Status Monitoring for Nuclear Steam Generator Using Neural Networks

In order to improve the capacity of nuclear steam generator (SG) status monitoring, a new monitoring approach based on neural networks (ANN) is investigated in this work. In this approach, a three-layer BP neural network was trained as the process model of SG. In the process of status monitoring, when the deviations between process signals measured from an actual SG and corresponding output signals from the ANN model of SG exceed the limits accepted, the abnormal events are thought to occur. The ANN modeling for the SG process is implemented by using of the monitoring data of the SG' important operation parameters, which are the steam flow rate, feed water flow rate, pressure and water level. The error back propagation algorithm with momentum factor and adaptive learning rate is employed to train the network. The typical operation patterns of SG were used to demonstrate the feasibility of the approach. The results reveal that employing ANN can improve the capacity of SG status monitoring.