1. Introduction
Structures used in engineering applications require the regular monitoring of their state, whether it be material defects, degradation or structural damage, to maintain safe operation and minimize overall costs. This requirement has led to the need to develop reliable methods of material evaluation and early detection of structural defects. Methods based on the propagation characteristics of ultrasonic waves have a well-established place among many available approaches. The analyses of the attenuation, reflections and scattering have led to the development and application of the methods used in field testing [
1]. It has allowed for the use of guided ultrasonic waves and their propagation characteristics for material testing and damage detection in many structures, i.e., plates [
2], thick-hollowed cylinders [
3] or composites [
4] to name a few.
In recent years, however, nonlinear characteristics of ultrasonic waves have drawn more research interest. Although the methods based on those features are more difficult for practical applications, they offer much higher sensitivity to early-stage defects, especially at a microstructural level [
5,
6,
7]. Overall, nonlinear features can be imposed over the propagating ultrasonic waves as a result of interacting with the following sources: material, i.e., inclusions, elastic and attenuation properties and grain structure; defects, i.e., cracks and delaminations; structural assemblies, i.e., joint friction; and intrinsic effects, i.e., overloads in the instrumentation chains. However, the last case relates to undesired experimental errors. In this paper, the first two sources, namely the material and defects, are of particular interest.
With such a range of nonlinear sources interacting with the propagating ultrasonic waves, a diversity of Non-Destructive Testing (NDT) and Structural Health Monitoring (SHM) approaches have been developed and deployed, represented by those using higher-order [
8,
9] or sub- [
10] harmonics, mixed-frequency responses [
11], shifts in resonance frequency [
12], vibro-acoustic modulations [
13] and the Luxembourg–Gorky effect [
14]. All the abovementioned methods have been reviewed comprehensively in [
15]. This work focuses on the higher-order harmonic generation in the guided ultrasonic waves. It has been evidenced as an energy transfer from the fundamental frequency
to the integer multipliers of fundamental frequency (
and so on). However, depending on the type of the chosen guided ultrasonic wave, different results can be obtained. If one uses Lamb waves to examine chosen structures, all higher harmonics should be present in the frequency response upon the interaction with the nonlinear source [
16]. However, if one focuses on the structure examination using the shear horizontal (SH) waves, due to the propagating characteristics of the aforementioned waves, only odd harmonics will be inflicted into the frequency response collected from the structure [
17]. The propagation characteristics of the latter type of guided ultrasonic waves are examined in the present work.
In the last decade, more intensified research development can be observed focusing on the nonlinear features inflicted over the propagating SH waves. However, very few works have considered the third-harmonic generation of the fundamental SH
0 wave mode. Li et al. [
18] investigated the cumulative characteristic of the third-harmonic generated due to the interaction of the fundamental SH
0 wave mode with the nonlinear material. Similar work was conducted by Wen et al. [
19], but in this case, material nonlinearity was enhanced by imposing thermal degradation over the investigated material. None of the abovementioned works considers a combination of the sources, i.e., nonlinear material and frictional motion of the crack. Osika et al. [
20] presented the numerical investigation of the combination of the two aforementioned sources interacting with the propagating SH
0 mode; however, no experimental results were shown. The present work aims to fill those gaps. Other works have either focused on the generation of the second-harmonic SH
0 mode due to the interaction of Lamb waves with nonlinear sources [
21] or wave mixing [
22].
The aim and novelty of the present work are the results of the interaction of the fundamental guided shear horizontal wave mode—SH0—with two nonlinear sources, i.e., nonlinear material and fatigue cracks executing frictional motion. A numerical study is prepared based on the Local Interaction Simulation Approach (LISA)—a finite-difference approach developed for considering sharp interface changes. The sources are considered separately and then combined for the final case. This numerical study is followed by an experimental study. In both investigational approaches, a location of the fatigue-generated crack is possible to point out in the presence of the material nonlinearity. The nonlinear gamma——parameter is introduced to enhance the location of the damage in the examined structure.
The paper is organized as follows. Shear horizontal wave theory is presented in the next section. The linear and nonlinear material definitions are considered. This is followed by the description of the local type of nonlinearity, namely Coulomb’s frictional law.
