Ultrasonic Multi-Hole Imaging Using Full Waveform Inversion

Abstract

Ultrasound computed tomography (USCT), in recent years, is becoming an increasingly popular method for structural health monitoring (SHM) and non-destructive evaluation (NDE). Full waveform inversion (FWI), a cutting-edge inversion method, utilizes all information in ultrasonic measurements in USCT. It iteratively reconstructs the model parameters (i.e., wave speed of materials) of the scanned specimen by calculating the gradient of waveform difference between the measured and synthetic signals through a partial derivative equation (PDE)-constrained optimization process. By reconstructing these model parameters, defects inside the specimen can be identified. This study aims to evaluate FWI’s performance in imaging defects (i.e., holes). FWI was implemented to process the numerically generated signals to simulate the data acquisition in a steel specimen with six holes of different sizes (diameters ranging from 0.4 mm to 5.2 mm). First, a shorter numerical model with the same number and size of holes was used to benchmark FWI’s performance with full coverage of source and receiver elements for the domain. To emulate large-scale structural component inspection, the numerical domain was then changed to the dimensions of the actual steel specimen, and the scanning setup was changed to a pair of linear transducer arrays with relatively shorter apertures located on the top and bottom of the specimen. The effect of the array aperture on the FWI performance was studied. The reconstructed model parameter, longitudinal wave speed (Vp) , and shear wave speed (Vs) maps showed that with the current setup, FWI can identify the locations of all holes. A new imaging condition was proposed using the inverted Vp and Vs maps to quantify the shapes and size of the holes. The ultrasound signals from the actual steel specimen with holes of the same dimensions were also analyzed and modeled. The results exhibit the potential of applying FWI on experimentally acquired data from the steel sample.