Blast and Impact Resistance of Laminated Glass Structures

Transcription

1 Proceedings of the IMPLAST 21 Conference October Providence, Rhode Island USA 21 Society for Experimental Mechanics, Inc. Blast and Impact Resistance of Laminated Glass Structures Mr P. Hooper, Mr H. Arora and Dr J. P. Dear Department of Mechanical Engineering, Imperial College London, SW7 2AZ, UK Abstract Glass fragments produced by explosions pose a significant risk of injury to those close by. Laminated glass can help mitigate these risks. Full scale blast testing of laminate windows was performed with charge sizes from 15-5 kg at ranges of 1-3 m. Full-field deflection measurements of the window pane were obtained using high-speed 3D digital image correlation (DIC) along with load measurements at the joint. A high-speed servo-hydraulic machine was used to replicate the high strain-rates seen under blast loading. Cracked laminated glass was loaded in tension at varying rates to determine the stress-strain response. Delamination between the interlayer to glass interface was observed using high-speed photoelasticity. These experiments can provide input data for models of blast response of laminated beyond the fracture of the glass plies. Assumptions made in current design standards were found to be not valid in some cases. Interlayers of thickness less than 1.52 mm were found to fail prematurely and should therefore be avoided in blast resistance designs. 1 Introduction Buildings with prominent glazed facades make ideal targets for terrorists aiming to maximise human casualties and perceived damage. Annealed window glass is a brittle material that offers little resistance to the shock waves produced by explosions and creates sharp fragments that can travel at high velocity if it fractures. Historically, the majority of injuries from bomb blasts have been from glass fragments[1]. To mitigate this laminated glass is used to protect building occupants by retaining fragmented glass to a polyvinyl butyral (PVB) interlayer. After the glass plies fracture, the PVB interlayer between the cracked glass continues to offer resistance to the blast wave. To be effective the laminated glass needs to be well held to a supporting structure, usually with a structural silicone joint. If the joint is not strong enough, the pane could detach from the structure and fly into the building, injuring occupants. The work presented in this paper is part of a research project at Imperial College London initiated by Arup Security Consulting. The aim of this research is improve the understanding of the behaviour of cracked laminated glass and loading on the joint so that current design standards can be improved, helping structural engineers optimise their glazing designs for a specific blast threat. 2 Background Three current standards address the design of glazing to reduce to hazards in a blast, UFC and ASTM F2248 in the US and the Glazing Hazard Guide in the UK. The UFC standard prescribes a minimum laminated glass make-up of a.75 mm PVB interlayer between two 3 mm annealed glass layers and a structural silicone bite depth of 9.5 mm[2]. This guidance is said to be valid for blast loading up to a 33 kpa peak pressure with a

2 283 kpa-ms impulse and a 4 kpa peak pressure with a 25 kpa-ms impulse[3]. For loadings above these values the design needs to be validated with blast testing or a detailed analysis. The ASTM F2248 standard allows the designer to calculate an equivalent 3 s duration wind load for a particular charge weight and standoff[4]. The calculated equivalent load is then used in conjunction with the ASTM E13 standard for selecting glass to resist wind loads[5]. The equivalent wind loads were determined from blast tests and correlate windows that broke safely with the charge weight and standoff distance used in the test. In this procedure the PVB interlayer is ignored and the laminated glass is considered as a monolithic pane with the same nominal thickness of glass. Both of these standards consider the glazing to have failed when the glass plies fracture, neglecting any additional resistance offered by the PVB interlayer. The Glazing Hazard Guide[6] adopts an approach that extends beyond the fracture of the glass plies by considering the laminated pane as a single degree of freedom (SDOF) system consisting of an equivalent load, mass and spring, as described by Smith[1]. The equivalent load and mass are determined from conversion factors for the particular dimensions of the glass pane. These were derived assuming that the deflected shape is the same as in the static case. The equivalent spring function is a non-linear function determined from a static analysis of the glass pane using large deflection plate theory. This analysis is performed up to the point where the maximum stress exceeds the breaking strength for the glass type considered. After this, a similar static analysis is performed on the cracked glass by treating the PVB interlayer as a membrane. An assumed high strain-rate modulus for the PVB is used to do this. The response of the SDOF system to a blast load is then analysed using a finite differencing scheme as described by Biggs[7]. A total 2 mm deflection limit is imposed on the analysis, after which it is assumed that the PVB interlayer would tear. This assumption is based on observations from blast tests on 1.5 m 1.2 m windows of 7.52 mm laminated glass. The guide compiles results together of common window sizes and a plots the limiting curves for glass breakage and a 2 mm deflection on a pressure-impulse diagram. This allows easy determination of the window size required to resist a specific blast threat. Several assumptions have been used in the development of the current design standards, namely the assumption that the deflected shape under blast load is the same as that in the static loading case and that the PVB response is linear elastic. Both of these are known to be untrue in many situations and the experiments conducted here attempt to address that. 3 Blast testing A series of eight open-air blast experiments were performed on a total of 12 laminated glass panes at RAF Spadeadam, Cumbria, UK. Charges ranging in mass from 15-5 kg (TNT equivalent) were detonated at stand-off distances from 1-3 m. For each test an explosive charge was detonated in front of a test cubicle housing the test window(s) and measurements of window deflection, edge reaction forces and blast pressure were made. The charge was positioned symmetrically in front of the cubicle at a standoff distance R, raised on foam blocks to height h as shown in Figure 1. B B ABC D C D EF Figure 1: Side elevation of test arrangement.

