Classification and Location of Neutron Noise Perturbations Using Convolutional Neural Networks
[EN] With the aging of the nuclear reactor fleet in Europe, and especially in Spain, monitoring these reactors through complex models has become of great interest to maintain the safety and operational capability of these nuclear power plants. It is of particular interest to locate the place where a...
| Autores: | , , , |
|---|---|
| Tipo de recurso: | artículo |
| Fecha de publicación: | 2025 |
| País: | España |
| Institución: | Universitat Politècnica de València (UPV) |
| Repositorio: | RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia |
| Idioma: | inglés |
| OAI Identifier: | oai:dnet:riunet______::a266ab2a3c34e194e2db93c910e48a00 |
| Acceso en línea: | https://riunet.upv.es/handle/10251/235310 |
| Access Level: | acceso abierto |
| Palabra clave: | Deep learning Neural networks Neutron noise Nuclear safety FEMFFUSION |
| Sumario: | [EN] With the aging of the nuclear reactor fleet in Europe, and especially in Spain, monitoring these reactors through complex models has become of great interest to maintain the safety and operational capability of these nuclear power plants. It is of particular interest to locate the place where a possible anomaly has occurred, as well as the type, to guarantee the safety of the reactor through the analysis of neutron flux fluctuations. Therefore, we propose a deep learning framework for the deconvolution of reactor transfer functions from perturbation-induced neutron noise sources. The main objective of this work is to develop tools based on deep learning techniques to classify the type and to locate the perturbation, working with simulated data with different noise levels, and to study the number of detectors that need to be active. In particular, the data used have been simulated for the BIBLIS 2D reactor using FEMFFUSION. This work has been carried out using the Keras library based on tensor flow, managing to develop two convolutional neural networks that adapt well to the data model. High-accuracy results are obtained both when predicting the type of the perturbation and when locating the place of the perturbation, with a low error rate even when only four to eight detectors are available. |
|---|