Structural Health Monitoring of Fiber-reinforced ... - CiteSeerX

1

Structural Health Monitoring of Fiber-reinforced Composite Plates for Low-velocity Impact Damage using Ultrasonic Lamb Wave Tomography B. V. Soma Sekhar, Krishnan Balasubramaniam* and C.V. Krishnamurthy Center for Non-Destructive Evaluation and Department of Mechanical Engineering, Indian Institute of Technology Madras, Chennai – 600 036, India The feasibility of imaging low-velocity impact damage in thin multi-layered composite plates using Lamb wave tomography is explored. Low-velocity impact damages (6–20 J) are induced by the dropweight method on graphite fiber epoxy matrix composite laminates of cross-ply and quasi-isotropic lay-up and the damage assessment is carried out by conventional C-scans and ultrasonic Lamb wave tomography. Lamb wave based tomography based on modified cross-hole sensor configuration is used to image the damage as a function of impact energy for graphite composite plates. The reconstructed tomograms are able to capture the lateral extent of the damage reasonably even for impact energies as low as 11 J in a 2 mm graphite epoxy quasi-isotropic lay-up sample that can be classified as barely visible impact damage. Keywords

structural health monitoring low-velocity impact damages Lamb wave tomography delamination composite materials barely visible impact damages

1 Introduction Composites have been used to achieve substantial reductions in the structural weight of both military and commercial aircrafts. The susceptibility of these materials to damage resulting from low energy impacts (e.g., from dropped tools, runway stones or hailstones) is a major problem. These events lead often to barely visible impact damage (BVID), which are primary sources of delamination and fiber cracking in composite structures. In BVID cases, the maximum damage has been found to be inside the structure (often as deep as 60–80% of the

thickness) while the surface shows no visible indication of damage. These delaminations in composites can cause a large reduction in residual compression strength. Ascertaining the structural integrity during operation therefore becomes an important requirement. Structural health monitoring (SHM) has been conceived to provide an early warning of the damage extent through development of embedded sensors. Ultrasonic guided Lamb waves offer a convenient approach to evaluate composite laminates because they can propagate long distances and are sensitive to the in-plane stiffness of laminates. With an array of guided wave sensors

*Author to whom correspondence should be addressed. E-mail: [email protected] Figures 1–8 and 10 appear in color online: http://shm.sagepub.com

Copyright ß 200? SAGE Publications, Vol 0(0): 0001–11 [1475-9217 (200?0?) 0:0;1–11 10.1177/147592170?067739]

1

+

[Ver: 7.51g/W]

[4.7.2006–2:13pm]

[1–12]

[Page No. 1]

FIRST PROOFS

{SAGE_FPP}SHM/SHM 067739.3d

(SHM)

Paper: SHM 067739

Keyword

2

Structural Health Monitoring 0(0)

mounted permanently on/in the structure, the ‘health’ of the entire area of the structure could potentially be monitored rapidly and continuously in terms of readily interpretable images by exploiting the developments of computerized tomography [1]. Several researchers, such as Rokhlin [2] in the case of homogeneous materials, have analytically studied the interaction of Lamb waves with damage in composite plates. While the nondestructive evaluation of composite structures using Lamb wave has been extensively reported, there are several additional issues that need to be addressed while considering the use of Lamb waves for SHM. These include 1. The limited number of sensors that must be used (in order to reduce the additional weight, due to the sensors and the associated electronics). 2. The sensor positions have to be predetermined and fixed. 3. The footprint of the sensors has to be small; hence mode selection may not be possible, as the signals will have several simultaneous modes. 4. The waves interact with the defective region and cause new wave modes to be generated due to mode conversion. 5. The reflections from the edges and corners will further complicate the signal analysis. 6. The high attenuation and the anisotropy of the composites lead to lower travel distances and regions with preferential coverage. 7. The sensors have to survive the operational conditions.

The latter is simulated using a gas gun or some other ballistic launcher. Machines currently used for simulating the low-velocity impact response of composite material include the Charpy and Izod pendulums, the drop-weight and hydraulic test machines. Studies of impact test techniques and effect of fiber stacking can be found in [3–5]. In the present work, the low-velocity drop-weight technique is employed.

