Experimental Ultrasonic Data Research Articles

Automated Weld Defect Detection in Industrial Ultrasonic B-Scan Images Using Deep Learning

Automated ultrasonic testing (AUT) is a nondestructive testing (NDT) method widely employed in industries that hold substantial economic importance. To ensure accurate inspections of exclusive AUT data, expert operators invest considerable effort and time. While artificial intelligence (AI)-assisted tools, utilizing deep learning models trained on extensive in-laboratory B-scan images, whether they are augmented or synthetically generated, have demonstrated promising performance for automated ultrasonic interpretation, ongoing efforts are needed to enhance their accuracy and applicability. This is possible through the evaluation of their performance with experimental ultrasonic data. In this study, we introduced a real-world ultrasonic B-scan image dataset generated from proprietary recorded AUT data during industrial automated girth weld inspection in oil and gas pipelines. The goal of inspection in our dataset was detecting a common type of defect called lack of fusion (LOF). We experimentally evaluated deep learning models for automatic weld defect detection using this dataset. Our assessment covers the baseline performance of state-of-the-art (SOTA) models, including transformer-based models (DETR and Deformable DETR) and YOLOv8. Their flaw detection performance in ultrasonic B-scan images has not been reported before. The results show that, without heavy augmentations or architecture customization, YOLOv8 outperforms the other models with an F1 score of 0.814 on our test set.

Frequency-dependent elastic properties of fracture-induced VTI rocks in a fluid-saturated porous and microcracked background

Fractures are widely distributed underground. The stiffness matrix of fractured rocks has been extensively investigated in a fluid-saturated porous background medium. However, the existing stiffness models only incorporate the attenuation mechanism of wave-induced fluid flow (WIFF). For macroscopic fractures, the elastic scattering (ES) of fractures cannot be ignored. To alleviate this issue, a frequency-dependent stiffness matrix model is developed, including the mesoscopic WIFF between fractures and background (FB-WIFF), the microscopic squirt flow, and the macroscopic ES from the fractures. By combining the far-field scattered wavefields of normal incident P and SV waves with the linear slip theory, the dynamic full-stiffness matrices for fracture-induced effective vertically transversely isotropic rocks in a fluid-saturated porous and microcracked background are derived. Then, the P, SV, and SH-wave velocities and attenuation can be obtained through the Kelvin-Christoffel equation. The results indicate that the FB-WIFF mechanism significantly affects the velocities and attenuation of the P and SV waves but has almost no effect on the SH wave, whereas the squirt flow and ES mechanisms affect the velocities and attenuation of all the P, SV, and SH waves. For validation, the model is compared with existing models and previous experimental ultrasonic data.

Ultrasonic time-reversal-based super resolution imaging for defect localization and characterization

In this paper, time reversal with multiple signal classification (TR-MUSIC) and its phase-coherent form (PC-MUSIC), the two typical ultrasonic time-reversal-based super resolution imaging techniques, have been introduced and explored to image defects in metallic samples. The principles for super resolution imaging technique and defect characterization based on ultrasonic image have been presented firstly. And then both TR-MUSIC and PC-MUSIC are tested with experimental ultrasonic array data acquired from the metallic samples including defects using the full matrix capture (FMC) process. Here, if the size of defect is smaller than the central ultrasonic wavelength, the defect can be regarded as point-like target. On the contrary, if the size of defect is larger than the central ultrasonic wavelength, it can be regarded as extended target. The comparison on super resolution imaging for point-like and extended targets has been investigated. For point-like targets, TR-MUSIC as well as PC-MUSIC can distinguish and locate the position, and the overall performance of TR-MUSIC is better than that of PC-MUSIC. For extended target, TR-MUSIC can only determine the area where the defect occurs. PC-MUSIC can distinguish and assess the length of extended target parallel to the array, the lowest error is 9.62% when the dimension of signal subspace is set to be 10 for the case considered. In addition, the defect characterization based on ultrasonic super-resolved image has been studied, and a sizing strategy is introduced. The length and angle of extended targets oblique to the array can be obtained from the PC-MUSIC images. The worst errors for assessing the length and angel are 9.14% and 13.3% respectively. And for the cases considered in the experiment, the longer the length of extended target, the smaller the error.

