UC San Diego

Inspection of structures is a critical task that needs to be performed in order to guarantee the safety of structural components during their service life. Different Nondestructive Evaluation (NDE) techniques can be used to inspect aerospace, civil, and biological systems to ensure their structural integrity and to identify the presence of damages and defects, which could impair the correct functioning of the overall structure.

The focus of this dissertation is the inspection of structures through the passive extraction of the acoustic transfer function of the medium under consideration, and the 2D and 3D characterization of defects by means of ultrasonic imaging. The first part of the dissertation addresses the issue of defect detection in railroad tracks by extracting the acoustic transfer function of rails through a normalized cross-correlation operator, which exploits the random acoustic vibrations generated by the train wheels. A technique to remove uncorrelated noise from the recorded signals is also introduced to make the transfer function reconstruction more robust. A statistical outlier analysis is used to detect any variation in the transfer function of the rail as the train moves along the track, in order to identify locations where discontinuities (joints, welds, defects) might be present. A prototype with multiple pairs of capacitive sensors was developed to perform the inspection in a non-contact, passive-only, high-speed manner. Results from fields tests performed at the Transportation Technology Center (TTC) in Pueblo, CO, will demonstrate the feasibility of the system for the reliable inspection of railroad tracks at speeds up to 80mph.

The second part of the dissertation is focused on the characterization of defects using ultrasonic imaging to create 2D and 3D images of the inspected medium. Imaging in bulk solids and plates is performed using sensor arrays and an improved beamforming algorithm that uses information about the structure of the propagating acoustic wave modes to improve the defect characterization process. Furthermore, the experimental application to railroad tracks and the implementation on a Graphics Processing Unit (GPU) shows the potential for the accurate real-time imaging of rail flaws.