4th International Symposium on NDT in Aerospace 2012 - Th.1.B.2 CFRP Bonds Evaluation Using Piezoelectric Transducer Paweł MALINOWSKI*, Łukasz SKARBEK*, Tomasz WANDOWSKI*, Wiesław OSTACHOWICZ* * Institute of Fluid-Flow Machinery, Polish Academy of Sciences 14 Fiszera Street, 80-231, Gdansk, Poland Abstract. The research was focused on CFRP an important material in aerospace structures. The quality of bonding was assessed using laser vibrometry and electromechanical techniques used as NDT tools. Three different causes of possible weak bonds were selected for the investigation: 1. Weak bond due to release agent contamination, 2. Weak bond due to moisture contamination, 3. Poor curing of adhesive. Piezoelectric transducer was used for specimen excitation. Combining the piezoelectric excitation with laser sensing a tool was obtained to measure precisely the propagating elastic waves. Whereas using the piezoelectric transducer with impedance analyzer a NDT tool based on electromechanical impedance was obtained. The excited waves were measured in defined points by the vibrometer obtaining the wavefield. In order to characterize the bonding condition an indicator based on wave velocity was proposed. The guided elastic wave velocity depends material parameters of the media (Young modulus, density, thickness, Poisson ratio). It was assumed that comparison of the velocities can provide an information about the bond condition. The wave velocity was extracted using Hilbert transformation of the time signals. The absolute value of this transformation gives an envelope of the signal. From the envelope the time of arrival of the wave at the measurement points was extracted. Knowing the distance between two chosen points, the velocity was calculated. Using the impedance analyzer the impedance spectrum was obtained. The conductance plot against the frequency showed the deviation due to considered defects. The root mean square deviation (RMDS) of the spectrum showed indication of the release agent contamination level. All the investigated scenarios showed deviation from the reference case. 1. Introduction In numerous examples guided elastic waves-based methods have been proven to be effective in damage detection, localization and quantification. Surface or subsurface damage influence the wave propagation and can be found using this method [1]. Structural elements made out of either isotropic or anisotropic materials can be inspected. Damage in the form of crack, delamination, lack of bolt or rivet can be detected. Using the Electromechanical Impedance (EMI) method it was possible to diagnose the piezoelectric transducer embedded in composite material [2,3]. The EMI method was a basis for crack detection in thin plates in research reported in [4]. In work [5] damage localisation method was developer using EMI and Artificial Neural Networks. Due to increasing utilization of fibre reinforced composites in aeronautics, the research interest is focused on adhesive joints of these composites. The presented work is Licence: http://creativecommons.org/licenses/by-nd/3.0 1
focused on the assessment of adhesive bonds. This topic was undertaken in the frame of FP7 ENCOMB project (www.encomb.eu). Three scenarios of weak bonds were investigated. Piezoelectric transducer was used for excitation of the specimens. The paper is structured as follows. In the second section the experiment is described. In the third and forth sections the results of measurements are presented. In the fifth section methods sensitivity is investigated. In the last section conclusion from research are drawn. 2. Experiment The investigated bonded samples comprised of two CFRP (Carbon Fiber Reinforced Polymer) 6-ply samples bonded together. The single sample size was 50 mm 200 mm 1.56 mm in the case of laser measurements and 100 mm 100 mm 1.56 mm in the case of EMI measurements. These samples were bonded with adhesive film. Three different scenarios of weak bonds were selected for the investigation: 1. Release agent contamination, 2. Weak bond due to moisture contamination, 3. Poor curing of adhesive. In the first and second case the contamination was introduced before bonding of the samples. This means that the investigated samples comprised of bonded contaminated and untreated parts. Fourth sample with healthy bond was also measured for reference. Piezoelectric transducer was used as an active element for elastic wave excitation and EMI measurement. The used piezoelectric transducer was a disc (diameter: 10 mm, thickness: 0.2 mm) made out of CeramTec SONOX P5 material. The laser sensing was realized by Polytec vibrometer (Fig. 1). The impedance measurements were conducted using HIOKI impedance analyzer (Fig. 2). Fig. 1. Laser vibrometer (scanning head and main unit) 2
Fig. 2. Impedance analyzer 3. Wave Propagation Measurements Samples were excited with a piezoelectric disc bonded to the surface using a wax for mounting the accelerometers. The disc was placed just at the middle of the shorter edge of the sample (Fig. 3). It was decided to use possibly short wavelength in order to obtain sensitivity to small anomalies. On the other hand, the vibrometer sensitivity varies depending on the frequency range that is measured. Consequently the frequency of excitation was chosen equal to 200 khz. The higher frequency would enforce the reduction of the sensitivity of the device. Due to dispersive nature of guided elastic waves that propagate in thin-walled structures a 5-cycle tone burst signal was chosen for excitation (Fig. 4) [6]. This type of signal has a relatively narrow frequency band. Fig. 3. Sample with piezoelectric disc and overlaid grid of measurement points Fig. 4. Tone burst signal 200 khz and 5 cycles 3
