Design of Sandwich Structures

Achilles Petras
Robinson College, Cambridge

Supervisor: Dr M.P.F. Sutcli e

Cambridge University Engineering Department

for the degree of Doctor of Philosophy December, 1998

A dissertation submitted to

To my parents and bebita

Declaration
This dissertation presents the results of research carried out in the Engineering Department of the University of Cambridge between October 1995 and October 1998. Except where speci c reference is made to the work of others, this dissertation is the original result of my own work and includes nothing which is the outcome of work done in collaboration. No part of this dissertation has been submitted for a degree at any other University. This dissertation is approximately 35,000 words long and contains 50 gures.

March 5, 1999

Achilles Petras Robinson College, Cambridge

ii

Acknowledgements
I would like to express my gratitude to my supervisor Dr Michael Sutcli e for all his kind help, encouragement and dedicated guidance, without which, this project would not have been possible. He gave me exibility to follow my research interests, but at the same time he kept me going in the right direction. His patience with proof-reading my thesis is highly appreciated. I am also grateful to Professor Norman Fleck for his invaluable advice and fruitful discussions. I would like to thank him for supporting nancially my project (US O ce of Naval Research grant #0014-91-J-1916), and my attendance at the 4th International Conference on Sandwich Construction in Stockholm. I acknowledge with gratitude the generous nancial support of the Greek State Scholarship Foundation (IKY). I would like to thank Professors Christos Panagopoulos and Constantinos Lascarides at the National Technical University of Athens and Dr Dimitris Niarchos at the National Centre for Scienti c Research `Demokritos' in Greece for supporting and urging me to pursue my PhD studies at Cambridge. I would like to thank Nigel Hookham and Peter Clayton from Hexcel Composites in Duxford for providing materials and valuable technical information. I wish to acknowledge the Cambridge University Engineering Department (CUED) and especially the Cambridge Centre for Micromechanics (CCM) for use of their facilities. I am also grateful to Alan Heaver and Simon Marshall for building the test rigs and helping me during my experiments. Special thanks to the CUED's librarians for their kindest assistance. I wish to thank my friend Antonios Zervos for helping me out on computational modelling topics. Many thanks to Roxy, Pia, Ingo, Sinisa, Guy, Dongquan, Kieran, Jide at the CCM for those stimulating discussions and entertaining moments, when there was a chance to see them due to my night shifts. Sta , students and fellows at Robinson College have helped make my time at Cambridge really enjoyable. Thanks to all my friends here in Cambridge for the enjoyable moments we have shared during these three years. Also to my friends in Athens, Kostas and Stefanos, for not forgetting me and for making my holidays in Greece a real fun. I am extremely grateful to my parents Violetta and Alexandros for their constant understanding, caring love and support. & " % ! " o o % . Finally but most importantly, I would like to thank Georgina for her love, patience and the happy life we share together. Te amo. iii

Abstract
Failure modes for sandwich beams of GFRP laminate skins and Nomex honeycomb core are investigated. Theoretical models using honeycomb mechanics and classical beam theory are described. A failure mode map for loading under 3-point bending, is constructed, showing the dependence of failure mode and load on the ratio of skin thickness to span length and honeycomb relative density. Beam specimens are tested in 3-point bending. The e ect of honeycomb direction is also examined. The experimental data agree satisfactorily with the theoretical predictions. The results reveal the important role of core shear in a sandwich beam's bending behaviour and the need for a better understanding of indentation failure mechanism. High-order sandwich beam theory (HOSBT) is implemented to extract useful information about the way that sandwich beams respond to localised loads under 3-point bending. `High-order' or localised e ects relate to the non-linear patterns of the inplane and vertical displacements elds of the core through its height resulting from the unequal deformations in the loaded and unloaded skins. The localised e ects are examined experimentally by Surface Displacement Analysis of video images recorded during 3-point bending tests. A new parameter based on the intrinsic material and geometric properties of a sandwich beam is introduced to characterise its susceptibility to localised e ects. Skin exural rigidity is shown to play a key role in determining the way that the top skin allows the external load to pass over the core. Furthermore, the contact stress distribution in the interface between the central roller and the top skin, and its importance to an indentation stress analysis, are investigated. To better model the failure in the core under the vicinity of localised loads, an Arcantype test rig is used to test honeycomb cores under simultaneous compression and shear loading. The experimental measurements show a linear relationship between the outof-plane compression and shear in honeycomb cores. This is used to derive a failure criterion for applied shear and compression, which is combined with the high-order sandwich beam theory to predict failure caused by localised loads in sandwich beams made of GFRP laminate skins and Nomex honeycomb under 3-point bending loading. Short beam tests with three di erent indenter's size are performed on appropriately prepared specimens. Experiments validate the theoretical approach and reveal the iv

Abstract nature of pre- and post-failure behaviour of these sandwich beams. HOSBT is used as a compact computational tool to reconstruct failure mode maps for sandwich panels. Superposition of weight and sti ness contours on these failure maps provide carpet plots for design optimisation procedures.

Keywords:

composite structures, sandwich structures, Nomex honeycomb sandwich beams, failure mode maps, honeycomb anisotropy, indentation, localised e ects, high-order sandwich beam theory, mixed failure criterion for honeycombs, biaxial testing of honeycombs

Contents
Preface and Declaration Acknowledgements . . . Abstract . . . . . . . . . Contents . . . . . . . . . List of Figures . . . . . . List of Tables . . . . . . Nomenclature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ii . iii . iv . vi . viii . xi . xii . 1 . 2 . 3 . 4 . 8 . 10 . 12

1 Introduction and Overview

1.1 Introduction to Composite Structures . . . . . . . . . . . 1.1.1 Fibre Reinforced Polymer Composites . . . . . . . 1.1.2 Structural Optimisation . . . . . . . . . . . . . . 1.2 Sandwich Structures . . . . . . . . . . . . . . . . . . . . 1.2.1 Previous Work on Indentation Analyses . . . . . . 1.2.2 Previous Work on Indentation Failure Prediction 1.3 Scope and Outline of the Thesis . . . . . . . . . . . . . .

