Sensor response monitoring in pressurized water reactors using time series modeling

Random data analysis in nuclear power reactors for purposes of process surveillance, pattern recognition and monitoring of temperature, pressure, flow and neutron sensors has gained increasing attention in view of their potential for helping to ensure safe plant operation. In this paper, application of autoregressive moving-average (ARMA) time series modeling for monitoring temperature sensor response charactersitics is presented. The ARMA model is used to estimate the step and ramp response of the sensors and the related time constant and ramp delay time. The ARMA parameters are estimated by a two-stage algorithm in the spectral domain. Results of sensor testing for an operating pressurized water reactor are presented. Since the estimation depends on the random signal characteristics, there are cases where the noise analysis approach fails to predict sensor characteristics accurately. In general, the approach is useful for sensor monitoring schemes, rather than to predict a quantitative value of the sensor response.