NUMERICAL STUDY OF THE EFFECT OF A CORE-ADJACENT LAYER OF POLYUREA ON THE BLAST ENERGY ABSORPTION OF A SANDWICH T-JOINT

Transcription

1 NUMERICAL STUDY OF THE EFFECT OF A CORE-ADJACENT LAYER OF POLYUREA ON THE BLAST ENERGY ABSORPTION OF A SANDWICH T-JOINT L.A. Louca 1, A. Soleiman Fallah 1, M. Saunders 2 and A. Groves 2 1 Imperial College London, Dept. of Civil and Environmental Eng., South Kensington, London, SW7 2AZ 2 Dstl (MoD), Physical Sciences Dept, Materials and Protection Science, Dstl Porton Down SP4 JQ ABSTRACT The present study investigates the effect of the energy absorption characteristics of a T-joint between 2 sandwich panels using a layer of polyurea between the core and the skin of the base sandwich panel. Finite element simulations using ABAQUS 6.6 have been conducted on a conventional T-joint and a modified configuration. The conventional T-joint comprises a leg and a base sandwich panel with balsa cores and GFRP skins, a Crestomer fillet and two overlaminates. The modified T-joint is similar to the conventional set-up except that a layer of polyurea has been introduced under the top skin of its base panel while preserving the original dimensions. Due to the presence of a rigid body mode (plunge), kinetic energy is not a reliable parameter as the basis of comparison between the two models. Study of partial strain energies and support transmitted forces shows the improvement in the behaviour of the T-joint due to the insertion of the polyurea layer. The support reaction has been abated by 43% at the cost of 7% increase in the weight of the T-joint. The possible mechanisms responsible for mitigation of partial strain energies and the amplitude of the force transmitted to supports are discussed. 1. INTRODUCTION Sandwich structures are extensively used in the construction of naval vessels due to their superior characteristics such as improved blast resistant capacity and low weight to stiffness or strength ratio compared to the conventional monolithic members. They usually comprise a balsawood core and two glass fibre reinforced plastic plates as skins in order for the constituents to be used most efficiently. Sandwich panels are bonded using adhesives and the so-called T-joint is formed where a base panel and a leg panel are joined. A crestomer fillet and two overlaminates are used to enhance the performance of the joint. Figure 1 depicts the schematic of a typical T-joint. The main advantages of adhesive bonding are that work can be done with relative ease and speed. Bolted joints alternatives result in local damage and stress concentrations. A number of studies [1-5] have paid attention to static and dynamic performance of T-joints and the effect of different parameters in this regard. Shenoi et al. [1-3] investigated the influence of geometry and material variations on the behaviour of 1

2 single composite skin T-joint under static loading. They concluded the strong dependence of the joint behaviour on geometry and material make-up and showed that for large fillet radii or small overlaminate thickness T-joint capacity increases when the loading is a 45 o pull-off traction. The importance of this loading direction (which induces shear, flexure, and tension) can be best perceived in the light of the requirements of high intensity loading and large deformations encountered in blast scenarios. Similar results were obtained by other researchers [6]. Fig.1. Schematic of a typical sandwich T-joint. Two main fracture modes for composite sandwich panel T-joints were identified by researchers [7; 8]. These are (1) the delamination between the leg and the base panel and (2) core shear failure in the base panel. Theotokoglou [9] stated that there was no significant difference in the ultimate strength of the two failure modes. As such it is expected that an increase in the capacity of the joint is achievable only when both the base core and the bonded surface between the base and the leg (including fillet and overlamiante bonded surfaces) are protected. One way to achieve this goal is by reduction of the level of forces transmitted through the connection. This is particularly of interest when dynamic loads are involved as the level of transmitted force is not directly related to the maximum input but also to its frequency content as well as to inertia and damping characteristics of the system. Numerical simulation using finite element software assists in evaluating the failure mode and ultimate load of a T-joint. This is particularly important when fracture and damage are involved as capturing crack initiation and propagation in dynamic tests is not straightforward. A 3D model can most accurately provide the analyst with the failure load as it takes into account all 3D effects in the constitutive formulation as well as in crack initiation and propagation. The failure load obtained as such is the ultimate load in a statistical sense. This is partly due to the intrinsic difference between the experimental model and the finite element model which is due to modelling assumptions and formulations such as the smeared crack model. Besides, manufacturing defects and inhomogeneity of the structure well contribute to its 2

