Data and Documents

REALISM -
In Engineering and Analysis of the Sources and Consequences of Radioactivity in Using Nuclear Technologies

ANS President’s Special Session:

Realism in Evaluating Nuclear Hazards

Contribution by Ian B. Wall

15 June 2004

·        Thank you, Ted.  Having been born and raised in Great Britain, I have never learned the lore and terminology of baseball.  In replacing Milt Levenson (former ANS President) in this session, I was going to describe myself as a ‘pinch hitter’ until my son informed me that pinch hitters often replaced softer part of the lineup.  Milt was one of the most distinguished bosses that I ever had.  Far from being a heavier hitter than Milt, you’re getting a bench warmer!  However, I am confident that Milt will agree with my remarks.  Am I batting 0.300 on baseball terminology?  Since I was only recruited 10 days ago, only one slide and no handout.  Copies of a related paper by Bob Henry and Frank Rahn are available.  My slide quotes Milt in a 1981 special issue of Nuclear Technology.  His statement is as pertinent today as it was 20 odd years ago.

·        I became involved in risk assessment in 1967, when GE assigned me to examine the use of probability in reactor safety.  Allocated small budget and few part-time staff to do a micro-WASH-1400.  GE policy required Technical Risk Evaluation for major sales; still true today.  Existing deterministic approaches were inadequate.  Specifically, GE had a health physicist/meteorologist estimate potential damages in the event its BWR had an accident.  He used a TID-14844 release and hypothesized that, for any site in Eastern US, the plume would pass over every major city during an inversion.  Wherever the plant was sited, the damages were equally large.  GE wanted to bid on a plant in Switzerland that was across a river from a major city.  Using his methodology, the damages were colossal and unacceptable to GE management.  I wrote a computer program, which modeled actual time-dependent weather, assuming the same TID-14844 source term.  What we found was that changes in the weather and the incidence of rain caused the damages to quite localized to the plant and that damages were much smaller and varied significantly from site-to-site.  GE never allowed me to publish those results.  When I joined the NRC in 1974, a major assignment was to correct a serious error in the draft WASH-1400 consequence model.  We wrote the more sophisticated CRAC-2 code, which is still used today around the world.  It gave a similar result.  Consequences are concentrated close to the plant, there is time to evacuate, and risk becomes very small at distances further away from the plant.  I remember a couple of colonels visiting me to discuss this phenomenon.  The Pentagon was interested with respect to use of tactical nuclear weapons on the battlefield!  The point is that realistic modeling changed our perspective about off­site consequences.

·        By introducing realism, WASH-1400 certainly changed our perspective on what is important to reactor safety.  Since its results are so widely known today, I won’t dwell on them.  I will merely observe that, prior to WASH-1400, the consensus of experts was that the probability of core damage was infinitesimally small and the consequences would be catastrophic.  WASH-1400 showed that the probabilities were larger than expected but that the consequences were finite.

·        Prior to TMI-2, safety experts assumed that released iodine would be elemental and gaseous and therefore a large fraction would be released to the atmosphere.  For the TMI-2 accident sequence and degree of fuel melting, the then current licensing models would have predicted the release of millions of Ci of iodine to the atmosphere.  In fact, only 15 to 30 Ci of I-131 and I-133 were released.  Dave Campbell, Tony Malinauskas, and Bill Stratton cogently argued that the iodine combined with cesium to form hygroscopic CsI, which dissolved in the huge quantities of water and steam present.  Therefore, it was less available for release.  Milt Levenson also had this insight and further argued that many fission products would form aerosols, which would plate out within the primary system and containment.  In November 1980, Chauncey Starr (another former ANS president), Milt, and I briefed the NRC Commissioners on these subjects.  In December, Milt and I briefed President Carter’s Nuclear Oversight Committee on the subject.  Again, realism about the chemical form of iodine has further changed our perspective on consequences of reactor accidents.

·        In those briefings, my contribution was to layout program of experiments to charac­terize and to measure the retention of radioactive material within the fuel, within the primary system, and within the containment.  I set as a goal to demonstrate a ten-fold reduction of the WASH-1400 source term.  This program of work was executed by EPRI during the 1980s by Dick Vogel’s and my staff including Ed Fuller, Mati Merilo, Frank Rahn, Bob Ritzman, Raj Sehgal, with the support of the late Miles Leverett (former ANS President) and Prof. Rudy Sher.  Many of these projects are briefly described and referenced in the paper by Henry and Rahn.  Most were done in conjunction with NRC Research.  Many were sponsored by international consortia of up to 17 countries, having most of the commercial nuclear power plants in the world.  Thus, these data are widely available within the nuclear community.

·        What has been accomplished with these data?

-         They have been used to benchmark the MAAP code that utilities used to do their IPEs and now use for their PRAs, when Level 2 is required.

-         NRC used them in doing NUREG-1150, an update of WASH-1400.  As shown in NUREG-1420 (Kouts’ Committee review of NUREG-1150), it reduced the WASH-1400 source term by a factor of 20.  A modest improvement over my goal of 10 years earlier.

-         NRC revised the TID-14844 source term.  This source term was specified in 1962 and has been the design basis for NRC’s Part 100 siting criteria ever since.  It has also been used as the design basis for equipment qualification, post-accident control room habitability, etc.  It assumed that iodine would be predominantly an elemental gas and that 50% of it would be instantaneously released to containment, of which 50% would be available for release to the atmosphere.  These assumptions significantly influenced the design of engineered safety features.

The 1995 design basis source term, as characterized in NUREG-1465 and Regulatory Guide 1.183, is much more realistic.  Although the overall release fractions to containment are little changed, they occur over periods of 2 or 3 hours.  More significantly, iodine is assumed to be combined with cesium as hygroscopic CsI.  Consequently, most iodine will be dissolved in water and what is airborne is likely to be in aerosols, which may be scavenged out by sprays.  The bottom line is that much less iodine is likely to escape the containment into the atmosphere than predicted by using a TID-14844 source term.

The impact of the updated design basis source term on existing and advanced LWRs has been significant.  In existing plants, the updated source term has been widely applied to show that control room and offsite design basis accident dose limits can be met while simplifying plant operations.  In advanced PWRs (e.g. AP600, AP1000), the updated source term has allowed credit for natural aerosol removal within containment, thereby eliminating need for a passive, safety-related containment spray system.

·        The practice of reactor safety has come a long way during the past 30 years.  The designers of existing nuclear power plants lacked data and operating experience so they used very conservative, bounding assumptions.  The TMI-2 accident validated this practice since, despite an event far exceeding the design basis accident, no member of the public received a significant radiation dose.  However, TMI-2 also highlighted flaws in such conservatism:  Then existing licensing codes were unable to predict the actual behavior of the reactor and safety analysts were not focusing on important contributors to risk.  Today, we have much more reactor operating experience and have accumulated considerable experimental data on severe accident phenomenology.  We can and are doing a much better job.  Reactor design should continue to be conservative.  But it should be supported by probabilistic risk as­sessments.  To be effective, these risk assessments should be as realistic as possible.

·        In closing, I quote Milt Levenson again.  “Whenever consequences are grossly over­stated, they result in actions being taken that introduce risks of their own - risks that would not otherwise have been taken.  They reduce safety.  Overstating consequences by orders of magnitude is not conservative.  It is wrong.”  Thank you.