Experimental investigation of the local bridging behaviour of the interlayer in broken laminated glass

Transcription

1 Experimental investigation of the local bridging behaviour of the interlayer in broken laminated glass Didier DELINCÉ Assistant Dieter CALLEWAERT PhD. Researcher Delphine SONCK Civil Engineer Rudy VAN IMPE Professor Jan BELIS Assistant professor Summary The post-breakage behaviour of laminated glass, an important feature in designing structural applications, is known to be ruled by three mechanisms : i) the stretching of the interlayer, ii) the delamination (debonding) between the glass pieces and the interlayer, and iii) the crushing of the glass in the contact zones. In this contribution, the focus is on the adhesion and delamination properties of glass-interlayer interfaces. Experimental results of so-called Through-Cracked Tensile tests (TCT tests) carried out on SGP-glass laminates are presented and discussed, and compared with results on PVB-glass laminates. The main conclusion of these preliminary tests results is that the adhesion of SGP to glass seems stronger compared to PVB to glass. However, a high adhesion does not automatically guarantee a higher safety, since a high adhesion can lead to a faster tearing of the interlayer material. Keywords: Laminated glass, interlayer, polymer, post-breakage, adhesion, delamination, material behaviour, structural application. 1. Introduction For some decades, laminated glass is successfully used as safety glazing. In addition, it appears to be a good technical solution to use in non-standard transparent structures due to its ability to fail in a safer way than simple monolithic glass (the interlayer material retaining the broken glass pieces bonded together). This ability is commonly designed under the concept of residual resistance or post-breakage behaviour, which is particularly important in case of structural applications. Generally a laminated glass element in the post-breakage state collapses under a bending effort, which can act parallel or perpendicular to its plane. Disregarding the important role of the configuration (geometry, boundary conditions of supports, etc.), the failure of the element can be due to one or to a combination of several of the following mechanisms, as explained with more details in [1] : i) the stretching of the interlayer under tensile (and/or shear) forces, eventually up to rupture; ii) the delamination (debonding) between the glass pieces and the interlayer under shear (and eventually normal) stresses; iii) the crushing of the glass in the contact zones where the glass pieces scrape against each other. In order to model the post-breakage behaviour of a laminated glass element for design purposes, the interlayer material properties and the interfacial adhesion properties are the main parameters involved in case a given breakage pattern and rigid broken glass pieces are assumed. The combined effect of the elongation and delamination of the interlayer material has been investigated by mean of a so-called Through-Cracked Tensile test (TCT test), a uniaxial tensile test on a laminated glass piece of which the two glass sheets are cracked perpendicular to the tensile direction.. Some experimental results for PVB-glass samples (PVB : polyvinyl butyral) have previously been reported on in [2].

2 However, alternative interlayer materials have been developed, which aim to improve the mechanical performances of laminated glass, especially for structural applications. SentryGlas Plus (SGP) by DuPont de Nemours is such a structural interlayer. An interesting characteristic of SGP is that it is much stiffer compared to PVB. This is particularly the case at ambient temperature (~20 25 C), and one important reason for that is that PVB and SGP, both thermoplastic polymers with similar density values, have got a different glass transition temperature T g. The glass transition temperature corresponds to an intermediate state typical of polymers between the glassy (solid) state (T < T g ) and the rubber state (T > T g ), due to a progressive change in the arrangement of the molecular chains. Indicative values of mechanical properties of both interlayers are presented in Table 1. Table 1 Mechanical properties of interlayer polymer : indicative values [1] Property unit PVB SGP Volumetric weight (density) kg/m³ Elastic modulus N/mm² Tensile strength N/mm² > 20 34,5 Deformation at breakage % > Glass transition temperature T g C ~10-15 C ~55-60 C Typical comparative tensile curves on PVB and SGP samples at ambient temperature (Fig. 1) present a shape difference similar to the one observed for a thermoplastic polymer tested above and below its glass transition temperature. Consequently, at ambient temperature and large strains, PVB mainly shows a (visco-elastic) hyperelastic behaviour, while the behaviour of SGP is better described as elasto-viscoplastic (see [1] for more details). In this contribution, first the analytical model for the TCT configuration proposed by Seshadri [2] is introduced. Subsequently, results of comparative TCT tests carried out at ambient temperature on four series of laminated glass samples with PVB and SGP interlayers are presented and discussed. Fig. 1 Schematic nominal stress-strain curves from uniaxial tensile tests on interlayer material at ambient temperature [1]