3 Steel subframe Silicone joint Laminated glass (a) High-speed camera setup. (b) Frame cross-section showing strain gauge pair. Figure 2: Details of instrumentation. 3.1 Image correlation High-speed 3D digital image correlation (DIC) was used to track the full rear-surface position of the window at 1 ms intervals during each blast test. Two synchronised high-speed cameras with a resolution of pixels were mounted inside the test cubicle at a working distance s w from the test window and centred on the window centre point. The camera setup used is shown in Figure 2a. A high-contrast speckle pattern was applied to the rear of the window using acrylic paint to allow correlation of position between the two cameras. Paint was also applied to the front of the window to block out light from the explosion. The system was then calibrated using a calibration grid before the test. After the test the captured images of the deformation were imported into the ARAMIS image correlation software (produced by GOM mbh) to compute 3D position and strain of the window. 3.2 Edge reaction forces Pairs of foil strain gauges were bonded to a steel window frame at the midpoint of each frame edge to measure edge reaction forces. The position of the gauges on the subframe cross-section is shown in Figure 2b. The strain readings from each gauge can be used to calculate the tension in the cracked laminate, F, at an angle of pull, θ, at the joint by considering the subframe as a built-in cantilever beam. Measurements for the angle of pull were made from analysis of the DIC results to allow the direct calculation of tension in the PVB. 3.3 Results from a 7.52 mm laminated pane The example results presented here are from a test on a m laminated pane using a charge weight of 3 kg (TNT equivalent) at 14 m. This charge weight and range created a peak reflected pressure and reflected impulse of 127 kpa and 413 kpa-ms respectively. The laminate used was constructed from two 3 mm annealed glass plies and a 1.52 mm PVB interlayer. A 2 mm deep single sided structural silicone joint was used the bond the glass to a steel window frame. Traces of central deflection, velocity and acceleration vs time are shown in Figure 3a. The pane began to move at 19 ms and rapidly accelerated up to a velocity of 29 m/s before failure of the joint at 26 ms. A peak acceleration of approximately 6 km/s 2 was recorded in this first period. A displacement of 14 mm was recorded at the time of joint failure. No tearing of the PVB interlayer was observed. The pane deflected over 25 mm before the DIC could no longer track it due to excess light entering around the failed joint. After this point the pane continued to travel inwards at approximately 3 m/s until it impacted a frame protecting the high-speed

