This paper proposes a framework of using deep learning-assisted methods for the prediction of interfacial conditions in coated plates using guided wave data. The coating-substrate interface is modeled as a linear spring layer of zero thickness, and the mechanical behavior of this spring layer is characterized by the spring compliance. Both tangential and normal spring compliances are introduced to characterize the bond quality. Numerical simulations are conducted for a wide range of spring compliances to generate the corresponding dispersion curves. A Long Short-Term Memory (LSTM) network is utilized to predict the interfacial conditions. In addition, we consider the delamination cases where the coating layer is completely separated from the substrate over the delaminated region. Finite element simulations are carried out to model guided wave generation, propagation, interaction with delamination, and reception. The time-space images are formed by measuring the time-domain signals by receivers at several locations downstream from the source transducer, which are then fed into the developed Convolutional Neural Network (CNN). Once trained, this Deep-Learning (DL) model enables the accurate prediction of delamination location and size. Results of this paper demonstrate that the proposed methodologies have tremendous potential for characterizing interfacial conditions in practical Nondestructive Evaluation (NDE) and Structural Health Monitoring (SHM) applications.
Predictive health monitoring will require the development of advanced sensing techniques capable of providing quantitative information on the damage state of structural materials. Second harmonic generation techniques can measure absolute, strength-based material parameters which can be coupled with uncertainty models to enable accurate and quantitative life prediction. Starting at the material level, this talk will examine a combination of sensing techniques and physics-based models to characterize damage in metals. These second harmonic techniques are acoustic-wave-based, so component interrogation can be performed with bulk, surface, and guided waves using the same underlying material physics. The talk will consider applications to characterize fatigue damage, thermal embrittlement, irradiation damage, and sensitization.
This research investigates the feasibility of measuring acoustic nonlinearity in aluminum with different ultrasonic guided
wave modes. Acoustic nonlinearity is manifested by generation of a second harmonic component in an originally
monochromatic ultrasonic wave signal, and previous research has shown this correlates to an intrinsic material property.
This parameter has been shown to increase with accumulated material damage - specifically in low- and high-cycle
fatigue - prior to crack initiation, whereas other ultrasonic nondestructive evaluation (NDE) techniques measuring linear
parameters are unable to detect damage prior to crack initiation. In structural components such as jet engines and
aircraft structures subjected to fatigue damage, crack initiation does not occur until ~80% of a component's life. Thus,
there is a need for structural health monitoring (SHM) techniques that can characterize material damage state prior to
crack initiation, and therefore nonlinear ultrasonic techniques have the potential to be powerful NDE and SHM tools.
Experimental results using Rayleigh and Lamb guided wave modes to measure acoustic nonlinearity in undamaged
aluminum 6061 samples are presented, and a comparison of the efficiency of these modes to measure acoustic
nonlinearity is given.
This study develops a technique to decompose a multi-mode, transient Lamb wave signal into individual Lamb mode signals. The previously-proposed technique (presented at the SPIE NDE conference 2005) showed an encouraging efficiency for numerically-simulated signals, but suffered when evaluating real experimental signals due to its high sensitivity to noises and experimental errors. The improved technique starts with the same assumption of known Lamb wave propagation characteristics (known propagation dispersion curves). However, a highly-strict signal model governing the development of the previously-proposed technique is relaxed to tolerate unavoidable presence noises and errors, and the problem is re-formulated. For actual experimental signals, some additional signal processing techniques are introduced in signal pre-conditioning. The entire implementation of the improved technique is first tested with simulated signals, and then applied to actual experimental signals. The final results with real experimental are presented and discussed.
