A neuro-fuzzy tool for CT-PT contact detection in a pressurized heavy water reactor

Abstract In a pressurized heavy water reactor (PHWR), contact between the calandria tube (CT) and the pressure tube (PT) makes them susceptible to delayed hydrogen cracking. Periodic inspection of the channels must be carried out to detect such contacts. As the number of channels in a PHWR is very large (306 in a 230 MW plant) periodic in-service inspection of all the channels leads to an unacceptable downtime. A non-intrusive technique that employs a system-identification method is presently used for contact detection. Attempts to identify all the contacting channels, without missing any, lead to overprediction of the number of channels in contact; i.e., many channels are diagnosed as contacting, while those channels are actually not in contact. This puts a large number of healthy channels in the at-risk list, reducing the efficacy of the method. Previously, the authors demonstrated the power of a neural post-processor in improving the strike rate of the system-identification tool. This paper demonstrates a stand-alone neuro-fuzzy tool for the detection of CT–PT contact. The network consists of a cascade of self-organizing artificial neural networks (ANNs), along with fuzzy processors. The performance of the network has been compared with that of the system-identification techniques. The noise tolerance of the network is also demonstrated.