Journal article Open Access

DEEP LEARNING-BASED ANOMALY DETECTION IN NUCLEAR REACTOR CORES

Thanos Tasakos; George Ioannou; Vasudha Verma; Georgios Alexandridis; Abdelhamid Dokhane; Andreas Stafylopatis

In this work, a methodology is proposed for the classification of different perturbation
types and their position in a nuclear reactor core. More specifically, it is based on a Convolutional
Neural Network architecture that identifies and locates specific perturbations,
given the spectrograms of detector signals as input. Training samples have been provided
by the SIMULATE-3K code, that simulates reactor core conditions. The different perturbation
types considered are (i) realistic fuel assembly vibrations at different positions
in the reactor core, (ii) fluctuations of inlet coolant temperature, (iii) fluctuations of inlet
coolant flow and finally, (iv) combinations of the above sources. A complementary robustness
analysis of the proposed architecture was performed to assess its performance in
the cases of noisy or missing data. The trained model has subsequently been utilized on
measurements obtained from the G¨osgen Power Plant in Switzerland, for an assessment
of its functionality.

Files (513.4 kB)
Name Size
2021_Tasakos_MC2021_paper_V1.pdf
md5:ce0f806c8a9d34942f1dd7944d4233d8
513.4 kB Download
10
5
views
downloads
Views 10
Downloads 5
Data volume 2.6 MB
Unique views 10
Unique downloads 5

Share

Cite as