Artificial neural network based direct non-destructive compositional analysis of (Th, U)O₂ fuel pellets by X-ray fluorescence technique

Mixed Oxide (MOX) fuels like (Th,U)O₂ and (Th,Pu)O₂ are the proposed fuels in Advance Heavy Water Reactor (AHWR) which is a forerunner to demonstrate the thorium fuel technology in the third state of Indian nuclear power programme. X-Ray Fluorescence (XRF) is a well known non destructive analytical technique with aided advantage of multielemental analytical capability. Here we have exploited the applicability of ANN based micro-XRF technique for direct, non destructive compositional analysis of MOX fuel pellets samples. We obtained from the present study accuracy for the U determination in the validation sample of 3.40%, 3.19% and 5.56% for region 1, 2 and 3 respectively. Similarly the precision's for the U determination in the validation sample are 4.40%, 1.90% and 6.52% for region 1, 2 and 3 respectively. Superiority of region 2 in terms of precession and accuracy of the results can be explained due to higher counting statics of U/Th Lα1 line