Detecting small flaws near the interface in pulse-echo

Autoencoder-based detection of near-surface defects in ultrasonic testing

Defect detection during pulse-echo ultrasonic testing (UT) is challenging when defects are located in a dead zone where the echoes from the defects are overshadowed by disturbances from the initial ringing signal of the UT transducer. The time-gate method is one of the most widely used approaches in UT to filter out such unwanted components, but defects in the dead zone are virtually impossible to detect using conventional methods. This paper proposes an autoencoder-based end-to-end ultrasonic testing method to detect defects within the dead zone of a transducer. The autoencoder is designed to predict the normal behavior of ultrasonic signals including disturbances, thus enabling the identification of even subtle deviations made by defects. To advance the performance of the autoencoder further with a limited amount of training data, a two-step training procedure is presented, involving training using pure normal signals measured from a defect-free specimen and re-training using pseudo-normal samples identified by the autoencoder with a smart thresholding strategy. This two-step procedure enables us to develop an adaptive autoencoder model that can be effectively employed to process the newly measured ultrasonic signals. For a demonstration of the proposed method, UT-based B-scan inspections of aluminum blocks with near-surface defects are conducted. The results suggest that the proposed method outperforms the conventional gate-based inspection approach with regard to its ability to identify the sizes and locations of near-surface defects.

Detection of micro inclusions in steel sheets using high-frequency ultrasound speckle analysis

With the increasing need for steel sheet quality assurance, the detection of micro-scaled inclusions in steel sheets has become critical. Many techniques have been explored to detect inclusions, e.g., visual inspection, radiography, magnetic testing, and ultrasound. Among these methods, ultrasound (US) is the most commonly used non-destructive testing (NDT) method due to its ease of use and deep penetration depth. However, ultrasound currently cannot be used for detecting the micro-scaled inclusions due to low spatial resolution, e.g., less than 30 μm, which are the key important factors causing the cracks in the high-quality steel sheets. Here, we demonstrate a high-resolution US imaging (USI) using high-frequency US transducers to image micro inclusions in steel sheets. Our system utilizes through-transmission USI and identifies ultrasound scattering produced by the inclusions. We first ultrasonically imaged the artificial flaws induced by the laser on the steel sheet surface for validating the system. We then imaged the real inclusions in the steel sheets formed during manufacturing processes and analyzed them to derive quantitative parameters related to the number of micro-scaled inclusions. Our results confirm that inclusions less than 30 μm can be identified using our high-resolution USI modality and has the potential to be used as an effective tool for quality assurance of the steel sheets.

Enhanced ultrasonic detection of near-surface flaws using transverse-wave backscatter

Diffuse ultrasonic backscatter measurements have been shown to enhance the detection capability of sub-wavelength flaws when combined with extreme value statistics. However, for a normal-incidence immersion measurement, a "dead zone" created by the ring-down of the front-wall echo will hide near-surface flaws. In this article, a pulse-echo transverse wave backscatter measurement is used to detect near-surface flaws under high gain. The approach is validated using a magnesium specimen with side-drilled holes. The confidence bounds of the grain noise from this specimen are given by a transverse-to-transverse scattering model, which takes the grain size distribution and the hexagonal crystal symmetry into account. The upper bound is then treated as a time-dependent threshold for the C-scan. Experiments show that the developed method has good performance for detecting sub-wavelength, near-surface flaws, and can suppress both missed detections and false positives.

A model of blind zone for in situ monitoring the solid/liquid interface using ultrasonic wave

To in situ monitor a solid/liquid interface to control metal qualities, the paper analysis blind models of the ultrasonic propagation in the solidifying molten metal with a solid/liquid interface in the Bridgman type furnace, and a mathematical calculation model of blind zone with different source locations and surface concavities is built. The study points out that the blind zone I is caused by ray bending in the interface edge, and the blind zone II is caused by totally reflection which is related with initial ray angle, critical refraction angle of solid/liquid media. A serial of simulation experiments are operated on the base of the model, and numerical computation results coincide with model calculated results very well. Therefore, receiver should locate beyond these blind zones in the right boundary to obtain time of flight data which is used to reconstruct the solid/liquid interface.

Simultaneous measurement of ultrasonic longitudinal wave velocities and thicknesses of a two layered media in the absence of an interface echo

A measurement technique has been developed to extract the phase information of successive echoes for the simultaneous estimation of thicknesses and ultrasonic velocities of individual layers in a two layered media. The proposed method works in the absence of an interface echo and requires the total thickness of the sample to be known. Experiments have been carried out on two layered samples of white cast iron and gray cast iron with layer thickness variation in the range of 2-8 mm for total thickness variation in the range of 12-13 mm. Comparison with micrographs of a few samples confirmed the model predictions. The model is found to be sensitive to the total sample thickness but fairly insensitive to noise in the data.