2 Failure Mode Maps for Honeycomb Sandwich Beams

2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2 Beam Theory for Sandwich Panels . . . . . . . . . . . . . . . . . . . . . 2.2.1 Skin Failure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2.2 Core Failure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2.3 Honeycomb Mechanics . . . . . . . . . . . . . . . . . . . . . . . 2.3 Construction of a Failure Mode Map . . . . . . . . . . . . . . . . . . . 2.3.1 A Failure Mode Map for Beams with a GFRP Skin and Nomex Core . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.4 Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.4.1 Experimental Results . . . . . . . . . . . . . . . . . . . . . . . . vi

13

13 14 16 17 19 21 23 24 28

Contents 2.5 Discussion . . . . . . . . . . . . . 2.5.1 Skin Failure . . . . . . . . 2.5.2 Core Failure . . . . . . . . 2.5.3 E ect of Ribbon Direction 2.5.4 Intra-cell Buckling . . . . 2.6 Conclusions . . . . . . . . . . . . 3.1 3.2 3.3 3.4 3.5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32 32 32 32 33 33

3 Indentation Resistance of Sandwich Beams

Introduction . . . . . . . . . . . . . High-Order Sandwich Beam Theory Surface Displacement Analysis . . . E ect of Spreading Stresses . . . . Contact Pressure Distribution . . . 3.5.1 Case Study . . . . . . . . . 3.6 Concluding Remarks . . . . . . . .

36

36 37 40 43 49 53 57

4 Indentation Failure Analysis

4.1 4.2 4.3 4.4 4.5

Introduction . . . . . . . . . . . . . . . . Failure Envelope for Nomex Honeycombs Failure Analysis with HOSBT . . . . . . Experimental Work . . . . . . . . . . . . Concluding Remarks . . . . . . . . . . . Introduction . . . . . . . . . . . Reconstruction of Failure Maps Optimisation Carpet Plots . . . Concluding Remarks . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . .

60

60 60 65 68 72 74 75 79 82

5 Failure Maps using HOSBT

5.1 5.2 5.3 5.4

74

6 Conclusions and Future Work

6.1 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83 6.2 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86

83

Appendix A: Experimental Results Bibliography

vii

88 95

List of Figures
1.1 Sandwich construction with honeycomb core . . . . . . . . . . . . . . . 5 1.2 Sandwich panels with (a) corrugated (b) foam and (c) honeycomb core 6 1.3 Typical hexagonal honeycomb with a set of doubled walls. t is the single wall thickness and is the honeycomb cell size . . . . . . . . . . . . . . 7 1.4 Test rig for indentation fatigue tests on aircraft oor panels . . . . . . . 11
0

2.1 2.2 2.3 2.4 2.5

2.6 2.7 2.8 2.9 2.10 2.11

(a) Simply supported beam, (b) Cross section on A-A . . . . . . . . . . Failure modes in the skin . . . . . . . . . . . . . . . . . . . . . . . . . . Failure modes in the core . . . . . . . . . . . . . . . . . . . . . . . . . . Comparison of theoretical and experimentally measured out-of-plane (a) compressive modulus, (b) compressive strength, (c) shear modulus and (d) shear strength of Nomex honeycombs . . . . . . . . . . . . . . . . . (a) Failure mode map. Solid lines refer to beams in which the honeycomb ribbon lies along the beam axis dashed lines for when the ribbon lies transverse to the beam axis. The symbol identi es experimental measurements described in section 2.4, (b) Failure load surface for ribbon lying along the beam axis . . . . . . . . . . . . . . . . . . . . . Failure mode map for three typical values of t=L and R=t = 26 . . . . . Layup details . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Test setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Photographs of the di erent failure modes (for transverse ribbon direction) and corresponding load de ection curves (for both ribbon directions) Variation of failure load Wo with skin thickness to span ratio t=L . . . Photographs and results of specimens with 13 mm cell size . . . . . . .

14 16 17 22

25 26 27 28 30 31 35

3.1 The behaviour of exible and rigid skins . . . . . . . . . . . . . . . . . 37 3.2 (a) Nonlinear displacement patterns, (b) Beam geometry and stresses. The origin of the z-coordinate is always taken at the top of the beam element, either skin or core, which is being considered . . . . . . . . . . 39

viii

List of Figures

3.3 Surface Displacement Analysis software window, showing the reference image frame of a side cross section (painted to have a random speckle pattern) and the `area of interest' where the analysis takes place . . . . 3.4 Experimental setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.5 Load-de ection and core compression curves. A and B correspond to the frames used for processing by SDA. The dashed line indicates the total midspan core compression at a line load of 6 kN/m . . . . . . . . 3.6 Vertical displacement eld in the core produced by (a) SDA and (b) HOSBT model. All contour values are in mm and the scaling in both plots is the same . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.7 Parametric Study change of transmission coe cient Cmzz with wavelength L=m for variations in the parameters t, Ef , c and c. The boxed labels show each time the parameters and the arrows indicate the direction of each parameter's increase. The position of in ection is marked by a symbol . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.8 Discrete contact pressure elements (a) uniform (piecewise constant), (b) overlapping triangles (piecewise linear) and (c) Geometric de nitions . 3.9 Distributions of contact stresses qt = cc and the corresponding normal stresses in the top skin-core interface zz = cc (normalised by the core's out-of-plane compressive strength) for beam A with = 0:8 mm . . . . 3.10 Distributions of contact stresses qt = cc and the corresponding normal stresses in the top skin-core interface zz = cc (normalised by the core's out-of-plane compressive strength) for beam B with = 3:7 mm . . . . 3.11 Dependence of spreading e ect on roller's radius R and determination on how exible or rigid are the skins with respect to the indentation resistance of the sandwich beam . . . . . . . . . . . . . . . . . . . . . . 4.1 4.2 4.3 4.4 4.5 4.6 4.7 4.8 Combined indentation failure mechanism . . . . . . . . . . . . . . . . . The Arcan-type rig . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Load paths and determination of failure envelope . . . . . . . . . . . . The three angle setups and the corresponding load-de ection curves (here for specimens with 128 kg/m3 core density) . . . . . . . . . . . . Failure envelopes for Nomex honeycombs. The dashed line corresponds to the linear failure criterion given by equation (4.1) . . . . . . . . . . . Combined failure criterion . . . . . . . . . . . . . . . . . . . . . . . . . Experimental setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Typical experimental results for sandwich beam with 29 kg/m3 core density loaded by a roller with diameter of 6 mm. Midspan bottom skin de ection, midspan core compression, and line load are plotted against midspan top skin de ection. Lines legend: (|) longitudinal and ({ {) transverse honeycomb ribbon direction . . . . . . . . . . . . . . . . . . ix

42 42 44 45

48 51 55 56 58 61 62 63 64 66 67 68

70

List of Figures

4.9 The pre-failure in uence of core density. Video images captured just before failure during 3-point bending loading with a 6mm diameter central roller . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.10 The post-failure in uence of core density. Video images captured after failure (2 mm total de ection) during 3-point bending loading with a 10mm diameter central roller . . . . . . . . . . . . . . . . . . . . . . . . 4.11 The post-failure in uence of core density on extent of damage. Video images captured after failure (2 mm total de ection) during 3-point bending loading with a 10mm diameter central roller . . . . . . . . . . . . . 4.12 Theory (lines) vs experimental results (symbols) for failure line load Wo 5.1 Improved failure map for Nomex honeycomb sandwich beams. Each contour represents sandwich beams of equal strength in N/m . . . . . . 5.2 Comparison with the experimental results from Chapter 2. Lines show the predictions of HOSBT failure analysis and the symbols represent the experimental data (c.f. Fig. 2.10). Solid lines and symbols correspond to the longitudinal ribbon direction, while dashed lines and hollow symbols correspond to the transverse ribbon direction . . . . . . . . . . . . . . . 5.3 Strength contour plot corresponding to Fig. 5.1 . . . . . . . . . . . . . 5.4 Sti ness contour plot . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.5 Optimisation carpet plot for t=L = 5 10 4 with c= s and c=L as design parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
;

