Table of Contents ICAC'98

Damage assessment in an SMC composite by means of ultrasonic techniques F. Aymerich and S. Meili Dipartimento di Ingegneria Meccanica Università di Cagliari, Piazza d'Armi - 09123 Cagliari, Italy Fax : 0039 070 6755717 Corresponding Author Contact: Email: meili@iris.unica.it

TABLE OF CONTENTS

Abstract Introduction Damage evaluation Experimental Material and procedures Damage indicators Results and discussion Conclusions References

Abstract

In this study the damage progression during tensile loading of an SMC composite was characterized by means of different ultrasonic parameters. A number of specimens were tension-tested with blocks of loading and unloading cycles at a constant maximum stress level, then scanned in pulse-echo mode at 5 MHz. Time- and frequency-domain parameters were C-mapped and averaged over the examined area to give integral indicators of the internal state of the material. A relationship was found to exist between such indicators and some mechanical and physical parameters, such as stiffness, cumulative damage energy and matrix crack density.

Introduction

The SMC (sheet moulding compound) consists of a combination of polyester resin, calcium carbonate filler, thermoplastic additive and random oriented glass reinforcing fibers. Good mechanical and physical properties, together with some advantages in processability and production, make SMC competitive with metals for automotive applications [1]. The relatively low fiber volume fraction and the uncontrolled filler and reinforcement distribution during processing lead to a highly heterogeneous structure with several kinds of stress concentrators such as fiber ends and interfaces, filler material and voids. The result is that at quite an early stage loading introduces various forms of localized damage, such as fiber-matrix interface debonding and matrix cracking. In tensile loading tests, matrix micro-cracks developing perpendicularly to the load direction represent the main form of damage, and this phenomenon can be observed as anelastic deformation in the stress-strain curve [2]. With increasing loads the size of the transverse cracks increases, while the cracking spacing decreases. This microdamage is responsible for mechanical properties degradation and leads to anisotropic effects due to the prevailing transverse orientation.
Ultrasonics is a suitable technique for direct, nondestructive evaluation of structural materials' degradation. Although the most common objective of ultrasonic testing is the detection and location of overt flaws, like cracks in metals or delaminations in composite laminates, ultrasonics can be used, in association with other techniques, to identify microstructural factors that alter strength or performance. However, while single large defects can be individually detected and characterized, widely dispersed discontinuites are impossible to resolve and only their global effect on bulk properties can be observed and measured.
Mechanical testing can give integral parameters such as Young's modulus decay or indicators of energy dissipation, while ultrasonic testing can focus on attenuation and velocity measurements, in that the ultrasonic wave interacts with the material, and the alterations experienced by the wave during the travel contain the information about the internal structure and its discontinuities.
The aim of this paper is to characterize damage progression of SMC during tensile loading by means of different ultrasonic procedures. Since the load-induced damage essentially consists of matrix microfractures difficult to detect by means of conventional non destructive techniques, various parameters obtained by different analyses of the ultrasonic signal have been evaluated as possible indicators of damage accumulation. Cyclic tests have been conducted to study material degradation under incremental loading and to correlate changes in integral damage parameters to the diffusion of local damage events. The results obtained by ultrasonic analysis are discussed and compared to the information drawn from mechanical indicators such as stiffness decay or hysteresis cycle area. Both these classes of data are correlated with the amount of damage, evaluated by means of radiographs taken after each loading step.

