Abstract

This paper presents a modified boundary element method (BEM) to solve the scattering problem of Love surface wave from a two-dimensional cavity defect. Because of the truncation of BEM models at a far distance from the cavity, spurious reflected waves are generated. In order to eliminate the unwanted reflections, the guided Love-wave displacement patterns are assumed on the far-field infinite boundaries previously omitted by model truncation, and they are incorporated into the BEM equation set as modified items. The numerical results are verified by theoretical solutions of far-field Green’s functions. Additional parametric studies are performed to find out the influence of truncation distance and defects’ geometric characters on the accuracy of scattered wave solutions.

1. Introduction

The ultrasonic nondestructive testing (NDT) techniques have wide applications for quantitative characterizations of mechanical properties and detection and characterization of cracks and defects. Traditional ultrasonic testing techniques using bulk waves are very time-consuming, since these techniques need an overall inspection of the structure. However, ultrasonic guided waves are attractive for inspection of long-range or wide area structures because they can travel considerable distances and therefore scan large regions for defects in shorter testing time [1, 2].

The current NDT applications of guided waves include pitch-catch [3] or pulse-echo [4], flaw detection method, phased array configuration [5], and diffraction tomography [6]. Generally, these methods make use of time-of-flight (TOF) of the reflected waves from inner defects to locate their approximate positions.

However, further information (e.g., defect shapes or depths) cannot be further utilized because of the complexity of guided waves. Hence, the quantitative nondestructive testing requires a thorough understanding of surface wave scattering in forward and inverse aspects. For the forward problem, we need to solve the near- and far-fields accurately and obtain the scattering coefficients for following inverse reconstruction [7–9].

The Love-wave is a special kind of guided waves that travels along the surface of elastic layer covered on top of an elastic half-plane. The scattered Love-waves are relied on to investigate underground information in geotechnique engineering, earthquake engineering [10], or detecting flaws and cracks at the bounding interface in nondestructive testing applications [11]. An effective utilization of the Love-wave requires a thorough understanding of its scattering phenomenon.

For the calculation of scattered wave field over a finite domain, various technologies can be implied, like finite element method (FEM) [12], BEM [13], mode-exciting method [14], matrix theory [15], and so on. However, for the forward analysis of a half-plane, the BEM is especially effective, since only the interfaces and flaw boundaries need to be meshed. There are BEM approaches using two kinds of Green’s functions: half-space and full-space. Using the former one, only the flawed portion needs to be meshed; however, Green’s function cannot be written in a closed form. Conversely, using the latter one, the whole interface should be meshed; however, Green’s function is much simpler. Thus, for the forward analysis of Love-wave, we adopt the latter one.

However, in traditional BEM approaches, due to the inevitable artificial truncation of BEM model, spurious reflected waves are introduced in the final results of scattered wave field, which causes considerable error. Another big challenge to solve the scattering problem is the existence of multiple dispersive modes of Love-waves at a certain frequency along with the modal conversion, due to the interaction at the damage location.

Here, a modified BEM for calculating scattered Love-waves is introduced. In this paper, the guided Love-wave displacement patterns are assumed on the far-field infinite boundaries previously omitted, and they are incorporated into BEM equation sets as the modified items. With this improvement, the spurious reflected waves are eliminated. The numerical results are verified by theoretical far-field Green’s functions [16, 17]. Furthermore, various parametric studies of the influence of defect locations and geometrical shapes and size on the calculations of Love-wave scattered fields are carried out in the later sections, which have potential values for investigating forward problem or inverse problem of flaw reconstruction based on surface waves.

2. Statement of the Problem

The Love surface wave propagates along the surface of elastic layer of thickness covered on top of a homogeneous, elastic half-plane, containing a cavity of arbitrary shape on the bonding interface of the - plane (see Figure 1(a)). Here, we consider an incident Love-wave propagating in the -direction, which interacts with the cavity generating forward-scattered and back-scattered surface wave.

Figure 1 

Linear superposition principle: (a) the total field; (b) the incident field; (c) the scattered field.

By virtue of linear superposition principle, the total field in the flawed structure defined by Figure 1(a) can be considered as the superposition of the incident and the scattered waves. The incident wave can be treated in the intact (or reference) structure without cavity, as shown by Figure 1(b), and the scattered field is analyzed in the flawed configuration in Figure 1(c). The scattered field is equivalent to the field generated by the contribution of the tractions exerted on the actual surface of the cavity. Furthermore, these tractions are equal in magnitude but opposite in sign to the corresponding tractions produced by the incident Love-wave field on the surface of the fictitious cavity as shown by Figure 1(b). Thus, these tractions can be obtained by calculating the stress components and the outward normal vectors along the fictitious cavity surface using the Cauchy’s formula from the incident field. The dynamic reciprocal theorem is then applied to calculate the scattered wave field equivalent to the radiated field generated by these tractions.