71 71 71 73 77

78 80 80 81

6.1 Vertical displacements calculated by Abaqus nite element analysis . . 87 A.1 For 6 mm diameter roller: midspan bottom skin de ection, midspan core compression, and line load variation curves with respect to midspan top skin de ection. Lines legend: (|) longitudinal and ({ {) transverse honeycomb ribbon direction . . . . . . . . . . . . . . . . . . . . . . . . A.2 For 10 mm diameter roller: midspan bottom skin de ection, midspan core compression, and line load variation curves with respect to midspan top skin de ection. Lines legend: (|) longitudinal and ({ {) transverse honeycomb ribbon direction . . . . . . . . . . . . . . . . . . . . . . . . A.3 For 20 mm diameter roller: midspan bottom skin de ection, midspan core compression, and line load variation curves with respect to midspan top skin de ection. Lines legend: (|) longitudinal and ({ {) transverse honeycomb ribbon direction . . . . . . . . . . . . . . . . . . . . . . . . A.4 Video snapshots during loading with a 6 mm diameter roller . . . . . . A.5 Video snapshots during loading with a 10 mm diameter roller . . . . . . A.6 Video snapshots during loading with a 20 mm diameter roller . . . . . .

89 90 91 92 93 94

List of Tables
1.1 An example of structural e ciency of sandwich panels in terms of weight 5 2.1 Summary of failure criteria . . . . . . . . . . . . . . . . . . . . . . . . . 23 2.2 Material properties of Nomex and laminate skin . . . . . . . . . . . . . 27 2.3 Experimental results. Photographs of failure given in Fig. 2.9 and 2.11 correspond to the entries in italics in the nal column of this table . . . 29 3.1 The range of the geometric and material parameters used in the parametric study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47 3.2 Change of spreading length . . . . . . . . . . . . . . . . . . . . . . . 49 5.1 Expressions for peak failure loads . . . . . . . . . . . . . . . . . . . . . 76

xi

Nomenclature
Greek symbols
Honeycomb cell size Length of contact between the central roller and the top skin or length of top distributed load Quantities de ned in section 2.2 cxz Out-of-plane honeycomb Poisson's ratio fxy Poissons's ratio of the skin material c Honeycomb core density s Density of honeycomb's constituent material cc Out-of-plane compressive strength of the honeycomb core fi In-plane compressive stress for intra-cell buckling of the skin fw In-plane wrinkling strength of the skin fx Maximum in-plane stresses in the skins (Chapter 2) fY In-plane yield strength of the skins bxx In-plane normal stresses in the skins zz Out-of-plane normal stresses in the core Out-of-plane shear strength of the honeycomb for transverse and 32 longitudinal ribbon direction respectively cs Out-of-plane shear strength of the honeycomb core cxz Maximum shear stresses in the core (Chapter 2) x Out-of-plane shear stresses in the core ' Setup angle in Arcan test rig In-plane rigidity of top and bottom skin Beam width Fourier coe cient Core thickness Flexural sti ness of the beam Flexural rigidity of top and bottom skin

txx
31

Latin symbols
At Ab b Cm c D Dt Db

Nomenclature

d E3 or Ec Ecx Ef (Efx) Es G31 G32 Gc (Gcxz ) Gs I If L long, trans M Pi qi qt R t (tt tb) t ut ub v W (Wo) wt wb wc x y z
0

Distance between the midplanes of top and bottom skin Out-of-plane Young's modulus of the honeycomb core In-plane Young's modulus of the honeycomb core in the x direction Young's modulus of the face material (in the x direction) Young's modulus of the honeycomb's solid material Out-of-plane shear moduli of the honeycomb for transverse and longitudinal ribbon direction respectively Out-of-plane shear modulus of the core (in the xz direction) Shear modulus of the honeycomb's solid material Second moment of area of the sandwich beam Second moment of area of the skins with respect to their own centroidal axes Span of the sandwich beam Longitudinal, transverse honeycomb ribbon direction Maximum beam bending moment Heights of triangular pressure elements (Chapter 3) Discrete pressure elements (Chapter 3) External distributed load applied on top skin or contact pressure distribution between the central indenter and top skin Roller's radius Skin thickness (top, bottom) Single wall thickness of honeycomb In-plane centroidal displacements of top or bottom skin Base width of triangular pressure elements (Chapter 3) Load per unit width (at failure) Vertical displacements of top or bottom skin Vertical displacements in the core Coordinates of sandwich beam

Note: Other symbols not mentioned are de ned in the text or gure when appearing

xiii

Chapter 1 Introduction and Overview

One day, towards the end of 1943, a circus proprietor called George May called to see me at Farnborough. After he had told me several Gerald Durrell-type stories about the di culties of keeping monkeys in travelling circuses, he produced something which looked like a cross between a book and a concertina. When he pulled on the ends of this invention, the whole thing opened out like one of those coloured-paper festoons which people use for Christmas decorations. It was in fact a sort of paper honeycomb of very light weight but of quite surprising strength and sti ness. Did I think that such a thing could be of any use in aircraft? The snag, as George May modestly admitted, was that, since it was only made from brown paper and ordinary gum, it had no moisture resistance at all and would fall to bits if it got wet. This must have been one of the relatively few occasions in history when a group of aircraft engineers have been seriously tempted to throw their collective arms around the neck of a circus proprietor and kiss him. However, we resisted the temptation and told May that there could be no serious di culty in waterproo ng the paper honeycomb by means of a synthetic resin. J.E. Gordon `Structures: or, why things don't fall down' 1]

1.1 Introduction to Composite Structures

Innovative, high performance design of load-bearing components is always sought in high-tech applications, such as aircrafts, spacecrafts, satellites or F1 racing cars. These structures should be as as light as possible, while having high sti ness, su cient strength, and some damage tolerance. This requires structurally e cient construction. Structural e ciency can be maximised by using the most e cient materials and optimising the structure's geometry. To produce an optimum design, both these factors need to be considered throughout the design process. The catalogue of materials is one of forbidding length. Thus a designer needs a systematic way to be guided through the maze of material classes so as to gradually narrow down the material choices and choose the optimum material. Ashby 2] proposed a material selection procedure, using material selection charts. Birmingham et al. 3] have introduced an integrated approach to the assessment of alternative 1

Chapter 1. Introduction and Overview

materials and structural forms at the concept stage of structural design based on the above methodology.

