Abstract
Monitoring the wall thinning in pipelines is crucial in the energy industry to guaranty its proper operation. Typically, in pipelines where fluids with particles or solid particles are transported, larger erosion is encountered which is usually the same along the length of the pipe. Conventionally, Non-Destructive testing techniques for wall thinning are based on a local point-by-point ultrasonic measurements, which is slow and costly. In contrast, Guided Wave Tomography (GWT) based on Full Waveform Inversion (FWI) is an accurate and high-resolution reconstruction method, that enables the monitoring of the whole pipe area by processing the waves excited and received by transducer arrays. Although FWI have been used successfully to reconstruct small localized defects in the past, the reconstruction of the elongated defects remains a challenge as the waves propagating through the defect may undergo cycle-skipping effect which breaks down the inversion algorithm. In this study, the limits of FWI are explored when reconstructing elongated defects on a steel pipe with 6.5 mm in thickness, 230.2 mm in diameter and 720 mm in length. Four elongated Hann-shaped defects were modeled in a three-dimensional (3-D) finite element model, and reconstructed using FWI based on an acoustic two-dimensional (2-D) forward model. The defects have a 70 mm width, 1.5 mm in depth, and longitudes of 25%, 50%, 75% and 95% each of the pipe length. Two initial velocity-thickness models were considered, one homogeneous along the reconstructed domain and the other one with a velocity variation. The velocity-thickness variation model was based on the time of flight (TOF) of the transmitted signal of the aligned transducers. For 25% and 50% defects’ longitudes good agreement was found between 3-D finite element model and the 2-D reconstructed model, when the background initial velocity model was constant. In contrast, for defects with a longitude larger than 75% of the pipe’s distance, good agreement was found between 3-D finite element model and the 2-D reconstructed model when selecting a suitable initial velocity-thickness model.