An experimental analysis of fracture mechanisms by acoustic emission of woven composite bolted assembly Analyse expérimentale des mécanismes de rupture par émission acoustique des assemblages boulonnés en composite tissé

This work is focused on the study of the evolution of damage mode and failure mechanisms of woven composite bolted assembly carbon fiber/epoxy. In the present paper three configurations are studied [0°,45°,0°,45°], [0°,45°,0°,45°]s and [0°,45°,0°,45,0°]s. In order to analyze a global mechanical behavior of the assembly, monotonous tensile tests are performed. The damage evolution is followed simultaneously by acoustic emission (A.E) and digital image correlation (D.I.C). Acoustic signatures of four modes of damage are identified, matrix cracking, fiber-matrix debonding, delaminating and fiber breakage, then confirmed by microscopic observations in scanning electron microscopy (SEM).


INTRODUCTION
Composite materials find more applications in the realization of structural parts of various dimensions in many industries such as aerospace, car manufacturing, nuclear, biomedical engineering.Due to their high specific properties, woven composites carbons (fiber/epoxy) are increasingly used in aerospace applications.These applications generally require the use of assemblies to load transfer between the laminates composite and other parts that are either metallic or composites.However, the design of assemblies remains a hard point for structural applications.Indeed, the use of bonded assemblies is often prohibited by industrial requirements of reproducibility and maintenance.
Mechanically-fastened bolted-joints under tensile loads frequently are damaged in five common failure modes (fig.1), namely cleavage, bearing, shear, tension, and pull-through.Cleavage failures are associated with both an inadequate end distance and too few transverse plies, Bearing failure occurs predominantly when the bolt diameter is a small fraction of the plate width, this mode of failure leads to an elongation of the hole.Shear-out failure can be regarded as a special case of bearing failure, this mode of failure can occur at very large end distances for highly orthotropic laminates.Net-tension failure occurs when the bolt diameter is a large fraction of the strip width; this fraction depends on the type of material and lay-up used... Pull-through failure occurs mainly with countersunk fasteners or when the thickness to diameter ratio is sufficiently high to precipitate failure [1-22.Many studies have attempted to develop methods for understanding these mechanisms using digital image correlation and acoustic emission.The spatial resolution of the aforementioned procedures is relatively poor.Digital image correlation (DIC) has improved the spatial resolution of optical full-field strain measurements.Parsons et al. [7][8][9] obtained strain fields on two sides of rectangular tensile specimens with this technique using one camera and a right-angle prism.De Almeida et al. [9] recently used one camera and a mirror to measure strain fields on both the front and the lateral side of a specimen simultaneously However, all the methods mentioned above are based on the assumption that the strains measured on the surface of the specimen are representative of the strains throughout the thickness of the specimen, i.e. the reduction of width measured on the front surface is representative for the width reduction throughout the thickness [6].The most significant parameter used in acoustic emission signals is the amplitude.The works of Chen Karandikar and al. 23 , Kim and Lee 24, Kargers-Kocsis and al 25, Kotsikos and al.26 ,Ceysson and al 27 , Benzeggagh and al.28 , Mouhmid and al 29, on different families of composites and on the tensile stress, on bending static or on fatigue show the interest of using this acoustic parameter.However, it is clear that the absolute amplitude values of involved areas vary from one test and material type to another.Uenoya 30 studied the signals in functions of their amplitude and rise time, four zones were identified; matrix cracking, fiber-matrix debonding, delamination and finally fiber breakage.Another parameter was analyzed by Hill 31 on composite glass/epoxy and carbon epoxy; in fact it is a matter of the influence of the energy and amplitude signals in a purpose of predicting failure in composite materials.Huguet and Godin 32-33 have conducted a conventional parametric analysis such as the amplitude, rise time, duration and energy of acoustic emission signals during a monotonous tensile test on unidirectional composites at matrix reinforced by glass fibers.This work concerns the study of damage and rupture under monotonic loading in tensile test of a carbon fiber reinforced epoxy laminated composite bolted joints with clearance between bolt diameter and hole diameter.The influence of stacking sequences and various geometric parameters of laminates will be taken into account for the study of global mechanical behavior of assembly.For this propose three different stacking sequences and geometries were studied and correspond to specimen A, B and C. Two experimental techniques have been used such as acoustic emission (A.E) and digital image correlation (D.I.C).These two techniques are coupled with the evolution of the load applied with the displacement for improvement identification of different phases as well as the chronology of damage.Complementary microscopic observations (S.E.M) were carried out on post mortem of specimen C.This enables the confirmation of results obtained by two techniques mentioned earlier.