Damage evaluation

The degradation state of a material can be drawn from different mechanical parameters. From static or cyclic tensile tests indicators such as ultimate stress or strain, Young's modulus decay, or indicators of energy dissipation, related to hysteresis cycle size and shape, can be obtained.
Non-destructive testing techniques are not always appropriate for evaluating the internal state of short fiber-reinforced composites due to the peculiar physical properties of these materials. For example, the detection of matrix cracks and fiber-matrix debonding by conventional X-radiography is difficult, because of the little decrease caused by these defects to the overall absorption. Penetrant-enhanced X-radiography gives better results and offers high resolution damage pictures, but since internal cracks cannot be reached by the radio-opaque liquid, inner damage can only be inferred from the distribution of analogous superficial damage.
As regards ultrasonic techniques, the resolution of the method depends on the beam size and on the wave frequency; if single cracks or delaminations are to be detected, only high frequency signals give the required resolution, but SMC composites, due to the highly heterogeneous structure and the presence of numerous scattering discontinuities, make it necessary to use lower frequencies, which set a serious limit to the detectability of single defects.
The ultrasonic attenuation of a solid containing a distribution of voids depends on the void fraction and on their ultrasonic scattering cross-section [3,4]. Quantitative correlations were observed between porosity and attenuation in metals and in morphologically more complex laminated composites, with relationships strongly dependendent on size, shape and orientation of voids. In particular, it was noticed that porosity fraction in laminates affects also the frequency content of the ultrasonic signal, with higher frequencies resulting more attenuated than lower frequencies [5-8]. In SMC composites damage starts with the initiation of matrix fractures from defects or stress concentrators; matrix cracks grow perpendicularly to the loading direction and therefore at an unfavourable orientation for ultrasonic testing, as compared to voids in composite laminates, which are generally flattened and elongated along the fiber directions in the laminate plane. Another drawback of SMC composites is that glass fiber bundles act as scattering areas which cause ultrasonic attenuation -similar to that from cracks and defects- that must be taken into account if proper damage-attenuation relationships have to be established.
In the present work different damage indicators, as obtained from mechanical testing and ultrasonic and radiographic analyses, are evaluated and compared to show the possibility of ultrasonic characterization of forms of damage typical of SMC materials.

Experimental

Material and procedures
The material investigated in this study was SMC LPR27 provided by IVECO and with properties as indicated in table 1.

Fibre content (volume)	19.5 %
Fibre content (weight)	27±2 %
density	1920 kg/m3
ult - tension	65 MPa

Table 1: Main properties of SMC LPR27

Specimens 210 mm long by 30 mm wide with a thickness of 3 mm were cyclically tested at a loading rate of 3 MPa/s in a servohydraulic testing machine. In these tests the specimen was subjected to 10 loading and unloading cycles at a constant maximum stress level with stress and longitudinal strain monitored by means of the machine load cell and a 50 mm extensometer. At the end of each block the specimen was removed from the machine for ultrasonic and radiographic evaluation; after non-destructive investigation the loading procedure was repeated with higher maximum stress. A typical stress-strain response of the material is shown in fig 1.

Fig 1: Cyclic stress-strain curve of SMC specimen.

Damaged specimens were X-rayed for 3 min with 30 kV tube voltage after infiltration of a zinc iodide solution; it is worth noting that the penetrant-enhanced radiographic technique is able to resolve only damage connected to the surface, while internal defects -impossible to fill with the dye- may remain undetected.

Fig 2: X-radiograph of damaged sample

X-radiographs, which offer a very detailed picture of matrix damage (fig. 2), were digitized using a flat bed scanner in order to obtain quantitative measures of superficial microcracks. The resulting 8-bit grayscale images, saved as TIFF files, were analyzed with an image processing program. Since in radiographic films fractures appear lighter than the surrounding undamaged material, thresholding was used to discriminate matrix cracks, which were displayed as black traces on a white background; a skeletonize procedure was then applied, by which edge pixels were removed from the traces until they were reduced to their medial line. At this point the total length of matrix cracks, represented by single-pixel width lines in a binary image (fig. 3), and the crack density could be obtained by simple pixel count operations.
As regards ultrasonic testing, damaged specimens were C-scanned in a water tank using a purpose-built X-Y scanning bridge and an ultrasonic system consisting of a pulser-receiver, a digital oscilloscope and a PC acting as a controller. The specimens were placed between the 5 MHz, 0.25 inch diameter, unfocussed transducer and a glass plate acting as a reflector. Tests were conducted by acquiring and storing, at each scanning point, the entire ultrasonic waveform reflected from the glass after passing twice through the sample; reference waveforms were also obtained, after removing the specimen, by acquiring ultrasonic pulses travelling in a water-only path.

Fig 3: Binary images of cracked specimen.

Damage indicators
A number of mechanical and ultrasonic parameters were selected to characterize the damage state.The decay of elastic modulus during incremental loading proved in previous studies [2] a reliable damage indicator, since stiffness is strictly related to the stress redistribution resulting from damage. The parameter adopted was (E-E₀)/E₀, where the Young's modulus E was evaluated in the range 5-20 MPa and E₀ was the modulus measured during the first loading step.
The area of the first hysteresis cycle of each loading block represents the energy dissipated for visco-plastic deformation, for new surface creation and by friction, while the areas of the following cycles represent mostly the friction energy. The sum of the differences between the areas of the first and the subsequent cycles (cumulative damage energy) gives the energy expended for damage extension, and can be assumed as an useful indicator of damage progression [2].

