Parameter and surface estimation in ultrasound non-destructive testing

Abstract

Ultrasound techniques play an important role in Non-Destructive Testing. The advantages of ultrasound are its low-cost, portability and safety. Ultrasound imaging facilitates the localization and characterization of internal defects, as well as the detection and mapping of corroded areas. The imaging process can be done in different ways, but common to all, inspection parameters such as sound speed in the different media, as well as the interface between these media, must be known. Wrong assumptions on these parameters lead to distortions in the resulting images that could, potentially, hinder the characterization of defects. Estimation of these parameters is usually done by calibration using specific materials and procedures. The objectives of this study are to develop (1) methods to estimate such parameters during inspection time, without the need for specific materials, as well as (2) surface reconstruction methods to address corrosion mapping in pipes. In the first objective, it is shown that commonly used imaging processes, such as the Total Focusing Method, can be used to estimate the sound speed in the imaged material with low uncertainty. The proposed method does a gridless search in a given interval and is able to produce an estimate in up to 23 iterations, from the 8000 iterations needed, for the same tolerance, in the methods present in the literature. In the second objective, a method for estimating a 1D surface profile is proposed. This new method weights the contribution of each surface point as a way to cope with different SNR levels. Also, by means of a second-order Total Variation regularization, the method promotes piecewise linearity in the solution while suppressing noise. The developed method is shown to be more robust in the presence of noise than other methods in the literature. When compared to the state-of-the-art, the proposed method obtained errors almost ten-fold smaller. Using the estimated profiles to produce interior images of the objects, the proposed method lead to more accurate images, enabling better detection of the flaws.