3 2. The Through-Cracked Tensile (TCT) test 2.1 Introduction A Through-Cracked Tensile (TCT) test device has been used by Seshadri [2] 1 to test and model the delamination process in PVB-glass laminates (see Fig. 2). Assuming a hyperelastic model for the interlayer material, he could derive values of interfacial adhesion from the measured force and displacement in steady state of TCT tests. The reported experimental part by Seshadri was quite limited, and the presented and discussed results were limited to only one test configuration. Because we expected some details of the experimental procedure to have a significant influence on the results, we present in this paper more details about the test procedure that was used for the realization of advanced orientation tests at the Laboratory for Research on Structural Models of. 2.2 Model of the TCT configuration Seshadri [2] analysed TCT test results and developed analytical models to derive measurements of interfacial adhesion, Γ 0, from a steady state, and for short crack (a<<h) and long crack (a>>h) conditions. Fig. 2 Schema of TCT configuration The steady state begins when the tensile force, P, is getting constant under a constant displacement rate applied between the two parts of the sample. The main assumption of the model is a weak interface, namely that delamination between the glass and the interlayer faster occurs than the tearing of the interlayer. The assumptions for this model are the rigidity of the glass sheets, the independency of the interfacial adhesion from time and temperature effects, perpendicular planestress / plane-strain conditions (plane strain in the stretched ligament in a plane parallel to the Fig. 2, plane stress in the interlayer s plane), unstrained interlayer material ahead of the crack tips, no friction between the glass and the delaminated part of the interlayer (the ligament). The interlayer is supposed to deform under constant volume conditions. Furthermore, the influence of the edges along the width of the sample is assumed to be negligible, in other words the delamination line is supposed to be straight and parallel to the crack along the width of the TCTsample. Values of interfacial adhesion were obtained by Seshadri from TCT tests on PVB-glass samples by using the analytical model based on the formulation of the energy release rate for long-cracks, written in the general form : ( ε ) = ( P. ε / 2b) U ( ) h Γ (1) 0 = ΓL L L ε L. where 2h is the thickness of the interlayer and b its width, U the elastic strain energy density, and P L and ε L resp. the force and the strain at steady state for a long crack configuration. The considered strain of the interlayer ligament is determined on basis of measures of the crack opening 2δ and of the delamination length a by the relation ε = δ/a. Seshadri derived from equation (1) practical equations for an elastic material and different hyperelastic material formulations [2]. There is an obvious influence of the choice of the interlayer material model on the value of adhesion derived from TCT tests. 1 The author s first name and name, Muralidhar Seshadri, seems to have been inverted in the referenced paper [2], in opposition to many other publications which mention Seshadri M. Consequently, we choosed to refer to the model of Seshadri in place of Muralidhar.

