An adaptive statistical model of nuclear rods temperature for the detection of total and instantaneous blockage

The monitoring of critical systems is of the utmost importance, especially when undetected malfunctions could lead to major accidents. This paper focuses on the temperature monitoring of fuel rod assemblies within nuclear power plants, with the goal of detecting total and instantaneous blockages as reliably and quickly as possible. First, we address the modelling of the temperature of the whole fuel rod assembly altogether. We propose a linear parametric model that is adaptive, incorporating previous temperature measurements to enhance its accuracy. This approach allows us to distinguish between regular, non-anomalous temperatures and the anomalous thermal event due to a blockage. The proposed sequential, or online, detection scheme is reliable, as the false alarm rate and detection power are analytically bounded. The model and subsequent statistical test are generic, making the methodology applicable to a wide range of nuclear cores. Numerical experiments, using real temperature measurements from the Superphénix power station, demonstrate the accuracy of the proposed model and the relevance of the detection procedure.