Approach to development of advanced expert diagnostics systems

The main requirements for an expert diagnostic system are for it to be able to: monitor the equipment's state in all operational situations, detect deviations from 'normal' operational conditions, classify the significance of phenomena relating to risk assessment, and recommend solutions to problems arising. The effort in building a knowledge base is concentrated on data validity problems and recognition and classification of anomalies and their dynamics and localization. The current investigation is oriented towards the problems of distinguishing between intrinsic anomalies due for example to mechanical vibrations of internals, hydrodynamic blockages or oscillations, thermodynamic changes (boiling), and those due to external origin, e.g. sudden load change. For tuning the diagnostics expert system real data are used. For intrinsic anomalies tests, the data received in the NPP in the form of measurements from an instrumented fuel assembly are used. Methods for distinguishing external events are tested in data from OECD-SMORN-V benchmark tests. In parallel, the knowledge base for vibration of reactor internals is being built. To identify the power of fluctuations and the physical parameters, the autoregression method is used. The application of autoregressive models in connection with physical interpretation of poles makes it possible to determine the physical nature of the mechanism of the failure process and its development, to carry out a computer simulation of anomalous states of the systems being diagnosed and to create databases of expert diagnostics systems. (author). 8 refs, 4 figs, 5 tabs