Material
The material studied is a laminated composite carbon fiber/epoxy G803/914, based on epoxy resin reinforced by a long carbon fibers fabric.Resin 914 is commonly used in aeronautic field for structures resisting at high temperatures.Mechanical characteristics are specified in table 1.The reinforcement G/803 is a balanced fabric of carbon fibers of high resistance strength (HR).There manufacture of the fiber by spinning and stretching process may influence the molecules of the material in preferred directions and thus create a mechanical anisotropy fiber.The behavior of the fiber in the longitudinal direction is elastic and brittle, its strain at break of up to 1.4%.Each layer of this dry unidirectional fabric has a thickness of 0.13 mm.
Table 2 shows the main features mechanical fibers, where x,y and z indices refer respectively the longitudinal, transversal and perpendicular direction respectively.The mechanical properties of the composite G803/914 are gathered in the table 3. The mechanical properties of the epoxy matrix, fibers T300 and the composite G803/914 cited previously are provided by the supplier Hexcel Composites.The properties of preimpregnated G803/914 are summarized in table 4. To determine the effects of joint geometry and stacking sequence on the failure behavior parametric studies were performed experimentally.Therefore, laminated plates were arranged as three stacking sequences named as Groups A, B and C (Tab. 5).The bolted joint of the specimen C is shown on figure 2. It is constituted from two substrates of carbon fiber/epoxy (Fig. 2 (1) and Fig. 2(2)) and assembled by bolt and screw (Fig. 2 (3) and Fig. ©UBMA -2016 73

Experimental technics
To determine the global mechanical characteristics, the monotonic tensile shear was carried out on an electromechanical tensile testing machine, with a displacement rate of the cross-piece of 1 mm/min.The load versus displacement curves for all composite configurations were drawn via a computer connected to the test machine.To check the reproducibility of the tests, five specimens were tested for each group until failure.Two techniques were used simultaneously A.  ©UBMA -2016 74

Digital image correlation
The DIC technique, developed in the 1980s [34-36], consists of displacement and strain determination.Since that, it has been used in various fields.At each load step, an image, corresponding to a given state of deformation, is taken with a CCD camera and digitalized.It is then compared to the undeformed state image (reference image) using a mathematical correlation function.The image processing uses the integral method (comparison of the images with the first image known as reference) with a step of five images per second and window size of 32 x 32 pixels (covering 50%).The pictures resolution used is 28 μm.Displacements and strains could then be calculated at every characteristic point of the studied region.Among full field measurement techniques, Digital Image Correlation (DIC) is fast emerging because of its versatility and simplicity of use [37].It consists in evaluating displacement fields corresponding to a series of (white light) pictures taken at distinct stages of loading.If the natural texture of the material is not sufficient for tracking accurately the displacements, a random speckle is usually sprayed onto the surface.Two gray level images I and h (I stands for the reference picture and h that corresponding to the deformed stage) are related through the local passive advection of the texture by a displacement field u: The problem consists in identifying the best displacement field by minimizing the correlation residual 2 dx   over the whole region of interest, where The minimization of φ is intrinsically a nonlinear and ill-posed problem.For these reasons, a weak form is preferred by adopting a general discretization scheme Where ψ n are the vector shape functions and u n their associated degrees of freedom.In a matrix-vector format,[ψ] is a row vector containing the values of the shape functions ψ n, and {u}the column vector of the degrees of freedom.After integration over the domain Ω, the global residual is defined as: At this level of generality, one may choose to decompose the displacement field on a ''mechanically meaningful" basis.When no simple behavior is expected, one may use a ''simple" Finite Element kinematic basis.Here, classical bilinear shape functions associated with quadrilateral 4-node elements (or Q4) are chosen.It is referred to as Q4 Digital Image Correlation (or Q4-DIC).The measured displacement fields are next used as inputs for an independent damage law identification procedure, based on the same kinematic description.