1.1.1 Fibre Reinforced Polymer Composites

Fibre Reinforced Polymer (FRP) composite materials are some of the most useful materials available to the designer of high performance structures in the aerospace or maritime industries because their high speci c strength and sti ness can lead to signi cant weight reductions. The aerospace industry has proved the e ectiveness of composite materials in reducing component weight and increasing fuel economy. In the marine eld, the use of composites has been growing steadily since the early 1950's, particularly driven by the low construction costs of Glass FRP's. Initial applications include small crafts, lifeboats and pleasure crafts. Potential applications are radar domes, masts and piping, ship hulls, ship superstructures, submersibles and o shore structure modules. Yachts participating in the America's Cup race have all-composite hull, keel and mast. Racing car bodies are made of composites providing more safety per unit weight to drivers. In the last decade composites have been used in applications outside the aerospace industry. Sporting goods such as tennis and squash rackets, golf shafts, bicycles and oars are a major outlet for composites materials. Future applications include high speed trains, subway cars and surface ships 4]. As well as weight reductions, other bene ts of composites include their ability to cope with extreme environment, reliability, maintainability, life cycle cost and service life extension. Despite all their advantages, composites are still regarded as expensive materials to purchase, tool up for and work with. When using composites, a designer has the freedom to tailor the properties. Although this can improve the structural e ciency, this also increases the complexity and number of design parameters. Many design tools have been developed to predict their behaviour and so decrease the level of empiricism, which has been used up to now in design procedures. Finding an e cient composite structural design that meets the requirements of a given application can be achieved not only by sizing the crosssectional areas and members thicknesses, but also by global or local tailoring of the material properties through selective use of orientation, number and stacking sequence of laminae that make up the laminate composite 5]. A comprehensive presentation of the equations which govern the mechanical behaviour of laminates can be found in Hull's book 6]. Stress analysis for the design of composites laminates is commonly implemented by the use of computer programs (i.e. Cambridge Composite Designer 7]) based on laminated plate theory (LPT).

Chapter 1. Introduction and Overview

This kind of software is convenient, e cient and user-friendly, but does not help determine the optimum laminate con guration or choose the best material system. For this purpose there are procedures which are based on the use of carpet plots, which are themselves generated using LPT principles, but also provide a graphic illustration of the whole range of elastic properties available with chosen system 8]. Miki 9] describes a highly practical tool for design optimisation of laminates based on a graphical procedure. Tsai and Patterson 10] have introduced the laminate ranking method for selecting the optimum ply angles. Recently Quinn 11] has published a design manual for composites, which provides practical information for engineers to facilitate the design of GRP, CFRP, A(ramide)RP composites. Quinn has also introduced a relevant nomogram 12, 13], which allows the costs of the constituent materials ( bres, matrix) in a composite to be quickly assessed.

1.1.2 Structural Optimisation

The design of full scale composite materials structures usually requires a building block approach to testing and design 14]. This approach involves increasing the testing complexity and size from coupon tests to full structural tests. The chemical and material tests in the rst stages are generally well de ned tests. As we move from the laminate level to sub-element, component and sub-structure level, the test become more application-dependent. Neither standard test methods nor databases exist for these larger tests. Thus there is a need to reduce cost and increase the e ciency of structural design and structural failure prediction by a global/local testing and analysis approach. The global/local approach involves supporting the global tests and analyses, used in traditional design approaches, by critical local sub-element tests. These tests are intended to be in between a coupon and sub-structure test. They are cheaper to manufacture and test and most importantly they are fully representative of structural con gurations undergoing the same manufacturing processes, and may even be cut from an actual structure. A major constraint to overcome in the design and construction of large bre-reinforced plastic (FRP) structures (e.g. in ship hulls or aircraft wings) is the exibility of the panels. Although this can be overcome by large increases in thickness, this causes molding di culties and is also very uneconomical. A more satisfactory alternative is the use of sti eners, which are commonly in the form of top-hats, I-beam, tee-joint, or by use of sandwich constructions 15, 16]. The sti eners may be an integral part of the shell or they may be attached by adhesive bonding (the latter is the usual case). Bonded or laminated connections between two structural members represent a zone of

Chapter 1. Introduction and Overview

potential weakness. Recent studies have resulted in computer codes for design of speci c sti ened panel con gurations subject to simple loadings (usually compression). In the 1980s, a computer code PASCO (Panel Analysis and Sizing Code) was developed by NASA 17], which has been widely used for composite optimization procedures 18]. The code has been designed to have su cient generality in terms of panel con guration, loading, and practical constraints so that it can be used for nal sizing of panels in a realistic design situation. The UWCODA (University of Washington Composite Optimization and Design Algorithm) design-analysis-optimization software tool was originally developed to optimise the lay-up of at composite panels. In the UWCODA analysis both ply orientation angles and sti ener geometries are treated as design variables. Optimum designs are sought 19] which minimise structural weight and satisfy mechanical performance requirements, for example maximum strain and minimum strength. Especially within the aircraft industry, nite element based optimisation methods are used to size complex structures for minimum weight taking into account a variety of constraints including strength, sti ness and aeroelasticity. However there is always a need for simple low-cost direct methods, which can assist in con gurations and weight studies at the initial design stage. Bartholomew 20, 21, 22] apply a simple optimization method known as geometric programming (GP) 23]. In this method the minimisation of a function representing weight or cost subject to non-linear constraints on the variables can sometimes be reduced to the solution of linear equations.

1.2 Sandwich Structures

Amongst all possible design concepts in composite structures, the idea of sandwich construction has become increasingly popular because of the development of manmade cellular materials as core materials. Sandwich structures consist of 1) a pair of thin sti , strong skins (faces, facings or covers) 2) a thick, lightweight core to separate the skins and carry loads from one skin to the other and 3) an adhesive attachment which is capable of transmitting shear and axial loads to and from the core (Fig. 1.1). The separation of the skins by the core increases the moment of inertia of the panel with little increase in weight, producing an e cient structure for resisting bending and buckling loads. Table 1.1 shows illustratively the exural sti ness and strength advantage of sandwich panels compared to solid panels using typical beam theory with typical values for skin and core density. By splitting a solid laminate down the middle and separating the two halves with a core material, the result is a sandwich panel. The

Chapter 1. Introduction and Overview

Figure 1.1: Sandwich construction with honeycomb core

new panel weighs little more than the laminate, but its exural sti ness and strength is much greater by doubling the thickness of the core material, the di erence is even more striking. Thus sandwich panels are popular in high performance applications where weight must be kept to a minimum, for example aeronautical structures, high-speed marine craft and racing cars. In the most weight-critical applications, composite materials are used for the skins cheaper alternatives such as aluminium alloy, steel or plywood are also commonly used. Materials used for cores include polymers, aluminium, wood and composites. To minimise weight these are used in the form of foams, honeycombs or with a corrugated construction (Fig. 1.2). As well as mechanical requirements, core
t
2t 4t

Relative Bending Stiffness Relative Bending Strength Relative Weight

1 1 1

7.0 3.5 1.03

37 9.2 1.06

Table 1.1: An example of structural e ciency of sandwich panels in terms of weight

Chapter 1. Introduction and Overview

materials may also be selected based on their re-resistance or thermal properties.

