Prediction of LOCA Break Size and Position Based on Deep Rectifier Neural Networks


Claudio Marcio do Nascimento Abreu Pereira
Filipe S. M. Desterro
Roberto Schirru


This report presents an approach for identification of break locations and sizes of loss of coolant accidents (LOCA) in nuclear power plant (NPP).  Under LOCA condition, it is quite important a quick identification the location and the size of the rupture. However, it is very difficult to carry out this identification only by observing the time trend of the information shown by sensors and panels in the control room. The identification of the rupture location is modeled as classification problem, in which a discrete set of possible locations is considered. The size of rupture is a regression problem.


Complex Systems Engineering