1.
Scott D. Portzline
Three Mile Island Alert
Harrisburg PA
September 2013
1

2.
If the Nuclear Regulatory Commission used
its faulty accident probabilities to run a Las
Vegas style casino, they’d have been
bankrupt in the first year.
2

3.
simply stated:
The NRC’s probabilities are an analysis of
an event’s likelihood of occurring over a
period of time.
3

4.
Las Vegas can bank on it’s accuracy
because there is far more certainty in the
mathematics of systems that do not
involve breakdowns, human errors,
electrical failures, mis-management etc..
Of course the NRC has been unable to
calculate those factors with accuracy.
4

5.
Governments need to know what risks are
associated with nuclear power. Therefore,
the nuclear industry creates favorable
probabilistic risk analyses in order to get
the governmental “green light.”
Here is an example….
a
5

6.
The NRC’s first probabilistic risk study,
the Rasmussen Report, was used to
persuade Congress to extend the
insurance policy for the nuclear power
industry.
6

7.
The Rasmussen Report gave the odds as
1 chance in 1 million per reactor per year
7

8.
A mathematician will use a probability
risk study to say something like ….
“You are more likely to be struck by a
meteor than to have a nuclear accident.”
(We were told this at Three Mile Island.)
8

9.
March 28, 1979
9

10.
Just two months prior to the TMI accident
the NRC was forced to disavow its
accreditation of the Rasmussen Report.
10

11.
However, the NRC could not say if the
odds of an accident were greater or less
than Rasmussen had predicted.
This amounted to an admission that it did
not understand its own study or which
side of the fence to fall on.
11

12.
The biggest problem was that the technical
data did not support the conclusions touted
in the executive summary.
This caused one NRC Commissioner to state
that he thought the Rasmussen report was
used for propaganda to persuade Congress.
12

13.
Then another analysis after TMI concluded:
“…about 20 factors contributed to the
damaging results at Three Mile Island. If any
one of these factors had been different -- in
a way which is common in other plants -there would have been no core damage and
no release of radioactivity.”
Electric Power Research Institute
March 1980 13

14.
In other words, all 20 factors came up against all odds.
Vegas would never let that happen.
14

15.
With long-shots like that, Vegas would be
“broke” many times over.
15

16.
After the TMI accident, still another analysis
showed that the odds as presented in the
Rasmussen Report were accurately predictive of
the occurrence of the TMI accident. Only, now
the odds were described as 1 chance in 7.7 of an
accident occurring.
Report of the Technical
Assessment Task Force Vol. 1
Kemeny Commission 16

17.
Interpretation of Rasmussen Report before TMI
vs. afterwards
17

18.
What government would green light a
project with a 13% risk of having a severe
nuclear accident?
18

19.
As you can see, interpretations of the
probabilities for the same incident are
“all over the map.”
19

20.
WANTED: MATHEMATICIAN
An aptitude test was given to three
applicants at a public relations firm.
Question: What does 1/3 mean to you?
The purest answered, .33333
The theorist answered, “one third”
The statistician answered, “What do you
want it to say?”
20

21.
PRAs do yield numerical estimates and are thus “quantitative.”
Some scientists believe the estimates are so imprecise and
subject to manipulation as to be virtually useless in decision
making.
Human Factors Reliability Benchmark Exercise,
Commission of the European Communities, 1989
21

22.
Estimates for the occurrence of “human
error” have varied by a factor of ten
thousand.
Human Factors Reliability Benchmark Exercise,
Commission of the European Communities, 1989
That means that older NRC accident PRAs could
be off by 4 orders of magnitude on the human
aspects alone.
22

23.
The more recent international HRA empirical
study in 2011, sponsored by the NRC has found
that, twenty years later, large variability appears
to remain.
Science-Based Simulation Model
Of Human Performance for
Human Reliability Analysis
Idaho National Laboratory
23

24.
The human contribution to the risk of
operation of complex technological systems
is significant, with typical estimates lying in
the range of 60-85%.
Case Study Report on Loss of Safety
System Function Events, AEOD/C504,
Washington, DC: U.S. Nuclear Regulatory
Commission, 1985
Therefore PRA’s cannot be accurately modeled,
since human behavior varies greatly, even for the
same event.
24

25.
The conclusion(s) are not always supported
by the data.
25

26.
Some data has been cherry-picked while
other data is ignored.
26

27.
The data is applied to faulty Modeling
27

28.
Incomplete processing
28

29.
All four of these deficits are found in the NRC’s
latest severe accident study called SOARCA.
They are described in detail with examples in
my report.
http://www.efmr.org/files/2012/SOARCA-2-22-2012.pdf
SOARCA is another example of a probability
study gone awry.
29

30.
The NRC provides false conclusions to Congress.
The NRC Commissioners testified that a
Fukushima-type accident can not happen in the
US. Flooding by tsunami triggered the accidents.
However, one of their own PRAs shows that there
is a 100% chance of a meltdown at a specific US
plant (unnamed here for security reasons) if a nearby
upstream dam were to fail.
30

31.
The confidence of the nuclear industry is falsely
bolstered by exaggerated probabilities.
Furthermore, these unwarranted confidences
morph into an attitude that an accident is possible,
but it won’t happen “here” or “any time soon.”
Mantra: “defense in depth.”
31

32.
These attitudes created conditions that led to
Three Mile Island and Fukushima.
Specifically: At TMI, the industry banked on
numerous backup systems which were not
available, did not work properly, or were disabled
prior to and during the accident. “Defense in
depth” did not succeed. (see slide #13)
At Fukushima, the industry and the government
banked on the probabilities of Tsunami height.
The entire chain after that point was defeated.
32

33.
There is a long chain of safety systems with
many variables. Too many to accurately quantify
in a probabilistic risk analysis.
The NRC is able to understand and quantify
certain (but not all) shorter segments of the
chain.
However the NRC cannot quantify the risks for
the entire chain.
33

34.
Insurance companies and Las Vegas are very
good at probabilities. Neither one is in the
nuclear game for good reason.
The NRC should not place a heavy reliance upon
PRAs in their licensing and safety analyses and
processes.
The NRC should likewise not use PRAs in
testimony to governmental bodies until the time
comes when they have a far better track record.
end
34