Section 3 introduces the numerical tool used in this work—LISA—followed by the definition of the model and the results of the investigated numerical cases.
Section 4 concerns the experimental study of the problem. Setup is presented followed by the results obtained from this study. Finally, conclusions are provided in
Section 5.
3. Numerical Simulation
3.1. Local Interaction Simulation Approach—LISA
The LISA method was first introduced in the 90s of the previous century by Delsanto et al. [
29,
30,
31] for the purpose of modeling wave propagation in media with sharp interfaces and inclusions of different material properties. This approach has also been used extensively for simulating crack–wave interactions of the propagating Lamb wave used for damage detection investigations [
32,
33]. The work in [
16,
28] presents the application of LISA for the modeling of Lamb wave propagation in a hyperelastic medium. This approach is based on the finite-difference (FD) formulation used for the discretization of the space derivatives in the governing differential equations. The explicit central difference formula is used for time-domain discretization. Thus, the method is well suited for parallel computations, as demonstrated in [
34]. For a medium described by two Lagrangian coordinates, the LISA method discretizes the geometrical models into a 2-D grid of rectangular cells. Material properties are assumed to be constant within cells but they may differ among these cells. As a result, LISA is an appropriate tool for wave propagation in complex media that are heterogeneous, anisotropic and nonlinear [
30]. The implementation of the LISA framework for SH wave propagation in a hyperelastic medium follows that given in [
28]; therefore, only a short description is presented below.
To obtain the iteration equations for nonlinear LISA, the spatial derivatives in Equation (1) are replaced by difference formulas. For a two-dimensional case, the iteration equations are derived for the point at the intersection of four adjacent cells—point
p shown in
Figure 3a. In this derivation, each cell is treated separately, and nodal point
p is replaced by four points
as shown in
Figure 3b, where
. The Navier elastodynamic equation—evaluated at the introduced points
—can be written as follows.
The finite-difference formulas
and
are evaluated for assumed stress component distributions for each cell 1–4 presented in
Figure 3b. To complete the set of four equations defined by Equation (18), a stress continuity needs to be enforced between the neighboring cells. This leads to the reduction in stress tensor components in Equation (18). After summing the equations from all four cells with the imposed stress continuity, the stress-based iteration equation for nodal point
P can be obtained as
It should be noted that the presented derivation using the LISA scheme is based on evaluating the elastodynamic equation and imposing continuity of selected stress tensor components across the interfaces of adjacent cells. As a result, this procedure can be conducted for the assumed form of the constitutive elastic relation and the geometrical definition of strains.
Due to the action of decomposing the point
p into four separate points, one for each cell, it is possible to model the stick–slip motion of the crack interfaces using the LISA framework. These nodes are grouped in pairs and are considered independent. First, the classic iteration equations are composed for each node as it was shown previously. Then, for node pairs, which are on the same surface of the crack, the constraints are imposed for the stick motion in the following form
where
is the displacement along the
axis, with superscripts corresponding to the
th node at the location of the defined crack and subscripts
corresponding to the current and next time step of the simulation. The signs
represent the surfaces of the crack, for instance, left and right.
Using the Lagrange multipliers method, it is possible to implement the abovementioned constraints, as well as determine crack node displacement , and reaction body forces . During each time step of the simulation, within each node pair, the values of the module of reaction body forces are compared with the body force equivalent of the maximum static friction force. If , then nodes at the location of the crack will move independently, i.e., in slip motion. Moreover, an additional external body force equivalent of the kinematic friction force is imposed on each of these nodes. Its direction and sense are dependent on the direction and sense of the relative velocity vector of the considered node pair.
3.2. Model Description
To scrutinize the influence of a nonlinear medium and its interaction with local defects on the propagation characteristics of the SH wave, a numerical model of the aluminum beam was prepared. A 2 mm thick beam was considered with its length equal to 1000 mm as shown in
Figure 4. Such a length was chosen to avoid any reflection of the wave from the opposite end to the excitation point. The model was discretized into a set of square elements, with a size of
. This setup led to obtaining 20 elements through the thickness of the considered beam. The time step for the simulations was set as 10 ns to maintain the numerical stability during calculations. The material properties of the model are presented in
Table 1. The Landau TOECs and FOECs values were calculated using the data presented in [
35].