3 complication. In the present study the results of numerical analyses on a T-joint tested under the auspices of the EUCLID programme have been presented. The T-joint has been modified by a layer of polyurea under the top skin of its base panel and analysed when subjected to the same shock load. The comparison of results from two analyses shows a considerable drop in the level of transmitted force through the joint as well as better protection of the core against damage when polyurea is used. This is at the cost of an increase in the weight but the trade-off favours the use of polyurea. The possible responsible mechanisms for this phenomenon are discussed and conclusions are drawn. 2. NUMERICAL MODELLING Failure of the balsawood core in a sandwich panel is due to out-of-plane shear stresses. A quadratic convex criterion is used here to account for the interaction between τ 13 and τ 23. Results from polyurea modified sandwich panels show a reduction in the level of kinetic and strain energies of the core and the skins of sandwich panels which depending on the boundary conditions can be related to stresses and strains. The results of the study on panels strengthened with polyurea with different set-ups and placement are shown in figures 2 and 3. Figure 2 shows the extent of damage to the core when no polyurea is used and the mitigation of damage when a layer of polyurea is used and is placed to act in through thickness and in-plane compression. Similar reduction in the number of damaged elements is observed when polyurea is used in different positions. Figure 3 shows a comparison of kinetic and strain energies for different constituents of the sandwich panel. It is expected that for a sandwich T-joint a similar reduction can be achieved through the use of polyurea. Fig.2a. The pattern of damage in the core of an unstrengthened sandwich panel (blue elements represent the failed areas) Fig.2b. The pattern of damage in the core of a polyurea strengthened sandwich panel with polyurea attached to the compression skin 3

4 Comparison of Kinetic Energies Kinetic Energy (N.mm) Time (msec) (a) KE-Core KE-Core-PC KE-Core-PT KE-Core-PCT KE-Core-PCO Kinetic Energy (N.mm) Comparison of Kinetic Energies KE-Skin KE-Skin-PC KE-Skin-PT KE-Skin-PCT KE-Skin-PCO Time (msec) (b) Fig 3. Comparison of kinetic and strain energy time histories for Sandwich panels of different set-up when subjected To pulse pressure of maximum 1 bar (PC=Polyurea attached to compression skin, PT= Polyurea attached to tension skin, PCT=Polyurea attached to both skins, PCO=Polyurea attached to the external face of the compression skin) in (a) Core (b) Skins (c) Core (d) Skins 4

5 Comparison of Strain Energies Strain Energy (N.mm) Time (msec) (c) SE-Core SE-Core-PC SE-Core-PT SE-Core-PCT SE-Core-PCO Strain Energy (N.mm) Comparison of Strain Energies SE-Skin SE-Skin-PC SE-Skin-PT SE-Skin-PCT SE-Skin-PCO Time (msec) (d) Fig. 3 Cont d Polyurea is both viscoelastic and viscoplastic however the effect of strain rate on yield is more pronounced than on elastic modulus. As such for the analyses of this section an elastic-viscoplastic material model is adopted with elastic modulus as the average value for strain rates encountered here. As for the balsa wood core there is no existing constitutive model in ABAQUS material library and a VUMAT subroutine is implemented with the constitutive model adopted from the work of Xue and Hutchinson [1] which is an anisotropic compressible material model and allows failure. The success of this constitutive model is verified [11]. Skins of sandwich 5

6 panels are modelled using an elastic transversally isotropic material model since no failure is observed in tests and the level of stresses are low compared to fail stress. A study on the effect of fillet radius [2] on T-joint capacity has shown the ascending functional dependence of capacity on fillet radius. As such, a triangular fillet has been selected for this study. The geometry of the joint is selected based on a test specimen tested in the EUCLID project. Figure 4 shows the set-up of the experiment. The interface between the faces and the core is an adhesive the properties of which depend on its type as well as the adherend faces. Fig. 4. Experimental set-up for shock load testing of a sandwich panel Solid elements are used in modelling the core and faces as well as overlaminates and Crestomer filler, however, the adhesive has been modelled using cohesive elements with direct traction vs. separation formulation. Boundary shock acceleration has been exerted to hardwood blocks at the edges of the base panel and the pull off force is the result of a clump mass at the top of the leg panel. This situation can be assimilated to ground motion in an earthquake scenario. The alternative set-up comprises a layer of polyurea under the tensile skin in the base panel. This layer acts as a cushion in a similar fashion with which it behaves in a sandwich panel. As such, it contributes to a wider distribution of shock waves. Figure 5 shows a comparison of the two cases at the same time (t 1 =5 msec). As it can be seen there is more pronounced failure in the core of the base panel in the original model at any particular instant of time. Besides, the damage is over a more extended area. Figure 6 is a comparison of partial strain energies of the two systems. Kinetic energies are not included here since the existence of a rigid body mode (plunge) hinders the reliability of this parameter as a basis for comparison. 6