4 4 35 z v a Pre-fracture Post-fracture Tension θ approx. θ Exp z (mm) v (m/s), a (km/s 2 ) F/b (kn/m) θ (deg) t (ms) (a) Central displacement z, velocity v and acceleration a t (ms) (b) Calculated tension in cracked laminate and angle of pull 3 z (mm) x (mm) (c) Out-of-plane displacement cross-sections. Lines are spaced at 2 ms intervals up to 3 ms. Figure 3: Results from a 3 kg charge at 14 m. camera equipment. Figure 3b shows the angle of pull at the frame and the the tension in the cracked derived from the strain gauge readings. It shows the laminate forms a 3 angle with the frame edge at the time of failure and that tearing of the joint started at approximately 25 ms with an edge load of about 2 kn/m. Tension in the laminate varied between 2-3 kn per unit width, corresponding to a stress in the PVB of between 13-2 MPa. Figure 3c shows a cross-sections of displacement taken horizontally across the centre of the window. Each line is plotted at 2 ms intervals ending with the line of largest deflection at 3 ms. The lines clearly show a relatively flat central region deflecting into the cubicle and deformed curved regions close to the edges. As the pane deflects further the flat central region becomes smaller until the whole profile is curved. This is due to the restraint at the edges causing transverse waves to propagate inwards towards the centre from each edge. The same effect is seen in the image sequences presented in Figure 4. The contour lines on the out-of-plane deflection plots are approximately rectangular in shape and are spaced tighter close to the window edges. This indicates that the deformed areas are concentrated around the window edges and that the centre region of the window is largely flat and undeformed. The maximum principal strain plots show how the strain was concentrated near the edges, reaching about 8% in the corners and 5-6% near the edges. Maximum strain-rates were also calculated and were in the order of 15 s 1. Under impulsive blast loading the window pane rapidly accelerates and quickly acquires an approximately uniform velocity field across its surface. If the blast wave duration is short the subsequent deflection occurs almost entirely due to the momentum of the pane. The restraint at the edge causes a transverse deceleration wave to propagate inwards from each edge towards the centre. The ratio between the transverse wave speed and the inward velocity is crucial to the response of the window. A fast inward velocity and slow transverse wave speed will cause a large undeformed central region, with strain and curvature concentrated near to the edges. Strain-rate in this region will also be high and could to lead to tearing of the PVB around the edges. A slow inward velocity and fast transverse wave speed will allow the strain and curvature to develop over a larger area and will be lower in magnitude.

5 t = 22. ms t = 22. ms Strain, ε 2 24 Displacement, δz (mm) t = 24. ms t = 24. ms Strain, ε 2 24 Displacement, δz (mm) t = 26. ms t = 26. ms Strain, ε 2 24 Displacement, δz (mm) t = 28. ms t = 28. ms x (mm) x (mm) Figure 4: Image correlation results showing: left - raw images, centre - displacement and right - strain. Strain, ε 2 24 Displacement, δz (mm).1 2

6 4 Tensile testing of cracked laminated glass Tension in the cracked laminated glass, as observed in the blast tests, was reproduced in the laboratory using a high-rate servo-hydraulic tensile test machine. Test samples were prepared from 15 6 mm laminated glass strips with 3 mm thick glass plies and PVB interlayer thicknesses ranging from.38 mm to 2.28 mm. The glass plies in the sample were fractured before testing to create fragments that were similar to the crazed glass seen in a blast test. This was achieved by scoring the glass at regular intervals with a purpose built jig and initiating cracks along the score line by gently tapping with a hammer. Using this method it was possible to produce a regular and controlled pattern, enabling the effect of fragment size to be investigated. Figure 5 shows an edge-on view of the cracked laminate. Under tension the PVB interlayer delaminates from the glass fragments and forms a ligament that bridges the gap between glass fragments. Figure 5: Edge-on view of cracked laminated glass under tension. A force vs extension curve for a cracked glass sample with a 1.52 mm interlayer is shown in Figure 6a. The glass fragment size in the sample was 1 mm and the test was conducted at 3 m/s, given a nominal strain-rate of 2 s 1. It can be seen from the graph that the tension rises quickly to a value of about 24 kn/m as the end of the sample was displaced. The tension was then relatively constant until the sample tore at a strain of 12%. This sharp rise F/b (kn/m) F nom /b (kn/m) Strain % Strain rate (s -1 ) (a) Typical force vs extension graph for a cracked laminate. (b) Nominal force vs strain-rate for a cracked laminate. Figure 6: Results of high-rate tension tests on cracked laminated glass with a 1.52 mm PVB interlayer.