The local type of nonlinearity in the form of the crack was located 250 mm from the left side of the beam. As the frictional motion of the crack interfaces is investigated, the static and kinetic friction coefficients were set to
and
, respectively. Six depths of the fatigue crack were investigated as the percentage of the beam thickness (from 0 to 25% with a step every 5%). For the numerical cases, where fatigue crack was considered, a distribution of compressive residual stress
was assumed over the crack interfaces. These stresses may result from manufacturing processes such as cold rolling. Following the work presented in [
36], a symmetrical stress distribution with respect to the center of the beam thickness was imposed. For
, the residual stresses were described by
, where
and
. By imposing the residual stresses over the crack surfaces, it was possible to determine the body forces between equivalents of static and kinetic frictional forces.
A fifteen-cycled sine signal multiplied by the Hanning window was used as an excitation signal. The central frequency was set as 200 kHz. The excitation signal was assigned to the left side of the beam as a displacement boundary condition and distributed uniformly over the beam thickness. The amplitude was equal to . The choice of the excitation frequency and the thickness of the modeled beam led to the excitation of only the fundamental SH wave mode, i.e., SH0 mode. Its non-dispersive characteristics significantly simplified the data analysis and interpretation of numerical results. Finally, the responses of the modeled structures were collected as displacements from the upper surface of the beam. Measurement points were selected over the entire length of the model with a step of every 20 mm.
3.3. Results
First, only the nonlinear material is considered. Results of the simulation presented in
Figure 5 are captured at three measurement points: 120 mm, 420 mm and 720 mm from the excitation point. In the time domain, only one non-dispersive wave packet is visible (for each point) that corresponds to the primary non-dispersive SH
0 mode. In the frequency domain, apart from the fundamental excitation frequency, one can observe the generation of the third harmonic. Moreover, the amplitude of this higher-order component increases with distance as shown in the bottom plot of
Figure 5. Due to the non-dispersive characteristic of the SH
0 mode, the internal response between the first and third harmonics is not dependent on the chosen frequency. Thus, at any chosen frequency, when only the SH
0 wave mode is analyzed, the cumulative effect with the propagation distance is observed due to the presence of global-type nonlinearity in the form of the nonlinear material definition based on the Landau material description.
The second case is focused on the interaction of the propagating wave with a local type of nonlinear source, namely the shear movement of crack surfaces. This motion was implemented through the incorporation of Coulomb’s friction model in the scope of the LISA model. The results of the performed simulations are presented in
Figure 6. The time-domain signals are collected from the measurement point at 720 mm from the excitation site for six different depths of damage. As can be seen in the top plot of
Figure 6, no significant difference can be observed among the wave packages obtained from the examined beam with different levels of damage. However, after transferring to the frequency domain, significant differences can be observed due to the increasing depth of the crack. As shown in the bottom plot of
Figure 6, the magnitude of the generated higher-order harmonics is proportional to the depth of the crack. Finally, when comparing the bottom plots of
Figure 5 and
Figure 6, it can be observed that the local type of nonlinearity, i.e., the frictional motion of the crack surfaces, generates much higher magnitudes and numbers of the higher-order harmonics upon its interaction with the propagating wave than the global type of nonlinearity being the nonlinear material definition.
In the last case, both nonlinear sources are implemented, and the results of such simulations are presented in
Figure 7. Similar to the case of only local nonlinearity, the displacement responses in the time and frequency domains from the measurement point are located at a distance of 720 mm from the excitation. No visible changes can be observed in the time-domain signals due to the increase in the depth of the frictional crack. However, the opposite can be said about the frequency representation of the analyzed signals.
As in the case where only the local type of nonlinearity is defined in the examined model, for both types of nonlinearities present, higher-order harmonics can be noticed with significant levels of magnitudes. Furthermore, following a deeper analysis, one can also observe that the levels of magnitude in the case of both types of nonlinearities present are higher than when only frictional crack motion is considered. This observation could lead to the conclusion that a certain level of the superposition of the higher harmonics generated from the nonlinear material and frictional motion of the crack has occurred.