4 2.3 Test description The principle of the test consists to apply a tensile force at a constant displacement rate on a TCT sample described in Fig. 2, and to measure the applied load P, the crack opening 2δ and the delamination length a during the whole test. The used test procedure [3] is described in short below Test samples The principle of a TCT configuration is illustrated in Fig. 2, a corresponding test sample is shown in Fig. 3 : it consists of a small piece of laminated glass with only one interlayer sheet, from which the two glass sheets are cracked previously to the test approximately in the middle of the length. By this way, the two parts of the sample are joined only by an interlayer ligament. Preliminary tests were realized by mean of friction clamps on samples with a 1.52 mm thick SGP interlayer, a rubber layer being placed between the steel clamp and the glass surface. This type of clamp however caused most of the time the sample to slip away from the clamps before a steady state was reached. We opted consequently for a stiffer and stronger type of clamps, shown in Fig. 3 : small pieces of 5 mm thick aluminium plate were glued on the glass surfaces at the ends of the sample, and were then clamped in the tensile machine. Fig. 3 TCT test sample A practical difficulty consists in making the two cracks in the same transversal plane Test device and data acquisition The samples are loaded in a universal electro-mechanic tensile test machine INSTRON For the tests results presented in this paper, a load cell of 2 kn or 50 kn has been used according to the test series. The load P and the displacement between the two clamping devices clamp were continuously measured and registered, at a frequency of 10 Hz. A constant displacement rate v was applied on the mobile clamp, and this was comprised between 2 and 60 mm/min according to the test. The crack opening 2δ and the delamination length a were measured by performing an image analysis on movies registered during the test. The measurement procedure is described here after Delamination measurement based on image analysis The deformation of the TCT samples round the initial cracks was registered during the tests using a common digital camera in video-mode (~30 fps, image resolution 640x480 pixels model: Sony Cyber-shot DSC-H1). The camera was placed at the greatest distance possible from the sample using the largest optical zoom (optical zoom 12x, distance camera-sample ~80 cm), to reduce the optical distortion in the zone of interest of the image. Hence, this last had a length round 8 cm along the long side (640 pixels), and thus a resolution of ~0.13 mm/dot. Using this configuration, the camera frame rate of 30 fps seems largely sufficient : on the highest loading rate of 60 mm/min, the displacement of the moving part is of 1 pixel on the 4 frames. The sharpness of the edges and the contrast of the digital picture were optimized by mean of a plain background and a light spot forming a horizontal angle spot-sample-camera round 45.

5 (a) (b) (c) (d) Fig. 4 Original image (ex.) Fig. 5 Steps of an image analysis of a TCT test A resulting image is shown on Fig. 4 (for a TCT sample with a PVB interlayer). The image analysis consists in transforming the video in a series of pictures, then detecting the different edges by means of Matlab-routines especially written for that purpose. The edges of the glass pieces and the delamination lines are identified by using edges detection techniques, in order to automate the measurements (Fig. 5 : (a) detection of the edges and the reference marks of the glass pieces, (b) crack edges, (c) width interlayer ligament, (d) delamination line). However, this automated technique is only useable for quite regular delamination patterns. The crack opening δ can be measured from the beginning of the test using applied reference marks on the glass surface, on the two sides of the crack (Fig. 5 (a)). The delamination lines can be detected on the images only when they are at a minimal distance from the crack edges, corresponding to an average value of 2.7 mm for the used configurations. A value of the delamination length a was calculated as an average distance between medians of the detected points and the edge of the initial crack, for the two delamination lines (Fig. 5 (d)). 3. TCT test results 3.1 Description of the test series Four main series of TCT samples were considered, with two different interlayer materials (PVB and SGP) and prepared in two different ways. Their length is around 120 mm, and their width is comprised between 25 and 50 mm. The two glass sheets have a thickness of 4 mm. The thickness of the interlayer of all the samples is 2h = 1.52 mm. All samples were tested at an ambient temperature of 23±2 C and a relative humidity of 50±5%. An overview of the main characteristics of the series is given in Table 2. The two series PVB-LMO and SGP-LMO were mostly intended to assess the test procedure and refine the picture acquisition procedure. These samples were taken from laminated glass beams with a height of 120 mm. The cut of the two glass sheets was made using a standard glass cutting technique. After the breakage of the two glass sheets, the interlayer was cut with a common cutter by opening the cracked section. The cut of the interlayer can be done at room temperature for PVBglass laminates, but SGP-glass laminates need to be heated to soften the interlayer. The lateral edges of the TCT samples made on this way were however relatively irregular ( rough cut in Table 2). The samples of the two last series were sawn from 300x300 mm laminated glass plates, excluding border zones of min. 15 mm to avoid edge effects. All the samples of one series originate from one single plate. The lateral edges obtained by this technique were straight and regular ( fine cut in Table 2).

