GROOVE SIZING USING A ROBUST NEURAL NETWORK APPROACH

摘要

The remote field eddy current technique is used to inspect conductive pipes from the inside. The problem is to calculate an estimation of groove dimensions from observed data. A first approach was previously developed using a two-step parametric inversion. Results from this first approach are produced using a new model. A second approach using a neural network is presented. This technique is known for the lack of robustness which may occur when precautions are not sufficient. This paper presents these precautions and the results of both approaches.