ARMA models in the analysis of temperature noise measurements in a super-phenix subassembly mock-up

The noise analysis methods play an important role in the early, reliable detection of local cooling disturbances in a fast reactor subassembly such as sodium boiling or blockage, which are considered among the initiating events of major disruptive core accidents. Box and Jenkins auto-regressive moving-average ARMA models are applied to the analysis of several temperature time-series measured by the Commissariat a l'Energie Atomique in the course of the CFNa experiment carried out at the nuclear center of Grenoble. The source of data is a thermocouple placed at the outlet of a Super-Phenix subassembly mock-up. The analysis shows that a simple ARMA (3,2) model adequately accounts for the observed fluctuations. This model provides methods for a continuous, in situ estimate of the thermocouple time constant, for the identification of a suitable boiling and blockage indicator and for the detection in real time of suddenly occurring disturbances. (author)