7 (a) Fig. 5a. T-joint under vertical shock load with balsa cores and GFRP skins (t=5msec) (b) Fig. 5b. T-joint under vertical shock load with balsa cores and GFRP skins and a layer of polyurea under the top skin (t=5msec) 7

8 Comparison of Strain Energies Strain Energy (N.m) Time (sec) SE_Balsa_1 SE_Balsa_2 (a) Comparison of Strain Energies Strain Energy (N.m) Time (sec) SE_Skins_1 SE_Skins_2 (b) Comparison of Strain Energies Strain Energy (N.m) Time (sec) SE_Overlam_1 SE_Overlam_2 (c) Fig.6. A comparison of strain energies in (a) balsa cores, (b) GFRP skins, (c) GFRP overlaminates (1) without polyurea (2) with polyurea 8

9 Comparison of Accelerations at the Centre of Mass of the Clump Mass Acceleration (m/sec 2 ) Time (sec) A2_ClumpMass1 A2_ClumpMass2 Fig.7. A comparison of accelerations at the centre of mass of the overhead (case1 without polyurea) (case2 with a layer of polyurea) Figure 7 shows the comparison of the accelerations induced in the clumped mass due to the transmissibility of the T-joint. 3. CONCLUSIONS From the present study the following conclusions can be drawn: 1. The polyurea layer acts as a cushion in a similar fashion with which it behaves in a sandwich panel. This phenomenon called cushioning contributes to wider distribution of shock waves which can abate the extent of damage. It is through this phenomenon that less damage is induced in the core and the core is protected. 2. There is a decrease in the level of forces exerted to supporting structures. This decrease can be related to cushioning, deformation of the polyurea layer and increase in the mass. The way it relates to cushioning is through the wider area of energy distribution. It is related to the through thickness deformation of the polyurea layer if a 2DOF analogy is assumed. Finally its dependence upon mass depends on the increase in the mass and on the frequency content of the input. There is a 43% decrease in the load level at the cost of 7% increase in the mass of the system. 3. An SDOF model of the system with increased mass is inadequate in this case and either a 2DOF model must be used or damping coefficient must be altered and adjusted based on the viscoplastic behaviour of the polyurea layer. 9

10 REFERENCES [1] H. J. Phillips and R. A. Shenoi, "Damage tolerance of laminated tee joints in FRP structures," Composites Part A: Applied Science and Manufacturing, vol. 29, pp , [2] R. A. Shenoi and G. L. Hawkins, "Influence of material and geometry variations on the behaviour of bonded tee connections in FRP ships," Composites, vol. 23, pp , [3] A. R. Dodkins, R. A. Shenoi, and G. L. Hawkins, "Design of joints and attachments in FRP ships' structures," Marine Structures, vol. 7, pp , [4] F. Dharmawan, R. S. Thomson, H. Li, I. Herszberg, and E. Gellert, "Geometry and damage effects in a composite marine T-joint," Composite Structures, vol. 66, pp , 24. [5] H. C. H. Li, F. Dharmawan, I. Herszberg, and S. John, "Fracture behaviour of composite maritime T-joints," Composite Structures, vol. 75, pp , 26. [6] A. R. Rispler, G. P. Steven, and L. Tong, "Failure Analysis of Composite T-Joints Including Inserts," Journal of Reinforced Plastics and Composites, vol. 16, pp , [7] E. E. Theotokoglou and T. Moan, "Experimental and Numerical Study of Composite TJoints," Journal of Composite Materials, vol. 3, pp , [8] U. V. R. S. Turaga and C. T. Sun, "Failure Modes and Load Transfer in Sandwich TJoints," Journal of Sandwich Structures and Materials, vol. 2, pp , 2. [9] E. E. Theotokoglou, "Strength of Composite T-Joints under Pull-Out Loads," Journal of Reinforced Plastics and Composites, vol. 16, pp , [1] Z. Xue and J. W. Hutchinson, "Constitutive model for quasi-static deformation of metallic sandwich cores," International Journal for Numerical Methods in Engineering, vol. 61,pp , 24. [11] D.W. Zhou, L.A. Louca, M. Saunders., Numerical simulation of sandwich T-joints under dynamic loading, Composites: Part B (28), doi: 1.116/j.compositesb