A spacing compensation factor for the optimization of guided wave annular array transducers

Cody Borigoa) and Joseph L. Rose

Fei Yan
Feature Based Systems Inc., State College, Pennsylvania 16801

(Received 29 August 2011; revised 30 August 2012; accepted 9 November 2012) Transducer arrays can be utilized in ultrasonic guided wave applications to achieve preferential excitation of particular points on a dispersion curve. These arrays are designed according to the principles of wave interference and the inuence of the wavelength excitation spectrum. This paper develops the relationships between the peak wavelength in the excitation spectra and the element spacing of linear comb and annular arrays. The excitation spectra are developed by applying Fourier and Hankel transforms to the spatial loading distribution functions of the comb and annular arrays, respectively. Although the peak wavelength of excitation of a comb array is typically assumed to be equal to the element spacing, it is shown that this can be an inaccurate assumption for annular arrays. The ratio of element spacing to the peak wavelength in the excitation spectrum is termed the spacing compensation factor, and is dependent on the number of array elements and the inner radius. It is determined that the compensation factor is negligible for comb arrays but is crucial for annular arrays in order to achieve optimal mode selection. Finite element analyses and experimental data are used to verify the calculations and demonstrate the signicance of the compensation factor. C 2013 Acoustical Society of America. [http://dx.doi.org/10.1121/1.4770238] V PACS number(s): 43.38.Hz, 43.35.Yb, 43.35.Zc [JAT] Pages: 127135

I. INTRODUCTION

Ultrasonic guided wave inspection techniques are powerful tools in the elds of non- destructive evaluation (NDE) and structural health monitoring (SHM). Guided wave techniques have several advantages over conventional bulk wave ultrasonic NDE/SHM techniques, including more rapid scanning of large structures,1,2 the ability to inspect hidden regions,2 and improved control over sensitivity parameters.2,3 However, in addition to the advantages of guided waves comes an inherent complexity beyond that of bulk wave mechanics. Fortunately, this complexity can be used to ones advantage if the guided wave mechanics are thoroughly understood and appropriately utilized. While there exist only three different wave modes (one longitudinal and two transverse) in bulk wave ultrasonics, there are an innite number of possible modes for guided waves. This means that there is theoretically an innite number of different mode-frequency combinations, and thus an innite number of potential wave structures that could be utilized to interrogate a structure.4 In reality, the frequency range of interest is limited by our ability to design and fabricate transducers, the size of the defects to be detected, the attenuation properties of the material, and the mode excitability for particular guided wave modes. Even with these limitations, there exists a plethora of options for designing a guided wave inspection system. It is up to the engineer to choose the optimal mode-frequency combinations for the
a)

Author to whom correspondence should be addressed. Electronic mail: [email protected]

task at hand and the consequences of mode-frequency selection can often determine the success (or failure) of the system.2,5 There are a number of factors that go into guided wave mode-frequency selection, including penetration power,6 mode excitability,1 defect sensitivity based on wave structure analysis,5 etc. Once desirable regions on the dispersion curve are identied, the question remains as to how these selected regions can be optimally excited. A variety of guided wave mode control techniques can be found in the literature. The most prevalent of these are mode-frequency tuning, angle beam excitation, and transducer array excitation. Mode-frequency tuning relies on the variation of wave mode excitability as a function of frequency. By varying the frequency of activation with a particular transducer such as a bonded piezoelectric wafer active sensor which applies a shear load on the surface, the relative excitation of various guided wave modes can be tuned.1 This technique relies on the wave mode excitability, which is a function of how well the forced displacements applied to the structure correlate with the individual guided wave mode velocity elds at the surface on which the transducer load is applied.1,7 Since the velocity wave structure is a function of frequency for each mode, so is the relative mode excitation. Angle beam activation relies on Snells Law to preferentially excite various guided wave modes at a particular phase velocity.4 The phase velocity excitation is a function of the phase velocity in the angle beam wedge material, as well as the angle from normal incidence at which the incident wave is introduced to the structure.8 Finally, transducer arrays can be utilized to inuence the excited guided wave eld in the wavenumber domain based on the spatial periodicity of the
C 2013 Acoustical Society of America V

