CHARACTERIZATION OF THE LAMB WAVES PRODUCED BY LOCAL IMPACT FRACTURE IN BRITTLE THIN PLATES

Transcription

1 PROGRESS in ACOU5nC EMISSION IX CHARACTERIZATION OF THE LAMB WAVES PRODUCED BY LOCAL IMPACT FRACTURE IN BRITTLE THIN PLATES YOSHIHffiO MIZUTANI, MIKIO TAKEMOTO Faculty of Science and Engineering Aoyama Gakuin University , Chitosedai, Setagaya, Tokyo 157, JAPAN HIDED CHO Faculty of Engineering Tohoku University Aobaku, Sendai, JAPAN KANJIONO Department of Materials Science and Engineering UCLA. Los Angeles, CA U.S.A. ABSTRACT Acoustic emission (AE) signal analysiswas used to detect local fracture ofa PMMA plate subjected to a steel ball impact. Main AE signal types in a 3-mm-thick PMMA plate are found to be Lamb waves. Next we examined the Lamb waves produced by impact of flying steel ball at low velocities. Waveform and relative peak amplitudes of the Lamb waves agreed well with those predicted from laser- and PZT-based simulation. Major peak amplitudes of the Lamb waves were quantified as a function of ball velocities. The fracture-produced Lamb waves were extracted by the time-shift subtraction method from the waves detected at higher ball velocities. This study demonstrates the feasibility of sepa.rating impact- and fracture-induced AE signals. KEYWORDS Lamb Wave; Impact-induced AE; Fracture-induced AE; Time shift subtraction; Brittle thin plate S115

2 INTRODUCTION Brittle thin plates such as glass, polymethlymethacrylate (PMMA) and CFRPs suffer local fracture when they are impacted by a flying object. Cross-ply and quasi-isotropic CFRP plates used for transportation equipment sustain complicated internal damages such as fiber fracture, delamination and transverse cracks from the collision with particulate and ice. Although the visualization of such internal damages in CFRP members is possible by ultrasonic microscopes, our final goal is to monitor the onset of internal damages in impacted CFRP members by acoustic emission (AE) methods. The critical condition to cause damages depends on the physical and mechanical properties of flying objects and structural members. AE monitoring for such damages also needs to account for special conditions. For instance, high velocity and massive flying objects produce large shock waves, which cause serious damage to AE sensors. Special mounting methods to protect them from impact/shock are required. However, the damage of sensor and members may be small when tough members are impacted by low velocity light objects. In such cases, AE signals from the damage have small amplitudes, and are embedded in large amplitude impact waves. A special signal processing method is needed to extract the fracture-induced signals from the impact waves. The AEs produced by impact and/or internal fracture were detected as Lamb waves in thin plates. Extraction and classification of the fracture-produced Lamb waves from those by impact appears to be difficult because of their multi-mode dispersive nature. No such research has been reported so far. We are exploring the fundamental approach for characterizing the Lamb waves produced by complicated internal damages in cross-ply CFRP coupons subjected to static point loading. The Lamb waves produced by different fracture modes were classified by the modal analysis via wavelet contour maps[i]. In this paper, we characterized first the Lamb waves produced by impact with a flying steel ball at low velocities in PMMA plates, followed by those due to impact and local fracture by higher velocity impact. The Lamb waves produced by impact can be simulated by using an experimental transfer function including the impulse excitation of a Q-switched laser in combination with bell-shaped impact source functions. Finally, we extracted the fracture-induced Lamb waves from impact-generated large amplitude Lanlb waves by the time-shift subtraction method. The waveform and frequency components of the extracted waves are analyzed in terms of local fracture. EXPERIMENTAL METHODS AND LAMB-WAVE GENERATION Figure 1 shows an impact test method. A rectangular PMMA plate of 3 mm thickness with 90 null width and 180 mm long was axed by circular steel flanges with an inner diameter of 50 mm. A 7-11ull-diameter steel ball was accelerated by a high-pressure nitrogen gas gun and hit the center of the plate. The ball velocity and crack behavior were monitored by a high 8116

