A Mobile System for Detecting Gamma-Radiation Sources: I. Physical and Statistical Background

Ch.B. Lenev³, M.G. Mitev², L.T. Tsankov¹

¹Dept of Nuclear Engineering, University of Sofia, Bulgaria
²Dept of Electronic Technique, Technical University of Sofia, Bulgaria
³Institute of nuclear research and nuclear energy, Bulgarian academy of science, Sofia, Bulgaria

A mobile (car borne) system for searching lost or 'orphan' radioactive sources is discussed. Theoretical limits for the minimal detectable gamma-source activity as a function of: the relative source-detector velocity; the minimal source-detector distance; background level; integration time of the signal etc. are determined. It is assumed that a scintillation detector based on a 3"x3" NaI(Tl) crystal is used and that the measured range is 60-3000keV.

All parametric characteristics of the system are considered for two registration modes: integral counting mode or spectral registration mode. The spectrometric version of the system, despite its complexity, yields to better analytical possibilities - not only to obtain information about the predominating gamma-rays energy emitted from the source (eventually to identify the radionuclide), but also its sensitivity is improved with respect to that of a equivalent radiometric system.

In order to overcome discrepancy between the requirement for a fast reaction time (within 1-2s) and the statistical reliability of the analysed counts, it is suggested to divide the whole energy range into a few (4-16) sub-ranges, i.e. to work in a semi-spectrometric mode.
A maximal sensitivity can be achieved if continuous measurements of the background and real time estimates of its level are performed in order to properly subtract the background from the whole signal. An optimal system corresponds to a maximal signal/noise ratio. Beyond the trivial dependence of signal/noise ratio on the source activity, detector surface and the minimal distance source-detector, a maximum can be searched also with respect to the length of the measurement time interval.

Adaptive algorithms are proposed for real time analysis of the measured data for both integral counting (single channel) mode and the spectral registration (multichannel) mode. It is shown that the problem of tuning the system sensitivity is reduced to classical statistical tests (paired T-test for single channel mode or chi-squared test for spectral registration mode).