3 Drop-weight Impact Tests A drop-weight impact-testing machine was designed to introduce low-velocity impact damage in carbon-epoxy laminates. Here a weight was allowed to fall from a pre-determined height to strike the test specimen supported on a horizontal plate placed on a frame with a hollow cut of 150 150 mm2. Thus, the specimen was sandwiched between the base plate and the frame on all four edges with the top surface open to receive the impact from the dropweight assembly (Figure 1). The required value of impact energy was achieved by an appropriate combination of the mass and its drop height. For instance, a drop-weight crosshead of a total mass m ¼ 3.74 kg was raised to height, h ¼ 0.30 m to deliver an impact of 11.01 J. The testing system was set to automatic mode with a special locking arrangement to ensure that each sample received only one impact. Similar tests were performed on a set of samples with difference in height to achieve different impact energy levels.

Vertical stud

Drop weight

2 Impact Test Techniques for Composite Materials The impact problem can broadly be divided into two separate conditions: (a) low-velocity impact by a large mass (dropped tool) and (b) high-velocity impact by a small mass (runway debris, small arms fire, etc.). The former is generally simulated in the laboratory conditions using falling weight or a swinging pendulum.

+

[Ver: 7.51g/W]

[4.7.2006–2:13pm]

[1–12]

[Page No. 2]

FIRST PROOFS

Sample holder Locking

Figure 1 Drop-weight impact testing machine.


(SHM)

Paper: SHM 067739

Keyword

Soma Sekhar et al. Ultrasonic Lamb Wave Tomography

4 Ultrasonic C-scan Results An ultrasonic C-scan examination performed after the impact tests reveals the extent of damage in the composite. The C-scan images of defects are shown in Figure 2. In the case of 2 mm thick cross-ply (CPLT) (0/90)S, the damage level was greater than that of quasi-isotropic plates (QILT) (45/45/0/90)S for the same intensity of load. There was a very small (3 mm diameter) indentation visible on the top face and an opening on the backside for quasi-isotropic

3

plates when subjected to an impact load of 11 J. In the case of a cross-ply plate, with the same energy level, there was a slightly larger (6 mm) indentation. For low energy values the delaminations were found to occur well below the thickness in both plates. When subjected to an impact load of 7–8 J in both cases, there was no surface indication. Figure 3 shows the damage progression in each of the plates by moving through the thickness. The size of the damage varies from layer to layer in both the cases.

(a)

(b)

(c)

(d)

Figure 2 Ultrasonic C-scan of impact damages at the maximum damage ply: (a) 7 J on CPLT; (b) 11 J on CPLT; (c) 8 J on QILT; and (d) 11 J on QILT.

+

[Ver: 7.51g/W]

[4.7.2006–2:13pm]

[1–12]

[Page No. 3]

FIRST PROOFS


(SHM)

Paper: SHM 067739

Keyword

4


(a)

(b) Figure 3 Ultrasonic C-scan of impact damages obtained by progressivel moving the gated time window through the entire RF wave form: (a) 11 J on CPLT and (b) 11 J on QILT.

The orientation of delamination was predominantly oriented along the fiber direction of the particular layer. For a cross-ply plate the damage spread was along the direction of fibers at 0 and 90 as observed in Figure 3(a). A detailed study varying the properties of the fiber, matrix, and interphases was conducted by earlier researchers [5,6].

5 Lamb Waves in Anisotropic Plates Lamb waves are elastic modes propagating in solid plates with free boundaries [7]. Longitudinal and shear waves combine in a specific manner determined by the boundary conditions and stiffness constants of the material resulting in the propagation of Lamb waves. For a plate of given thickness, a finite number of Lamb wave modes can propagate depending on the frequency bandwidth of the excitation [7–9]. It is primarily important to know the number of modes that can be generated in the structure. With general pulsed transducers, however, many modes are excited and often superpose to produce a wave packet in the time domain. When a mode is dispersive, an initial narrow time-domain wave packet broadens as a function of the propagation path

+

[Ver: 7.51g/W]

[4.7.2006–2:13pm]

[1–12]

[Page No. 4]

