Optimising full waveform inversion with inhomogeneous transducers: Parameters and considerations for successful implementation

Abstract

Guided wave tomography (GWT) based full waveform inversion (FWI) is an emerging technique for structural health monitoring applications, primarily for plates and pipeline structures. Generally, FWI employs a two-dimensional (2-D) forward model to circumvent the high computational cost associated with the inversion scheme. Consequently, a re-scaling step is implemented to compensate for any potential discrepancies between the 2-D model and the observed data. Druet et al., (2019) introduced the autocalibration method, which utilises the information from the healthy rays to calibrate those rays that pass through the defect. In this method, only the phase information is re-scaled, given that phase information is the dominant factor in FWI. However, overlooking amplitude discrepancies might lead the inversion scheme to become trapped in a local minimum. In this study, we propose to include the amplitude information as well, following the autocalibration method. We use an updated autocalibration method to reconstruct a 100 mm wide defect on an 8 mm thick steel straight pipe with traditional GWT using the $A_0$ mode. This novel approach provides a more accurate representation of the defect and avoids becoming trapped in a local minimum, thereby improving the reliability and effectiveness of FWI. Furthermore, we offer guidance for the successful implementation of this method in the presence of inhomogeneous transducers, a common challenge in practical applications.