A model for secure nuclear power is studied at Enea

An advanced artificial intelligence model to improve the calculations of the simulations of ‘severi’ accidents in nuclear reactors. To study it is the Aeneas as part of the European project of 4 million euros, coordinated by the French Asnr, which sees the participation of 14 partners of EU countries, Switzerland and Ukraine. The model aims to combine temporal analysis and relationships between variables to optimize the calculations of the Astec code, the European reference software that allows you to simulate the phenomena generated by a severe accident in a water -cooled reactor, from the initial event to any possible Release of radioactive materials outside the containment. The final goal is to strengthen safety in case of nuclear accidents adopting ready and experienced response strategies. This will be possible thanks to the collaboration between Inter national researchers specialized in severe accidents and automatic learning experts.

“Nuclear reactors simulators assist the formation of operators, the design and evaluation of safety. However, only a few simulators in the world modelize severe accidents that can involve the merger of the core. The Assas project aims to create a prototype of simulator based on the Astec code to model, calculate and view the main phenomena of an accident in real time through an interactive graphics interface “, explains the manager for Aeneas of the Fulvio Mascari project, researcher of the nuclear department.

“The goal – he continues – is to achieve one Code calculation speed such as to offer simulator users a realistic training experience. As a demonstration of the commitment in the formation of the younger generations, Enea finances a doctorate bag with the University of Bologna in this area “.

The upgrade of the Astec code will allow you to create engineering and royal scale simulators that can be used to develop guidelines on the management of severe accidents, develop new safety systems and form operators.
“The next steps – concludes Mascari – provide for the validation of the models and the integration of physical models e Machine Learning techniques to further increase the accuracy of the simulations “.