FIRST PROOFS

length, making mode identification a challenging task. Special transducers are required to generate and sense particular modes. For an isotropic plate, the solution of the so-called Rayleigh– Lamb equation gives the phase velocity (Vph), and dispersion characteristics of the Lamb waves as a function of frequency–thickness product (called dispersion curves). In homogeneous isotropic plates, the Lamb waves are separated as symmetric and anti-symmetric depending on the variation of the particle displacement component in the propagation direction along the thickness of the plate [7–9]. Composite plates are normally made of fibers in a resin matrix. These fibers have specific orientations, which make composite plates anisotropic in their elastic properties [11–15]. For multi-layered anisotropic plates, the solution should satisfy the Christofell equation for each layer, the continuity condition at the interfaces, and the stress-free boundary conditions at the plate surfaces to give the direction-dependent phase velocity. A typical Lamb wave dispersion curve shown in Figure 4 for the cross-ply plate was plotted using DISPERSE [9]. These dispersion curves show that for the frequency–thickness product value of up to 1 MHz-mm, there exist two modes,


(SHM)

Paper: SHM 067739

Keyword


5

So

Excitation Region

Ao

Figure 4 Normalized phase velocity of Lamb wave modes in cross-ply (0/90)S, 2 mm thick plate.

the S0-mode and A0-mode. Owing to anisotropy, the ultrasonic energy tends to travel faster along the fiber direction, thus creating a noncircular spatial energy distribution function [11–14]. Consequently for a cross-ply plate, it is observed that Lamb waves travel faster along 0- and 90 - directions [1]. In addition, for proper imaging, these effects must be addressed in the reconstruction algorithm.

6 SHM Using Lamb Wave Tomography Structural health monitoring of composite structures can be achieved by using several sensors attached or embedded in the structure. Owing to practical limitations, the number of sensors that can be attached or embedded has to be limited, particularly where the additional weight due to the SHM system is a constraint, for example, aerospace structure. This leads to small number of sensors monitoring a large area of the structure. Hence, tomographic reconstruction with limited Lamb wave data must be performed.

+

[Ver: 7.51g/W]

[4.7.2006–2:13pm]

[1–12]

[Page No. 5]

Reconstruction of tomograms has been demonstrated with many different kinds of sensor–receiver configurations, such as parallel, fan beam, and cross-hole (CH) for isotropic materials, such as aluminum and plexiglas [2]. Iterative algebraic reconstruction techniques (ART) based on the Kaczmarcz algorithm [15] is used in the present work for reconstructing the tomographic image. The present work employs narrow band longitudinal PZT crystals directly in contact with the structure. An approach to analyze Lamb wave signals was proposed by calculating the energy of the received lamb wave signal [1,2]. The energy of the received signal should be characteristic of the plate, and should change when defects are encountered. The image reconstruction method found most suitable for the rapid acquisition and processing of lamb wave data is based on algorithms originally developed for cross-borehole tomography [16–18]. This technique is routinely used in geophysical applications in which the transducer geometry consists of parallel arrays of transmitter–receiver locations. Since composite materials are highly anisotropic in nature, the conventional CH configuration was replaced by a modified cross-hole

FIRST PROOFS


(SHM)

Paper: SHM 067739

Keyword

6


(MCH) configuration with transducers placed on the four sides of the plate [1].

7 Simulation of the Effect of Different Sensor Configurations Two types of sensor configurations were considered; the first one was a standard CH configuration with 20 20 sensors. The second one was a recently proposed configuration with 5 sensors per side [1]. To visualize the coverage of the structure during SHM, ray diagrams connecting rays from each sensor to all other sensors were used. The ray picture for the CH and modified cross-hole (MCH) configurations are shown in Figure 5(a) and (b). It can be observed from the ray diagrams that the MCH configuration produces a more uniform coverage with fewer sensors per side when compared to the traditional CH configuration. To verify that tomographic reconstruction would reflect this feature, a simulation study was carried out for both the configurations for an isotropic medium. The medium was represented as a 20 20 cell grid structure. The Gordon-ART algorithm was employed for the reconstruction using the time of flight as the projection data. The details of the algorithm can be found in [1,18]. Figure 6(a) depicts the isotropic medium