Estimation of the Squirt-Flow Length Based on Crack Properties in Tight Sandstones

Abstract Tight sandstones have low porosity and permeability and strong heterogeneities with microcracks, resulting in small wave impedance contrasts with the surrounding rock and weak fluid-induced seismic effects, which make the seismic characterization for fluid detection and identification difficult. For this purpose, we propose a reformulated modified frame squirt-flow (MFS) model to describe wave attenuation and velocity dispersion. The squirt-flow length (R) is an important parameter of the model, and, at present, no direct method has been reported to determine it. We obtain the crack properties and R based on the DZ (David-Zimmerman) model and MFS model, and how these properties affect the wave propagation, considering ultrasonic experimental data of the Sichuan Basin. The new model can be useful in practical applications related to exploration areas.

A parametric model for global thermodynamic behaviour of ultrasonic attenuation in magnetic field

The ultrasonic attenuation and velocity variations are theoretically investigated near the Curie temperature of ferromagnet under an application of magnetic field. The temperature, frequency and magnetic field dependence of acoustic properties of ferromagnet near a critical point is given. A parametric representation is used to describe a crossover from asymptotic critical behaviour to classical regime far away from the critical point. The crossover scaling functions are determined for sound attenuation coefficient and dispersion. We compare the proposed crossover model with experimental ultrasonic data for manganese phosphide MnP and find good agreement between theory and experiment.

Capability of Advanced Ultrasonic Inspection Technologies for Hydraulic Turbine Runners

This paper presents the results of a project aimed at evaluating the performance of ultrasonic techniques for detecting flaws in Francis turbine runners. This work is the first phase of a more ambitious program aimed at improving the reliability of inspection of critical areas in turbine runners. Francis runners may be utilized to supply power during peak periods, which means that they experience additional load stress associated with start and stop sequences. Inspection during manufacturing is then of paramount importance to remove as much as feasible all flaw initiation sites before the heat treatment. This phase one objective is to collect initial data on a simplified mock-up and then to compare the experimental ultrasonic data with the results of simulations performed by CIVA, a computer simulation package. The area of interest is the region with the highest stress between the blade and the web. A welded T-joint coupon made of UNS S41500 was manufactured to represent this high-stress area. During the FCAW welding process, ceramic beads were embedded in the weld to create discontinuities whose size is in the critical range to initiate a crack. Inspection of the material was carried out by various nondestructive testing (NDT) methods namely conventional pulse-echo, phased array, total focusing method (TFM). With these results, detection rates were obtained in order to compare the effectiveness of each method.

Elastic properties of transversely isotropic rocks containing aligned cracks and application to anisotropy measurement

Currently, most rock physics models, used for evaluating the elastic properties of cracked or fractured media, take into account the crack properties, but not the background anisotropy. This creats the errors of in the anisotropy estimates by using field logging data. In this work, based on the scattered wavefield theory, a sphere-equivalency method of elastic wave scattering was developed to accurately calculate the elastic properties of a vertical transversely isotropic solid containing aligned cracks. By setting the scattered wavefield due to a crack equal to that due to an equivalent sphere, an effective elastic stiffness tensor was derived for the cracked medium. The stability and accuracy of the approach were determined for varying background anisotropy values. The results show that the anisotropy of the effective media is affected by cracks and background anisotropy for transversely isotropic background permeated by horizontally aligned cracks, especially for the elastic wave propagating along the horizontal direction. Meanwhile, the crack orientation has a significant influence on the elastic wave velocity anisotropy. The theory was subsequently applied to model laboratory ultrasonic experimental data for artificially cracked samples and to model borehole acoustic anisotropy measurements. After considering the background anisotropy, the model shows an improvement in the agreement between theoretical predictions and measurement data, demonstrating that the present theory can adequately explain the anisotropic characteristics of cracked media.

Experimental and Numerical Study on Characterization and Evaluation of Surface Cracks with Wireless Ultrasonic Sensor

Adopting both wireless ultrasonic sensing and numerical simulation techniques, this research investigates the interaction between Rayleigh wave and artificial surface cracks of varying depths. When analyzing experimental ultrasonic data collected by a wireless sensing node, the signals are enhanced through a two-step procedure including signal reconstruction and envelope extraction. The waveforms are interpreted in detail by analyzing wave components through time-of-flight technique. A finite element (FE) model is devised to properly simulate the experimental testing. The simulated waveforms are consistent with experimental results and corroborate the analysis and explanations of experimental waveforms. Based on both experimental and numerical waveform analysis, a relationship between ultrasonic characteristic parameter and crack size is established for the quantitative estimation purpose. The proposed model shows a good agreement with data from both test and literatures.