As concerns ultrasonic parameters, the simple examination of back plate reflection maximum amplitude provides a measure of attenuation and can reveal the presence of material changes and discontinuities. The changes in maximum echo amplitude from the plate -after passing twice through the test object- were normalized to the glass reflection peak through a water-only path for the same test conditions and then C-mapped, thus revealing more attenuative zones.
The total attenuation was then considered in the frequency domain. The signal loss through the composite specimen was obtained as a function of frequency by normalizing the frequency spectrum to a calibration spectrum obtained from a water path trace; the frequency dependence of the attenuation contains information about the size of flaws and voids. The acquired signals, with and without specimen, were Fourier transformed and the signal loss as a function of frequency f was defined as,

Total Attenuation(f) = 20 log( Awithout specimen / A with specimen)

where A is the amplitude spectrum, and C-mapped for all the load steps.
The normalized data can be analyzed by evaluating the attenuation slope around the center frequency of the useful bandwidth (fig. 4); within this bandwidth, the attenuation is frequency-dependent with a linear law. The curve slope increases with the degradation of the material and several studies have shown that the attenuation slope is proportional to the void content [4-6].

Fig 4: Attenuation slope resulting from power spectra with and without specimen.

Also the curve shape can be considered, as the presence of cracks affects the spectral content of the broadband pulse; one can then consider the centroid frequency < f >, defined as

as a significant parameter to measure degradation.
This quantity can be made dimensionless, in order to make it independent of the probe central frequency; some researchers have shown that the shift of the centroid frequency < f >/< f ₀>, when normalized to the centroid frequency of the virgin specimen < f ₀>, shows a linear correlation with porosity [5,8]. The frequency dependence of the attenuation is, once again, the physical mechanism for the centroid frequency shift.

Results and discussion

Fig 5: C-scans of maximum peak voltage (darker areas correspond to higher peak values).

Fig 6: Maps of damage propagation between subsequent load steps.

Fig 7: Attenuation maps at 3 and 7 MHz (maximum stress = 25 MPa; darker areas correspond to more attenuative zones).

C-scans of the selected parameters are useful to follow the history of damage spreading. Fig. 5 shows C-scans (array size = 40x90 - scanned area = 30x70 mm²) of time domain peak values for different load steps, while the maps of fig. 6 have been obtained by subtracting the map of the specimen at the current load step from the map at the preceding loading step, in order to highlight the onset of new damage zones (lighter areas). It is possible to observe that at high load levels matrix micro-cracking tend to concentrate in critical regions with final fracture, which leads to the separation of the sample in two pieces, occurring in these critical areas.
Maps of total attenuation at selected frequencies are visible in fig. 7. From the observation of these maps it emerges that higher frequency waves offer higher resolution than low frequency waves in detecting the internal structure of the material.

While C-scans can point out more attenuative zones, the mean value over the map can be adopted as an integral indicator of the degradation state in the chosen region. The curves representing ultrasonic indicators in the diagrams of figs. 8-12 have been obtained by averaging the selected parameters over a 25x60 mm² region of the C-scan. Ultrasonic parameters have been normalized by a reference value corresponding to the undamaged material, making it possible to compare variations in material properties.
As the peak value changes inversely with attenuation it could be used as a first degradation indicator. The decay of its value, averaged over 3 different 25x25 mm² regions and the whole 25x60 mm² region, and normalized to the average value of the undamaged specimen, is plotted in fig. 8(a) versus the maximum stress reached at each load block; in the same graph the peak values decay is compared with the stiffness decay measured on the cyclic stress-strain curve. The two parameters appear to follow the same trend, i.e. they hold approximately constant before a stress level of about a 30 MPa, decrease in a linear fashion from that level, and fall suddenly just before failure. The same indicators, plotted versus crack density -as obtained by image processing procedures-, are shown in fig. 8(b). We can observe a first abrupt drop of both parameters (stiffness decay and maximum echo amplitude), apparently unrelated to damage crack progression, and a second phase in which they decrease with increasing damage. The initial decrease is probably associated with the onset and propagation of internal damage, starting from pre-existing voids and stress concentrators and impossible to reveal by radiographic techniques, sensitive to surface-connected damage only.

Fig 8: Peak amplitude and Young's modulus decay vs stress (a) and crack density (b).