(a)

(b)

(c)

Figure 1.2: Sandwich panels with (a) corrugated (b) foam and (c) honeycomb core
The most common and some unorthodox techniques employed to manufacture sandwich components for structural applications, as well as the recent developments and future trends in terms of both materials and processing routes are comprehensively reviewed by Karlsson 24]. Sandwich panels will have sti ness and strength criteria to meet. The sti ness of honeycomb sandwich panels is straightforward to predict, but it remains di cult to estimate the strength. Typical modes of failure are face yielding, face wrinkling, intra-cell dimpling, core shear or local indentation (where the load is applied to the panel). These are described in detail in section 2.2. The critical failure mode and the corresponding failure load depend on the properties of the face and core materials, on the geometry of the structure and on the loading arrangement. Proper analysis of sandwich structures demands a thorough understanding of the mechanical behaviour of both the skins and the core. The skins behave in a relatively simple manner and, in case of composite laminates, the aforementioned methods of

Chapter 1. Introduction and Overview

analysis (i.e. laminated plate theory) facilitate the modelling procedure. However, the mechanical modelling of the core material, particularly for foams or honeycombs, is less straightforward. The response of the core to shear loading from the skins or loading normal to the plane of skins is required. The behaviour depends both on the materials used in the core and on the core relative density, which is the ratio of the core density to that of the solid material constituting the core. Gibson and Ashby 25] give a thorough overview of the literature on cellular materials, quoting many results for foam cores. Figure 1.3 illustrates the honeycomb structure. One of the most common core materials is Nomex honeycomb1 , because it possesses an extremely high strength-toweight ratio. It is also electrically and thermally insulating, chemically stable, selfextinguishing and corrosion as well as shock and fatigue resistant. As Nomex is used extensively in this thesis, the make-up of this core is described in detail here. Nomex is constructed from ribbons of aramid paper running in the 2 direction (the longitudinal ribbon direction). These are glued together at intervals along the ribbon and the stack of ribbons is then expanded into a honeycomb by pulling in the 1 direction (transverse). The paper substrate is nally dipped into phenolic resin to build up the walls of the honeycomb. Because of this construction method, the honeycomb is anisotropic with respect to out-of-plane shear sti ness and strength.

Figure 1.3: Typical hexagonal honeycomb with a set of doubled walls. t is the single wall
0

thickness and is the honeycomb cell size

Nomex is a Du Pont trademark

Chapter 1. Introduction and Overview

Zhang 26] and Ashby 27] model the elastic and collapse behaviour for Nomex honeycomb materials under shear and out-of-plane compression. Their models agree well with experiments that they made on a wide range of Nomex honeycombs. Zhang and Ashby 28] have also investigated the in-plane biaxial buckling behaviour of Nomex honeycombs. Shi et al. 29] and Grediac 30] model the transverse shear modulus of a honeycomb core. Considerable e ort has been devoted to the analysis of beams, panels and struts of sandwich construction and the results have been summarised in the books of Allen 31] and Plantema 32]. Modelling a sandwich panel as a beam, with the simplifying assumptions that the skins are thin relative to the core and that the core material is homogeneous and much less sti than the skin material, was presented by Allen 31] and developed by Gibson and Ashby 25]. Trianta llou and Gibson 33] have developed an optimisation procedure which can determine the optimum values of skins' and core's thicknesses that satisfy the sti ness constraint at minimum weight. Although most of research work in literature is concerned with bending loading of sandwich beams, Kwon et al. 34] or Pearce 35] have investigated the overall buckling and wrinkling of sandwich panels under in-plane compression. A recent comprehensive review to the subject of sandwich construction and the development of theoretical analyses up to now is given in the book of Zenkert 36]. Holt and Webber 37] summarise recent developments and analyse the elastic behaviour of honeycomb sandwich beams, assuming linear elastic behaviour for the skin and the core. Mechanical, thermal, and hygrometric loading on a sandwich beam with a honeycomb core and laminated facings are included in reference 38]. Failure mode maps have been derived by various authors for sandwich panels with exible cores 39, 40, 41]. These authors have been particularly concerned with beams with ductile foam cores, making appropriate assumptions about the elastic and plastic behaviour of the core and skin. However there appears to be little work on failure of panels with honeycomb cores, whose shear anisotropy can reveal the important role of core shear in the bending of sandwich beams. Furthermore resistance to indentation failure is rarely considered as an important factor, although it can be important in determining the durability and service life of the structure.

1.2.1 Previous Work on Indentation Analyses

The use of exible foams and non-metallic honeycombs as core materials has introduced a new design di culty they are transversely exible compared to the e ectively incompressible metallic honeycombs or rigid foam cores. This leads to signi cant lo-

Chapter 1. Introduction and Overview

cal de ections of the loaded skin into the core material, hence to indentation failure. Indentation failure of sandwich structures has been primarily investigated for aircraft oor panels. These panels need to be su ciently sti to avoid passengers perceiving that the oor is unsafe because of its de ection. High heel shoes or dropped objects are typical causes for local indentations on the top skin of these panels in areas like the aisles, thresholds, toilets/galley. However, a sandwich panel would continue to give good service until a number of such indentations had been made. Therefore, the time from incipient to nal failure should be su cient to allow maintenance operators to change a defective panel at the next convenient check. A good source of information on indentation analyses of sandwich beams are the books by Allen 31] and Plantema 32]. They cover the development of the theoretical analyses based on a so-called `splitted rigidity' model. This approach assumes that the beam consists of a component with only bending rigidity and a component with only shear rigidity. They are connected through equilibrium, assuming that the shear resultants in the two components are the same. These models o ers an adequate level of accuracy for sandwich structures with incompressible cores, but are inadequate for non-metallic honeycomb sandwich panels, since they neglect localised e ects, which play an important role for these exible cores. To model indentation failure of transversely exible cores in sandwich structures requires the use of models that take into account the compressibility of the core in the vicinity of the applied loads. A common approach is the elastic foundation model. Selvadurai 42] presents the principles of these models. Their application in sandwich structures analysis is described in Zenkert's book 36]. The simplest of these is the one-parameter Winkler foundation model, which treats the core material as a set of continuously-distributed linear springs. However, its main drawback is that it neglects shear interactions between the loaded skin and the core. The elastic foundation approach adopted by Thomsen 43, 44] was based on the use of a two-parameter elastic foundation model including the shear interaction between the skins and the core. Typical two-parameter models are those of Pasternak or Vlazov (see Selvadurai 42]). Nevertheless these elastic foundation models neglect the interaction between the top and bottom skin. A di erent approach by Frostig and Baruch 45] consists of treating the beam as an ordinary incompressible sandwich substructure interconnected with a special elastic foundation substructure, which provides the localised e ects due to the di erent displacements of the upper and lower skins. The non-planar deformed cross section of the sandwich beam which is observed by experiments suggested the need for a model which allows non-linear variations of in-plane and vertical displacement