6 Table 2 Overview of series of tested TCT samples (+ PVB-Sesh = series Seshadri [2]) Series 2h [mm] b 3 [mm] lat. edges v [mm/min] # tests 1 PVB-LMO rough cut 2 6/ rough cut 2 1/ rough cut 60 0/5 PVB-Dup fine cut 2 0/4 SGP-LMO rough cut 2 4 5/9 SGP-Dup fine cut 2 5/ fine cut 2 1/3 PVB-Sesh n.a number of tests with steady state reached (before tearing of the interlayer) / total number of tests in the series 2 probably fine cut as well 3 nominal value of sample width The displacement rate v = 60 mm/min corresponds to the value mentioned by Seshadri, and the other values v = 2 mm/min were chosen based on preliminary tests on samples of SGP-LMO type in order to obtain long crack deformation conditions (v = 2 mm/min). It is important to mention that no specific adhesion control of those orientation test samples was performed nor discussed with the producers. However, the adhesion level of the series PVB-LMO and SGP-LMO are likely to be lower than a new commercial product, due to the storage conditions and cutting technique of the samples described above. The samples PVB-Dup and SGP-Dup on the other hand have been manufactured in laboratory conditions and tested relatively quickly, making the adhesion grade more likely to be rather top-level. 3.2 Analysis of the test results As introduced by the 6 th column of Table 2, not all the tests have reached the expected steady state. If we exclude the tests where a test procedure fault occurred and not a sample problem, the shortage of steady state was generally due to the tearing of the interlayer in the ligament part before a steady state could be reached. In most cases, the crack initiation in the ligament was not at the edges of the TCT-sample, but in the central part of the width. The crack initiation in the interlayer ligament began at a nominal stress σ f = P/(2h.b) significantly smaller than the reference values of uniaxial tests mentioned in Table 1 : the dispersion is limited to 19.0 σ f 21.9 N/mm² for the SGP-samples with a width b = mm, and is greater than 22 N/mm² for the wider samples; for the PVB- Dup samples (no tearing occurred for the PVB-LMO samples), the dispersion is higher and its average value is σ f = 9.8 N/mm². For the samples PVB-LMO, no tearing of the interlayer ligament occurred, even at large deformations (2.δ ~ 40 mm!). Consequently, all the TCT-samples of the series reached long-crack deformation conditions, but a steady state never occurred for the samples loaded at a rate of v = 60 mm/min. Besides, the steady states identified for the narrower samples (b = 25 mm) loaded at a slower rate of v = 2 mm/min occurred on a limited time/deformation range. A steady state seems to develop more obviously for wider samples (b = 50 mm), but no general conclusion can be drawn on basis of the limited number of successful tests. The corresponding delamination patterns were irregular in many cases, and we suppose this is due to irregularities along the edges or even in the adhesion level at the interfaces of the sample, among others due to the history of the samples and the rough cutting technique. We admit that this is the explanation to the few occurrences of steady states for the series, due to straying from some hypotheses of the model of Seshadri, such as the plain stress conditions. In comparison, the test results of the series PVB-Dup, with the same testing configuration, were rather different. The tearing of the interlayer occurred for an opening 2.δ of about 4 mm. No steady state ever occurred for the samples of this series. Long crack deformation conditions were clearly not reached, the interlayer ligament always began to neck in an early deformation stage, up to tearing. Limited delamination around the crack occurred, and the delamination pattern was almost always irregular : this seemed to be due to a different balance between the delamination and the stretching of the interlayer along the width of the TCT sample. The delamination length a at the breakage of the interlayer ligament was about 1 to 2 mm.

7 By comparing the two PVB-series, the adhesion level is clearly higher for the series PVB-Dup, leading to the breakage of the interlayer ligament at a shorter crack opening. Furthermore, for all the PVB-samples, no cracks occurred in the glass sheets during the test. From the orientation tests SGP-LMO, half the samples developed long cracks deformation conditions, and developed correspondingly a steady state (Fig. 6). For the other samples of the series, the interlayer ligament began to tear up for smaller deformations in a similar way as the samples PVB-Dup, before any steady state could eventually be reached. Finally, glass splinters and some glass cracks occurred during the opening of the initial crack near the crack tips. Fig. 6 Long cracks deformation pattern for a sample SGP-LMO (S20, b = 25 mm) Fig. 7 Long cracks deformation pattern for a sample SGP-Dup (A9-60, b = 25 mm) The last series of TCT-samples, SGP-Dup, showed a slightly different deformation pattern : cracks in the glass sheet near the initial crack opening and then forward the delamination line occurred together with delamination throughout the TCT test, until the complete tearing of the interlayer ligament. As a consequence, some glass splinters stayed attached to the central part of the ligament (Fig. 7). The start of the tearing of the ligament began at an opening 2.δ comprised on an average between 2 and 3.3 mm, corresponding to a strain value ε = δ/α comprised between 70 and 90 %.