To confirm the abovementioned statement, all collected signal responses have been analyzed from the perspective of the generated third higher harmonic for the following cases: an intact beam with the linear material definition; an intact beam with the nonlinear material definition; a damaged beam (frictional motion of the crack surfaces) with linear material definition and a damaged beam (frictional motion of the crack surfaces) with nonlinear material definition. Obtained magnitudes of the third harmonics are plotted in the propagation distance domain and presented in
Figure 8. It can be observed that in the case of the intact beam with linear material definition, no third harmonic is visible. The value obtained from the frequency representation remains constant through the whole distance of wave propagation. Next, when the beam with the nonlinear material definition is scrutinized, a linear relationship between the third-harmonic magnitude and propagation distance can be seen. This observed phenomenon corresponds to the synchronism between the first and third harmonics, causing a continuous power transfer between the two harmonics—due to the presence of the global type of nonlinearity—and agrees with the available literature as well [
18,
19]. The third curve from the bottom refers to the model with the linear material definition and the presence of the crack with induced frictional motion among the asperities. One can observe that as the wave approaches the vicinity of the crack, a high-level of magnitude third harmonic is generated upon the crack–wave interaction. As the wave propagates further from the position of the damage, the value of the magnitude remains almost constant with respect to the propagation distance. When both types of nonlinearity are present in the analyzed numerical model, an increase in the third-harmonic magnitude can be seen in the area before the crack. Then, upon the interaction of the wave with the friction interfaces, a significant increase is observed, which is followed by the linear increase in the magnitude with the propagation distance in the area after the damage.
The observations obtained from the analysis of the results presented in
Figure 8 confirm that in the case of the SH
0 wave mode propagation, a superposition of the third harmonics generated from different nonlinear sources is possible to obtain. In the next section, an experimental validation is presented to confirm the numerical findings.
5. Conclusions
The characteristics of the generated higher harmonics resulting from the interaction of the shear horizontal wave with different sources were investigated. This work involved numerical simulations and experimental validations. The focus was on the identification of damage using the nonlinear characteristics of the shear horizontal waves. This objective was successfully achieved with the help of numerical and experimental investigations. The major conclusions of this work can be summarized as follows.
The results from the numerical simulations confirm that the fundamental shear horizontal wave mode—SH0—is synchronous with the higher-order harmonics of the same mode. That leads to the presence of cumulative effects when the wave propagates in the nonlinear medium. The increase in the third harmonic with the propagation distance is confirmed. Furthermore, it is also shown that the local type of nonlinear source, i.e., the frictional motion of the cracks’ surfaces, inflicts nonlinear features upon the propagating wave. However, the magnitude of the higher harmonics does not increase with the propagation distance. When considering the presence of both nonlinear sources in the examined medium, a superposition-like behavior of the third harmonics generated from the global and local nonlinear sources is observed. Nonetheless, numerical simulations also confirm that frictional motion induced by crack–wave interactions generates higher values of the third and other higher-order harmonics than the continuous interaction of the propagating wave with the nonlinear material. Therefore, the generation of higher-order harmonics can be used to localize the fatigue-generated damage in the investigated structure in the presence of the globally distributed material nonlinearity.
This experimental investigation shows that the cumulative effect is difficult to obtain. The increase in the third-harmonic magnitude with the propagation distance is only possible to observe when a first-order curve fitting is applied. However, analyzing the third-harmonic characteristics over the propagation distance allowed us to identify the location of the fatigue-induced crack, which has a bigger influence on the generation of higher harmonics than the material’s nonlinearity, thus validating the results obtained from numerical simulations. Moreover, through the calculation of the parameter as the ratio of the third-harmonic magnitude to the cube of the first-harmonic magnitude, it is possible to point out the location of the damage with higher precision than in the case of analyzing only third-harmonic magnitudes. Finally, by analyzing the area after the crack, it is still possible to identify the damaged beam from the undamaged one.
In summary, the nonlinear propagation characteristics of the shear horizontal wave help identify the source of nonlinearity—global or local—and in the case of the local type, it is also possible to find its location.
Future work in the area should focus on the experimental identification of the depth of the crack. In addition, further narrowing down the location of the damage would be advantageous.