7 Figure 7: Face-on view of delamination in a 1.52 PVB cracked laminate sample at 3 s 1. Figure 8: Delamination and tearing of a.76 mm PVB interlayer at 2 m/s. followed by a nominally constant force was typical of most tests. The initial sharp rise is due to the elastic extension of the PVB bridging cracks in the glass plies. Here, the effective length of the sample is very small, and the strain in the PVB is concentrated around the cracks, giving local strain-rates that are much higher than the nominal strainrate. Delamination of the PVB from the glass changes the length PVB that is able to extend and nominal force is reached when the force required to progress the delamination is reached. Figure 6b shows the nominal force reached as the strain-rate is increased. In the blast test shown in Section 3.3 a strain-rate of 15 s 1 was observed. The nominal force recorded using this test method was between 2-25 kn/m at that strain-rate and agrees well with the tension values calculated from the blast test. PVB is birefringent and a colour high-speed camera combined with a polariscope was used to observe the delamination of the PVB from the glass fragments. Figure 7 shows a cracked laminate sample with a 1 mm fragment spacing and a 1.52 mm PVB interlayer tested at 3 s 1. The sequence shows a face-on view of the glass fragments, the progression of delaminated area around the cracks and the area of the fragment that is still bonded to the PVB. In this test the delamination fronts progressed far enough so that the glass fragments were no longer bonded to the PVB before the PVB failed. This is clearly not a desirable case in a blast event, defeating one requirement of the interlayer, and therefore the maximum nominal strain reached in any model needs to be limited. Complete debonding of the was not seen as often in interlayers thinner than 1.52 mm. Figure 8 shows a crack on a laminate sample with a.76 mm PVB interlayer. The sequence is focused on a single crack and shows how the

8 delamination front does not propagate far into the fragments before the the PVB tears. The reason for this is that the delamination front propagates slower due to the reduced force exerted by the thinner interlayer for a given strain. The length of PVB that is able to extend does not increase quickly enough to relieve the build up of strain in the bridging ligaments. Therefore the PVB reaches its failure strain quickly and tears. In a blast event this would result in the laminate tearing near the edges and the whole laminate would enter the building as one piece at high velocity. For this reason PVB interlayers below a thickness of 1.52 mm should be avoided in blast resistant designs. 5 Conclusions In this paper selected results from a research project into the post-fracture behaviour of laminated glass under blast loading have been presented. It was shown that some of the assumptions made in current design standards are not valid in some cases. Specifically the deflection profile under blast loading can differ significantly from the assumed static deflection profile. Measurements of the deflection profile of a window during a blast test, made using high-speed digital image correlation, showed that the window was undeformed across the centre with deformation and strain concentrated near the edges. Maximum principal strain and strain-rate reached approximately 8% and 15 s 1. Tension values between 2-3 kn/m were measured in the cracked laminate during the blast. High-rate tension tests on cracked laminates recorded a nominal tension value of 24 kn/m at similar strain rates. It was shown that thin interlayers can fail a low extensions due to a concentration of strain in the interlayer bridge between fragments. This was caused by a slow moving delamination front between the PVB and glass fragments. Use of PVB interlayers below 1.52 mm in thickness should therefore be avoided because of this effect. Acknowledgements We thank the Engineering and Physical Sciences Research Council (EPSRC) and Arup Security Consulting (Mr D. Hadden, Mr D. Smith and Mr R. Sukhram) for supporting Mr P. Hooper and Office of Naval Research (Dr Y. Rajapakse) for supporting Mr H. Arora. References [1] Smith, D. Glazing for injury alleviation under blast loading: United Kingdom practice. In Glass Processing Days Conference Proceedings, (Tampere, Finland, 21). [2] Department of Defense. Unified Facilities Criteria: DoD minimum antiterrorism standards for buildings. UFC (23). [3] Norville, H. S. & Conrath, E. J. Blast-resistant glazing design. Journal of Architectural Engineering 12, (26). [4] ASTM. Standard practice for specifying an equivalent 3-second duration design loading for blast resistant glazing fabricated with laminated glass. F (29). [5] ASTM. Standard practice for determining load resistance of glass in buildings. E13-9a (29). [6] Security Facilities Executive Special Services Group - Explosion Protection. Glazing hazard guide (RESTRIC- TED). Cabinet Office, London (1997). [7] Biggs, J. M. Introduction to structural dynamics (McGraw-Hill, 1964).