Principal Component Analysis applied to neutron dosimetry based on PADC detectors and FNTDs

Abstract

Solid State Nuclear Track Detectors (SSNTD) have been used as passive dosimeters for many decades. Fast neutron detection is performed by detecting recoil protons generated in a hydrogenous material (neutron converter). The dose is estimated by multiplying the recoil protons track-spots density present on the detector surface by a calibration coefficient obtained from reference irradiations. Therefore, one needs to automatically and reliably discriminate between the track-spots induced by recoil-protons and the spurious signal engendered either by material imperfections or by other ionising particles interactions. The objective of this work is to demonstrate and apply a method, based on the multivariate statistical tool named Principal Component Analysis (PCA), aiming at identifying and filtering out the spurious signal in track detectors. The advantage of this approach is that it can be applied to any type of SSNTD, provided apt definition of initial physical variables describing each track-spot. To show that, we applied the method to poly-allyl diglycol carbonate (PADC) detectors, to filter the signal generated by material imperfections, and to fluorescence nuclear track detectors (FNTDs), to remove the signal induced by other radiation field components. After filtering, dose assessments were compared with reference values of exposures, showing satisfactory agreement. In case of the FNTDs, the procedure proved to be effective regardless of the crystal colouration, which is known to affect signal-rejection techniques based on track-spots fluorescence intensity. The technique also offers the advantage of self-adjusting the filtering parameters based on the currently analysed detector set, as long as calibration and unirradiated detectors are included in the dataset.