Chapter 1. Introduction and Overview

10

eld through the core. Frostig et al. 46, 47] used variational principles to develop the high-order sandwich panel theory, which includes the transverse exibility of the core. `High-order' refers to the non-linear way in which the in-plane and vertical displacements are allowed to vary through the height of the core, in contrast to simple beam theory where the core in-plane displacements are assumed to vary in a linear way through the depth, and the out-of-plane displacements are assumed to be constant. The high-order theory enables the prediction of localised e ects under conditions such as concentrated loads, delamination and diaphragms 48], curved sandwich panels 49], hygothermal (environmental) e ects and discontinuous skins 50]. A comprehensive review and comparison between this method and conventional beam theories was presented recently by Frostig 51]. In the literature the importance of the core behaviour in a ecting indentation failure has been considered however, the in uence of the skin's exural rigidity is generally overlooked. This omission derives from the fact that in practical applications the skins are quite sti and the in uence of skin exural rigidity can be neglected. One of the main aims of this thesis is to focus on the role of skin rigidity in indentation failure and give a better insight in the mechanism of indentation.

1.2.2 Previous Work on Indentation Failure Prediction

The aircraft industry, in collaboration with sandwich panel manufacturers, has conducted research to establish design codes for indentation resistant and cost e ective sandwich panels 52]. However most of the design speci cations have been based on experimental results rather than accurate theoretical models. In manufacturers' data sheets 53] and handbooks 54] the prediction of indentation failure load is derived as a product of the out-of-plane compressive strength of the core and the area over which the load is applied. However, such an approach is at best approximate. Special experimental setups have been introduced depending on the application static blunt indentation tests for freight oors or dynamic fatigue tests for cabin oors. Bonded Structures Division CIBA (now Hexcel Composites) have devised the `fatigue wheel' machine (Fig. 1.4) to provide data on the relative performance of various sandwich panels. Similar indentation damage accumulation can be observed in sandwich panels used as a part of control surfaces ( aps, ailerons), in this case because of hail or stone impacts. A strength analysis of the damaged sandwich panels is essential to predict the in uence of the multi-site damage. Razi et al. 55] have introduced an analytical method to determine the stress distribution in sandwich panels with damage that is

Chapter 1. Introduction and Overview

11

Figure 1.4: Test rig for indentation fatigue tests on aircraft oor panels
elliptical or circular in shape and at an arbitrary location. The low-velocity or quasi-static impact behaviour of sandwich panels is also another aspect of the whole indentation problem that has been investigated by several authors. The lack of accepted test methods for measuring impact damage resistance of composite sandwich structures led Lagace et al. 56] to propose a new methodology based on static indentation and impact tests. Mines et al. 57, 58] have tested the dropped weight impact performance of sandwich beams with woven/chopped strand glass skins and polyester foam and aluminium honeycomb cores and have simulated the upper skin post-failure energy absorption behaviour with an elastic-plastic beam bending model. Low velocity damage mechanisms and damage modes have been investigated in sandwich beams with Rohacell foam core by Wu and Sun 59] and with Nomex honeycomb core by Herup and Palazotto 60]. The pure static indentation response of composite sandwich beams has been modelled as linear elastic bending of the top skin on a rigid-perfectly plastic foundation (the core) by Soden 61] and Shuaeib 62]. Olsson and McManus 63] introduced a theory for contact indentation of sandwich panels the model is based on the assumption of axisymmetric indentation of an in nite elastic face sheet bonded to an elastic-ideally plastic core on a rigid foundation. Despite the usefulness of such models for analysing sandwich beam indentation behaviour, they are restricted only to the loading case where a sandwich beam is indented on a rigid foundation. However this loading case is not normal in practice. Most sandwich beams in service are simply supported or

Chapter 1. Introduction and Overview

12

clamped and are indented whilst also su ering bending loads. These loading conditions make the in uence of the bottom skin an important factor in the overall behaviour of a sandwich beam under localised loads.

1.3 Scope and Outline of the Thesis

The use of laminate composites as skins and low density cellular materials as cores in sandwich constructions has enabled a very good utilisation of the constituent materials, providing structural components with high sti ness and strength-to-weight ratios. Non-metallic honeycombs are popular in high performance applications because of their good mechanical properties, excellent resistance to hostile environments and better fatigue resistance than aluminium. However, because of the low sti ness of the core, they are susceptible to indentation failure. The need to take into consideration the indentation resistance of sandwich panels when optimising their design and the importance of having a robust computational tool to predict indentation failure in sandwich beams under bending, have been the main motivations for this thesis. To address design including indentation failure, we rst need to investigate the other failure mechanisms described in Chapter 2. Failure maps for loading under 3point bending are constructed and the region, where indentation failure is important, is identi ed. This work uses sandwich panels made of GFRP laminates and Nomex honeycomb cores of various densities. Experimental measurements are compared with simple theoretical predictions. In Chapter 3 analytical modelling which can be used for indentation failure is addressed. This uses an existing beam theory developed by Frostig et al. 46, 47] and applies it to the GFRP - Nomex construction considered in the thesis. The e ect of material properties and the beam and indenter geometry on the beam response are considered. Chapter 4 presents the methodology to predict more accurately indentation failure loads. This is achieved by introducing a new failure criterion for honeycomb core, which takes into account both the out-of-plane normal compression stresses and the shear stresses in the core. Theoretical predictions are compared with experimental data. Re ned failure maps including the new indentation analysis are described in Chapter 5. Carpet plots based on these failure mode maps provide a preliminary optimisation methodology for mimimum weight design of sandwich beams. The thesis concludes in Chapter 6 by summarising the main results of this research together with a discussion of new research directions.