The excited wave was measured in more than 11 000 points by the vibrometer (Fig. 3). The propagating wave is presented in Fig. 5. O the right hand side one can notice a slower guided wave propagating from the source. On the left hand side the faster wave already reached the edge and converted to the slower mode (Fig. 5a). a) b) c) Fig. 5. Registered wavefield 60 μs (a), 80 μs (b) and 128 μs (c) after wave excitation In order to characterize the bonding condition an indicator based on wave velocity was proposed. The guided elastic wave velocity depends on material parameters of the media (Young modulus, density, thickness, Poisson ratio). It was assumed that comparison of the velocities can provide an information about the bond condition. The wave velocity was extracted using Hilbert transformation of the time signals [7]: 1 xr ( ) H( x( t)) dr (1) t r where x is the time signal and t is the time. Having H calculated the analytical signal is defined by A() t x() t ih( x()) t. (2) The envelope of x signal is obtained by the following relation: 2 2 E() t A() t x () t H ( x()) t (3) 4
The envelope was used to extract the time of arrival of the wave at the measurement point. The wave velocity was calculated taking two of these points and knowing the distance between them. One point was kept unchanged while for the second ten points were chosen in order to calculate the mean value of the velocity. The obtained velocities were gathered in table 1. For all the samples with weak bonds the velocity is higher than for the reference sample. The highest velocity was obtained for weak bond sample with moisture contamination. Table 1. Calculated mean wave velocity for the considered weak bond scenarios c [m/s] Reference 1339.2 Poor curing 1414.2 Release agent 1455.4 Moisture 1808.0 4. EMI Measurements The piezoelectric transducer was mounted at the middle of each sample surface using Hysol adhesive. Measurements were conducted using HIOKI Impedance Analyzer IM3570. Investigations were executed in wide range of frequencies (4 Hz 5 MHz). Conductance (G) and susceptance (B) were analyzed that are, respectively, the real and imaginary part of admittance: Y G ib (4) where Y is the inverse of impedance Y Z 1 (5) Conductance for the four samples is presented in fig. 6. Rest of the graphs (Fig.7-9) presents the characteristic of electric parameters after processing. Firstly, the trend was removed and energy of whole spectrum was calculated. Every sample was normalized by this value, so energy of final signal was equal to 1 (G n normalized conductance, B n normalized susceptance). 5
Fig. 6. Conductance of investigated samples Firstly, interval from 200 khz to 300 khz was investigated (Fig. 7). For reference, moisture and release agent contaminated samples four peaks were obtained (local decreases of conductance) while for the poorly cured sample two peaks can be observed. Fig. 7. Logarithm of conductance in the bandwidth of 220-280 khz for all four samples. Fewer peaks for poorly cured sample can be noticed Interval from 2.9 MHz to 3.3 MHz contains only one peak for moisture contaminated sample and three peaks for the remaining samples (Fig. 8). Fig. 8. Logarithm of conductance in the bandwidth of 2.9-3.3 MHz. Only one peak for moisture contaminated sample 6
In the interval 3.7 MHz 4.1 MHz susceptance of release agent contaminated sample shows two additional peaks (Fig. 9). Fig. 9. Logarithm of susceptance in the bandwidth of 3.7-4.1 MHz for all four samples. Additional peaks for release agent contaminated sample There are few frequency intervals in which different response depending on the investigated scenario could be seen. Globally, most of peaks of release agent contaminated sample are shifted to higher frequencies, while peaks of moisture contaminated sample are shifted down. 5. Methods Sensitivity In the second step of research the methods sensitivity analysis was conducted. Four new samples were investigated that were prepared in the frame of ENCOMB project. The new bonded samples comprised of one 6-ply (thickness: 1.56 mm) and one 10-ply sample (thickness: 2.6 mm) bonded together. The total dimensions of the bonded samples were: 100 mm 100 mm 4.4 mm At this stage sensitivity analysis was focused on release agent contamination case. The thinner sample were dip-coated with release agent solution. It allowed to achieve four different levels of Si content: 2.1, 6.5, 8.2, 10.1 at%. For the measurements the piezoelectric disc was bonded at the middle of the samples surface (Fig. 10). In the case of laser measurements the retroreflective tape was used to enhance the vibrometer signal. Moreover the frequency of excitation was reduced to 50 khz because the wave amplitude was lower in this thicker samples. The wave velocity was measured horizontally (to the left and to the right from piezoelectric transducer) and vertically (above the piezoelectric transducer) fig. 10. The results were gathered in Tables 2-4. All the contamination cases showed deviation in relation to reference measurement. However no trend in velocity value (increasing or decreasing) was observed in relation to increasing amount of contamination in the bonded sample. 7
