Pitfalls in the experimental recording of ultrasonic (backscatter) polar scans for material characterization.

Ultrasonics xxx (2014) xxx–xxx

Contents lists available at ScienceDirect

Ultrasonics journal homepage: www.elsevier.com/locate/ultras

Pitfalls in the experimental recording of ultrasonic (backscatter) polar scans for material characterization Mathias Kersemans a,⇑, Wim Van Paepegem a, Koen Van Den Abeele b, Lincy Pyl c, Filip Zastavnik c, Hugo Sol c, Joris Degrieck a a b c

Department of Materials Science and Engineering, Ghent University, Technologiepark-Zwijnaarde 903, 9052 Zwijnaarde, Belgium Department of Physics, Catholic University of Leuven Campus Kortrijk – KULAK, Etienne-Sabbelaan 52, 8500 Kortrijk, Belgium Department Mechanics of Materials and Construction, Vrije Universiteit Brussel, Pleinlaan 2, 1500 Brussel, Belgium

a r t i c l e

i n f o

Article history: Received 5 June 2013 Received in revised form 8 April 2014 Accepted 8 April 2014 Available online xxxx Keywords: Ultrasonic (backscatter) polar scan Experimental pitfalls NDT Material characterization

a b s t r a c t The ultrasonic polar scan (UPS), either in transmission, reflection or backscatter mode, is a promising non-destructive testing technique for the characterization of composites, providing information about the mechanical anisotropy, the viscoelastic damping, the surface roughness, and more. At present, the technique is merely being used for qualitative purposes. The limited quantitative exploration and use of the technique can be primarily ascribed to limitations of current theoretical models as well as the difficulty to perform accurate, and more importantly, reproducible UPS experiments. Over the last years, we have identified several potential pitfalls in the experimental implementation of the technique which severely deteriorate the accurateness and reproducibility of a UPS. In this paper, we make an inventory of the most important difficulties, illustrate each of them by a real experiment and present a feasible mediation, either numerical or experimental in nature. Once the experimental set-up is fine-tuned to overcome these pitfalls, it is expected that the recording of high-level UPS experiments, in combination with numerical computations, will facilitate the technique to become a fully quantitative non-destructive characterization method. Ó 2014 Elsevier B.V. All rights reserved.

1. Introduction The increasing use of composite materials in primary components of wind turbine blades, automotive and aerospace industry, etc., incessantly demands the development of highly sensitive and reliable non-destructive testing (NDT) tools to monitor the structural performance of these materials, preferably using an in situ measurement technique. Within the wide class of NDT tools, the classical ultrasonic C-scan technique has already proven its usefulness in qualitatively visualizing defects, voids, delaminations, and material inhomogeneities [1–4]. The so-called UPS technique, which insonifies a sample surface with ultrasonic waves from all possible angles of oblique incidence w, can serve as a first-class successor for the more classical methods [5]. An UPS is created by simply gathering the reflected or transmitted ultrasonic signal and mapping the maximum amplitude for all angles of incidence (see Fig. 1). The oblique insonification and recording from all directions results in a large amount of information about the local

⇑ Corresponding author. Tel.: +32 (0)9 331 04 25.

wave propagation characteristics. Typically, ring-shaped contours can be observed which are linked, depending on the applied acoustic beam (pulsed or quasi harmonic), to the generation of critically refracted bulk waves or to the generation of guided plate waves [6– 9]. The stimulation of these waves and the resulting occurrence of the contour lines in a UPS are from a physics point of view directly connected to the mechanical properties of the sample material at the targeted spot [10,11]. Hence, each UPS immediately provides a ‘fingerprint’ of the local stiffness tensor Cij. At present, it is most common to perform the UPS experiments in what is called the double-through transmission (2T) regime [6,7,12]. In this configuration, the ultrasonic transducer, located in the transmission field, is replaced by an acoustic mirror which rotates together with the sending transducer. As a variation on the amplitude UPS, an equivalent fingerprint of the material properties can be obtained when recording the time-of-flight (TOF) values instead of the amplitude. Moreover, it has been suggested that the sensitivity to damage notably increases in case of TOF measurements [12]. On the other hand, TOF measurements significantly add complexity to the dataacquisition. Above and beyond, the interpretation of TOF–UPS data

E-mail address: [email protected] (M. Kersemans). http://dx.doi.org/10.1016/j.ultras.2014.04.013 0041-624X/Ó 2014 Elsevier B.V. All rights reserved.

Please cite this article in press as: M. Kersemans et al., Pitfalls in the experimental recording of ultrasonic (backscatter) polar scans for material characterization, Ultrasonics (2014), http://dx.doi.org/10.1016/j.ultras.2014.04.013

2

M. Kersemans et al. / Ultrasonics xxx (2014) xxx–xxx

Fig. 1. Principle of the UPS (a) and numerically computed UPS in transmission for a cross-ply [02, 902]S carbon/epoxy laminate insonified by an acoustic pulse with center frequency f = 2 MHz.

is not at all straightforward since, at present, no numerical technique is available to simulate a TOF recorded UPS. In addition to the reflected and transmitted fields, researchers have observed that a small part of the incident acoustic energy can be non-specularly backscattered (coherent as well as incoherent), even for a large incident angle h [13–18]. By recording this backscattered signal for every possible angle of incidence w (see Fig. 2), a complementary data set can be obtained, which is known as the ultrasonic backscatter polar scan (UBPS) [19,20]. For the backscatter results, we use the term time-of-flight-diffraction (TOFD) to denote the time value associated to the maximum backscatter amplitude. Over the last years, research on the UPS method has demonstrated its excellent ability to (i) determine the fiber direction of a unidirectional laminate, (ii) quantify the fiber volume fraction, (iii) detect internal damage, (iv) determine the specific mass, etc [6,7,12,21,22]. Likewise, recent research on the UBPS showed its potential to (i) detect and even locate a hidden surface breaking crack, (ii) quantify a microscopic multi-directional surface corrugation located at the front-or backside of the structure, (iii) determine the stacking sequence of commonly used fiber reinforced composite and (iv) semi-quantify hidden metallic rust in an early stage [19,20]. These capabilities, together with the ability to measure characteristic features at the backside of the structure in a single-sided setup, immediately opens up a variety of industrial applications for both the UPS and the UBPS. Nevertheless, despite the tremendous potential for several critical NDT & SHM applications, only limited in-depth experimen-

