Copper based alloy tubing utilised in power plant heat exchangers is susceptible to pitting on the inside tube surface. More recently, an increasing number of stainless steel tubing is also affected by microbiologically influenced corrosion in the form of ID pits. Localised pitting is most damaging since it reduces load-carrying capacity and increases stress concentration by creating holes in the tube wall. To assess tubing integrity, eddy current testing has been employed as a primary inspection tool. Detection of pitting has not been a problem, but the sizing of ID pits produced mixed results based on using either phase angle or amplitude alone to estimate pit depths. In general, the overall depths of clustered or multiple pits are overestimated. This paper presents results of optimised neural networks to estimate ID pit depths by using selected phase angle and signal amplitude information at various operating frequencies. The paper will provide comparative results of the neural network-predicted pit depths versus the field analysis results. In addition, the network results will be obtained from a removed tube section containing service-induced pits, and the predicted results compared directly with the destructive analysis results.
Publication Source: First International Conference on NDE in Relation to Structural Integrity for Nuclear and Pressurised Components , 20 - 22 October 1998, Amsterdam, Netherlands. Held by the Joint Research Centre of the European Commission. Publisher:Woodhead Publishing Limited
NDTnet
