 |
| Titel |
Bayesian estimation of a source term of radiation release with approximately known nuclide ratios |
| VerfasserIn |
Ondřej Tichý, Václav Šmídl, Radek Hofman |
| Konferenz |
EGU General Assembly 2016
|
| Medientyp |
Artikel
|
| Sprache |
en
|
| Digitales Dokument |
PDF |
| Erschienen |
In: GRA - Volume 18 (2016) |
| Datensatznummer |
250126452
|
| Publikation (Nr.) |
 EGU/EGU2016-6176.pdf |
|
|
|
|
|
| Zusammenfassung |
| We are concerned with estimation of a source term in case of an accidental release from a
known location, e.g. a power plant. Usually, the source term of an accidental release of
radiation comprises of a mixture of nuclide. The gamma dose rate measurements do not
provide a direct information on the source term composition. However, physical
properties of respective nuclide (deposition properties, decay half-life) can be used
when uncertain information on nuclide ratios is available, e.g. from known reactor
inventory.
The proposed method is based on linear inverse model where the observation vector y
arise as a linear combination y = Mx of a source-receptor-sensitivity (SRS) matrix M and
the source term x. The task is to estimate the unknown source term x. The problem is
ill-conditioned and further regularization is needed to obtain a reasonable solution. In this
contribution, we assume that nuclide ratios of the release is known with some degree of
uncertainty. This knowledge is used to form the prior covariance matrix of the source term x.
Due to uncertainty in the ratios the diagonal elements of the covariance matrix are considered
to be unknown.
Positivity of the source term estimate is guaranteed by using multivariate truncated
Gaussian distribution. Following Bayesian approach, we estimate all parameters of the model
from the data so that y, M, and known ratios are the only inputs of the method. Since the
inference of the model is intractable, we follow the Variational Bayes method yielding an
iterative algorithm for estimation of all model parameters.
Performance of the method is studied on simulated 6 hour power plant release where 3
nuclide are released and 2 nuclide ratios are approximately known. The comparison with
method with unknown nuclide ratios will be given to prove the usefulness of the proposed
approach.
This research is supported by EEA/Norwegian Financial Mechanism under project
MSMT-28477/2014 Source-Term Determination of Radionuclide Releases by Inverse
Atmospheric Dispersion Modelling (STRADI). |
|
|
|
|
|
|