Ultrasonic multi-frequency time-reversal-based imaging of extended targets

In this paper, two multi-frequency time-reversal-based imaging methods, time reversal with multiple signal classification (TR-MUSIC) and its phase-coherent form (PC-MUSIC), are explored for application to the non-destructive testing (NDT) imaging of extended target machined in solids. The size of which is on the order of the ultrasonic wavelength or larger. The principle for multi-frequency TR-based imaging method is presented, and the performance is estimated in terms of point spread function (PSF). Both TR-MUSIC and PC-MUSIC are tested with experimental ultrasonic array data acquired using the full matrix capture (FMC) process. Two experiments in metal solids with different kind of defect, side-drilled hole (SDH) as well as electrical discharge machining (EDM) lines that can be considered as extended targets, were implemented. Here, the EDM lines have different positional relationships with array. For SDH, it is shown that both TR-MUSIC and PC-MUSIC can locate the position, but fail to distinguish the shape. For EDM lines, TR-MUSIC can only detect the defect, but PC-MUSIC can assess the length based on the imaging results under some dimension of signal subspace. In addition, the effect of dimension of signal subspace on the imaging methods is also investigated, and it is shown that TR-MUSIC is sensitive to the dimension when compared with PC-MUSIC.

Estimation of Wave Velocity for Ultrasonic Imaging of Concrete Structures Based on Dispersion Analysis

Abstract Wave velocity is the most critical parameter used by the synthetic aperture focusing technique for the ultrasonic imaging of concrete structures. However, it has been challenging to estimate wave velocity with the traditional time-domain method that requires the picking of direct waves. The dispersion trend of direct waves is theoretically a horizontal line with frequency as the x-axis and phase velocity as the y-axis, and the horizontal line is due to the nondispersive phenomenon of direct waves in an elastic, homogeneous material. To overcome the challenge of picking direct waves in the time domain, this study proposes a frequency-domain method to first obtain the dispersion trend of direct waves based on the nondispersive phenomenon and then estimate the shear-wave velocity from the average phase velocity of the nondispersive trend in the frequency domain. This frequency-domain method was verified with the numerical and experimental ultrasonic data from concrete specimens with embedded delamination. A comparison of the results obtained using the two methods indicates that the frequency-domain method can significantly improve the accuracy of estimating the shear-wave velocity and eventually improve the accuracy of ultrasonic imaging.

A sphere-equivalency approach for calculating the elastic property of an isotropic solid containing aligned inclusions: Modeling and experiment data verification

Accurate modeling of elastic properties of cracked rocks in the earth’s shallow crust has long been an important topic in the field of geophysics. A sphere-equivalency approach of elastic wave scattering was used to model the elastic moduli of an isotropic solid containing aligned cracks. The results were compared with those of the existing Eshelby-Cheng and Hudson’s theories for dry and fluid saturation conditions, showing remarkably improved stability and accuracy for high crack concentrations, especially for Hudson’s second-order model. The stability and accuracy of the new approach were determined for varying solid and crack parameters. Finally, the new and existing theories were applied to model the laboratory ultrasonic experimental data measured on artificially cracked samples with varying wave frequencies and crack concentrations. Compared to Hudson’s theory, the new model agrees significantly better with the data. Specifically, the root-mean-square errors of theoretical fitting to data from our model are generally smaller than those from the other two models. We have thus developed an effective tool for modeling elastic properties of cracked rocks.

Non‐destructive testing method of micro‐debonding defects in composite insulation based on high power ultrasonic

It is crucial to ensure the adhesive quality of the interface in composite insulation equipment for power grid security. Due to viscoelasticity of silicone rubber and roughness of interface, the thickness of the local air layer in the defective interface is <100 μm. This kind of defect is called micro-debonding defect in this paper, and it is significantly smaller than the millimetre-scale air gaps which could be defected by the x-ray, THz, and traditional ultrasonic method. It remains challenging to detect the micro-debonding in composite insulation by non-destructive testing methods. In this paper, the acoustic model and bi-linear stiffness model for micro-debonding in different stages are established. Accordingly, a conjecture is proposed that the bonding property of composite insulation is related to the acoustic impedance of interface and non-linear distortion of its constitutive relation. The result obtained from numerical simulation with and without defect is compared with the high-power ultrasonic experimental data to validate the correctness of the theoretical model. In addition, it can be concluded that the non-linear distortion of high-power ultrasonic wave can be effectively used to diagnose micro-debonding defect at the interface of composite insulation in its early stage (1–20 μm).

