Simulation Study of System for Diagnosis of Nuclear Reactor Operation

Abstract

The authors propose a system for diagnosing the state of operation of nuclear reactors.The system comprises the block TWODIS, EST and MEMORY.
The signals received from a number of detectors in the reactor are led sequentially into the block TWODIS. The input signals are analyzed statistically, to identify the actual state of operation with one of several reference modes prescribed beforehand. The algorithm used here for the identifying operation is based on the linear discriminant function.
When TWODIS identifies the state of the reactor with one of the abnormal reference modes, the disturbance estimating block EST set to work. This block statistically analyzes the signals depicting past history retrieved from MEMORY, and estimates the quantitative characteristics of the abnormality source (disturbance). Laguerre function expansion and multiple regression analyses are used in EST.
As a numerical example, a single channel model sodium coolant fast reactor is represented by a reactor simulator, in which only the coolant outlet temperature and the coolant flow rate are chosen as state variables to be detected.
The reference modes of reactor operating state prescribed for the identifying operation by TWODIS are; (1) Normal state, (2) Increasing power density, and (3) Decreasing flow rate.The EST estimates the magnitude and time constant of the causative disturbance.
Trial experiments made with simulated disturbances prove the practicality of the proposed diagnostic system.