J. Acoust. Soc. Am. 133 (1), January 2013

0001-4966/2013/133(1)/127/9/$30.00

127

Downloaded 07 Jan 2013 to 14.139.180.10. Redistribution subject to ASA license or copyright; see http://asadl.org/terms

array elements.4,9,10 These elements can be distributed in onedimensional (1D) or two-dimensional (2D) congurations in a variety of geometries. The most commonly used transducer arrays for guided wave mode control are likely comb transducers, which are comprised of a linear array of parallel elements. This allows for mode control in directions perpendicular to the length of the elements. On the other hand, an annular array is necessary for omni-directional mode control, due to the axisymmetric design of the transducer. This paper will focus on annular arrays and how their spacing should be optimally determined in relation to the number of elements and the inner radius of such arrays; the undesirable effects that can result from improper design will also be discussed. The earliest mention of comb transducers is most likely that by Viktorov,4 in which a periodic loading of a waveguide is utilized to generate Rayleigh and Lamb waves with a wavelength equal to the spacing of the elements. Following this initial mention in the literature, comb transducers received little study until the 1990s; at this point Demol et al.9 utilized a multi- element array of thin piezoelectric actuators to generate Lamb waves with wavelengths equal to the spacing of the elements. This was followed by the work of Rose et al.,8,11 who performed parametric studies on several design parameters of comb transducers by studying the signals generated by various array designs. Li and Rose12 later analyzed the excitation of individual comb elements by utilizing a normal mode expansion over the area of transducer loading, which led to the conclusion that maximum mode excitation could occur at wavelengths that were integer multiples of the element spacing. Rose and Quarry13 later utilized this normal mode expansion method in pipes to develop a phase velocity excitation spectrum for a comb transducer. Capineri et al.14 utilized a Fourier transform of the spatial pattern of comb elements to determine the acoustic response in the spatial frequency domain and determined that the optimum element spacing is that equal to the wavelength of the excited guided waves. Some of the earlier work into annular arrays can be considered from the work done by Wilcox et al.15 in studying the excited acoustic elds from angular segments of curved inter-digital transducers (IDTs); here the spacing equal to wavelength assumption is used. Similar curved segmented IDTs were studied by Gao et al.,16 who utilized the same wavelength equal to spacing comb assumption for guided wave mode generation from angular IDT segments of 90 . A more recent work on curved segmented IDTs has been carried out by Salas and Cesnik17 who applied a 2D spatial Fourier transform to the wave equation and loading distribution; mode control was not attempted by means of specifying the IDT spacing but by utilizing IDT segments that were split at a given radius, allowing the activation of curved segmented IDTs with either a predetermined smaller or larger radius to preferentially excite and receive particular wave modes. Glushkov et al.18 utilized a Fourier transform of the wave equation to achieve selective Lamb wave mode excitation by varying time delays and amplitudes among the elements of a bonded annular array applying radial shear loading at the element edges; however, no mention is made of optimizing the geometric parameters of the annular array.
128 J. Acoust. Soc. Am., Vol. 133, No. 1, January 2013

The scope of this work is to apply spatial Fourier transforms to the loading distribution functions of comb and annular array transducers in order to develop and analyze the excitation spectra of these arrays in the wavelength domain to optimize the relationship between transducer spacing and the wavelength of the guided wave mode to be generated. Upon derivation of these excitation spectra, an analysis of the compensation factor and its importance will be provided, followed by nite element calculations and experimental results to verify the compensation factor calculations.
II. EXCITATION SPECTRA ANALYSIS A. Transducer array spatial loading distribution functions

The excitation spectrum, or wavenumber-domain inuence, of a transducer array is a direct result of the spatial distribution of the array on the structure to which it is applied, which holds true for both generating13,14 and receiving19 guided waves. Thus, to develop relations for the excitation spectra of transducer arrays, the spatial loading distribution functions of the arrays must rst be developed. Two general array geometries will be considered: Linear comb arrays (Fig. 1) and annular arrays (Fig. 2). The linear comb arrays are examined in a plane strain formulation and can thus be assumed to extend to z 61. The geometric parameters of these arrays are the center-to-center element spacing, s, the element width, w, and the inner radius, r0 (for the case of an annular array), as shown in Figs. 1 and 2. Assuming a piston-type loading (although other loading patterns could be studied using a similar approach), the spatial distribution of the load applied by the comb and annular arrays can be dened in terms of rectangle functions, denoted by Pa(x) and Pa(r), respectively. These rectangle functions are dened in Eq. (1), in which a is any positive value  1; a x a Pa x (1a) 0; x < a or x > a0 ; ( 1; 0 r a Pa r (1b) 0; r > a: Thus the spatial loading distribution function for an N-element comb transducer with generalized element spacing and width parameters s and w can be described as fc x
N 1 X n0