Thermodynamic and acoustic properties of binary mixtures of diisopropyl ether, benzene and alkanes at 298.15, 308.15 and 318.15 K: Prigogine-Flory-Patterson theory and graph theory

Densities and speed of sound for the binary mixtures of diisopropyl ether (1) + n‑hexane, n‑heptane, benzene (2) and benzene (1) + n‑hexane, n‑heptane, n‑octane (2) and n‑hexane (1) + n‑heptane (2) were measured from 298.15 K to 318.15 K. The measured data were used to calculate excess molar volume VmE, deviation in ultrasonic speed Δu, isentropic compressibility KsE, excess intermolecular free length LfE and excess available volume VaE. All the derived properties were fitted to Redlich-Kister equation. For theoretical interpretation of VmE values, Prigogine-Flory-Patterson theory and Graph theory were used at 298.15 K. Various empirical correlations like Nomoto, Van-Dael and impedance dependence relation were applied to predict the experimental ultrasonic speed data. Schaaff's collision factor theory was used for prediction of experimental ultrasonic speed data at 298.15 K. The excess intermolecular length (LfE) was calculated from Jacobson free length theory at 298.15 K. The effect of temperature for KsE and VmE was also discussed in terms of intermolecular interactions.

Ultrasonic Analytic-Signal Responses From Polymer-Matrix Composite Laminates.

Ultrasound has been used to inspect composite laminates since their invention but only recently has the response from the internal plies themselves been considered of interest. This paper uses modeling techniques to make sense of the fluctuating and interfering reflections from the resin layers between plies, providing clues to the underlying inhomogeneities in the structure. It shows how the analytic signal, analyzed in terms of instantaneous amplitude, phase, and frequency, allows 3-D characterization of the microstructure. It is found that, under certain conditions, the phase becomes locked to the interfaces between plies and that the first and last plies have characteristically different instantaneous frequencies. This allows the thin resin layers between plies to be tracked through various features and anomalies found in real composite components (ply drops, tape gaps, tape overlaps, and out-of-plane wrinkles), giving crucial information about conformance to design of as-manufactured components. Other types of defects such as delaminations are also considered. Supporting evidence is provided from experimental ultrasonic data acquired from real composite specimens and compared with X-ray computed tomography images and microsections.

Finite-Temperature Behavior of PdHx Elastic Constants Computed by Direct Molecular Dynamics

ABSTRACTRobust time-averaged molecular dynamics has been developed to calculate finite-temperature elastic constants of a single crystal. We find that when the averaging time exceeds a certain threshold, the statistical errors in the calculated elastic constants become very small. We applied this method to compare the elastic constants of Pd and PdH0.6 at representative low (10 K) and high (500 K) temperatures. The values predicted for Pd match reasonably well with ultrasonic experimental data at both temperatures. In contrast, the predicted elastic constants for PdH0.6 only match well with ultrasonic data at 10 K; whereas, at 500 K, the predicted values are significantly lower. We hypothesize that at 500 K, the facile hydrogen diffusion in PdH0.6 alters the speed of sound, resulting in significantly reduced values of predicted elastic constants as compared to the ultrasonic experimental data. Literature mechanical testing experiments seem to support this hypothesis.

Attenuation and dispersion of ultrasound in cancellous bone. Theory and experiment

The paper presents theoretical and experimental issues related to modeling of elastic wave propagation in cancellous bone. The Biot’s theory, commonly used for modeling of ultrasonic wave propagation in cancellous bone, is discussed in context of its potential applicability for theoretical prediction of wave parameters. Particular attention is focused on the analysis of physical mechanisms of ultrasonic wave propagation in cancellous bone that govern phase velocity and attenuation coefficient as function of frequency and porosity. The analysis of the model is focused on the absorption and scattering mechanisms responsible for attenuation of ultrasonic waves in cancellous bone, which based on the ultrasonic experiments presumably play a predominant role in the total attenuation. The suitability of the model is discussed and verified by comparison of results of sensitivity analysis of the model with experimental ultrasonic data for bone mimicking phantoms and human cancellous bones.