3 Fig. 1 AE monitoring from PMMA plate impacted by flying steel ball. speed camera with 27,000 flames/so Initially, we used PMMA plates without slit, and often observed radial cracks hitting AE sensors, damaging them. Crack behavior was unpredictable, making it difficult to monitor AE signals. Therefore, we made a shallow straight slit (6 mill length and 0.3 mm depth by cutting with a razor blade) on the opposite surface of an impact point such that the slit direction is normal to the AE sensor. The AE waves propagated along the length of the plate. As the length is twice the width, the wave reflection from plate edges was separated from the segment of interest. Sensor mounting needed special attention, since the sensors tend to fly out upon impact. A small AE sensor (PAC PICO) was affixed on the impact plane at 20 mm from the slit. It was pressed to the plate via a screw on a slender aluminum beam, whose ends were firmly attached to the flange. Sensor's outputs were attenuated by 20 db using a high impedance attenuator and digitized by an A/D converter at 100 ns sampling interval with 4096 sampling points at 10 bit. Digitized data were analyzed using a signal processing system we have developed previously [2],[3). 'Wave types in a PMMA plate were examined using the laser/ae system shown in Fig. 2. vve launched broad-band AE signals due to the break-down of a silicon grease film by a pointfocused Q-s"'itched YAG laserat the platecenter, and monitored the transmitted waves by two PICO sensors on both surfaces of the plate at 20 mm from the source. Two typical waveforms are shown in Fig. 2, which are triggered by the laser pulse detected by a photo-diode. The first arriving wave with the positive polarity was at 7.8 J.lS (corresponds to the P-wave velocity of 2.74mm/J.ls), and was followed by So-mode Lamb waves at 8.36 J.lS (with 2.39 nun/j.ls velocity) $117

4 and Ao-mode Lamb at around 13 ps (note the opposite phases of the Ao-waves). Arrival times agreed well with those calculated from the velocity dispersion data except the Ao-arrival time (faster than the calculated one of 14.5 J..lS). Signal characterization in the following sections was mainly performed for the So-mode Lamb waves because the trailing So waves were overlapping with the Ao-mode waves. Point focused VAG laser S I'I' Icon./ grease 20mm Pica Sensor 3mm PMM Q) i II, Ao o Time, ~s Fig. 2 Characterization of AE signals in 3 mm thick PMMA plate AE Signals by Ball Impact RESULTS AND DISCUSSION We first examined AE signals produced by a steel-ball impact at the ball velocities below 13 mis, during which no fracture occurred. Six waveforms detected at ball velocities from 7 to 13 mls are compared in Fig. 3(a). $118

5 6 Ball velocity 4 (a) A,-1/ 13m1s 12m.s > 2,P 11m/s >.,; ~ ::l 0 10mls :::l.e- 9m1s Co ::l 7rn1s -:::l (b) Fig. 3 Comparison of waveforms produced by ball impact (a) and simulated one for a bell-shaped impact source function of 16JLs long duration (b). The time a.xis was shifted so that P-wave arrivals were matched. We observed weak noise before the P-wave arrival, possibly due to the high-pressure gas blast since its amplitude increased with the gas pressure (or ball velocity). Waveforms by impact agree well from 10 to 20 JLS where P- and So-components arrived, but developed velocity-dependent differences at longer times when Ao-waves arrived. It is noted that the frequency components of the Lamb waves are unchanged over the velocity range studied. As shown in (b), the waveform simulated to bell-shaped impact source fuction of 16JLs long duration agrees well with the measured ones in (a). The squares of major peak amplitude of the Lamb waves (indicated by So, Ao-1 and Ao-2) shown in Fig. 4 increased almost linearly with the ball velocity. However, the slope for So peaks is much smaller than for Ao-waves. These relations can be utilized for predicting the wave amplitudes at higher ball velocities. Above 14 mls ball velocity, local fracture occurs. Thus, we can obtain fracture-induced AE signals by subtracting the predicted impact waveform from the actual measured wave. :> a) 40 -g 35 ~ 30 ~ 25 co 20 - o 15 co ~ So en ~::::...~~II="*::dE*~:J:;;:;==::::;:~( 1.2) Ball velocity I m/s (6.6)2 at 21 rnis Fig. 4 Relationship between the ball velocities and the squares of major peak amplitudes. S119

6 Signals Produced by Local Fracuture in Impacted Plate 'Ve extracted the fracture-produced wave components from the signals monitored at ball velocities of 14 and 21 mls by the time-shift subtraction method. Figure 5(a) represents the signal detected at ball velocity of 14 m/s. This contains two overlapping signals from impact and local fracture. The signal shown in (b) is the impact signal at ball velocity of 13 m/s. Ignoring the peak amplitude changes of impact-based Lamb waves, we matched the P-wave arrivals and subtracted the wave (b) from (a). The subtracted wave (c) is expected to represent the fracture-produced wave component. In this case, a 15-mm-Iong crack was produced by the impact..: ao \ rack length 15mm (a) 14m1s > 6 (b) 13m1s 4.: ao ~ > 6 (c)=(ahb) Time,lJs Fig. 5 Extraction of fracture-produced waveform at ball velocity 14m/s by time-shift subtraction method. (a) Signal detected by ball impact at 14m/s (b) Signal detected by ball impact at 13m/s (c) Extracted waveform due to local fracture 8120