Pw=2 x ns:

(2a)

FIG. 1. Plane strain geometry of a generalized N-element comb array transducer with inner element width w and center-to-center spacing s. Borigo et al.: Spacing compensation factor for annular array

Downloaded 07 Jan 2013 to 14.139.180.10. Redistribution subject to ASA license or copyright; see http://asadl.org/terms

To describe the excitation spectrum of the annular array, the 2D spatial Fourier transform of the annular array geometry function must be taken due to the 2D loading distribution. However, since the geometry function of the annular array is axisymmetric, the 2D spatial Fourier transform can be simplied as a 0th-order Hankel transform17,20
FIG. 2. Axisymmetric geometry of a generalized N-element annular array transducer with inner radius r0, element width w, and center-to-center spacing s.

Fa k H0 ffa rgk

1 0

fa rrJ0 kr dr:

(8)

Similarly, for an N-element annular array, the mathematical description of the spatial loading distribution function is f a r
N 1 X Pw r ns r0 ; n0

(2b)

which can be written alternatively as f a r

Here H0 ff g represents the 0th-order Hankel transform of the function f and J0(x) represents the 0th-order Bessel function of the rst kind. Applying the 0th-order Hankel transform [Eq. (8)] to the annular array spatial loading distribution function [Eq. (2c)], utilizing Bessel function integral recurrence relations, and integrating each term in the summation over the N array elements yields F a k
N 1 1X fr0 ns wJ1 r0 ns wk k n0

(2c)

in which H(r) is the Heaviside step function.

r0 nsJ1 r0 nskg;

(9)

In order to analyze the excitation spectra of the arrays, the spatial loading distribution functions must be transformed into the wavenumber, k, or wavelength, k, domains. Since the loading is distributed in a single dimension in the comb array case, the 1D spatial Fourier transform [Eq. (3)] can be utilized to transform the distribution function into the wavenumber domain 1 f xeikx dx: (3) F ff xgk
1

which is the excitation spectrum of the generalized N-element annular array, Fa (k), in the wavenumber domain.
C. Excitation spectra results

Therefore the excitation spectrum of a generalized N-element comb array, Fc (k), in the wavenumber domain is  1 X N 1 Fc k Pw=2 x ns eikx dx: (4)
1 n0

Although the k-domain functions are complex, we are interested only in the magnitude of the spectrum. Thus, ~ denotes the complex Eq. (10) can be invoked, in which F conjugate of F q ~k: jFkj Fk F (10) Using the relation k 2p=k the excitation spectra can be plotted as a function of wavelength. Examples of the excitation spectra for a particular comb and annular array in the wavelength domain are given in Fig. 3. Note the peak in the spectrum for the comb array near k 1 s and the split main peak that replaces it in the annular array spectrum. This split peak is due to the interference of the outward- and inwardpropagating wave spectra for the annular array, which can be

However, this can be simplied by rst taking the Fourier transform of a single element centered about the origin   1 kw ikx ; (5) Fc0 k Pw=2 xe dx w sinc 2 1 and using the modulation property of the Fourier transform, by which this element can be translated along the x-axis by a distance Dx according to FcDx k Fc0 keikDx : (6)

Thus, by assuming constant element spacing and width, any element n can be described by combining Eqs. (5) and (6), and Eq. (4) can be written more simply as   kw sinksN =2 iks=2N 1 e : F ffc xg w sinc 2 sinks=2 (7)
J. Acoust. Soc. Am., Vol. 133, No. 1, January 2013

FIG. 3. (Color online) Example 4-element comb and annular array excitation spectra with w 0.5 s and r0 0.5 s. Borigo et al.: Spacing compensation factor for annular array 129

Downloaded 07 Jan 2013 to 14.139.180.10. Redistribution subject to ASA license or copyright; see http://asadl.org/terms

described individually by replacing the Bessel function J1 in 1 2 Eq. (9) with the Hankel functions H1 and H1 , respectively.
III. SPACING COMPENSATION FACTORS A. Determination of the compensation factor