tal research on the subject of the U(B)PS technique is available at present. In particular, the ultimate application of the U(B)PS as a fully quantitative and online experimental tool is still far ahead. Based on our experience, we believe that the barrier in transferring the UPS concept to a real application is mainly due to the difficulty associated with the accurate recording as well as the reproducibility of an U(B)PS experiment. In this paper, we present and discuss several pitfalls for the recording of an U(B)PS. These experimental obstructions have been gradually identified and investigated during the last several years at our laboratory. Where applicable, several UPS recordings for an aluminum sample have been analyzed by a system identification procedure to invert its elastic characteristics [23,24]. The material choice is mainly motivated by its homogeneous and uniform nature, which makes that the mechanical properties are well known within small bounds. The different types of pitfalls were considered in the different experiments, in order to demonstrate their influence on the inversion procedure. The system identification procedure has been applied 10 times to a UPS experiment in order to determine a standard deviation. As such, we obtain a measure for the robustness and uniqueness of the inverted characteristics. The first row of Table 1 lists the elastic parameters for the aluminum according to literature. The second row represents the inverted material parameters for a UPS experiment in which each identified pitfall has been avoided as well as possible. It can be readily verified that the second row shows good agreement with literature, while the standard deviation is small, indicating the good convergence and robustness. The exact parameters

Fig. 2. Principle of the UBPS (a) and experimental recording for a cross-ply [45, 45]S glass/epoxy composite laminate insonified by an acoustic pulse with center frequency f = 4 MHz. The backscatter pattern is related to the stacking sequence of the glass laminate.


3


Table 1 Elastic material parameters (mean value ± standard deviation) extracted from UPS recordings of an aluminum sample, exposing the influence of different experimental pitfalls. C11 (GPa)

C12 (GPa)

C22 (GPa)

C44 (GPa)

C55 (GPa)

C66 (GPa)

LITERATURE NO PITFALLS

107.74 108.72 ± 0.13

55.50 56.92 ± 0.25

107.74 109.19 ± 0.58

26.12 26.20 ± 0.00

26.12 25.99 ± 0.00

26.12 26.02 ± 0.00

PITFALL PITFALL PITFALL PITFALL

104.26 ± 11.04 101.40 ± 11.22 105.13 ± 22.28 106.79 ± 6.14

58.64 ± 7.11 54.23 ± 9.41 56.08 ± 6.10 52.53 ± 6.44

104.23 ± 11.02 100.05 ± 13.26 100.72 ± 12.19 106.41 ± 6.68

23.43 ± 3.01 27.04 ± 1.41 27.79 ± 0.08 27.05 ± 0.28

23.42 ± 3.04 26.75 ± 1.17 26.94 ± 1.89 26.86 ± 0.18

22.80 ± 2.61 24.08 ± 2.85 24.54 ± 3.61 26.78 ± 0.01

1 2 2 BIS 5

considered for the experiments, together with a brief discussion of the essential parts of the in-house developed experimental facility, can be found in the next section. In the third section, several phenomena and sources are identified and discussed which severely decrease the accuracy and reproducibility of the U(B)PS. Each time, a feasible solution to overcome these difficulties is presented and discussed. Taking into account these tips and tricks, U(B)PS experiments with a high level of accuracy can be obtained, and only then, the proper application of the U(B)PS in industry can be considered. Certainly, combining robust and reproducible experiments with high-end numerical simulations, the characterization of the mechanical stiffness of anisotropic composite laminates, the inspection of layer adhesion, or the detection of incipient delaminations comes into view. 2. Experimental setup To perform UPS experiments, a high-performance mechanical multi-axes scanner has been developed in-house. The scanner is programmed to insonify a predefined material spot from every possible angle of incidence while maintaining a constant distance between the acoustic transducer and the sample. Three translational axes (X, Y, Z) and two rotational axes (R1, R2) allow to implement this movement. The scanner is aligned automatically in an accurate and precise way in order to enhance the quality of the experiments, and has full flexibility to perform multiple types of acquisitions, e.g. pulse echo (back-scatter), through-transmission, double through-transmission, reflection, etc. Different mechanical holders for different types of ultrasonic transducers can be used, each of them having specific physical dimensions. Each axis of motion is driven by a brushless DC-motor which is programmed using LABVIEW in order to fully automate the movement involved in a U(B)PS. A schematic of the path of movement is shown in Fig. 3 in polar coordinates, with each point defining a unique incidence angle w(h, u). The solid arrows represent the polar angle sections where measurements are performed. In this schematic only 7 sectors are illustrated for clarity, in reality the number is typically 360. The dotted arrows represent the compensation in both X-direction and Y-direction while changing the polar angle u. This compensation is essential in order to ensure the insonification of a single predefined material spot. In this configuration, the scanning is done in a continuous way in the vertical incidence plane which corresponds to the h-direction (the scan axis), while the u-direction (the index axis) proceeds in discrete steps. Upon changing the polar angle u, the direction of movement for the scan axis is inverted. Besides reasons of convenience for the mechanical implementation, this particular movement of the UPS has been chosen as it increases the performance as well as the control. Indeed, the first run for 90° 6 h 6 90° at u = 0° should yield symmetric results, since at a fixed polar angle u, position h = x° is acoustically equivalent to position h = x° (at least in the case of an undamaged material sample). Furthermore, note that the first run at u = 0° insonifies the material spot at exactly the same incidence angles as the last run at u = 180°, and consequently should give the same recording. The circles on Fig. 3 represent a standstill (in the order of

Fig. 3. Schematic of the typical movement involved in a U(B)PS experiment. The solid arrows represent the polar angle sections where measurements are performed. In this schematic only 7 sectors are illustrated, in reality this number is typically 360. The dotted arrows represent the compensation in X- and Y-direction in order to insonify a single spot. The circles represent a short waiting time to allow processing of the measurements and data storage.

