Sensor Validation Using Auto associative Neural Network

Sensor Validation Using Auto associative Neural Network. Sensor signal accuration plays a significant role in safety and operation of a nuclear reactor. These sensor performance could be decline even become totally fault when the reactor in operation. This paper demonstrates the application of auto associative neural network (AANN) to validate signals from various correlated sensor so that the sensor performance deterioration can be earlier detected. An AANN model produces predicted signal, which is used as a measured signal validator. The faulty of a sensor will not disturb the reactor operation since the predicted signal can be used as redundant signal and replace the faulty signal. This method is applied to a set of data from Borselle Nuclear Power Plant and corresponds to various operation modes. Result showed that the system could be used to detect 3% drift on one of the input channel. (author)