In fig. 9 the total attenuation through the specimen is shown for different frequencies within the transducer bandwidth. Two remarkable features of these curves can be pointed out:

• the attenuation increases as frequency increases, thus confirming the frequency dependence of attenuation;
• for high load levels, i.e. as crack number grows, the attenuation increase is steeper for higher frequencies.
This fact can be explained by the frequency dependence of scattering and by the inhomogeneous size distribution of defects, with smaller cracks only detectable by shorter wavelengths, and suggests considering the slope of attenuation versus frequency as a significant indicator of the internal state of the material

Fig 9: Parametric plot of attenuation vs maximum cyclic stress.

The attenuation slope, plotted versus crack density in fig. 10(a), exhibits a linear dependence on the crack density except again for the first load steps, during which the presence of internal microcracks and defects cannot be detected by radiographs. It can be inferred, as already mentioned, that the only form of damage occurring at this early stage is that starting from stress concentrators typical of the heterogeneous nature of SMC which even for low loading levels introduce various forms of inner micro-damage. Fig. 10(b) shows a comparison between attenuation slope and cumulative damage energy as a function of the load steps. These two parameters agree reasonably, making it possible to think of using any of them to locate a critical load level after which mechanical properties deteriorate faster.

Fig 10: Attenuation slope vs crack density (left) and compared to damage energy as a function of maximum applied stress (right).

The average centroid frequency shift over the scanned area is plotted versus crack density in fig. 11 and as a function of applied stress in fig. 12; again the ultrasonic indicator proves to be proportional to crack density, after a first step of X-ray undetectable damage. According to mechanical and physical parameters such as cumulative energy or matrix crack density, the material starts to deteriorate at a fast rate at a maximum applied stress of about 30 MPa; the frequency shift conforms quite satisfactorily to this trend, thus suggesting the potential of using frequency related parameters as reliable damage indicators.

Fig 11: Centroid frequency shift vs crack density.

Fig 12: Comparison among centroid frequency shift, damage energy and crack density as a function of maximum applied stress.

Conclusions

The damage progression during tensile loading of an SMC composite was characterized by means of different ultrasonic indicators. On the basis of the results of the analysis the following conclusions can be drawn:

• Mapping attenuation values is a reliable way of identifying weak regions characterized by diffuse matrix cracking;
• the adopted ultrasonic parameters (total attenuation, attenuation slope, centroid frequency shift) are sensitive indicators of damage accumulation in SMC composites;
• attenuation slope and frequency shift indicators show a linear correlation with matrix crack density; a first phase of internal damage onset, impossible to detect by radiographic techniques, can be identified with ultrasonic analyses.

References

1. Handbook of Composites, G. Lubin Ed.,Van Nostrand Reinhold, New York (1982).
2. Aymerich, F., P. Priolo, "Characterization of an SMC Material for Bumper Applications", Proc. Experimental Techniques and Design in Composite Materials, Cagliari, October 30-31, 1996, Key Engineering Materials, Vol. 144 , pp. 43-52 (1998).
3. Birks, A.S., R.E. Jr. Green, P. McIntire, Ultrasonic testing (Nondestructive testing handbook, 2nd ed.; v. 7), American Society for Nondestructive Testing, (1991).
4. Jeong, H., D.K. Hsu, "Experimental analysis of porosity induced ultrasonic attenuation and velocity change in carbon composites", Ultrasonics, Vol 33 No 3 (1995).
5. Hsu, D.K. and S. M. Nair, "Evaluation of porosity in graphite-epoxy composite by frequency dependence of ultrasonic attenuation", Review of Progress in Quantitative Nondestructive Evaluation, Vol. 6B (Thompson, D.O. and Chimenti, D.E., Eds) Plenum Press, New York, pp. 1185-1193 (1986).
6. Hughes, M.S., S.M. Handley, J.G. Miller, and E.I. Madaras, "A relationship between frequency dependent ultrasonic attenuation and porosity in composite laminates.", Review of Progress in Quantitative Nondestructive Evaluation, Vol. 7B (Thompson, D.O. and Chimenti, D.E., Eds) Plenum Press, New York, pp. 1037-1044 (1987).
7. Long, E.R., Jr, S.M. Kullerd, P.H. Johnston and E.I. Madaras, "Ultrasonic Detection and Identification of Fabrication Defects in Composites", First NASA Advanced Composite Technology Conference, NASA CP 3104, pp. 705-720, (1991).
8. Akers, D., C. Vaccaro, S. Ellworth and D. Pettit, "The effect of porosity density and configuration in composite materials on the ultrasonic waveform", Review of Progress in Quantitative Nondestructive Evaluation, Vol. 15 (Thompson, D.O. and Chimenti, D.E., Eds) Plenum Press, New York, pp. 1239-1246 (1996).