2 s) in order to process and store the recordings of the last scan that was performed. Besides the proper recording of the different ultrasonic quantities (amplitude and/or TOF), information regarding the actual direction of insonification has to be accurately known and stored. In our set-up, high-precision position feedback is assured by the use of both linear and rotational relative position encoders, with an accuracy of Dr = 0.007 mm, respectively Da = 0.01°. Besides accurate position feedback, the encoders further provide a means for time-triggering the acoustic circuit. In most applications, it is sufficient to set the resolution of the U(B)PS experiment to Du = 0.5° for the index axis and to Dh = 0.05° for the scan axis. For obvious mechanical considerations, it is not quite possible to scan the complete hemisphere. Here, the scanning is limited to h 2 [70°, +70°] and u 2 [0°, 180°]. At each angle w(u, h), a timed sequence is carried out in which an acoustic pulse is launched by the emitting transducer to the investigated material spot and subsequently recorded in one or multiple modes, i.e. transmission, reflection or backscatter. With these parameters, a single U(B)PS experiment involves more than 1 million independent acoustic measurements at unique incidence angles w(u, h). Exploiting the symmetry of the investigated material, it is sufficient to scan only part of a spherical surface, e.g. h 2 [0°, +90°] and u 2 [0°, 45°] for orthotropic symmetry [25], thereby significantly reducing the experimental scanning time. However, this is not considered here since in reality composite materials rarely exhibit perfect symmetry, and the anisotropic nature of a sample often is a priori unknown. The probing ultrasonic beam is generated by means of a piezoelectrically driven shockwave transducer, generating a broadband


4


signal (http://www.ge-mcs.com/en/transducer-selection-guide/ type-selection/probe-search.html). Fig. 4 illustrates the acoustic pulse and the frequency content for such a typical transducer which operates at a center frequency f = 5 MHz. The signals are received with an identical transducer as used for the excitation. Water is used as an immersion liquid to enhance the transfer of acoustic energy into the solid sample. To limit the amount of recorded data during the experiment, only the maximum amplitude and/or its associated TOF(D) value within the received signal is recorded, employing a sampling rate of 400 MHz with 12-bit resolution. The combination of the automated mechanical scanner with the dedicated data-acquisition allows to keep the experimental U(B)PS time to less than 15 min, which is quite acceptable considering the large amount of valuable information which is obtained. A schematic representation of the entire experimental setup is shown in Fig. 5.

UPS and UBPS experiment. During intensive experimental research over the last few years, several of these parameters have been gradually identified as responsible for a serious decrease of the accuracy and reproducibility of a U(B)PS experiment. In order to obtain accurate experiments, which is crucial for a fully quantitative method, a proper setting of these parameters should be taken into consideration. The current section classifies several pitfalls for the recording of a U(B)PS, and each of them is illustrated by a real experiment. Wherever applicable, a feasible mechanical mediation or numerical solution to anticipate or correct these issues is presented and discussed. Table 2 shows an overview of the importance of the identified pitfalls for the different operation modes, i.e. transmission, backscatter and double-through transmission mode.

3. Pitfalls

During the last decades, researchers have consistently applied the double-through transmission method (also referred to as 2T) for the recording of a UPS [6,7,12,22] as well as for other applications [26,27]. In this implementation, the transducer in the

Many controlling parameters within the set-up, both of physical and mechanical nature, affect the quality of the results of both the

3.1. Pitfall 1: difficulties in the double-through transmission recording

Fig. 4. Broadband signal in time domain (a) and frequency content and (b) for a typical shock wave transducer operating at a center frequency f = 5 MHz.

Fig. 5. Schematic of the ultrasonic setup.


M. Kersemans et al. / Ultrasonics xxx (2014) xxx–xxx Table 2 Overview of the relevance of the different pitfalls for the various operation modes.

PITFALL PITFALL PITFALL PITFALL PITFALL PITFALL PITFALL

1: 2: 3: 4: 5: 6: 7:

twin rings coordinate system translational shift sloshing clearance air edge effect

Transmission

Backscatter

Double-through transmission

No Yes Yes Yes Yes No No

No Yes No Yes Yes Yes Yes

Yes Yes No Yes Yes No No

transmission field is replaced by an acoustic mirror. Doing so, the transmitted ultrasound gets back-reflected from the mirror and can be recorded by the same emitting transducer. One of the major advantages of this configuration is that only one transducer is required, limiting the cost of the implementation. Furthermore, as the transmitted sound beam is totally reflected by the mirror, back to the emitting transducer, the ultrasonic wave is forced to interact a second time with the solid, resulting in more defined and detailed polar contours. It is even possible to extend this idea to work in multi-through transmission recording (4T, 6T) [28] by exploiting the flat surface of typical immersion transducers. Of course, at some point the amplitude of the (2n)th transmission becomes too low to make a proper measurement. Another

5

advantage of the 2T method is that the acoustic mirror neutralizes the lateral beam shift of the ultrasonic beam, which is useful in case thick samples are investigated. Fig. 6a displays the UPS results for a [0]8 carbon/epoxy laminate, recorded in double-through transmission. At first sight, no peculiarities or anomalies are observed when compared to the numerically computed UPS displayed in Fig. 6c. As input for this simulation, we used the complex C-tensor of the carbon/epoxy laminate which was obtained through inversion of UPS data [23]. A more in-depth inspection of individual scanning lines extracted from the experimental data reveals the presence of closely spaced amplitude dips which are persistently observed at all polar angles u. To substantiate this, the scan section at u = 45° and u = 60° for the [0]8 carbon/epoxy laminate are shown in Fig. 6b, respectively Fig. 6d, for the experimental, respectively numerical, results. There is not a single indication of double amplitude dips to be observed in the numerically computed curves. The presence of local amplitude dips corresponds to a high local curvature value in the amplitude versus angle recording. Hence, in order to accentuate these amplitude dips, we adapted the data acquisition and recorded the difference between subsequent amplitude recordings in the scan axis direction. In doing so, a representation is obtained of the gradient of the transmission coefficient, and by extension to the second order difference, of the curvature of the transmission coefficient [29]. The ‘curvature UPS’, obtained in double-through transmission recording, is shown

Fig. 6. Experimentally obtained UPS results for a [0]8 carbon/epoxy laminate recorded in double-through transmission regime (a) and measured UPS sections at polar angles u = 45° and u = 60° (b). Numerically computed UPS for a unidirectional carbon/epoxy laminate (c) and numerically computed UPS sections at polar angles u = 45° and u = 60° (d). The black arrows around 40° accentuate the topological difference between the experiment and the numerical computation.