This research proposes a technique to decompose a transient, multi-mode, Lamb wave, time-domain signal into its individual Lamb wave modes. The technique is derived for a Lamb wave signal consisting of two Lamb wave modes (double-mode Lamb wave signal). The extension to the general multi-mode signal is straightforward, but requires additional computations. The proposed technique assumes knowledge of the dispersion characteristics of Lamb waves, which can be theoretically calculated or obtained by other signal processing techniques. The proposed technique is verified for simulated (by eigen-expansion) signals to demonstrate the method's accuracy in a well-controlled environment. The use of eigen-expansion signals for the simulation allows for the calculation of the total response, as a combination of responses due to all existing Lamb modes. The comparison between the decomposed and simulated signals shows good agreement and demonstrates the validity of the proposed technique. The paper concludes with a discussion of the extension of this technique to the more general multimode, Lamb wave signal in leaky conditions, and its use in attenuation calculations.
This research uses a combination of laser ultrasonics, signal
processing and analytical modeling techniques to examine the
propagation of transient Lamb waves in absorbing plates -- in particular an isotropic plate with a lossless fluid on one side.
The motivation is to develop a non-contact, point source-receiver
technique capable of measuring attenuation of Lamb waves in lossy
situations in general. The theoretical model enables an understanding
of loss mechanisms and enables prediction of attenuation. The laser
ultrasonic techniques enable broadband, point source-receiver
idealization. The proposed signal processing technique, the
short-time Fourier transform, resolves a signal into the
time-frequency domain and enables Lamb mode separation and
calculation of energy distribution at specific frequency-velocity
points on the dispersion curves. The experimental procedure uses a stress-free plate as a reference for system variation and other loss mechanisms. Then, a plate with fluid on one side is tested to validate the signal processing algorithm. There is satisfactory agreement between the attenuation measured and predicted in the frequency range of interest.
This research examines the propagation of guided Lamb waves in
bonded components, establishing the effectiveness of combining
laser ultrasonic techniques with a time-frequency representation
(TFR) to experimentally measure the dispersion curves of a layered
medium. The specific layered medium examined is a fiber-reinforced
polymer (FRP) plate bonded (with an adhesive layer) to concrete. A
TFR is used to operate on experimentally measured, guided Lamb
waves to resolve individual Lamb wave modes and to generate the
system's dispersion curves. The objective of this research is to
demonstrate that it is possible to develop the dispersion curves
of FRP bonded components from a single, experimentally measured
guided wave signal. The experimental results show that, by
examining the characteristics of the system's dispersion curves,
the stiffer the bond, the more deviation from the behavior of a
free plate case, and the less modes that are present.
This paper considers the problem of wave propagation in a nonlinear elastic medium with a quadratic stress-strain relationship. The paper is limited to one-dimensional wave propagation. Under these conditions, the initial value problem is formulated into a hyperbolic system of conservation laws. The Riemann problem due to an initial step function excitation is considered first. Analytical solutions to the Riemann problem are obtained by solving the corresponding eigenvalue problem. In addition, a computer program is developed based on the high-resolution central scheme of Kurganov and Tadmor. The accuracy of this numerical procedure is verified by comparing the numerical results with the exact solutions. The second part of the paper considers several different types of initial excitations in order to determine special characteristics of the wave propagation due to material nonlinearity.
Pile tip elevations are unknown for nearly half of Georgia's 14,500 bridges. Since the length of a pile is directly related to its capacity, and ultimately, to a bridge's load capacity, the need for a straightforward, effective, and inexpensive method to determine pile embedment lengths is apparent. If the top of the pile is exposed and free to be impacted and instrumented, the task is relatively simple. The exposed pile top is impacted axially, and the resulting longitudinal wave motion is monitored and used to calculate the pile lengths. However, such piles rarely exist in bridge structures, and the more common case is that the tops of the piles are cast into the bent cap. This lack of access poses a challenging problem when trying to nondestructively ascertain the length of pile that is embedded in the surrounding soil. Although soil borings and other intrusive tests are capable of determining pile tip elevations, the time and cost of performing these tests on a large number of bridges is prohibitive. The research project described here uses flexural waves, induced by exciting piles laterally, to determine the unknown embedment lengths. Modal analysis techniques are employed to quantify the difference in modal characteristics (natural frequencies, damping, and mode shapes) of piles with differing lengths, to create a modal model of the system, and finally to back-calculate the unknown pile embedment lengths.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.