A diagnostic system of the nuclear power plant operation has been proposed and the applicability of this system to the actual plant has been verified by computer simulation. A typical pressurized water reactor plant simulator was made by an analog computer and the diagnostic system was made by a digital computer. The observed signals obtained from the actual plant are simulated by superposing the equivalent observation noises generated by the digital computer on the sampled signals obtained from the plant simulator. The diagnostic system of a nuclear power plant, using a digital computer, is composed of three steps:

  1. detection and classification of the anomalous state
  2. estimation of the primary cause of anomaly
  3. storage of past observed signals.
The linear discriminant function technique is applied in phase 1, and the Kalman filtering technique is applied in phase 2.