6


Fig. 7. Curvature UPS for a [0]8 carbon/epoxy laminate (same material as considered in Fig. 6a) recorded in (a) double-through transmission regime and (b) transmission regime. Note the appearance of twin rings in (a).

in Fig. 7a for the same [0]8 carbon/epoxy laminate as considered in Fig. 6a. The double amplitude dips observed in Fig. 6b are clearly visible in the curvature UPS in the form of two closely spaced polar rings, as if each ring has its own twin brother/sister. As polar scan measurements were merely used for qualitative purposes in the past, this feature has been overlooked (or neglected) for several years. Nevertheless, the importance of unambiguous transmission minima for the characterization of materials cannot be understated since the mechanical stiffness of the insonified spot is encrypted in the exact location of the polar contours as they correspond to the location of the critical bulk wave angles. Any dubious feature in the recording imposes a significant enlargement of the error on the stiffness values. This is demonstrated by inverting the UPS recording of the aluminum sample when twin rings (1° spacing) are present (see third row (PITFALL 1) in Table 1). Besides a deterioration of the accuracy of the inverted elastic parameters, also the standard deviation significantly increases, indicating that the inversion procedure get stuck in different suboptimal sets of data. The exact physical origin of the twin-effect can be identified by considering the ray path involved in the double-through transmission regime. Indeed, even the smallest misalignment between the ultrasonic transducer and the acoustic mirror will cause an erroneous recording since the back-and-forth ray paths differ from each other. In order to overcome this, a high precision mounting mechanism was devised in order to self-align the transducer with respect to the fixed acoustic mirror. In principle, this solution should be sufficient, however, in reality we observed that the twin effect still remains present now and then. Further investigation of

the persisting problem led to a dedicated investigation of the beam pattern of the employed acoustic transducers. Fig. 8 displays the measured radiation field of a typical immersion transducer along a plane containing its axis, together with a cross section of the amplitude distribution parallel to the transducer surface. The recording reveals that the radiation pattern is not completely symmetric with respect to the physical center line of the transducer. Similar skewing behavior has been obtained for various other transducers. Although the beam skewing is quite limited, it is bothersome for the UPS application. The inherent beam skewing of transducers implies that the twin-effect simply cannot be neutralized by means of a high-precision mounting mechanism only, which brings us to the conclusion that the double-through transmission regime irrefutably leads to the formation of twin rings, and as a consequence this modus operandus should be avoided for any UPS application directed toward a proper characterization of the laminate. As a comparison, the curvature UPS, recorded in single transmission, for the [0]8 carbon/epoxy laminate is shown in Fig. 7b. It can be readily verified that the twin effect is not anymore present. Further experimental results shown in this study are therefore recorded in single transmission. 3.2. Pitfall 2: misalignments in the material coordinate system and the scanner coordinate system The U(B)PS insonifies the material for all possible angles of incidence w(h, u), where the angle of normal incidence w(0, u) is defined as the reference. It is very tempting to incrementally adjust

Fig. 8. Amplitude of the radiation field of a typical shockwave transducer operating at a frequency f = 5 MHz (a) YZ-section and (b) XY-section at z = 130 mm (dotted line in (a)). The dashed lines guide the eye. Note that small skewing of the beam can be observed.



the angle of incidence w(h, u) in order to maximize the intensity of the surface echo, and consequently consider this as the correct zero position. However, the experiment is then obtained in the material reference system X0 Y0 Z0 . The mechanical movement of the scanner on the other hand always takes place in its own reference system XYZ, and thus a systematic error in the measurement of the position w(h, u) can be induced. For the scanning path of the robot considered in our set-up, the consequences of this error should become evident when comparing the recorded scan results at w(h, 0) (first scan) with the recordings at w(h, 180°) (last scan). As illustrated in Fig. 9 a clear horizontal shift between the first and the last run of a UPS experiment can be observed. The complete UPS experiment obtained in the incorrect X0 Y0 Z0 coordinate system is shown in Fig. 10a (for viewing purposes an exaggerated misalignment was used). Obviously the presence of the discrete jump in the polar contours will impede the accurate determination of the C-tensor through inverse modeling, as can be verified in the fourth row (PITFALL 2) of Table 1 for a 1° misalignment. To overcome this problem, one could opt to numerically compensate for this error. However, the more correct approach is of course to record the experiment ab initio in the correct scanner coordinate system XYZ. This can be simply achieved by exploiting the horizontality

7

of the free surface of the immersion liquid in combination with the bottom transducer which is used in pulse echo. Maximizing the echo at the water-air interface leads to the identification of the upward vertical direction w(180°, u) in the scanner coordinate system XYZ. Hence, rotation over h = 180° defines the direction of normal incidence w(0, u) which should be used as a reference. Taking into account this procedure, each UPS experiment is automatically obtained in the same scanner system XYZ. One can clearly observe that the discrete jump between the contour lines in the upper and lower halfspaces has been neutralized by the use of a correct reference (see Fig. 10b). Another peculiarity that can be attributed to the mismatch between the scanner system XYZ and the material system X0 Y0 Z0 is that the contour lines in the fingerprint are not always exactly centered (see Fig. 10a). Indeed, the direction of normal incidence in the XYZ system does not necessarily correspond to the direction of normal incidence in the X0 Y0 Z0 material system. The influence of a mismatch of 1° between the material and scanner coordinate system to the inversion results can be verified in the fifth row (PITFALL 2 BIS) of Table 1. To mediate this problem, the above discussed alignment of the ultrasonic transducer along the zero axis of the scanner reference system should be followed by a positioning of the sample under investigation in such a way that its surface echo gets maximized. This can be done by placing the sample in a holder which can be rotated by means of micro screws. When the surface echo is maximized, both the ultrasonic transducer and the sample under investigation are positioned in the correct scanner coordinate system XYZ. Proper positioning of the transducer and the sample within the scanner coordinate system results in the UPS recording shown in Fig. 11b. Alternatively, in cases where it is impossible to position the investigated sample in line with the scanner XYZ system, one can numerically search for the center of the recorded UPS, and use a translation in polar coordinates Dw(Dh, Du) to neutralize the coordinate system misalignment.

