be assumed.
A complimentary cumulative distribution was then constructed from the smoothed density function and a risk curve derived.
A computer program was developed to execute the numerous calculations required in the analysis and to plot the results.
The procedure is generally applicable to all random phenomena requiring only that the distribution of the phenomenon be specified.
To facilitate identification of the distribution of choice, another computer program was developed to perform several goodness of fit tests for some Example results of the analysis are displayed in Figure 1.6 common distributions.
886 061
'f
~
.A.;
HIGH WINO RISK RNRLYSIS FOR Bt.H LEECHBERO FUEL PROCESSING SITE SM00THE0 CC0F 100 7
5 i
L 10-1 7
5
-E-3 10-2 c.
m 5
2 J
t 10-3 :
s i
1 s
o J
l 5
l 3
i s
a 10-4 i g
i i
5 3
3-10-5 i 7
~~
5 g
3 I
10-8 7
5 3
10-7 7
5 5
i l
3 10-8 i, 3
3 3
3 3
05 50YRDOSECOMMITMEN@kP[ BON-REM)