To design an annular or comb array with optimal excitation at the desired wavelength, it is important to understand the relationship between the wavelength of peak excitation and the array spacing, as well as how this ratio depends on N and r0. This ratio can be considered a spacing compensation factor, a, when designing an array transducer. Note that a compensation factor less than 1 corresponds to a transducer with spacing smaller than the desired wavelength. It has often been approximated that for a comb array in which w 0.5 s, the wavelength of peak excitation is equal to the spacing of the array elements, i.e., the compensation factor is unity. The calculations show that this approximation is accurate for comb arrays with a large number of elements but slightly less accurate with fewer than approximately four elements. The relationship between the wavelength of peak excitation and the array spacing for an annular array is slightly more complex since it depends on both the number of elements, N, and the inner radius, r0, of the array, as is shown in Fig. 4. For the comb array case, the main peak is within 3% of k 1 s for N > 3. Even with only two elements, the main peak is still within 10% of k 1 s. However, for the annular array case, the compensation factor can be as low as 0.7 or as high as 1.06, depending on the inner radius and the number of elements, as shown in Fig. 4. The strong variations with r0 in the trend of the compensation factor for an increasing number of elements is due to the role that the inner radius plays in determining the interference of the outwardpropagating waves and the inward- propagating waves, which pass through the center of the array. This is also the cause of the split primary peak in the annular array excitation spectrum (see Fig. 3). It is clear that for particular inner radii, such as near 0.2 s and 0.7 s, the compensation factor undergoes a dramatic shift. This corresponds to the points at which the two peaks that comprise the split primary peak in

the excitation spectrum switch in magnitude. Further analysis of the individual outward and inward wave spectra can be performed by replacing the Bessel function J1 in Eq. (9) with 1 2 the Hankel functions H1 and H1 , respectively; however, this is beyond the scope of this paper. Although ones instinct may initially be to minimize the compensation factor, or choose an inner radius with a compensation factor that varies minimally with changes in N, this is of no real advantage. In fact, designing an annular array with an inner radius such that the compensation factor is minimized may, in some cases, be detrimental due to the change this incurs in the properties of the excitation spectrum such as lower order peak amplitudes and primary peak bandwidth. As long as a plot such as that shown in Fig. 4 is referenced to determine the compensation factor for s, any annular array, regardless of inner radius or number of elements, can be designed to optimally excite a particular wavelength.
B. Significance of the compensation factor

In addition to analyzing the numerical value of the compensation factors, it is equally useful to evaluate the effect of not utilizing the compensation factors, which are related to the width of the primary peak in the excitation spectrum. This can be done by evaluating the value of the excitation spectrum at the point k s and comparing that to the peak value at the wavelength with the compensation factor k s/a. The ratio of these two values, converted to the decibel scale, is hereafter referred to as the excitation loss for a particular array. The excitation loss for annular arrays with various values of N and r0 is shown in Fig. 5. The calculations show that the excitation losses for the comb array are quite small, on the order of 0.1 dB in the worst cases. Thus, in nearly all cases, the compensation factor is not critical with comb array applications. This is due to the combined factors of rather small compensation factors and a rather broad primary peak in the excitation spectra for comb arrays. This single broad primary peak is not present

FIG. 4. (Color online) Correction factor as a function of N and r0 for annular arrays. 130 J. Acoust. Soc. Am., Vol. 133, No. 1, January 2013

FIG. 5. (Color online) Decrease in the excitation spectra for annular arrays with N elements and inner radius r0 at k s. Borigo et al.: Spacing compensation factor for annular array

Downloaded 07 Jan 2013 to 14.139.180.10. Redistribution subject to ASA license or copyright; see http://asadl.org/terms

in the annular array excitation spectrum due to the added complication of the interfering inward- and outwardpropagating waves, which cause a split in the primary peak that is dependent on the inner radius of the array. The conclusion that the compensation factor is not critical for comb arrays is in good agreement with the fact that comb transducers have been successfully utilized with the k s design assumption for decades.4,8,9 However, as shown in Fig. 5, this is not the case with annular arrays. Depending on the inner radius, the effect of the compensation factor can be quite extreme, even when the value of the compensation factor is rather small. However, the excitation loss for an annular array can be beyond 30 dB. Here it can be seen that the variation of the excitation loss does not vary strongly with the number of elements in the annular array but it does vary signicantly with the inner radius. At some r0 values, the use of the compensation factor will not be critical since the value of the excitation spectrum without it is nearly identical to that with it. Yet at other r0 values, the excitation loss is dramatic. Examples of some extreme cases at r0 0.37 s and r0 0.63 s, as well as the case of inner radius r0 0.5 s, are given in Fig. 6.