3.3. Pitfall 3: translational shift of the centerline of the ultrasonic probes

Fig. 9. The first and last run in a UPS experiment for an isotropic and homogeneous sample with incorrect reference coordinate system, showing a systematic horizontal shift of one curve relative to the other.

Any translational misalignment of the sender and receiver pair may cause unwanted intensity variations in the polar scan. The reason for this can be easily understood by considering the schematic representation shown in Fig. 12. Indeed, as a result of the oblique insonification, a characteristic beam shift is present after traversing the solid layer. Hence, for a given transducer misalignment and depending on the polarity of the incident angle h, the transmitted signal will be recorded at a different position at the

Fig. 10. UPS for an isotropic aluminum sample. Measurements taken in an incorrect material coordinate system (a) and in the correct laboratory system (b).


8


Fig. 11. UPS for an isotropic aluminum sample with a slight misalignment between laboratory and material coordinate system (a) and with a proper alignment between laboratory and material coordinate system (b).

Fig. 12. Influence of translational misalignment of the transducers for two equivalent incidence angles w: (a) w(+h, u) and (b) w(h, u). The transmitted signal is captured at different locations of the receiver’s surface, resulting in a different amplitude recording.

surface of the receiving transducer. Since the sensitivity of the membrane of the transducer decays gradually toward the edges of the circular surface because of the housing, the recorded amplitude value will be affected. An example of the translational misalignment effect can be observed in Fig. 11, in which a drop in amplitude can be observed for the equivalent incidence angles w(+h, u) and w(h, u). Although this alignment error does not affect the exact position of the characteristic contours within the UPS plot (and consequently has no or negligible influence on the characterization of the mechanical elasticity), it does critically affect the exact amplitude levels which are linked to the damping properties of the material at the insonified material spot. Hence in view of extending the inversion analysis to the characterization of the viscoelastic damping tensor, it is crucial to record the transmission amplitude with as high precision as possible (see Fig. 13). As a consequence of the physically induced lateral beam shift in transmission, these results furthermore indicate that the UPS in transmission should be interpreted with caution whenever the thickness of the solid layer, and thus the lateral beam shift, becomes comparable to the physical dimensions of the receiving transducer. The samples under investigation in the present study

Fig. 13. UPS for an isotropic aluminum sample, with correctly aligned transducers: misalignment D = 0.



had a maximum thickness of 2 mm, which is sufficiently thin to avoid the above concern. In case thicker samples are investigated, a receiver with larger lateral dimensions has to be employed, or the measurement has to be performed in reflection. Regarding the latter option, however, to our knowledge such a measurement in reflection has not yet been performed because of additional experimental difficulties. 3.4. Pitfall 4: Sloshing of water and redundant vibrations caused by rotational movement To obtain an accurate U(B)PS recording, the sample has to be insonified at a sufficiently large number of incidence angles w(h, u). However, in order to keep the experimental time as low as possible, the measurements should be performed relatively fast. Our current laboratory set-up allows to cover around 1000 incidence angles w(h, u) each second. Since several parts are moving through the water, appropriate experimental settings for the control of the mechanical axes have to be taken into account to avoid sloshing of the water. Although, the sloshing is of no importance for the acoustic wave itself (Doppler effects does not play any role), the water movement can affect the mounted sample because it is never

9

completely rigid. When the sample starts waving with the water, it will obviously influence the UPS recording. An example of this effect on the UPS is shown in Fig. 14. The white arrow indicates an inaccurate recording as a consequence of slight water sloshing. Among the control parameters for the mechanical axes that cause water sloshing, the far most important ones concern (i) the acceleration profile, (ii) the deceleration profile and (iii) the maximum rotation speed. The settings should be fine-tuned such that the sloshing is kept within reasonable limits. This might be different depending on the components used to create the scanner. Moreover, apart from sloshing, the use of excessively high scanning speeds could cause overlap of two subsequent ultrasonic signals within one time gate. In addition, redundant vibrations, inherent to any mechanical construction with moving parts, may be a source of degradation for the experimental recording as well. Since the ultrasonic circuit is triggered by the installed position encoders, peripheral vibrations could be interpreted as legitimate encoder pulses, causing a walk-off between the recorded and the actual position of the ultrasonic transducer. To cope with this, an electronic axis jitter suppression has been applied in order to separate the legitimate encoder pulses from the axis vibrations. 3.5. Pitfall 5: gear clearance

Fig. 14. UPS for a [0]8 carbon/epoxy laminate, showing the effect of sloshing of the water as indicated by the white arrows.

The use of gears cannot be avoided in order to allow for the complicated movement involved in the scanning process. Even more, gears are indispensible to increase the scanning resolution. Unfortunately, common gears always exhibit a non-negligeable backlash, which has obvious consequences for the position w(h, u) during the reversal of the movement. As shown in Fig. 15a, a high clearance manifests itself in the form of crenulated results in h-direction, making the experimental results less accurate for quantifying purposes. This is shown in the last row of Table 1, considering a clearance of 0.7° for the UPS experiment of the aluminum sample. To overcome this, one could opt for the installation of high-precision gears. However, such gears are very costly, and furthermore, only a few types can be immersed in water. To save costs, a customized gear was produced in our lab. Basically, we transformed two commercial gear boxes (type bonifoglio) in to a single gear box with upgraded properties. In order to decrease its clearance as well as extend its applicability in submersion experiments, several additional high-precision parts have been milled and installed. In this way, it can be immersed in water without further additional problems, while the clearance could be reduced to around 0.1° which is significantly lower compared to the 0.65°

Fig. 15. UPS for a [0]8 carbon/epoxy laminate: Results using a commercial gear with a large clearance (a) and in-house developed gear with a smaller clearance. The inset shows a magnified portion of the UPS experiment.