Acoustic emission technique (A.E)
The acoustic emission (AE) is a technique that allows studying the phenomena of energy release in the form of transient elastic waves, resulting from internal micro local displacement in a material subjected to stresses (AFNOR NFA 09350 Standard) [32].The exploitation of the various A.E parameters as showed in figure 4 allows the identification of different phases and propagation of defects.EA software locates the origin of the events, this technique requires specific instrumentation.The device consists of an acquisition system, four sensors piezoelectric and a preamplifier.The threshold value is fixed at 40 dB AE .Acoustic emission signals are registered by four differential sensors piezoelectric type Micro-80, the bandwidth is ranging between 100KHZ to 1 MHz.The sensor diameter is 2 mm, stuck to the specimen with silicone grease.The signals are then amplified with 40 dB, converted through the software and finally scanned to obtain various parameters of acoustic emission.

Location of acoustic events
The location is to determine the coordinates of the area where acoustic event has occurred.Thus, the use of sensors allows the location of acoustic output source (Fig. 5 (1)).The most commonly used method is to measure the differences in arrival time (ΔT) of the same signal to a plurality of sensors distributed over the structure.The measurement of (ΔT) is usually triggered by the arrival of the acoustic emission sensor reaches the first wave and closed by its passage of last sensor [38].When the difference in time of arrival of the same signal is given for four sensors, the locus of points at which the source belongs is defined by: Where, v is the velocity of wave propagation, assumed to be constant whatever the direction and distance of propagation.
For elastic wave propagation in substrate, with the increase of propagation distance, the amplitude of vibration and the energy reduce gradually.The acoustic signals were acquired at four spatially separate points using acoustic sensors mounted on the substrate in rectangular array as shown in figure 5.The sensors captured signals are mathematically modeled as: where S1(t) -S4(t) are sensors outputs, F(t) is the AE source signal, g1(t) -n4(t) are un-wanted signals,  is attenuation factor caused by the acoustic path differences and D1-D3 are time differences.In this study, the geometrical coordinates of S1 position Is considered as (0, 0) and others are chosen following S1.The position of AE source is O(x, y).The sensor-source distances can be obtained by: V is the value of propagation velocity, it is the time of propagation of waves from acoustic emission source to sensor Si.The time-delay of propagation of acoustic waves from acoustic source to S1 and S2 can be determined as: in AE source localization propagation velocity and wave arrival times are the most important parameters.The time delay between two collected signals (P1(t), P2(t)) can be determined using cross correlation-function as: The parameter ( ) (which maximizes the cross-correlation function () pp R  provides an estimation of time delay.In Eq. ( 3) it is found that a constant wave velocity in substrate must be known in order to locate acoustic emission source. (2) ©UBMA -2016 76