3. Equations

3.1. The Elastodynamic Reciprocal Theorem

The dynamic reciprocal theorem relates two elastodynamic states 1 and 2 of the same bounded or unbounded body, which can be stated aswhere , , and represent body forces, displacements, and stresses, respectively, and is the k component of unit vector outward surface normal to .

Let us consider two-dimensional elastodynamic problems in an isotropic half-plane with a different homogeneous and isotropic layer covered with boundary . The boundary integral equation of antiplane motion for a source point taken on , in the absence of body forces, is developed from (1) and derived aswhere the factor 1/2 is valid only if the boundary is smooth at point and and are the full-space frequency domain elastodynamic antiplane fundamental solution displacement and traction tensors, respectively, which are derived [18] aswhere is the Hankel function of the order of the first kind; , , , and stand for the elastic constants and the wave-numbers of the shear wave at current frequency, for the upper and lower materials, respectively, where , in which and are material densities; represents the distance between and ; and are the displacement and boundary traction, respectively, at the point , respectively, due to a unit line force exerted at . For current antiplane problem, both the line force and Green’s function - - and only have the component.

Let us assume that, except the flaw region and , both the free-traction surface and the interface are flat. Let and be the free upper surface and the interface between upper-layer and half-plane, respectively, and and represent the remaining infinite part of upper and lower boundary, respectively, which will be omitted by truncation in traditional BEM (see Figure 1).

By substituting all boundaries divided in Figure 2 into (2), the BIE of the layered media and half-plane are derived asrespectively, where the superscripts and indicate the Green functions of half-plane and the layer, respectively.

Figure 2 

Schematic definition of the computational domain of the Love-wave scattered field, with divided interface boundaries .

3.2. Far-Field Assumption

Since body waves geometrically attenuate in the propagating direction, the far-field displacement solution can be approximated by a series of Love surface waves, neglecting the contribution of body waves.

Therefore, we assume that if the truncated points are located far enough from the source regions, the displacement solutions of the infinite boundary at each side can be expressed aswhere the coordinate vector is in the form of , are defined as the unknown complex amplitudes of the far-field solutions of the th order mode Love-wave. Here, is the number of modes, and represent the th order mode displacement of unit amplitude Love-wave propagating in the positive and negative direction of axis . (Note that Love surface waves are dispersive.)

By virtue of assumptions in (6), (4) and (5) can be rewritten as respectively. From (7), we definewhich represent the corrected items accounting for the contribution of the omitted boundary. Thus, (7) are simplified as

Note that unknown parameters are introduced into the BIEs, which will add degrees of freedom to the final BEM system of the BIEs.

3.3. Correction over the Omitted Part of the Infinite Boundary

In traditional BEM approaches, the contribution of integral terms on the infinite boundary, that is, the fourth term on the right-hand side of (4) and the third term on the right-hand side of (5), is omitted, which introduces considerable error. In order to separately determine the integral terms over infinite boundaries such as and , a multidomain approach is applied, which involves the division of the whole interfaces and boundaries into four regions by introducing two fictitious boundaries and , as shown in Figure 2. Here, an incident Love-wave mode with unit amplitude is introduced propagating along the upper free surface in the positive or negative direction of , respectively (see Figure 3).

Figure 3 

Schematic diagram of the multidomain approach: (a) calculate ; (b) calculate ; (c) calculate ; (d) calculate .

Let us choose the Love surface wave of unit amplitude as elastodynamic state 1 and the full-space fundamental solution as elastodynamic state 2. For instance, by virtue of reciprocal theorem seen from (2), the BIE for region 1 is given as

By simplifying (10), we arrive at

By implying an analogous approach for other regions, we can get , and , which are expressed as

Note that for the calculation of and , the incident Love-wave is assumed to propagate in the positive direction, while for and the propagating direction is opposite. The fictitious boundary is, in principle, infinite. However, since the integrand over attenuates rapidly in exponential form away from the interface of half-plane, without loss of accuracy, we consider the boundary as a distance of about two Love wavelengths.

4. Numerical Computation

Numerical solutions of (9) require the discretization of the boundary into elements. After the discretization of the boundary and interpolation of the displacements and tractions, the discretized BIEs for the layer and half-plane can be written for each of nodes and , respectively, aswhere , are the total number of nodes for the layer and half-plane, respectively, is the number of nodes per element, is the same shape function for each element, and represents the intrinsic coordinate of the parent element. It is noted that the calculation of corrected coefficients and is performed in the previous section.

Equations (15) can be expressed in a more concise manner by definingwhere the subscripts represent the collocation points and with the node of element . Then the above equations are rewritten as

Then, let us assemble the local elements , into global matrices , , the node displacement and node traction into global matrices , , and the correction and the unknown amplitudes into the correction matrices and the amplitude matrices . Equation (18) can be written aswhereConveniently, the corrected BEM system can be rewritten aswhere , are block matrices of , and , are the node displacement vectors and node traction vectors corresponding to , respectively.

Analogously, (19) can be expressed in matrix form:wherewhere , are block matrices of , and , are the node displacement vectors and node traction vectors corresponding to , respectively.