Automatic Detection of Pipeline Construction Features with RFEC technology
In-line inspection with Remote Field Eddy Current (RFEC) tools requires detection of construction features such as joints, elbows and off-takes. We propose to automate this process using supervised learning. Firstly, signal processing techniques are used to detect features in the RFEC recorded data, where features in general refer to both defects and construction characteristics. Secondly, a machine learning algorithm is employed to classify all the detected features into construction features or defects. Over 800 meters of RFEC data recorded from the Strathfield research test-bed, established as part of this collaborative project, have been used to evaluate the proposed approach.