RESULTS AND ANALYSIS
Figure 6 illustrates the evolution of the load versus displacement of three types of samples with photos of post-mortem tested specimen.The load displacement curve of the test of piece A is characterized by a drop in rigidity of the assembly just after the first peak at 4632 N. The load generally reaches the value ranging between 1 and 5 mm displacements.The observed failure mode of specimen A is the fastener shear out (Fig. 7.( 1)).This mode of failure is influenced by the form of screw heat and weak thickness of the substrate.The initiated damage according to the specimen A is shown at 83% of final bearing failure.The curve shape of the load versus displacement according to the specimens B is characterized by an increase in the rigidity of the joint just after the first peak.FPS followed by a further stress build-up.This response is characteristic for a progressive failure.According to the definition of failure modes (Fig. 1), the observed failure of laminate B starts as a bearing failure.During the stress build-up between the FPS and the second peak stress (SPS), the damage propagates in a progressive manner.At the SPS, the damage has reached beyond the head of screw and the stress starts declining continuously.The failure mode of the specimen B consists from a combination of bearing and net-tension with a maximum force around 9800 N and a corresponding local displacement of 2.63 mm.(Fig. 7. ( 2)).More energy appears to be required to propagate the shear cracks through the 45°, presente in this laminate B. Experimental test has shown initial damage occured at 61% of final bearing failure.According to the specimen C, the maximum force is around 11800 N with a corresponding local displacement of 3.41mm.Consequently the mode of failure observed is the net tension Figure 7. (3).Experimental test has shown initial damage in the specimen C occurred at 72% of final bearing failure.As for laminate B, the bearing response shows a small drop at the FPS point followed by a stress build-up to a SPS.But beyond the peak stress, the stress drops rapidly, faster than for laminate B. The rapid stress drop beyond the SPS, is caused by a tension failure that spreads across.
In conclusion, the diversity in laminating sequences led to equally diverse failure patterns.There are however general trends that go beyond this diversity.Perhaps the most salient general feature was that there was a first peak stress at about 4632 N, irrespective of coupon geometry and reinforcement architecture, for all the coupons that had an ultimate bearing strength greater than 4632 N.They also observed that the damage load was independent of the coupon geometry.Both bearing and shearout were observed at the damage load for the present laminates.The bearing damage caused an instant load drop and hole elongation.A steady stress build-up was therefore possible, up to the second peak load.For the standard coupon geometry, the bearing damage could only develop when enough offaxis reinforcement was present.By picking three samples geometry types, a global picture is created for the bearing behaviour of typical fiber of Carbon/epoxy laminates.The standard substrate geometry showed the most diverse failure behavior.By adding off-axis reinforcement the bearing behaviour changed from a catastrophic failure with relatively low strength to a progressive high strength failure.The development of a progressive failure was always associated with macroscopic bearing damage at the hole edge [39-40].
©UBMA -2016 77 The mechanical properties obtained from the tensile test are the Young's modulus E, the maximum load and the ultimate load are presented in Table 6.The Young's modulus is taken as a slope of the strain-deformation curve between 0.04% and 0.15% of strain.The distribution of events as function of time is presented in figure 8.More important acoustic events are ranging from 40 to 65 dB, it is the matrix cracking which is characterized by short duration waveforms of long rise time and low energy 24.The amplitude is ranging between 65 and 70 dB, less acoustic events is observed compared to the first range.From 80 dB Acoustic events become less important.It is the final rupture of fibers characterized by higher energy26.©UBMA -2016 78  ©UBMA -2016 79  (1) ©UBMA -2016 81

Localization of events
To specify the positions of the localizations in the mark, we proceeded to a temporal division.Collected information is displayed for the intervals time corresponding to the acoustic activity of one or several zones.During the elastic zone, for a while t inferior to 63, 28 seconds, severs stress concentration appear around the hole location (Fig. 13. ( 1  ©UBMA -2016 82