10


clearance of a typical commercial gear. The insets of Fig. 15a and b clearly illustrate the improvement of the quality of the UPS when recorded with the currently installed in-house developed high-precision gear as opposed to a standard commercial gear. 3.6. Pitfall 6: distortion by air bubbles Both the surface of the sample and the acoustic transducer are susceptible to the accumulation of air bubbles due to degassing of the immersion liquid. Because of the large difference in acoustic impedance at the liquid-air interface, the air bubbles work as

scatterers for the incident acoustic wave. The effect of air bubbles is illustrated in Fig. 16, in which both the transmission amplitude UPS (left column) and the UBPS (right column) for an isotropic homogeneous material are presented for various levels of air bubble presence (no bubbles, few bubbles, lots of bubbles). The typical dimension of the air bubbles is less than 1 mm in diameter. In Fig. 16(a and b), the results are shown in absence of air bubbles. For the UPS recording, as expected, circular patterns are formed which represent the isotropic and homogeneous nature of the insonified sample. The UBPS on the other hand, shows an intense amplitude near normal incidence which obviously corresponds to

Fig. 16. UPS (left column) and UBPS (right column) for an isotropic aluminum sample: (a and b) without distortion effect by air bubbles, (c and d) with distortion by a few small air bubbles and (e and f) with distortion by many small air bubbles. The typical diameter of the air bubbles is less than 1 mm.



the specular reflection of the acoustic beam at the sample. The influence of the presence of a few small air bubbles for the different polar scan recordings is shown in Fig. 16(c and d). For the transmission UPS, no significant changes are observed compared to the result in Fig. 16a. However, for the UBPS, the air bubbles are a significant source of increased backscatter which manifests itself in the form of a more or less chaotic backscatter fingerprint. Finally, Fig. 16(e and f) reveals that the presence of many small air bubbles highly distorts the results of the UPS as well as the UBPS. In this case, we artificially added air bubbles to create the large amount of bubbles. The multitude of air bubbles at top and bottom surfaces of the sample prevents the acoustical energy to be efficiently transmitted through the aluminum sample, hence distorting the UPS measurement. For the UBPS recording, enhanced scattering further disturbs the backscatter data. The above results clearly indicate that the presence of air bubbles mainly influences the UBPS, whereas only an extremely large amount of air bubbles also affects the UPS recording. 3.7. Pitfall 7: distortion by edge effects of the sample To illustrate the distortion due to edge effects, we consider the recorded UBPS for an aluminum sample of limited dimensions as shown in Fig. 17. The particular material spot under investigation is marked by the X on the photograph of the sample, and is located a distance d of approximately 30 mm from the edges. Despite the isotropic and homogeneous nature of the scanned aluminum

11

sample, increased backscatter can be observed for certain directions which cannot be attributed to specular reflection. The formation of ‘parasite’ backscatter images can be readily understood by considering that the polar scan principle basically relies on the stimulation of guided waves in the solid [6–9]. Hence the physical edges of the sample cause the stimulated waves to be reflected, and in some cases to be mode-converted. If the guided wave impinges perpendicularly on the edge, the converted wave leaks its energy into the immersion fluid toward the acoustic transducer, making the edge a source of increased backscatter [30–32] In Fig. 18, the UBPS results for the amplitude and the corresponding TOFD analysis are shown as obtained on the small size aluminum plate. To reduce the scatter in the TOFD based result, only those results are shown which correspond to a certain threshold level of backscatter amplitude. The free edges of the rectangular aluminum sample near the insonified spot, can be identified in the backscatter map at u = {0°, 90°, 270°}. Basically, a free edge produces two ‘parasite’ backscatter spots. The inner ‘parasite’ spot corresponds to the inplate reflection and subsequent backscattered leaking of the longitudinally polarized bulk wave in the water, the outer backscatter spot to the in-plate reflection of the shear polarized bulk wave and subsequent backscatter leakage into the water. The fact that the propagation speed of the longitudinal component is faster than the shear polarized wave can be verified in the TOFD based backscatter map. Moreover, the TOFD measurements can be used to estimate the physical distances to the edges when taking into

Fig. 17. Photograph of an aluminum sample, its dimensions, and area under investigation (a) and recorded UBPS (b). Despite the isotropic and homogenous nature of the aluminum sample, increased backscatter is observed at certain incidence angles w(u, h).

Fig. 18. Recorded UBPS for an isotropic and homogeneous aluminum sample: (a) amplitude recording (arbitrary units) and (b) TOFD recording (units in ls).


12


account the mechanical properties of the insonified sample. For this case, the TOFD values at the backscatter spots at u = {0°, 90°, 270°} predict edges distances of d {28.0 mm, 29.7 mm, 26.4 mm}, which is fairly close to the insonification spot identified in Fig. 17a. In addition, a closer inspection of the backscatter map reveals another backscatter spike around the direction w(13°, 180°). In accordance with the TOFD map, this spike originates from a feature at a distance d 94.5 mm, which corresponds to the most remote edge. The reduced intensity of the backscatter spot at u = 180° can be explained by the fact that the acoustic wave is damped upon propagation (material damping as well as damping as a consequence of energy leakage to the surrounding fluid). Apart from the parasite backscatter spikes associated to the perpendicular free ends of the sample, backscatter spikes can also be observed as a result of a corner interaction (as indicated by white circles in Fig. 17). Elimination of the edge and corner effects can be obtained by respecting a sufficient distance between the targeted spot and the edges of the insonified structure, i.e. by using large enough samples. However, as this might not always be possible in practice, the most practical and probably the only workable solution involves the complementary analysis of the TOFD value in addition to the amplitude recording. In this way, the backscatter signals which correspond to an edge or corner reflection can be identified and eliminated. Again, the importance of this observation cannot be overrated since in reality every structure has limited dimensions and thus could potentially produce a ‘ghost’ backscatter image. For industrial applications, such as the quantification of corrosion, misinterpretation of the backscatter data could incorrectly lead to needless, and often very expensive, replacement of a structural component.

