Speaker
Description
Authors: Miroslav Hýža 1, Lenka Dragounová 1, Mahulena Kořistková 1
Affiliation: 1 National Radiation Protection Institute (NRPI), Czech Republic
For atmospheric radioactivity monitoring, the emphasis is on two conflicting requirements: sensitivity, and timeliness of reported results. This paper describes a sensitive method for quick screening measurements of aerosol filters, using a HPGe detector and accounting to high and variable natural background.
During a routine monitoring regime, filters from a high volume sampler (900m3/h) are being exchanged in intervals of several days. The withdrawn filters are then set aside for several hours prior to the laboratory measurement to let the 222Rn/220Rn daughters to decay. Not doing so significantly reduces the sensitivity of the measurements.
Delaying the measurement by several hours has an effect especially on the activities of 214Pb and 214Bi with half-lives of 26.9 min and 19.8 min. Thoron decay products 212Pb, 212Bi and 208Tl decay on the filter with a half-life of 10.6 h, which is not as significant considering the approx. 8-hour deadline for result reporting. This delay before placing the filter on a detector can therefore take up a significant part of the total time available for analysis.
For this reason, the possibility of exploiting also the time reserved for the decay of daughter radionuclides was investigated. Experimental measurements were performed using a 100% HPGe detector (FWHM 2 keV) whose dead time with a fresh filter reached up to 25 %. The high natural background is also very variable – affected by the concentration ratios of 222Rn/220Rn decay products (DP) in the air, and the sampling time.
The proposed method is based on the precise and reliable DP background subtraction, using a full-spectrum analysis instead of the classical peak search algorithm. Our approach is based on several machine-learning algorithms - mainly the Principle Component Regression (PCR) and Autoencoder neural network, which fully exploit the autocorrelation structure of HPGe spectra.
Apart from the higher sensitivity, a major feature of this method is the ability to deal efficiently with spectral interferences, i.e. 137Cs (661.6 keV) and 214Bi (661.1 keV, γ=0.054 %), or with the contributions of β+ active radionuclides to the 511 keV annihilation peak.