8 Fig. 8 Test results of series SGP-Dup (b = 25 mm) The registered force-displacement diagrams (Fig. 8) also show a steady state like form, however the involved fracture mechanism are here rather differrent from the ones described by Seshadri and observed for the tests on PVBsamples, for which the fracture propagates at the interface between the glass sheets and the interlayer; in contrast, the fracture in the samples of the SGP-Dup series propagates also in the glass sheets. The assumption of a weak interface was not verified anymore in this case. Considering this last observation, it does not seem useful trying to use the model of Seshadri to obtain an adhesion value between glass and SGP interlayer at ambient temperature. Eventually, other models should be more appropriate, but this has not been investigated so far. Additionally, for the TCT tests for which a steady state was observed (namely P is getting constant), a strain steady state was also observed in most cases, but this appeared after the beginning of the load steady state. This observation seems however biased by the used definition of the ligament strain (ε = δ/a) : in steady state conditions, the strain is obviously not uniform on the considered length, namely the distance between the delamination lines on both sides of the opening. 4. Conclusions and perspectives A detailed description of the experimental procedure of TCT tests was presented, and then comparative orientation TCT test results on PVB- and SGP-laminates were described and discussed. It appears clearly that some hypotheses of the model of the TCT test by Seshadri were not really matched, especially for the narrower TCT configurations. This may partly explain the low occurrence of a steady state in the test results, besides the tearing of the interlayer ligament in short cracks conditions. This early breakage of the interlayer ligament could perhaps be due to its damaging during the making of the initial cracks in the glass sheets, and/or by an unfavourable high interfacial adhesion level between the glass sheets and the interlayer. Whatever the effective cause, the corresponding tested samples did not fulfil the assumption of a weak interface : the interfacial adhesion was too high with respect to the strain required to initiate damage in the interlayer ligament. Consequently, no representative value of interfacial adhesion between glass and SGP could be determined from the presented TCT tests based on the model of Seshadri. This is partly due to the geometric configuration of the samples, leading to a greater deviation from the assumptions of the model of Seshadri; but an adapted material formulation for the polymer in a glassy state is also needed to use in place of proposed hyperelastic model formulations for polymers in rubbery state. In addition, further experimental research on the characterization of the adhesion-delamination properties of glass laminates should consider in parallel a description or a control of adhesion grades.

9 5. Acknowledgement This research is supported by the Fund for Scientific Research-Flanders (FWO-Vlaanderen, grant nr. 3G018407). The authors gratefully acknowledge the support of Johanna Louwagie from the Department of Textiles of regarding the experimental aspects. The test samples SGP-Dup and PVB-Dup were kindly provided by DuPont de Nemours Belgium, Mechelen. 6. References [1] DELINCÉ D., CALLEWAERT D., BELIS J., VAN IMPE R., Post-breakage behaviour of laminated glass in structural applications, Challenging Glass - Conference on Architectural and Structural Applications of Glass, Delft University of Technology (TU-Delft), 22/5/ /5/2008, Bos, Louter, Veer (Eds.), Delft, The Netherlands, 2008, 668 pp. [2] MURALIDHAR S., JAGOTA A., BENNISON S.J., SAIGAL S., Mechanical behaviour in tension of cracked glass bridged by an elastomeric ligament, Acta Materialia, Vol. 48, Issue 18-19, 2000, pp [3] SONCK D., Delaminatie en lokaal mechanisch gedrag van gebroken glas/ionomeerlaminaten, Master thesis,,, 2008 (not published).