Chapter 2 Failure Mode Maps for Honeycomb Sandwich Beams

2.1 Introduction
In this chapter we consider loading under three-point bending of sandwich beams made with honeycomb and skins, both of which fail in a brittle manner. We apply the analysis framework and failure mode map technique used by Trianta llou and Gibson 41] for ductile foam cores and ductile skin materials. However here we investigate the failure modes of panels with honeycomb cores, whose shear anisotropy can reveal the important role of core shear in the bending of sandwich beams. Furthermore the constructed failure mode maps include indentation failure and identify for what beams this mechanism will be important. In section 2.2 we review beam theory for sandwich panels. Section 2.2.3 describes existing work on honeycomb mechanics. By combining the analysis for sandwich beams with the honeycomb mechanics, we derive in section 2.3 failure loads for the various failure mechanisms. This information is used in section 2.3.1 to draw up failure mode maps, in which the mechanisms of failure and the corresponding failure loads are plotted as a function of the core relative density and skin thickness/span ratio. Theoretical results are illustrated using the commercially-popular combination of laminated glass bre reinforced plastic (GFRP) skins and Nomex core. Theory is compared in section 2.4 with experiments on sandwich panels of various core relative densities and skin thickness/span ratios.

13

Chapter 2. Failure Mode Maps for Honeycomb Sandwich Beams

14

2.2 Beam Theory for Sandwich Panels

In this section we outline the elastic analysis of sandwich beams in three point bending. This will be used to evaluate the stresses in the core or skin and hence the failure loads due to the various mechanisms. Consider a simply supported sandwich beam of span L and width b loaded in 3-point bending with a central load W per unit width as illustrated in Fig. 2.1. The skins each have thickness t and are separated by a thick layer of honeycomb core of thickness c.

Figure 2.1: (a) Simply supported beam, (b) Cross section on A-A
We assume that the skins remain rmly bonded to the core, that the beam bends in a cylindrical manner with no curvature in the yz-plane and that cross-sections remains plane and perpendicular to the longitudinal axis of the beam. The exural rigidity D of the sandwich beam is then given by
3 3 btd2 D = Efxbt + Efx2 + Ecxbc (2.1) 6 12 where d is the distance between the midplanes of the upper and bottom skins. Efx and Ecx are the in-plane Young's moduli of the skin and core respectively for loading in the x direction (along the axis of the beam). Subscripts `f ' and `c' denote the face material and the honeycomb core respectively. Subscript `s' is used in later expressions for the solid material from which the honeycomb is made. The three terms on the right hand side of (2.1) correspond to bending of the skins about their centroidal axes, bending of the skins about the centroid of the whole beam, and bending of the core, respectively. We can simplify this equation by assuming that bending of the skins about the centroid of the beam is the dominant term. The contributions of the rst

Chapter 2. Failure Mode Maps for Honeycomb Sandwich Beams

and third terms amount to less than 1% of this when d > 5:77 and Efx t d 2 > 16:7 t Ecx c c respectively, so that (2.1) becomes

15

(2.2)

where I is second moment of area of the cross-section of the sandwich beam. With 3-point bending the maximum bending moment M is at the mid-span and the corresponding maximum stress fx in the skins is given by MEfx d = WL (2.4) fx = D 2 4dt However, the above theoretical model neglects the e ect of shear de ection in the core, which becomes signi cant for low density cores. Inclusion of this e ect also allows a prediction of observed di erences in beam strength for di erent orientations of the honeycomb ribbon (see section 2.2.3 for further details). For the above reasons we follow the suggestion of Allen 31] for the maximum axial stresses in the faces WbL c + 2t + WL t 1 (2.5) fx = 4 2I 4 2If
2 2 3 3 2 (2.6) I = bt + btd If = bt where = L Gcxz c 1 + 3d c 2Efx t t2 6 2 6 Gcxz is the out-of-plane shear modulus of the core, I is the second moment of area of the sandwich with respect to its neutral axis and If is the second moment of area of the faceplates with respect to their own centroidal axes1 . Equation (2.6) shows that depends on the relative sti ness of the skin and the core. Finally (2.5) gives t W = 4 fx L (2.7)
1

D = Efx btd = Efx I 2

2

(2.3)

t =9 t =3 = ht3 ( ; 1)=++ d4 =3 + t2d2 (2.8) 3 t As the span length L or the core shear sti ness Gcxz approach in nity, (2.7) tends to the simple beam model (2.4). In the case study presented in section 2.3.1, maximum deviations from the simple beam model due to the nite thickness of the skins and the e ect of nite shear sti ness in the core amount to a maximum of 26%.
where
5 3 2
1

Note that If is not negligible in this approach

Chapter 2. Failure Mode Maps for Honeycomb Sandwich Beams

16

2.2.1 Skin Failure

Section 2.2 gives an expression for the maximum stress fx in the skins. This can be used to predict beam failure due to the skin failure modes of face yielding, intra-cell dimpling or face wrinkling, as illustrated in Fig. 2.2.

(a) Face yielding

(b) Intra-cell dimpling

(c) Face wrinkling

Figure 2.2: Failure modes in the skin

Face Yielding
Failure occurs in the top skin due to face yielding when the axial stress in either of the skins equation (2.5)] reaches the in-plane strength fY of the face material for loading along the beam axis,
fx

fY

(2.9)

It is assumed that the skin behaves in a brittle manner. With a symmetrical beam the stress is the same in the tension and compression faces. For composite face materials the compressive face is generally the critical one.

Intra-cell Dimpling
A sandwich with a honeycomb core may fail by buckling of the face where it is unsupported by the walls of the honeycomb (Fig. 2.2b). Simple elastic plate buckling theory can be used to derive an expression for the in-plane stress fi in the skins at which intra-cell buckling occurs as 2Efx 2t fi = 2 1 ; fxy
2

(2.10)

where is the cell size (i.e. the diameter of the inscribed circle) of the honeycomb and Efx and fxy are the elastic modulus and Poisson's ratio for the skin for loading in the axial direction. A similar expression, veri ed experimentally by Kuenzi 64], has been given by Norris 65]. Equations (2.9) and (2.10) can be used to derive the value of cell

Chapter 2. Failure Mode Maps for Honeycomb Sandwich Beams

size above which there is transition from face yielding to intra-cell buckling as = 2t 1 ;2 2 Efx fxy fY

17

(2.11)

Face Wrinkling
Face wrinkling is a buckling mode of the skin with a wavelength greater than the cell width of the honeycomb (Fig. 2.2c). Buckling may occur either in towards the core or outwards, depending on the sti ness of the core in compression and the adhesive strength. In practice, with 3-point bending, inward wrinkling of the top skin occurs in the vicinity of the central load. By modelling the skin as a plate on an elastic foundation, Allen 31] gives the critical compressive stress fw that results in wrinkling of the top skin as 3 E 1=3 E 2=3 (2.12) fw = 2 (1 + (12(3 ; cxz ) )2) 1=3 fx 3 cxz where cxz is the out-of-plane Poissons ratio and E3 the out-of-plane Young's modulus of the honeycomb core (see section 2.2.3).
;

2.2.2 Core Failure

Honeycomb sandwich structures loaded in bending can fail due to core failure. Pertinent failure modes are shear failure or indentation by local crushing in the vicinity of the loads, as illustrated in Fig. 2.3.