arrays, with and without taking the compensation factors into consideration, are given in Table I. Abaqus/Explicit commercial nite element software was utilized for the analysis with linear axisymmetric stress elements with general dimensions of 0.10 mm 0.10 mm. The annular array loading was modeled by the application of out-of-plane surface forces over the active regions of the annular array transducer design. In certain cases, such loading would not be highly realistic but in many cases this is a sufcient representation of out-of-plane loading. In either case, more complex loading distributions could be analyzed (by not assuming piston-type loading while deriving the excitation spectra) and modeled. However, the point of this model is to test the validity of the calculations, in which pistonloading was assumed. The out-of-plane displacement was extracted at a distance 500 mm from the center of the array. To account for the disproportionate surface area of loading for the various arrays (due to different geometries), the pressure load applied to the transducer in the model was scaled to yield equivalent excitation for both the compensated and uncompensated arrays.
B. Analysis of the finite element results

IV. FINITE ELEMENT ANALYSIS VERIFICATION A. Design of the finite element model

To verify these predictions, several nite element models were created to simulate different annular array designs taking into account the predicted compensation factors. Each array was tested for its ability to excite the A0 mode in a 1-mm aluminum plate at 300 kHz, for which the wavelength is 5.167 mm, with a 5-cycle Hanning-windowed tone burst. The rst annular array was a 4-element array with inner radius 0.5 s, while the second array also had 4 elements but had an inner radius of 0.7 s. For both arrays, the element width was equal to half the element spacing. The specications for these

The antisymmetric mode was extracted from the signals for the corrected and uncorrected array designs, and they were plotted and compared, as shown in Figs. 7 and 8. The results show that the arrays with the compensation factor produce less-distorted wave packets with higher amplitude. To estimate the total energy recorded in each wave packet, the sum of the squares of the extracted displacement amplitude over the time duration of the wave packet were taken for each time-domain signal in Figs. 7 and 8. This resulted in an estimated increase in energy of 25.7% for the 0.5 s-inner-radius array and an increase in energy of 12.8% for the 0.7 s-inner- radius array when the spacing compensation factors were taken into account during transducer design. Perhaps a more meaningful method to analyze the excited guided wave energy at the mode and frequency of interest (A0 mode at 300 kHz in this case) is to compare the powerspectral density (PSD) time-frequency plots of the signals.21 The time-frequency PSD plots, generated with a short-time Fourier transform method, are shown in Figs. 9 and 10 for the annular arrays with inner radii r0 0.5 s and r0 0.7 s, respectively. Analysis of the time-frequency plots makes it apparent that the signals produced by the arrays without the compensation factor are distorted in the frequency domain. For instance, the annular array with r0 0.5 s and no compensation factor
TABLE I. Dimensions for compensated and uncompensated annular arrays with inner radii 0.5 s and 0.7 s used for spacing compensation factor FE verication. Annular array with r0 0.5 s no compensation factor s 5.167 mm w 2.5835 mm r0 2.5835 mm Annular array with r0 0.7 s no compensation factor s 5.167 mm w 2.5835 mm r0 3.6169 mm

FIG. 6. (Color online) Decrease in the excitation spectra for annular arrays with N elements and inner radius r0 0.37 s, 0.63 s, and 0.50 s at k s. For some cases, the correction factor has a negligible effect on the excitation spectrum amplitude, while for others, the effect is severe. J. Acoust. Soc. Am., Vol. 133, No. 1, January 2013

a 0.9526 s 4.922 mm w 2.461 mm r0 2.461 mm

a 0.8939 s 4.6188 mm w 2.3094 mm r0 3.233 mm

Borigo et al.: Spacing compensation factor for annular array

131

Downloaded 07 Jan 2013 to 14.139.180.10. Redistribution subject to ASA license or copyright; see http://asadl.org/terms

FIG. 7. (Color Online) A0 mode generation from 0.5 s inner radius annular arrays with uncompensated spacing (dashed line) and compensated spacing (solid line).

FIG. 9. (Color Online) Time-frequency power spectral density plots of the compensated (top) and uncompensated (bottom) arrays, with inner radius r0 0.5 s.