with two defective regions. Figure 6(b) shows the tomogram reconstructed using the CH configuration. Artifacts arising from the use of ART algorithms, observed early on [16,18] can be seen. These artifacts adversely influence any automated flaw evaluation algorithm that may be used in an autonomous SHM. Figure 6(c) shows the reconstructed tomogram using the MCH configuration. It is quite clear from these images that the use of MCH configuration leads to significantly improved tomograms. It was further surmised that the uniform coverage provided by the MCH configuration may lead to improved tomograms even when the media had anisotropy. To verify this aspect, simulations were carried out with angle-dependent projection data representing the anisotropy. The defects were as shown in Figure 6(a). The angle dependence of the projected data was modeled from the time of flight determined from a non-circular slowness curve. Figure 7(a) shows the tomogram reconstructed using the CH configuration. The anisotropy in the projection data appears to deteriorate the quality of the tomogram so severely as not to indicate one of the defects. Figure 7(b) shows the reconstructed tomogram using the MCH configuration. The definite improvement in the tomogram quality with the use of MCH configuration is clearly noticeable. It can be observed from the results

(a)

(b)

Figure 5 Coverage based on rays for (a) cross-hole configuration and (b) modified cross-hole configuration.

+

[Ver: 7.51g/W]

[4.7.2006–2:13pm]

[1–12]

[Page No. 6]

FIRST PROOFS


(SHM)

Paper: SHM 067739

Keyword


7

(a)

(b)

(c)

Figure 6 Tomographic reconstruction of an isotropic medium with simulated projection data: (a) schematic of the defective medium; (b) cross-hole configuration with 20 20 sensors; and (c) modified cross-hole configuration with five sensors per side.

0

(a)

(b)

Figure 7 Tomographic reconstruction of an anisotropic medium with simulated projection data: (a) cross-hole configuration with 20 20 sensors and (b) modified cross-hole configuration with five sensors per side.

that the directional artifacts have been removed, but the ‘good’ regions are still ‘noisy’. The next step was to see if there could be an improvement in the signal to noise ratio (SNR).

+

[Ver: 7.51g/W]

[4.7.2006–2:13pm]

[1–12]

[Page No. 7]

The data set corresponding to the undamaged medium was taken as baseline data and factored out of the data set corresponding to the damaged medium. The factored projection values

FIRST PROOFS


(SHM)

Paper: SHM 067739

Keyword

8


PC

ADC card SENSORS PULSER / RECEIVER

MULTI-PLEXING UNIT COMPOSITE PLATE

Figure 8 Compensated modified cross-hole tomogram for anisotropic case.

Figure 9 Schematic diagram for Lamb wave tomography.

were used as the data set for the reconstruction. The reconstructed image for the anisotropic case is shown in Figure 8. It is found that the noise, in the ‘good’ region of the sample was reduced considerably and the defect was imaged with greater clarity.

8 SHM Experiments on Composite Plates A MATEC PR5000 pulser–receiver was used in a tone burst mode to improve the efficiency of transduction in the narrow-band PZT crystals. It was found that the tone burst excitation was more efficient and less noisy. The schematic of the experimental setup is shown in Figure 9. Eleven PZT crystals of 25 mm diameter in size, with a center frequency of 500 kHz, were used for each side of the plate for MCH configuration. Signals were digitized and stored for post-processing. Quasi-isotropic cross-ply plates 300 300 mm2 in size and 2 mm in thickness were chosen and impact loaded to produce delaminations as defects (to simulate the lowvelocity impact 6–20 J damage). The C-scan images of the impact damages are shown in Figure 2(b) and (d) for cross-ply and quasiisotropic plates. The sensor arrangement for the MCH configuration with the damage location is shown in Figure 10(a)–(e). The reconstructed tomograms for defect free and defective quasiisotropic plate are shown in Figure 10(b) and (c), respectively. The energy of the first arrival signal

+

[Ver: 7.51g/W]

[4.7.2006–2:13pm]

[1–12]

[Page No. 8]

FIRST PROOFS

burst (fastest wave mode S0) was used as the projection value. To account for the anisotropy, the ratio of the projection values in the damaged sample with that in the undamaged sample was computed. This ratio was then used to generate the anisotropy-compensated tomograms shown in Figure 10(d). Figure 10(f) and (g) shows the reconstructed tomograms for undamaged and damaged crossply samples. The anisotropy-compensated tomogram for the cross-ply plate is shown in Figure 10(h). It can be seen from these tomograms that anisotropy-compensation leads to considerable improvement in the image quality. In particular, the image quality improvement is significant for the case of the quasi-isotropic plate.