7 Figure 6 shows another example of detected waves at. ball velocity of 21 m/s. Here, a 32-mm-Iong crack was formed upon impact. The impact waveform (a) to be subtracted was constructed by using the peak amplitudes extrapolated by the data offig. 4. As a comparison, the 13m/s-impact wave is shown by broken line. The wave in (c) is the result of subtraction representing the waves generated by local fracture. >...: :::J Q. -:::J 0 1 (a) 13m1s irnpactwave >.: :::J Co -:::J Detected wave Crack length: 32mm >...: :::J Co -:::J 0 Extracted wave due to local fracture Time, J.ls Fig. 6 Extraction of fracture-produced waveform at ball velocity 21m/s by time-shift subtraction method. The present results indicate that we can separate the wave components from two different sources. The waves generated by local fracture are delayed by 6-7 p,s following the impact. Simulation of Impact Waves In order to simulate a crack opening, a dipole source is needed. This is provided by a PZT element, excited by a ramp function and sandwiched between a larger plate and the subject plate, as shown in Fig

8 (a) Detected wave (b) 5 Cross section 25mm.JIi~mm ~m PMMA plate PZT Element Fig. 7 Overall transfer function for dipole source. The source displacement was determined by a laser interferometer. The transfer function inclusive of the PICO sensor is shown in Fig. 7(c), while the detected wave and source volumetric displacement (accounting for the contact size of 2 x 6 mm) are given in (a) and (b). With this transfer function and a ramp input of 12 J.LS rise time and crack volume of 2.0 x m 3, we obtain a simulated waveform illustrated in Fig. 8. The broken curve also shown in Fig. 8 is that of the extracted fracture wave given in Fig. 6(c). 8r-----~r=:::::=======" ATr=12J.ls 6 Simulated A V=3mmX32mmX2.11J \ =2.0Xl0 10 m 3 S Q. 0I--1'"Ili. os 0_ 4 6, \. i \ '\ ' ~ /\.J.. "... '-'v" \ Extracted Time,IJS Fig. 8 Comparison of simulated waveform due to crack generatiion and extracted one (Fig. 6(c)). 5122

9 The initial part matches very well and the peak positions at around 18, 21 and 28 p,s agree with experiment. However, the experimentally observed peak amplitudes at the third peaks are lower than the simulated waveform. In spite of some differences, the present simulation procedure provides an excellent result, giving the quantitative description of the underlying fracture process of the fracture wa,:,e extracted. Note that the crack length was 32 mm in this case. The average crack propagation speed is 1.3 mm/j1.s, assuming that the crack extends in two directions over the rise time of 12J1.s. This speed is about one-half ofthe P-wave velocity and is comparable to the Ao- Qr shear wave velocity. The observed velocity is two to three times faster than those found in PMMA by Schardin [4]. Our high-speed camera results also indicated the crack velocity of mm/j.ls in the above experiment. Thus, more detailed analysis of crack propagation and AE generation is clearly needed to resolve the observed discrepancy. Disucussion The procedures developed here with the aid of laser instrumentation and the use of a piezoelectric element allow one to characterize fracture-induced AE signals hidden in overlapping impact waves. This method will be useful for analyzing more complicated fracture phenomena expected in CFRPs subjected to impact. Such a study has been initiated. We have also examined AE signals due to quasi-static fracture of similar PMMA plates. In these cases, the fracture induced plate vibration and we were unable to separate fracture waves. Further work is needed to ascertain why fracture-induced AE cannot be singled out in the quasi-static cases. CONCLUSION Fracture in 3-mm-thick PMMA plate impacted by a flying steel ball was detected and analyzed using quantitative AE methods. The AE signals due to impact and local fracture in impacted plates were separated by the time-shift subtraction method following the characterization of impact-only signals. AE signals produced only by ball impact are dominated with low frequency Lanlb waves and can be simulated by using a bell-shaped source function of 16 J.lS duration. Fracture-inducedwaves due to local fracture in an impacted PMMA plate correspond to the ramp-type source function of 12 J1.S rise time, according to the simulation analysis with the aid of laser interferometer. The estimated crack velocity of 1.3 mm/j.ls was faster than those in other experiments and points up the need of detailed crack radiation modeling. The present method can be applied to the fracture analysis of various composite materials subjected to impact. ACKNOWLEDGMENTS A part ofthis research was supported by the fellowship from the Japan Society for Promotion of Science and the Japan Science Society to Y. Mizutani. K. Ono gratefully acknowledges the support of.japan Soc. For Promotion of Sciences for an extended stay at Aoyama Gakuin University during his sabbatical leave from UCLA. 8123

10 REFERENCES (1] Mizutani,Y., Nagashima,K., Takemoto,M., Ollo,IC, Proceeding of AECM, Texas, ASNT. to be published. [2] Suzuki,H., Kinjo,T., Takemoto,M., Ono,K., Proceeding of PROGRESS in ACOUSTIC EMISSION VII, The Japanese Society for NDI, Tokyo, 1996, pp [3J Suzuki,H., Kinjo,T., Hayashi,Y., Takemoto,M. and Ono,K., J.Acoustic Emission, Vol. 14 No.2, 1996, pp {41 Schardin,H., Velocity Effects in Fracture", in Fracture, eds. By B.L. Averbach et ai., MIT Press, Cambridge, MA, 1959, pp $124