4. Conclusions Several pitfalls in the experimental recording of an U(B)PS have been identified, some of which have been persistently overlooked for several years. Each of these difficulties has been illustrated with an experiment. A solution, whether it is numerical, mechanical or a combination of both, has been proposed and discussed. One of the most notable conclusions that can be deduced from the here reported pitfalls concerns the inappropriateness of the doublethrough transmission recording for high-quality polar scan applications. The downside of using double-through transmission can be best seen in the ‘curvature UPS’, in which twin rings became apparent, causing ambiguity in the identification of critical bulk angles. In addition, the presented UPS pitfall study clearly reveals that several weaknesses of the current version of the polar scan technique are related to mechanical movements and alignments. This strongly suggests that the next generation of polar scans should be preferably based on a system without mechanical components, providing a fixed alignment, and involving mature phased matrix technology to allow flexibility in excitation and recording. In view of this, we are currently considering the construction of a hemisphere containing a large number of transducers that can be individually activated. With this, the full immersion of the structure can be replaced by a custom-made fluid-filled bag providing local coupling to the sample under consideration. In addition, and most relevant for industrial applications, the phased matrix technology easily permits measurements in reflection. In attendance of such technology to be developed, the tips and tricks gathered in this report can help to largely reduce the experimental difficulties inherent to the current polar scan implementation for obtaining accurate and reproducible experiments. This may lead to a revival of the U(B)PS as a quantitative, contactless NDT technique on laboratory scale. Furthermore, it is expected that the combination of high-level experiments with numerical

simulations will eventually facilitate the inverse characterization of the local mechanical stiffness (elasticity and attenuation) of anisotropic plates. Currently, our research team focuses on the implementation of a combined experimental/numerical inversion technique for this purpose. Acknowledgment The funding of the FWO - Vlaanderen (Fonds voor Wetenschappelijk Onderzoek in Vlaanderen) through grant G012010N is highly appreciated. References [1] S.R. Iyer, S.K. Sinha, A.J. Schokker, Ultrasonic C-scan imaging of post-tensioned concrete bridge structures for detection of corrosion and voids, Comput. Aided Civil Infrastruct. Eng. 20 (2005) 79–94. [2] M. Amenabar I., A. López-Arraiza, M. Lizaranzu, J. Aurrekoetxea, Comparison and analysis of non-destructive testing techniques suitable for delamination inspection in wind turbine blades, Compos. Part B: Eng. 42 (2011) 1298–1305. [3] J.C. Pandey, M. Raj, T.K. Roy, T. Venugopalan, A novel method to measure cleanliness in steel using ultrasonic c-scan image analysis, Metall. Mater. Trans. B – Process Metall. and Mater. Process. Sci. 39 (2008) 439–446. [4] F. Aymerich, S. Meili, Ultrasonic evaluation of matrix damage in impacted composite laminates, Compos. Part B: Eng. 31 (2000) 1–6. [5] W.H.M. Vandreumel, J.L. Speijer, Non-destructive composite laminate characterization by means of ultrasonic polar-scan, Mater. Eval. 39 (1981) 922–925. [6] J. Degrieck, Some possibilities of nondestructive characterisation of composite plates by means of ultrasonic polar scans, in: D. VanHemelrijck, A. Anastassopoulos (Eds.), Emerging Technologies in Nondestructive Testing (ETNDT), A.A. Balkema, Patras, Greece, 1996, pp. 225–235. [7] J. Degrieck, N.F. Declercq, O. Leroy, Ultrasonic polar scans as a possible means of non-destructive testing and characterisation of composite plates, Insight 45 (2003) 196–201. [8] N.F. Declercq, J. Degrieck, O. Leroy, Ultrasonic polar scans: numerical simulation on generally anisotropic media, Ultrasonics 45 (2006) 32–39. [9] N.F. Declercq, J. Degrieck, O. Leroy, Simulations of harmonic and pulsed ultrasonic polar scans, NDT and E Int. 39 (2006) 205–216. [10] J.L. Rose, Ultrasonic Waves in Solid Media, Cambridge University Press, 1999. [11] A.H. Nayfeh, Wave Propagation in Layered Anisotropic Media: With Applications to Composites, North-Holland, New York, 1995. [12] L. Satyanarayan, J.M. Vander Weide, N.F. Declercq, Ultrasonic polar scan imaging of damaged fiber reinforced composites, Mater. Eval. 68 (2010) 733– 739. [13] V.K. Kinra, A.S. Ganpatye, K. Maslov, Ultrasonic ply-by-ply detection of matrix cracks in laminated composites, J. Nondestr. Eval. 25 (2006) 39–51. [14] H.C. Kim, J.K. Lee, S.Y. Kim, S.D. Kwon, Influence of the microstructure on the ultrasonic backscattered energy from a liquid/solid interface at the Rayleigh angle, Jpn. J. Appl. Phys. Part 1 – Regular Pap. Short Notes Rev. Pap. 38 (1999) 260–267. [15] G. Bechtold, K.M. Gaffney, J. Botsis, K. Friedrich, Fibre orientation in an injection moulded specimen by ultrasonic backscattering, Compos. Part A – Appl. Sci. Manuf. 29 (1998) 743–748. [16] J. Schuster, K.V. Steiner, Ultrasonic backscattering using digitized full-waveform scanning technique, J. Compos. Tech. Res. 15 (1993) 143–148. [17] Y. Bar-cohen, R.L. Crane, Acoustic-backscattering imaging of subcritical flaws in composites, Mater. Eval. 40 (1982) 970–975. [18] Y. Bar-Cohen, R.L. Crane, The acoustic backscattering method as a means of characterizing composites and imaging flaws. Patent no. 4,457,174, in, July 3 1984. [19] M. Kersemans, W. Van Paepegem, K. Van Den Abeele, L. Pyl, F. Zastavnik, H. Sol, J. Degrieck, Ultrasonic Characterization of Subsurface 2D Corrugation. J. Nondestruct. Eval., 2014, doi 10.1007/s10921-014-0239-7 (in press). [20] M. Kersemans, W. Van Paepegem, K. Van Den Abeele, L. Pyl, F. Zastavnik, H. Sol, J. Degrieck, The pulsed ultrasonic backscatter polar scan and its applications for NDT and material characterization. Exp. Mech., 2014, doi: 10.1007/s11340013-9843-1 (in press). [21] M. Kersemans, I. De Baere, J. Degrieck, K. Van Den Abeele, L. Pyl, F. Zastavnik, H. Sol, W. Van Paepegem, Nondestructive damage assessment in fiber reinforced composites with the pulsed ultrasonic polar scan, Polym. Test. 34 (2014) 85– 96. [22] N.F. Declercq, J. Degrieck, O. Leroy, On the influence of fatigue on ultrasonic polar scans of fiber reinforced composites, Ultrasonics 42 (2004) 173–177. [23] M. Kersemans, A. Martens, N. Lammens, K. Van Den Abeele, J. Degrieck, F. Zastavnik, L. Pyl, H. Sol, W. Van Paepegem, Identification of the elastic properties of isotropic and orthotropic thin-plate materials with the pulsed ultrasonic polar scan. Exp. Mech., 2014, doi: 10.1007/s11340-014-9861-7 (in press). [24] M. Kersemans, N. Lammens, G. Luyckx, J. Degrieck, W. Van Paepegem, Quantitative measurement of the elastic properties of orthotropic