9 Discussion Usually the material attenuation, at the specific frequency used, would limit the size of the component to be inspected. In addition, the frequency is selected based on the size of the defect/damage to be detected. Since the composite materials have significantly high levels of attenuation at the frequency used here (500 kHz), the method could handle sizes as high as 2 2 m2. In cases with low material attenuation, such as in metals like aluminum, detection using guided waves over distances of the order of 5 m at these frequencies using post-processing


(SHM)

Paper: SHM 067739

Keyword


9

Sensor arrangement and

(a)

(e)

Modified cross-hole Tomogram for defect free plates

(b)

(f)

Modified cross-hole Tomogram for defective plates

(c)

(g)

Compensated modified cross-hole Tomograms

(d)

(h)

Figure 10 (a) Sensor arrangement for QILT; (b) MCH tomogram for defect free QILT; (c) MCH tomogram for defective QILT; (d) compensated MCH tomogram for QILT; (e) sensor arrangement for CPLT; (f) MCH tomogram for defect free CPLT; (g) MCH tomogram for defective CPLT; and (h) compensated MCH tomogram for CPLT.

of the received signals using digital signal processing techniques has been reported. For larger structures, lower frequencies may be used. However, as the complexity of the structure

+

[Ver: 7.51g/W]

[4.7.2006–2:13pm]

[1–12]

[Page No. 9]

increases and if the boundary conditions are influenced by the presence of materials that are compatible for wave leakage, the region of monitoring could reduce significantly.

FIRST PROOFS


(SHM)

Paper: SHM 067739

Keyword

10

+


The ability of the technique for complicated geometries is also limited by the frequency and the type of complexity in the structure. In curved structures, the guided tend to follow the curved geometry of the component as they propagate and hence the algorithm developed here may be applied. However, when dealing with structural features, such as stiffeners, bolt holes, etc., additional work is required and is currently being pursued. At this juncture, it is not possible to discriminate damage as a function of depth. However, it may be feasible in the future to explore methods using the mode specific displacements (as a function of thickness) to do this. In field situations, the sensors must be permanently attached-on or embedded-in to the specimen. The data collection is expected to be controlled using a multiplexer and the data will be processed with a computer processor. Using current computation power of a Pentium processor, the data collection and the damage detection could be carried in 30–45 s. This technique could be extended for other applications in the area of SHM where shell structures are used. These applications could include structures, such as the storage tank floor monitoring, ship hull monitoring, pressure vessels, pipes, rails, etc. However, each of these applications would require application-specific development.

were generated and received with PZT crystals in an MCH configuration, and images were reconstructed with the collected data using iterative tomographic algorithms. Reconstruction was carried in two steps: first, tomograms for undamaged and damaged plates were generated with the energy of the Lamb wave signal as projection values, later the ratios of the defective to defectless data were used to reconstruct the compensated tomograms. Specifically, it has been demonstrated that Lamb wave tomography with MCH setup can assess low-velocity impact damages with reasonable accuracy. Further, it has been shown that a simple procedure for anisotropy-compensation employed in the tomographic reconstruction algorithm captures the damage features better. It is believed that there is scope for improvement in the anisotropy-compensation procedure since the tomograms for the cross-ply sample continue to show artifacts due to strong anisotropy. Increasing the number of sensors further, it is believed, will lead to improved resolution of the tomogram. Use of other features, particularly from the time–frequency domain (using wavelets) must be explored. The present study was carried out with surface-mounted PZT crystals. However, for certain aerospace applications, a study based on embedded sensors may become necessary, particularly on understanding the effect of the sensors on the structural performance of the component.

10 Conclusions

Acknowledgments

Low-velocity impact damages (BVIDs) were generated and imaged in composite plates using SHM concept based on Lamb wave tomography. Composites of different lay-ups (cross-ply and quasi-isotropic) were subjected to low-velocity impacts to simulate the damages like delaminations. Conventional C-scan was carried to confirm the presence of damage, its location, and size. It is observed that the orientation of delamination was influenced by the fiber direction. For the same intensity of load, the damage extent was greater in cross-ply than quasi-isotropic lay-ups. Ultrasonic Lamb waves

The authors would like to acknowledge the Aeronautical Development Agency for sponsoring this research work and the National Aerospace Laboratories, Bangalore for providing the composite laminates. The authors also thank Mr. Mahadev Prasad and Mr. R. Jaganathan for the technical discussions and the support in carrying out the experimental work.