is actually exciting the A0 mode at 280 kHz instead of the desired 300 kHz due to the misalignment of the wavelengthdomain excitation spectra, which occurs due to the fact that the compensation factor was not utilized. Similarly, for the annular array with r0 0.7 s and no compensation factor, the wave packet in the time-frequency domain is split into two distinct time-frequency peaks; this is again due to the fact that the excitation spectrum is not maximized at the desired wavelength. Thus, other wavelengths are preferentially excited within the frequency bandwidth of the 5-cycle windowed tone burst activation. Although the excited wave energy may still be in the A0 mode, it is generated at a frequency signicantly different from that which was originally desired. This leads to a different wave structure from that which was desired,6 as well as a different group velocity, as can be seen in the temporal

separation of the two peaks in the time-frequency spectrum in Fig. 10. To further compare the waves generated at the desired mode and frequency, a cross-section of the time-frequency PSD plots at 300 kHz can be analyzed; these are shown in Figs. 11 and 12. Here the signicance of the compensation factor can be clearly seen. Note that the wave packets corresponding to the uncompensated array design are signicantly lower in amplitude and are distorted in the time domain, while the wave packets from the properly-designed arrays are of higher amplitude and are not distorted. Thus, the arrays with the compensation factor taken into account generate more energy into guided waves with the desired mode and frequency. Since the plots in Figs. 11 and 12 are PSD vs time, integrating under the curves yields an energy approximation for each wave packet at 300 kHz (although it should be noted that the time-domain signals were extracted from the surface

FIG. 8. (Color Online) A0 mode generation from 0.7 s inner radius annular arrays with uncompensated spacing (dashed line) and compensated spacing (solid line). 132 J. Acoust. Soc. Am., Vol. 133, No. 1, January 2013

FIG. 10. (Color Online) Time-frequency power spectral density plots of the compensated (top) and uncompensated (bottom) arrays, with inner radius r0 0.7 s. Borigo et al.: Spacing compensation factor for annular array

Downloaded 07 Jan 2013 to 14.139.180.10. Redistribution subject to ASA license or copyright; see http://asadl.org/terms

TABLE II. Annular array transducer dimensions for experimental verication. Here fres indicates the thickness resonance of the annular array. Element number 1 2 3 4 Inner radius (mm) 2.625 7.875 13.125 18.375 Outer radius (mm) 5.250 10.500 15.750 21.000

N 4 r0 2.625 mm, w 2.625 mm, s 5.25 mm, fres 250 kHz.

FIG. 11. (Color Online) Cross-section of the time-frequency PSD plot of the compensated (solid line) and uncompensated (dashed line) arrays, r0 0.5 s, at 300 kHz.

and thus cannot be used to represent the full wave energy through the thickness of the waveguide). Through this analysis, it is calculated that the arrays which were designed with the compensation factor in mind produce 45.2% and 38.7% more energy in the desired mode and frequency than the uncorrected arrays with inner radii of 0.5 s and 0.7 s, respectively. This fact, along with the lack of temporal distortion associated with the correctly-designed arrays, leads to the conclusion that the compensation factor is critical to optimum annular array design.
V. EXPERIMENTAL VERIFICATION A. Experiment design

To further validate these results, an experiment was carried out to directly measure the effect of the compensation factor on annular array guided wave mode generation. In lieu of designing several transducers with and without the compensation factor taken into account, which would be

expensive and difcult, validation tests were done with a single existing transducer with 4 elements, s 5.25 mm, w 2.625 mm, and r0 2.625 mm. The geometric details of this transducer are also outlined in Table II. The experiments were conducted on a 3.175-mm aluminum plate with dimensions 1.12 m 1.22 m to avoid edge reections. The dispersion curve for the aluminum plate was examined and the wavelengths were plotted corresponding to the theoretical peak wavelength of excitation if the array had been designed with no compensation factor (kunc s), as well as the wavelength of peak excitation if the compensation factor a 0.9526 (kcom s/a) had been utilized, as shown in Fig. 13. The solid black line corresponds to the uncompensated peak wavelength of excitation, and the dashed black line corresponds to the compensated wavelength. These two wavelengths fall at different points on the A0 mode, specically at 480 and 458 kHz for the uncompensated and compensated wavelengths, respectively. The annular array was driven at a number of frequencies between those corresponding to the uncompensated and the compensated wavelengths for the A0 mode. A 500-kHz normal beam transducer was used to collect the data. The data were then analyzed in the time-frequency domain and the energy was estimated for each wave packet from the PSD functions. It is expected, according to the spacing compensation factor theory developed here, that this annular array would be optimally designed to excite the A0 mode at 458 kHz. According to the traditional k s design

FIG. 12. (Color Online) Cross-section of the time-frequency PSD plot of the compensated (solid line) and uncompensated (dashed line) arrays, r0 0.7 s, at 300 kHz. J. Acoust. Soc. Am., Vol. 133, No. 1, January 2013