[25] [26]

[27]

[28]

composites by means of the ultrasonic polar scan method, JEC Composites 75 (2012) 48–52. B.A. Auld, Acoustic Fields and Waves in Solids, second ed., Krieger publishing company, Florida, 1990. S.I. Rokhlin, L. Wang, Ultrasonic waves in layered anisotropic media: characterization of multidirectional composites, Int. J. Solids Struct. 39 (2002) 5529–5545. L. Wang, S.I. Rokhlin, Ultrasonic wave interaction with multidirectional composites: modeling and experiment, J. Acoust. Soc. Am. 114 (2003) 2582– 2595. M. Kersemans, W. Van Paepegem, D. Van Nuffel, G. Luyckx, I. De Baere, F. Zastavnik, J. Gu, H. Sol, K. Van Den Abeele, J. Degrieck, Polar scan technique for material characterization and identification of new operating regimes, in: J.F.

[29]

[30] [31]

[32]

13

Silva Gomes, M.A.P. Vaz, (Eds.), 15th International Conference on Experimental Mechanics (ICEM15), INEGI, Porto, Portugal, 2012. M. Kersemans, W. Van Paepegem, N. Lammens, K. Van Den Abeele, F. Zastavnik, J. Gu, H. Sol, J. Degrieck, Gradient polar scan technique for material characterization, in: IEEE International Ultrasonics Symposium (IUS2012), Dresden, Germany, 2012. B. Morvan, N. Wilkie-Chancellier, H. Duflo, A. Tinel, J. Duclos, Lamb wave reflection at the free edge of a plate, J. Acoust. Soc. Am. 113 (2003) 1417–1425. M.E. Elkettani, P. Pareige, F. Luppe, J. Ripoche, Experimental study of the conversion of lamb waves at the end of an immersed plate, Acustica 82 (1996) 251–259. J.M. Galan, R. Abascal, Lamb mode conversion at edges. A hybrid boundary element-finite-element solution, J. Acoust. Soc. Am. 117 (2005) 1777–1784.


Simplified image processing for ultrasonic scans.

Contribution of double scattering in diffuse ultrasonic backscatter measurements.

Three-dimensional characterization of human ventricular myofiber architecture by ultrasonic backscatter.

Novel characterization method for fibrous materials using non-contact acoustics: material properties revealed by ultrasonic perturbations.

Material characterization of dual-energy computed tomographic data using polar coordinates.

Soft material adhesion characterization for in vivo locomotion of robotic capsule endoscopes: Experimental and modeling results.

Effect of intervening tissues on ultrasonic backscatter measurements of bone: An in vitro study.

Ultrasonic grey-scale recording with rapid processing film.

Recording Mouse Ultrasonic Vocalizations to Evaluate Social Communication.

Non-invasive evaluation of breast cancer response to chemotherapy using quantitative ultrasonic backscatter parameters.

On-line assessment of ventricular function by automatic boundary detection and ultrasonic backscatter imaging.

Experimental evaluation of myocardial fibrosis in a rapid atrial pacing model in New Zealand rabbits using quantitative analysis of ultrasonic backscatter.

Experimental studies of the ontogeny of ultrasonic vocalizations in bats.

Backscatter-difference Measurements of Cancellous Bone Using an Ultrasonic Imaging System.

Beamforming approaches for untethered, ultrasonic neural dust motes for cortical recording: a simulation study.

Artifact characterization and removal for in vivo neural recording.

Simplified Models for the Material Characterization of Cold-Formed RHS.

The accuracy of caridac function indices derived from ultrasonic time-position scans.

Multislice CT scans in patients on extracorporeal membrane oxygenation: emphasis on hemodynamic changes and imaging pitfalls.

Experimental comparison of empirical material decomposition methods for spectral CT.

Potential Pitfalls of the Mx1-Cre System: Implications for Experimental Modeling of Normal and Malignant Hematopoiesis.

The ultrasonic characteristics of high frequency modulated arc and its application in material processing.

Review of air-coupled ultrasonic materials characterization.

Evolution-based hierarchical feature fusion for ultrasonic liver tissue characterization.