Microscopic observations
In order to confirm mechanisms damage identified by acoustic emission during tensile tests, we conducted a microscopic analysis using a scanning electron microscope (SEM) on post-mortem specimens of the specimen C (Fig. 14).Some specimens are additionally cutout through their thickness.The cutting is made with a diamond disk, along the symmetry plane of the specimen.To avoid material losses and local modifications during the cutting, specimens are preliminary moulded in a coloured resin, sufficiently fluid to infiltrate inside the cracks by capillarity effect.Many mechanisms damages are observed; in the clamping area under both shoulders of the fixtures, most damages concern only the plies at 0° which are subject to plane micro-buckling.
An inter-ply delamination area can also be noticed.Outside the clamping area, plies at 0° are present out-of-plane micro-buckling.Fiber ruptures by compression can be noticed, as well as fragmentations consecutive to multiple successive micro-buckling.Due to multiple transverse cracking, an important fragmentation zone of the plies at 45° is observed.These defaults can be at the origin of inter-ply delamination once they reach the interface.Figure 15 shows enlargement of defaults as matrix cracking, delaminating and specimen C fibers breakage.In conclusion, the main rupture mechanisms giving rise to the bearing process, the load develop in front of the circular clamping area and delamination initiated by the transverse cracks in the plies at 45° can widely spread in direction of the specimen edges.©UBMA -2016 83

CONCLUSION
The aim of this work is within the scope of the analysis by acoustic emission the mechanisms of damage and failure mode of woven composite carbon fiber/epoxy.The coupling of the acoustic emission technique (A.E) and (C.I.D) has highlighted the different phases describing the global mechanical behavior of bolted assembly.
Experimental test has shown initial damage in woven composite bolted assembly (carbon fiber /epoxy) of the specimen A, B, and C occurring respectively at 83%, 61% and 28% of final bearing failure.The damage was apparent in laminated as fiber micro buckling and throughthickness matrix cracking.The damage was concentrated on the bearing plane at the hole of edge and with increasing the load spread on the bearing side of the fastener.The fiber micro buckles were seen in the tows oriented at 0° and +45° to the applied load and the matrix cracks in the tows oriented at 45° to the applied load.Delamination appears as a direct consequence of fiber buckling and its progression generates the ruin of the laminate.Failure occurs in a hard way when stratification favors the propagation of delamination (interfaces 0°/45°).Failure modes and bearing strengths are completely influenced from stacking sequence of composite laminated plates.The analysis is supported by microscopic visualizations showing the failure mechanisms involved in the bearing process, including fragmentation, micro-buckling, delaminating and fibbers breakage.
E and C.I.D. Figures. 3 (1) and 3(2) show details of the two techniques as well as the experimental devices.The coupling of A.E. and D.I.C. allows us to identify the various phases of the global behaviour and the chronology of damage in bolted joint assemblies.

Figure. 6 :
Figure.6: Loading function displacement of the three types of test specimen

Figure. 8 .Figure 9
Figure.8.Distribution of events as function of time during the tensile test of the specimen C

Figure. 9
Figure.9 Load applied against time of the specimen C followed by the acoustic parameters as energy cumulated count number the energy, duration of the signal and rise.

Figure. 10 :-
Figure.10: Images given by (D.I.C) techniques at specific point of the curve load versus time during the tensile test until failure of the specimen C -Zone 3 is represented by the segment [b, d].The beginning of the non-linearity of the curve at point b coincides with the occurrence of damages in the edge of the hole.It is about matrix cracking in plies 0° and micro-buckling of fibers at 45°.These defects propagate and eventually give rise to macrocracks.This phase is described by a strong acoustic emission activity marked by high values of amplitude, energy, cumulated count number, and rise time, hence more failure points and more severe damage areas were observed.
)). the amplitude of signals at other location is only within 40-60 dB for fibers breakage.for the time interval 206.48  t  223.68 corresponding to the last zone (zone 5), more locations are introduced around the hole and spread rapidly in the four area of the marker (Fig.13.(2)).These localizations confirm the presence of propagation of defaults during the tensile test until failure [22].

Figure 14 .
Figure 14.Microstructure of composite specimen C after fracture by scanning electron microscope.

Figure. 15 .
Figure.15.Magnification of microstructure of composite specimen C after fracture by scanning electron microscope.

Table .
Table.4: the properties of pre-impregnated G803/914 Table.5:Stacking sequences with the following geometric parameters D (D hole diameter) and q thickness of substrate).

Table 6 :
Mechanical properties of the specimen C with statistical parameters.