FIG. 13. (Color Online) Dispersion curve for the 3.175-mm aluminum plate with the uncompensated (solid line) and compensated (dashed line) wavelength activation lines plotted. The circle shows the region of interest on the A0 mode. Borigo et al.: Spacing compensation factor for annular array 133

Downloaded 07 Jan 2013 to 14.139.180.10. Redistribution subject to ASA license or copyright; see http://asadl.org/terms

FIG. 14. (Color online) Time-domain power spectral density plots showing the A0 mode as a function of excitation frequency.

approximation, the optimum excitation of the A0 mode would occur at 480 kHz.
B. Analysis of experimental results

Several of the time-domain PSD plots for the signals taken at various frequencies are shown in Fig. 14. Note that away from the optimal frequency, the A0 wave packet (which arrives just after 200 ls) splits into two peaks of lower amplitude, while it is a single substantial peak when excited near the optimum frequency; this is similar to the results witnessed in the nite element analyses. The results were then analyzed by integrating the time-frequency PSD curve at the center frequency for each data point in Fig. 15 (i.e., each actuation frequency). However, since these results are taken over a range of frequencies, the wave mode excitability of the A0 mode must be taken into account. After the mode excitability as a function of frequency is taken into

account, the nal wave mode energy results are yielded, as shown in Fig. 15. The optimum frequency of excitation is near 450 kHz, while the predicted optimum frequency was 458 kHz. This difference could be attributed in part to transducer resonance of both the actuator (250 kHz) and receiver (500 kHz), or possibly due to inexact material properties for the aluminum plate, which would alter the dispersion curves slightly. Either way, it is clear that using the uncompensated wavelength, which would correspond to exciting the transducer at 480 kHz, would be far from optimal in terms of power and single wave packet generation for the A0 mode. Thus the results agree much better with the compensation factor theory, as it can be seen that a higher amplitude less-distorted signal is generated when using this theory as opposed to the k s design approximation, for which the optimum excitation frequency would have been 480 kHz.
VI. CONCLUDING REMARKS

FIG. 15. (Color online) Total energy in the A0 mode for each frequency of excitation, after the A0 mode excitability is taken into consideration. 134 J. Acoust. Soc. Am., Vol. 133, No. 1, January 2013

In this paper, the excitation spectra of comb and annular transducer arrays were developed by applying Fourier and Hankel transforms, respectively, to the spatial loading distribution functions of the generalized arrays. Such an approach has been used indirectly by other researchers by applying spatial Fourier transform methods to solve the wave equation but no explicit mention of applying such transforms to the spatial loading distribution of annular arrays in order to determine the wavenumber inuence on guided wave mode excitation could be found in the literature. These excitation spectra were analyzed in the wavelength domain and a relationship between element spacing and wavelength of peak excitation was developed as a function of the total number of elements in the array, as well as the array inner radius, in the case of an annular array. This relationship was used to develop a spacing
Borigo et al.: Spacing compensation factor for annular array

Downloaded 07 Jan 2013 to 14.139.180.10. Redistribution subject to ASA license or copyright; see http://asadl.org/terms

compensation factor to be used in the design of annular arrays for guided wave mode control in lieu of the common assumption applied to comb transducers that the spacing should be equal to the desired wavelength. By analyzing the broadness of the primary peaks in the excitation spectra, the signicance of the spacing compensation factor was predicted for comb and annular arrays with various numbers of elements as well as various inner radii. It was predicted that the compensation factor is negligible for comb array design but could be quite critical for the design of an annular array in order to optimally excite the desired guided wave mode and frequency. Finite element models as well as an experiment with an annular array transducer on an aluminum plate were conducted to analyze the validity and signicance of these predictions. In both cases, time-frequency power spectral density plots were utilized to compare the excited wave mode energy in the time-frequency domain. In the nite element analysis, this yielded a comparison of the waves excited by the annular arrays designed with and without the compensation factor, and it was determined that the utilization of the compensation factor led to an increase in excited wave mode energy at the desired frequency of up to 45% in the two cases considered. The time-frequency PSD analysis of the experimental results showed that the excited A0 wave mode energy at the frequency of excitation is signicantly greater near the frequency corresponding to the compensated wavelength than near the frequency corresponding to the uncompensated wavelength. Therefore, by applying the spacing compensation factor to determine the element spacing in relation to the number of elements and the inner radius of an annular array transducer, the excitation spectrum can be optimized to reach a maximum at the desired wavelength and guided wave signals with maximum amplitude and minimal distortion can be generated at the appropriate mode and frequency.
ACKNOWLEDGMENTS

The authors would like to acknowledge the support of Air Force Ofce of Scientic Research (AFOSR) for their support of this work through funding under STTR No. FA9550-10-C-0020.
1

V. Giurgiutiu, Structural Health Monitoring with Piezoelectric Wafer Active Sensors (Academic Press, Burlington, 2008), pp. 18, 321330.

