Probability of detecting the deep defects in steel sample using frequency modulated independent component thermography

Abstract

Active thermography is a widely used nondestructive testing and evaluation technique (NDT&E) for evaluating the properties of materials without impairing its future usefulness. In this work, a mild steel sample made of artificial flat-bottom holes at varied depths, was examined with the emerging non-stationary thermal wave imaging (NSTWI) technique,i.e. frequency modulated thermal wave imaging (FMTWI). The pulse compression favorable of NSTWI technique is eminent for compressing the applied thermal energy into a narrow-compressed pulseto enhance the depth resolution and sensitivity. In this work, pulse compressed thermographic data generated from FMTWI experimentation is analyzed with the unsupervised learning approach independent component analysis (ICA) to test their mutual return in the detection of the deep defects in a mild steel sample and this proposed technique was referred to as frequency modulated independent component thermography (FMICT). In comparison, the effect of FMICT was contrasted with othermethodsi.e. pulse compression of time domain and ICA of feature space by considering the signal-to-noise ratio (SNR) as a figure of merit. Furthermore, a probability of detection (POD) analysis framework based on the minimum threshold SNR criteria for apparent visibility of the defects has been presented to assess the probability of identifying defects at various depths using such approaches. The influence of the SNR threshold value for the above strategies on the POD curves has also been presented.