PhD Defense | Victor Casas Molina | Bayesian Machine Learning for Accurate Characterization of Radioactive Waste

Title: Bayesian Machine Learning for Accurate Characterization of Radioactive Waste

Nuclear energy and nuclear technologies provide an efficient source of low-carbon energy and support many important industrial and medical applications. However, their use also generates radioactive waste, which may pose risks to human health and the environment if it is not properly characterized and managed.

Although radioactive waste management practices have improved significantly over the past decades, an important challenge remains in the characterization of legacy waste. In many cases, this waste was produced long ago and the available historical information is incomplete, uncertain, or missing. This makes it difficult to accurately determine the nature and quantity of the radioactive material contained in waste drums.

My doctoral research focused on methods that combine X-ray imaging with gamma spectrometry for the characterization of radioactive waste drums. These techniques provide complementary information: transmission imaging helps estimate the internal mass distribution, while gamma spectrometry identifies the radionuclides present in the drum.

The combination of both techniques makes it possible to account for the spatial distribution of the radioactive sources inside the drum. Since the measured gamma signal depends on both source location and material attenuation, this additional information is necessary to reduce biases in the estimation of the total radionuclide content.

In this way, the work developed in this thesis contributes, even if modestly, to a more reliable characterization of radioactive waste. Better estimates of radioactive inventories support safer decisions in waste handling, storage, and disposal, and therefore help reduce the environmental and societal burden associated with legacy radioactive waste.

Promoters:

Prof. Ivo Couckuyt, Ghent University 

Prof. Tom Dhaene, Ghent University

SCK CEN mentors:

Dr. Eric Laloy

Dr. Bart Rogiers