Fig. 10. Sample for sensitivity analysis with bonded piezoelectric disc and retroreflective tape Table 2. Calculated mean wave velocity for leftward propagation in samples with release agent contamination c [m/s] Reference 802.37 2.1 at% of Si 519.56 6.5 at% of Si 519.24 8.2 at% of Si 339.58 10.1 at% of Si 392.94 Table 3. Calculated mean wave velocity for rightward propagation in samples with release agent contamination c [m/s] Reference 828.2 2.1 at% of Si 1359.4 6.5 at% of Si 1194.8 8.2 at% of Si 2380.6 10.1 at% of Si 1466.7 8
Table 4. Calculated mean wave velocity for upward propagation in samples with release agent contamination c [m/s] Reference 217.22 2.1 at% of Si 348.14 6.5 at% of Si 286.09 8.2 at% of Si 415.44 10.1 at% of Si 254.82 In the case of EMI measurements the differences were noticed in the measured conductance in 3.0-4.5 MHz range (Fig. 11). Fig. 11. Conductance measured for samples with release agent contamination In order to quantitatively asses the results a Root Mean Square Deviation (RMSD) index was utilized: RMSD ( y x ) j j j % 2 x j j 2 100% (6) where x j is the conductance vector for reference sample and y j is the conductance vector for samples with weak bond. The calculated RMSD values were gathered in table 5. One can notice that with the increasing amount of silicone the RMSD index is increasing. For further comparison the relative conductance G/G 0% was calculated in the investigated frequency interval (Fig. 12). Again the RMSD index was utilized to assess the adhesive bond of the samples. It was found that the RMSD values follow the same trend as before, and the index is increasing with increasing amount of the Si contamination (Table 6). 9
Table 5. Calculated RMSD values for conductance measured for samples with release agent contamination RMSD for conductance 2.1 at% of Si 12.66% 6.5 at% of Si 15.99% 8.2 at% of Si 31.67% 10.1 at% of Si 48.48% Fig. 12. Relative conductance of samples with release agent contamination Table 6. Calculated RMSD values for relative conductance of samples with release agent contamination RMSD for relative conductance 2.1 at% of Si 12.01% 6.5 at% of Si 13.73 % 8.2 at% of Si 30.53% 10.1 at% of Si 40.63% 6. Conclusions This paper summarizes the initial characterization results of adhesively bonded CFRP using the laser scanning vibrometry and EMI techniques. Three cases of weak bonds were considered. In first approach chosen indicator was based on measured velocity of elastic guided waves that propagate in the investigated samples. All the investigated scenarios showed deviation in wave velocity. The biggest difference was obtained for moisture contaminated sample. In the investigation with different amount of release agent results allowed to clearly 10
distinguish the reference case from the contaminated cases. However there was no clear trend in the velocity values that can be related to increasing amount of the contaminant. Differences of velocity for the two horizontal paths might be due to differences in moisture content. Moreover one has to remember that the first measurements showed little influence of release agent contaminated bond on the wave velocity. In second approach the indicator was based on conductance and susceptance as a function of frequency. In selected frequency intervals each of the contaminated samples showed deviation from the remaining ones. The differences were manifested with fewer peaks for moisture contaminated and poorly cured samples. On the other hand, the release agent contaminated samples showed two more peak in suspectance plot. In the investigation with different amount of release agent results allowed to clearly distinguish the reference case from all the contaminated cases. Measurements showed deviation in the conductance plot due to contamination. Changes were visible in 3-4.5 MHz interval. Damage index was calculated that shows increasing trend with the increasing silicon content. The investigation will be continued. In the case of laser vibrometery other frequencies of excitation will be considered and the tests will be repeated with dried samples. In the case of EMI method the influence of adhesive layer will be investigated. Acknowledgements The research leading to these results has received funding from the European Union Seventh Framework Programme (FP7/2007-2013) under grant agreement n 266226. Pawel Malinowski would like to acknowledge the support from the Foundation for Polish Science. References [1] Z. Su, L. Ye, Y. Lu, Guided Lamb waves for identification of damage in composite structures: A review, Journal of Sound and Vibration 295 (2006) 753 780. [2] J. Pohl, G. Mook, SHM of CFRP-structures with impedance spectroscopy and Lamb waves, International Journal of Mechanical and Materials in Design 6, p. 53 62 [3] S. Park, G. Park, C.-B. Yun, C.R. Farrar, Sensor Self-diagnosis Using a Modified Impedance Model for Active Sensing-based Structural Health Monitoring, Structural Health Monitoring An International Journal 8 (1),. 71 82 [4] A.N. Zagrai, V. Giurgiutiu, Electro-mechanical impedance method for crack detection in thin plates, Journal of Intelligent Material Systems and Structures 2001, 12 (10): 709 718 [5] V. Giurgiutiu, C. Kropas-Hughes, Comparative study of neural-network damage detection from a statistical set of electro-mechanical impedance spectra, SPIE's 10th annual international symposium on smart structures and materials and 8th annual international symposium for on NDE for health monitoring and diagnostics, San Diego 2002, p. 1 12 [6] V. Giurgiutiu, Structural Health Monitoring with piezoelectric wafer active sensors, Elsevier, 2008. [7] S. Braun, Discover Signal Processing, An Interactive guide for engineers. Wiley and Sons, 2008. 11