Methodology for predicting flow-accelerated corrosion wear using unreferenced multiple inspection data

The lack of fixed spatial referencing in inspection data is a confounding factor in the assessment of wall thinning due to flow-accelerated corrosion (FAC). Because of the loss of wall thickness over time, it is impossible to match any two unreferenced points precisely between inspection outages, making the use of point-to-point comparison methods highly uncertain. In this study, we develop a new methodology for estimating the FAC rate and component lifetime from spatially unreferenced inspection data. The method is based on a small circular patch that represents the minimum wall thickness as a random variable. This allows the predictions to be performed in a probabilistic setting, hence providing valuable risk information for long-term replacement planning and life-cycle management. The uncertainty and errors introduced by inspection and grid coverage are minimized in the analysis by fitting a continuous surface through the inspection data using the method of kriging. The developed methodology is applied to the assessment of carbon steel feeder piping in CANDU reactors, where the economic and safety consequences of FAC can be highly significant. Based on the analysis of a large industry dataset, FAC in feeder piping appears to be a relatively uniform and stationary process. The results of the study demonstrate the key advantages of the developed methodology and how it can be used to support risk-informed decision making with respect to FAC at power plants.