In CANDU reactors, the feeder pipe system supplies the primary coolant from the reactor core to steam generator. The flow-accelerated corrosion (FAC) of the extrados of the first bend of the outlet feeder is a life limiting factor. Therefore, accurate prediction of FAC rate is needed to determine the scope of inspection and replacement of feeder pipes during an outage. To estimate the FAC rate, the initial wall thickness of the extrados of the bend is required. However, initial wall thickness is not known precisely, because the bending of pipe during installation causes the thinning of extrados section. Another difficult is the sizing error or noise in the wall thickness measurements introduced by the inspection probes.
The objective of this paper is to develop a sound probabilistic method for the FAC rate estimation considering uncertainties in the initial wall thickness of bend and the sizing error (or noise). A Bayesian model is developed and applied to predict the end of life of the outlet bend on a risk-informed basis. A practical case study is presented to illustrate the use of the proposed method.