J. L. Rose, Ultrasonic Waves in Solid Media (Cambridge University Press, New York, 1999), pp. 12, 101, 113126. 3 J. L. Rose, A. Pilarski, and J. J. Ditri, An approach to guided wave mode selection for inspection of laminated plate, J. Reinf. Plast. Compos. 12, 536544 (1993). 4 I. A. Viktorov, Rayleigh and Lamb Waves: Physical Theory and Applications (Plenum Press, New York, 1967), p. 83. 5 J. J. Ditri, J. L. Rose, and G. Chen, Mode selection criteria for defect detection optimization using Lamb waves, Rev. Prog. Quant. Nondestr. Eval. 11B, 21092115 (1991). 6 P. D. Wilcox, M. J. S. Lowe, and P. Cawley, Mode and transducer selection for long range Lamb wave inspection, J. Intell. Mater. Syst. Struct. 12, 553565 (2002). 7 B. A. Auld, Normal mode theory for acoustic waves and its application to the interdigital transducer, IEEE Trans. Electron Devices 18(10), 898 908 (1971). 8 J. L. Rose, S. P. Pelts, and M. J. Quarry, A comb transducer model for guided wave NDE, Ultrasonics 36, 163169 (1998). 9 T. Demol, P. Blanquet, and C. Delebarre, Lamb waves generation using a at multi-element array device, Proc.-IEEE Ultrason. Symp. 1, 791794 (1995). 10 D. N. Alleyne and P. Cawley, Optimization of Lamb wave inspection techniques, NDT & E Int. 25(1), 1122 (1992). 11 J. L. Rose, S. P. Pelts, and M. J. Quarry, A comb transducer for mode control in guided wave NDE, Proc.-IEEE Ultrason. Symp. 2, 10331036 (1997). 12 J. Li and J. L. Rose, Implementing guided wave mode control by use of a phased transducer array, IEEE Trans. Ultrason. Ferroelectr. Freq. Control 48(3), 761768 (2001). 13 M. J. Quarry and J. L. Rose, Phase velocity spectrum analysis for a time delay comb transducer for guided wave mode excitation, Rev. Prog. Quant. Nondestr. Eval. 20, 861868 (2001). 14 L. Capineri, A. Gallai, and L. Masotti, Design criteria and manufacturing technology of piezo-polymer transducer arrays for acoustic guided waves detection, Proc.-IEEE Ultrason. Symp. 1, 857860 (2002). 15 P. D. Wilcox, P. Cawley, and M. J. S. Lowe, Acoustic elds from PVDF interdigital transducers, IEE Proc.: Sci., Meas. Technol. 145(5), 250259 (1998). 16 H. Gao, M. J. Guers, J. L. Rose, G. Zhao, and C. Kwan, Ultrasonic guided wave annular array transducers for structural health monitoring, Rev. Prog. Quant. Nondestr. Eval. 25, 16801686 (2006). 17 K. I. Salas and C. E. S. Cesnik, Guided wave excitation by a CLoVER transducer for structural health monitoring: theory and experiments, Smart Mater. Struct. 18, 075005 (2009). 18 E. V. Glushkov, N. V. Glushkova, O. V. Kvasha, and R. Lammering, Selective Lamb mode activation by piezoelectric coaxial ring actuators, Smart Mater. Struct. 19, 035018 (2010). 19 M. Senesi and M. Ruzzene, A frequency selective acoustic transducer for directional Lamb wave sensing J. Acoust. Soc. Am. 130(4), 18991907 (2011). 20 C. Hu, Fourier-Hankel transform used in image raw data processing, in Applications of Articial Neural Networks in Image Processing VI, Proc. SPIE 4305, 4548 (2001). 21 A. V. Oppenheim and R. W. Schafer, Discrete Time Signal Processing (Prentice Hall, Upper Saddle River, NJ, 2010), pp. 811836.

J. Acoust. Soc. Am., Vol. 133, No. 1, January 2013

Borigo et al.: Spacing compensation factor for annular array

135

Downloaded 07 Jan 2013 to 14.139.180.10. Redistribution subject to ASA license or copyright; see http://asadl.org/terms