Applcability of Data cont'd

DOE-HDBK-3010-94

1.0 Introduction

for probabilistic assessments. The overall collection of data available for a wide variety of

stresses will not support fine statistical resolution as a technically meaningful activity, and

this handbook specifically rejects citation as a defensible basis for such attempts.

The generation and suspension of particles is the result of the interaction of multiple

physiochemical variables that have not been completely characterized as the majority of the

experiments performed were designed in an attempt to reflect reasonably bounding conditions

for specific industrial situations of concern. Accordingly, the data obtained are more

accurately characterized as selected points from multiple distributions against multiple

parameters than as different values from a common distribution. Even if this point is

neglected, there are still practically intractable problems in attempting to generate statistical

distributions. While the data are presumed to be bounding for the purpose intended, it is

largely unknown whether the data values are truly 90th percentile, 99th percentile, 99.9th

percentile, etc. Further, in many cases it is considered likely that accident specific ARFs are

actually distributed in a highly irregular manner (i.e., multi-modal or truncated distributions).

Assuming a typical distribution (i.e., log-normal, Poisson) using standard deviations will

produce seriously distorted values that may have little or nothing to do with reality.

The available data do, however, cover a range of conditions that typify the energy sources

associated with nonreactor nuclear facilities. The data cover a more complete range of

phenomenological concerns than the data upon which nuclear reactor source terms have been

estimated. They are at least the equal of reactor source term data in overall quality, and a

number of the experiments performed were very close to actual scale for the type of

operations conducted at nonreactor nuclear facilities. In general, scaling effects, while not to

be trivialized, are less of an issue with this data than with comparable reactor source term

estimation data. The NRC has already accepted a considerable amount of this information as

a basis for source term estimation in NUREG-1320, "Nuclear Fuel Cycle Facility Accident

Analysis Handbook." DOE-HDBK-3010-94 serves a similar function for DOE.

The values for ARFs and RFs taken from experimental measurements are reasonably well

defined. It is noted, however, that they are dependent upon the types and levels of stress

imposed, the initial state (physical form, chemical composition, particle size distribution,

degree of dispersion of the material-of-concern), and the response of the material-of-concern

and other materials present. In most cases, the materials chosen for the experiments were

selected to bound the behavior of materials under accident conditions for a specific location

or process. The applicability of the experimental conditions to the complete spectrum of

processes and potential accident conditions is, however, uncertain. For this reason, the

discussions of the data have tried to indicate where typical industrial accident phenomena

(e.g., fires, explosions, spills) are considered covered by the data.

Page 1-8