[Ver: 7.51g/W]

[4.7.2006–2:13pm]

[1–12]

[Page No. 10]

FIRST PROOFS

2

References 1. Mahadev, Prasad, S., Balasubramaniam, K. and Krishnamurthy, C.V. (2004). Structural health


(SHM)

Paper: SHM 067739

Keyword


2.

3.

4.

5.

6.

3

7. 8.

9.

4

10.

11.

+

monitoring of composite structures using Lamb wave tomography. Smart Mater. Struct., 13, N73–N79. Rokhlin, S.I. (1980). Diffraction of Lamb waves by a finite crack in an elastic layer. J. Acoust. Soc. America, 67, 1157–1165. Baker, A.A., Jones, R. and Callinan, R.J. (1985). Damage tolerance of graphite/epoxy composites. Composite Structures, 4, 15–44. Cantwell, W.J. and Moron, J. (1991). The impact resistance of composite materials – a review. Composites, 22, 347–362. Cantwell, W.J. and Moron, J. (1989). Comparison of the low and high velocity impact response of CFRP. Composite, 20(6), 545–551. Victorov, I.A. (1967). Rayleigh and Lamb Waves, New York: Plenum Press. Rose, J.L. (1999). Ultrasonic Waves in Solid Media, Cambridge University Press. Kessler, S. (2002). Piezoelectric based in-situ damage detection of composite materials for structural health monitoring systems. PhD Thesis, Cambridge, USA: Massachusetts Institute of Technology. Disperse-User’s Manual (2001). Disperse Software User’s Manual, UK: Imperial College. Luangvilai, K., Punurai, W. and Jacobs, L.J. (2002). Guided Lamb wave propagation in composite plate/ concrete component. Journal of Engineering Mechanics, 128(12), 1337–1341. Sullivan, R., Balasubramaniam, K. and Bennett, A.G. (1994). Experimental imaging of fiber orientation in multi-layered graphite epoxy composite structures, In: Thompson, D.O. and Chimenti, D.E. (eds),

[Ver: 7.51g/W]

[4.7.2006–2:13pm]

[1–12]

[Page No. 11]

12.

13.

14.

15.

16.

17.

18.

11

Review of Progress in QNDE, Vol. 13, pp. 1313–1320, N.Y.: Plenum Press. Balasubramaniam, K. and Ji, Y. (1995). Guided wave analysis in inhomogeneous plates, In: Thompson, D.O. and Chimenti, D.E. (eds), Review of Progress in QNDE, Vol. 14, pp. 227–234, N.Y.: Plenum Press. Sullivan, R., Balasubramaniam, K. and Bennett, A.G. (1996). Plate wave flow patterns for ply orientation imaging in fiber reinforced composites. Materials Evaluation, 54(4), 518–523. Ji, Y., Sullivan, R. and Balasubramaniam, K. (1996). Guided wave behavior analysis in multi-layered inhomogeneous anisotropic plates. In: Thompson, D.O. and Chimenti, D.E. (eds), Review of Progress in QNDE, Vol. 15, pp. 217–222, N.Y.: Plenum Press. Kak, A.C. and Slaney, M. (1988). Principles of Computerized Tomography, http://www.slaney.org/pct/ pct-toc.html. Malyarenko, E.V. and Hinders, M.K. (2001). Lamb wave diffraction tomography. Ultrasonics, 39, 269–281. Malyarenko, E.V. and Hinders, M.K. (2001). Comparison of double cross-hole and fan beam Lamb wave ultrasonic tomography. In: Thompson, D.O. and Chimenti, D.E. (eds), Review of Progress in QNDE, Vol. 20, pp. 732–739, N.Y.: Plenum Press. Subbarao, P.M.V., Munshi, P. and Muralidhar, K. (1996). Performance of iterative tomographic algorithms applied to non-destructive evaluation with limited data. NDT&E, 30(6), 359–370.

FIRST PROOFS


(SHM)

Paper: SHM 067739

Keyword

+

[Ver: 7.51g/W]

[4.7.2006–2:13pm]

[1–12]

[Page No. 12]

FIRST PROOFS


(SHM)

Paper: SHM 067739

Keyword