Sep 7 – 11, 2026
Buenos Aires, Argentina
America/Argentina/Buenos_Aires timezone

Frontiers of Artificial Intelligence: From Automation in Scientific Instrumentation to Predictive Safety

Sep 9, 2026, 4:45 PM
15m
Buenos Aires, Argentina

Buenos Aires, Argentina

IFIByNE (CONICET-UBA) - Intendente Güiraldes 300 Buenos Aires Autonomous City of Buenos Aires
Oral Presentation Data management (Software, AI)

Speaker

Leonardo Ibáñez (CNEA)

Description

The integration of Artificial Intelligence (AI) in the nuclear sector ranges from the optimization of scientific instruments and their auxiliary systems to the strengthening of operational safety in reactors. This work presents advances in the development of artificial intelligence models applied to two critical areas: neutron control and instrumentation through synthetic training, and reactor safety diagnostics based on real-world data.

In the first line of research, Artificial Neural Networks (ANNs) were designed for a polarized neutron reflectometer with multiple operating modes and for a gas delivery system used as an autonomous sample environment equipment (SEE). Due to the lack of historical records, physical simulation environments were developed to generate synthetic training data. In the SEE, the ANNs perform the automatic operation of flows and pressures, suggest optimal configurations, and run simulations, offering a modular design transferable to other industries. In the reflectometer, the networks assist in the dynamic optimization of the instrument and the analysis of complex experimental data.

The second line of research focuses on nuclear safety through the application of ANN for the analysis of digitized sensor signals from real reactor data. Here, these neural networks are deployed to solve shutdown classification tasks (distinguishing between controlled and abnormal events) and for time-series forecasting. Unlike the simulation-based approach, this direct application of ANN to real operating data faces two fundamental challenges in the sector: the extreme scarcity of data associated with failures or anomalies and the imperative need for model interpretability to ensure auditable and safe decision-making.

In conclusion, these developments, currently in the design and testing phase, demonstrate the versatility of AI to adapt to diverse scenarios and requirements within the nuclear sector. The ability to transition effectively from the automation and simulation of scientific instrumentation to predictive diagnostics in reactors consolidates AI as a transversal and scalable technology, capable of providing solutions both in frontier experimental research and in the critical operation of safety systems.

Key Words

AI, ANN, Instrumentation, Reactors, Automation, Classification, Interpretability, Simulation, Safety

Other Safety

Authors

Leonardo Ibáñez (CNEA) Mr Facundo Silberstein (CNEA) Mr Federico Montenegro (CNEA)

Co-authors

Mr Federico Casella (CNEA) Mr Christian Dacal (CNEA) Mr Santiago La Greca (UTN) Mr Ricardo Mateucci (CNEA) Mr Agustín Nieto (CNEA) Mr Luis Romero (CNEA) Mr Manuel Suarez Anzorena (UTN) Dr Marina Tortarolo (CNEA-COCINET)

Presentation materials

There are no materials yet.