Hardware and software architectures for computationally efficient three‐dimensional ultrasonic data compression

Ultrasonic industrial and medical imaging applications involve acquisition of large amount of volumetric data in real time. Therefore, data storage becomes critical in many current day applications which utilise ultrasound technology. Compressing the acquired data allows possessing minimal storage and also helps to rapidly transmit information to remote locations for expert analysis. The objective of this study is to design computationally efficient architectures for implementing discrete wavelet transform-based ultrasonic three-dimensional (3D) data compression algorithm on a reconfigurable ultrasonic system-on-chip (SoC) hardware platform. In this study, hardware and software architectures of the 3D ultrasonic compression algorithm are realised using Xilinx Zynq all programmable SoC. This study demonstrates that, compressing 33 MB of experimental ultrasonic 3D data into 0.42 MB (98.7% compression) requires only 84 ms for hardware architecture, and 1 min for software architecture, making both designs highly suitable for real-time ultrasonic imaging applications. Furthermore, the 3D compression is implemented by using Open Computing Language (OpenCL) targeted on Nvidia GT 750M graphical processing unit. OpenCL implementation of ultrasonic 3D compression algorithm completes the execution in <1 sec. This approach provides improved computational performance as that of hardware architecture, and comparable flexibility as that of software implementation.

Ultrasonic measurements of undamaged concrete layer thickness in a deteriorated concrete structure

Ultrasonic wave propagation in deteriorated concrete structures was studied numerically and experimentally. Ultrasonic single-side access immersion pulse-echo and diffuse field measurements were performed in deteriorated concrete structures at 0.5MHz center frequency. Numerically and experimentally it is shown that the undamaged layer thickness in a deteriorated concrete structure is measurable using pulse-echo measurements when the deterioration depth is larger than the wavelength. The signal overlapping, which occurs in the thin deteriorated layers, can be overcome using diffuse field measurements or a pattern matching technique. The ultrasonic experimental data were shown to be in good agreement with the widely used phenolphthalein test for concrete degradation.

Ultrasonic study on molecular interactions in binary mixtures of formamide with 1-propanol or 2-propanol

Ultrasonic speeds have been measured at 298.15K and 308.15K for mixtures of formamide+1-propanol or 2-propanol. For an equimolar mixture, excess molar compressibility follows the sequence of 1-propanol>2-propanol. The ultrasonic speed data are correlated by various correlations such as Nomoto's relation, van Dael's mixing relation and impedance dependence relation, and analyzed in terms of Jacobson's free length theory and Schaaff's collision factor theory. Excess isentropic compressibility is calculated from experimental ultrasonic speed data and previously reported excess volume data. The excess molar ultrasonic speed and isentropic compressibility values are fitted to Redlich–Kister polynomial equation. Other properties such as molecular association, available volume, free volume, and intermolecular free length are also calculated. The excess isentropic compressibility data are also interpreted in terms of graph theoretical approach. The calculated isentropic compressibility values are well consistent with the experimental data. It is found that the interaction between formamide and propanol increases when hydroxyl group attached to a carbon atom has more –CH3 groups.

Multi-frequency time-reversal-based imaging for ultrasonic nondestructive evaluation using full matrix capture.

In this paper, two multi-frequency time-reversal (TR)-based imaging algorithms are explored for application to the nondestructive evaluation (NDE) imaging of defects in solids: time reversal with multiple signal classification (TRMUSIC) and a related phase-coherent form (PC-MUSIC). These algorithms are tested with simulated and experimental ultrasonic array data acquired using the full matrix capture (FMC) process. The performance of these algorithms is quantified in terms of their spatial resolution and robustness to noise. The effect of frequency bandwidth is investigated and the results are compared with the single-frequency versions of these algorithms. It is shown that both TR-MUSIC and PCMUSIC are capable of resolving lateral targets spaced closer than the Rayleigh limit, achieving super-resolution imaging. TR-MUSIC can locate the positions of scatterers correctly, whereas the results from PC-MUSIC are less clear because of the presence of multiple peaks in the vicinity of target. However, an advantage of PC-MUSIC is that it can overcome the elongated point spread function that appears in TR-MUSIC images, and hence provide enhanced axial resolution. For high noise levels, TR-MUSIC and PC-MUSIC are shown to provide stable images and suppress the presence of artifacts seen in their single-frequency equivalents.