A real-time acident tracking calculation technique is investigated for in-depth diagnoses of internal state variables of a pressurized water reactor(PWR) plant in the case of a small-break loss-of-coolant accident (SBLOCA). The technique is composed of two parts:(a) a faster-than-real-time open-system simulator TOKRAC for calculating postaccident thermal-hydraulic behavior in the primary system of a PWR plant, and (b) several real-time state estimators for supplying the unobserved external input to TOKRAC. Observed and unobserved various are defined as variables that are or are not directly measured by exisitng plant instrumentation. The external input to the open-system TOKRAC is assumed to be composed of several observed signals and several unobserved state variables: The former are the injection conditions of the emergency core cooling system and the operator actions, while the latter include the total heat transfer rate from the primary to the secondry side in the steam generator(SG) (SG heat transfer rate), the break size and approximate break location inth case of a primary pipe break, or the leak flow rate through an SG tube hole (SG leak rate) in the case of SG tube rupture. The SG heat transfer rate and SG leak rate are first estimated in real time using the observed signals of the SG secondary side. The Kalman filtering technique is applied to a simplified SG secondary model. Due to the estimation technique, the primary side of the PWR plant is separated from the secondary side with respect to heat and mass transfer; then the primary side thermal-hydraulic behavior is rapidly computed using TOKRAC. The unobserved external input,ie., break size and approximate location of the pipe break in the primary side, is estimated within a short time after the pipe break using the estimated SG heat transfer rate and several observed signals of the primary side. Additional Kalman filters derived by a simplified primary system model are needed to ensure the whole accident tracking calculation.
The validity of these estimation techniques is examined through computer experiments. The reference data are taken from the calculation results for several accidents of a Westinghouse-type PWR plant using RELAP4/MOD6 and the experimental data of loss-of-fluid testL9-3. The validity of the whole accident tracking calculation technique is also examined by a computer experiment of a 1.5% cold-leg SBLOCA of a Westinghouse-type PWR plant.