In order to be able to control and mitigate an accident, the operators should have at any time a realistic and correct picture of the accident and its progress. Hence signal validation is a main issue under severe accident conditions, especially if instruments are working beyond their specification range. The objective of the AMS project is to define, investigate and develop means and methods providing reliable information and diagnostics as well as support tools for accident management. Regarding the signal validation methodologies it has been agreed to emphasize the existing instrumentation rather than new instrumentation needs. As a result some efforts are dedicated to conceptual definitions of the degradation of the instrumentation under accident conditions.
During the year of reference emphasis has been placed on the preparation of two state-of-the-art reports (SOARs, namely on 'Instrumentation and Signal Validation in Accident Situations' and on 'Operator Assisting Systems for Accident Management'). The exchange of information was concentrated mainly on national concepts and used practices in applying methods and procedures to assist the operator in an accident situation.
Signal Validation Methodologies
Incipient and actual failure detection using signal validation methods are mainly based on model-based monitoring procedures (e.g. functional redundancy) as well as on advanced noise diagnostic methods. At NNC for this purpose a linearised model has been identified, and compensated for non-linear effects by the used of Kalman filters. The model has been shown to detect simulated signal failures effectively and rapidly. Simulated signal failures included the effects of steps on a signal and ramps. The methodology of model reduction and model operation has been successfully demonstrated.
At ISTec mathematical models for the relief tank, the preheater, and the condensate water heat exchanger have been implemented. IFDI-(Instrumentation Fault Detection and Isolation)-modules, based on analytical redundancy methods, were developed and tested using data, provided by the simulator ATLAS. Several sensor faults have been simulated.
Sensor Modelling and Signal Processing
A fission ionisation chamber impulse response model has been developed at CEA/LETI/DEIN and validated using signals from an experimental nuclear reactor ('Ulysses'). A very fast electronic has been used in order to obtain all the information contained in the signal. After this, several comparisons between the spectral densities estimated from the model and from the real measurements have been made. An excellent agreement was achieved. Additionally, for a boron ionisation chamber a new model has been developed, which takes into account phenomena such as space charge effects. The model is based on complex differential equations and was derived from Maxwell and Poisson equations. The first numerical results confirmed the fact that even a slight oxygen intrusion could have fatal consequences for the chamber. Therefore, an experiment was prepared which uses a chamber identical to that in commercial PWRs. For signal validation, methods for signal-noise and neutron-gamma separation have been developed. The separation was achieved by minimising a heuristic distance function. This method has been extended using a maximum likelihood technique.
The two originally envisaged projects on "Instrumentation and Signal Validation" and on "Operator Assistance" have been combined. As a result detailed investigations are being conducted in signal validation methodology and sensor modelling/signal processing. Two State-Of-the-Art-Reviews (SOARs) are being drafted on the current AMS practices in European NPPs, namely "Instrumentation and Signal Validation in Accident Situations" and "Operator Assisting Systems for Accident Management". Specific attention is given to identifying the safety function of the different classes of operator support systems.
There are two basic approaches to match the requirements of reliable information about the validity of a plant signal: model-based methods realizing functional redundancy and noise diagnostic methods, using the signatures of inherent signal noise as "finger prints" of specific sensors and plant conditions. Sensor modelling activities performed at CEA-DTA/Saclay are aimed at the improvement of a fission chamber model in a an extended range of faulty operating conditions. Experiments are foreseen in the Ulysse test reactor of Saclay.
In the field of operator assisting systems for accident management, investigations about real-time monitoring and decision-making techniques, using neural networks, advanced modelling and noise analysis techniques, are carried out both by IST/Garching and by ECN/Petten. A validation of these techniques is foreseen on real data, e.g. those of the Borssele reactor.
FRAMATOME is developing expert system based strategies for design and maintenance of emergency guidelines in connection with an "automatic operator model". CEA-DTA/-Cadarache investigates the operators role with respect to automation.
The development of a knowledge-based operator aid system, able to interact visually with the operator, is developed by ANSALDO/Genova. ECN/Petten works also on computational aids for technical support centres to predict critical milestones. NNC/Knutsford is developing operating instruction displays to optimise command/control under emergency conditions.
SIEMENS-KWU/Erlangen is developing the SIROG (SItuation Related Operation Guidance) procedure which is based on dynamic HYPERTEXT documents for the operator, using adaptative algorithms and extrapolations from recorded plant data. An advanced expert system, called "OPerator Advisor" (OPA), is also developed by TRACTEBEL/Bruxelles with the aim of assisting the operator team for optimal recovery following an accident.
GRS/Garching together with IST/Garching is working on fast predictive models which help the operator in providing immediate answers for the evaluation of emergency operating procedures.
Funding SchemeCSC - Cost-sharing contracts
91191 Gif Sur Yvette
WA16 8QZ Knutsford
92084 Paris La Defense