Re: MRFA

[RuthWeiner@AOL.COM]

This is for your information -- please do not take it as an argument for or against anything.  It is to inform RADSAFERs. briefly, how RAM transportation accident  risks are analyzed at present, and what the reasoning is behind this analysis.  Clearly, since there have been no accidents with Type B casks that involve releases, we do not have direct evidence of what would happen in an accident like that.  However, as in any risky endeavor (which any transportation is) we are frequently required to make some projections or estimates of risk.  The method, which I outline below, was pioneered in NUREG-0170 (1977)and is, to my mind, a clever way to obviate arguments about specific  accident scenarios. Please excuse the length of this post, but it may answer some questions.


Accident releases, and consequent radiation dose risks, for accidents whose severity exceeds that of 10 CFR Part 71 Subpart E are estimated as follows:


1.  Consider the entire universe of possible accident scenarios.  From general truck and rail accident data, better than 99%  of the possible accidents have no impact on the cargo at all, and would have no release or loss of lead shielding.  Current data and studies indicate that at least 99.99% of possible accidents fall into this category.

2.  The remaining 0.001% are classified into classes defined by two parameters: mechanical force exerted during the accident and temperature if there is a fire.  When accidents are defined by a scheme like this, any scenario can be represented by a combination of mechanical force and temperature, and one can dispense with conjuring up scenarios.

3.  Each mechanical force/thermal stress class has an associated conditional probability.  

4.  NUREG-0170 used engineering judgment to arrive at a matrix of accidents.  The 1987  Modal Study (Fischer et al,  NUREG/CR-4829) analyzed a large number of heavy truck and rail accidents and developed an event tree and a two-dimensional matrix, with dimensions of strain and temperature, and estimated releases under these conditions, using cask designs and cask tests as a basis for the estimates.   NUREG/CR-6672 (Sprung, et al, 2000) re-examined the Modal Study, and using more recently developed accident and cask test data as well as finite element analysis, refined the Modal Study matrix, used impact speed and temperature as the two parameters, and developed a matrix of 19 cells for truck and 21 cells for rail, for four "generic" Type B casks.  Each cell has an associated conditional probability of occurrence and release fractions for volatiles, Ru, CRUD, particulate matter, and inert gas. (I can send anyone interested a copy of these !

matrices electronically).  The conditional probabilities and release fractions may be found on pp. 7-73 to 7-76 of NUREG/CR-6672. Slightly more than 99.99% of the accidents defined in this way have neither release nor loss of lead shielding.

4.  The accident analysis then calculates the risk associated with each cell by  multiplying each release by the conditional probability of the combination of impact speed and temperature that could have produced that release.  These conditional risks are then summed, multiplied by accident frequency (e.g., from Saricks and Tompkins, 1999) and a "dose risk" calculated from the resulting "release risk" by applying a dispersion and deposition model, dose conversion factors, etc.  The details of this calculation may be found in the RADTRAN 5 Technical Manual (Neuhauser, Kanipe and Weiner, 2000) and in the Transportation Health and Safety Calculation Package of the Yucca Mountain EIS.

5.  All of the conditional probabilities of accidents that would include a release are estimated at less than 1E-4; the largest is the fire-only scenario, which has an estimated conditional probability of about  6E-5.  However, the matrix of accidents does include at least one accident of conditional probability about 1E-7, and the associated combination of impact speed and temperature  may be considered the "maximum reasonably foreseeable accident."   Another way of arriving at this accident is to make a plot of accident conditional probabilities and see where the 1E-7 cutoff point is.  At any rate, the potential release and consequent dose can then be calculated


OK it's not a perfect method, but in the absence of actual accident  release data, I think it's pretty good.  And (please take note, Sandy!) we cannot say that such accidents will never happen.  There are combinations of impact and temperature that can be estimated to be so unlikely that their conditional probability is zero, but the method also considers conditional probabilities of the order of 1E-14, 1E-10, etc.


Before anyone launches into a broad-scale critique of this method, I suggest you familiarize yourselves with the details, using the cited references.    What I have given is a necessarily brief overview.  However, I would certianly welcome suggestions for improving this method, making it more realistic, and so on.


Finally, the limit of "reasonably foreseeable" as 1E-7 is admittedly arbitrary, but is related to the EPA credibility limit and is parropriat4e, considering the latter.



Ruth Weiner, Ph. D.

ruthweiner@aol.com