(a) Core shear

(b) Local indentation

Figure 2.3: Failure modes in the core

Core Shear
Assuming simple beam behaviour, the shear stress varies through the face and core in a parabolic way under 3-point bending. If the faces are much sti er and thinner than the core, the shear stress can be taken as linear through the face and constant in the

Chapter 2. Failure Mode Maps for Honeycomb Sandwich Beams

18

core. Neglecting the contribution from the skins, the mean shear stress in the core is given by W (2.13) cxz = 2d Assuming brittle behaviour, failure occurs when the applied shear stress xz equals the shear strength cs of the honeycomb core in this direction.
cxz

cs

(2.14)

Low density Nomex cores are particular susceptible to this failure mode. Due to the anisotropy of the honeycomb structure (section 2.2.3) the shear strength of the core depends on the loading direction.

Local Indentation
Failure of sandwich panels in 3-point bending can occur at the load point due to local indentation. Failure is due to core crushing under the indenter. The bending sti ness of the skin and the core sti ness determine the degree to which the load is spread out at the point of application. It is important here to mention the main failure characteristic by which indentation di ers from skin wrinkling. In indentation the top skin de ects after failure with a wavelength of the same scale as the indenter-top skin contact length, whereas in skin wrinkling the de ection of the top skin after failure exhibits wavelengths that are larger than the contact length between the indenter and the top skin. Indentation failure has not been adequately modelled for honeycomb sandwich panels. To include this important failure mechanism, we use a simple empirical approach used in handbooks on sandwich panel construction 53]. In the next chapters a more accurate model will be presented for indentation failure prediction. Here we assume that we know the length2 of contact between the central roller and the top skin. It is further assumed that the load is transferred uniformly to the core over this contact length, so that the out-of-plane compressive stress zz in the core is given by
zz

= W=

(2.15)

Failure is then predicted when this compressive stress equals the out-of-plane compressive strength cc of the honeycomb core.
zz
2

cc

(2.16)

This length derives from the assumption that the skins are `transparent' enough to equate the contact length with the length of initial damage in the top skin-core interface

Chapter 2. Failure Mode Maps for Honeycomb Sandwich Beams

19

The above approach is de cient in three respects (i) the contact area must be estimated in some way - in the experiments described in section 2.4 this is measured, (ii) load transfer from the roller to the core is over-simpli ed this will depend on the relative skin and core sti nesses, (iii) failure in the core will not be governed solely by the compressive stress in the core but will also be in uenced by the local shear stress. A more rigorous stress analysis of the contact region can be found in Chapter 3 and its implementation to predict local failure in honeycomb panels is presented in Chapter 4.

2.2.3 Honeycomb Mechanics

To evaluate the failure mechanisms described in section 2.2, sti ness and strength properties for the honeycomb core are required. In this section we use the results of references 25, 27] to express the properties of the honeycomb as a function of the properties of the solid material from which the honeycomb is made and the relative density c= s of the honeycomb. Although the theory is applicable to any honeycomb, in practice we will focus on the Nomex honeycomb core used in the beam failure experiments of section 2.4. Expressions are compared with published experimental data to evaluate the applicability of the theoretical models to the Nomex honeycomb core and to nd the most suitable expressions for use in the beam calculations. The following notations `1', `2', `3' for honeycomb's main axes refer to those illustrated in Fig. 1.3. The honeycomb Poisson's ratio cxz required for the failure analysis (section 2.2.1) is 13 or 23 for in-plane Poisson strains due to out-of plane loading in the 3 direction. To a rst approximation its value can be taken as that of the solid material (eq. (4.64) in ref. 25]), i.e. 13 = 23 = s. The Young's modulus of the honeycomb in the out-of-plane 3 direction is given by the rule of mixtures expression E3 = c (2.17) Es s In honeycombs, failure under out-of-plane compressive stresses occurs due to fracture of the cell walls or due to elastic or plastic buckling of the cell walls 26]. For Nomex honeycombs failure is due to a `crushing' mechanism, initiated by elastic buckling and developing as a plastic buckling process. The relevant collapse strength cc can be simply estimated using the rule of mixtures expression, cc= sc = c= s, where sc is the compressive strength of the solid from which the core is made. Wierzbicki 66] gives an alternative expression for the failure stress based on a plastic collapse model. For a

Chapter 2. Failure Mode Maps for Honeycomb Sandwich Beams

cc = 3:25 c s
5=3

20

honeycomb with regular hexagonal cells this approach predicts the collapse strength
sc

(2.18)

Zhang and Ashby 27] show that the out-of-plane shear strength and sti ness of honeycombs are independent of height and cell size. Honeycomb cores exhibit slight anisotropy in their out-of-plane shear strength and sti ness, due to the set of doubled walls. By using simple mechanics models based on an array of regular hexagons and considering the double wall e ect approximate expressions for the shear strengths 31 and 32 are derived as

Es = 1:7
31

c s c s c s

(2.19a) (2.19b)

Es = 2:6 and for the shear moduli G31 and G32 as G31 = 0:375 Gs
32

(2.20a)

G32 = 0:6 c (2.20b) Gs s The core shear modulus Gcxz used in equation (2.6) to calculate the skin stress should be taken as either G31 or G32 depending on the orientation of the ribbon direction in the honeycomb. This anisotropy leads to a dependence of skin failure loads on the honeycomb orientation. Similarly the core shear strength cs depends on the honeycomb orientation.

Experimental Evaluation of Honeycomb Mechanics

The theoretical relations detailed in section 2.2.3 are those that we use for our calculations in section 2.3. In this section we compare the theoretical expressions with experimental data from reference 27] and manufacturers'3 data sheets 53] for honeycombs made of Nomex (aramid paper impregnated in phenolic resin). Often it is observed that there is a wide variation amongst data from di erent sources re ecting the wide manufacturing tolerances in the constituent aramid sheet (particularly s) and the di culties in making accurate measurements.
3

Hexcel Composites formerly Ciba Composites

Chapter 2. Failure Mode Maps for Honeycomb Sandwich Beams

21

The measurements of out-of-plane compressive properties are made by testing honeycombs under stabilised4 compression. As depicted in Fig. 2.4(a) the prediction of Equation (2.17) for Young's modulus lie between the two sets of data and close to that of reference 27]. Fig. 2.4(b) shows that Wierzbicki's 66] equation (2.18) for compressive strength ts the experimental data better than the usual mixture's law, re ecting the plastic collapse mechanism of Nomex under compression. Figures 2.4(c,d) show plots of shear modulus and strength. Manufacturers' measurements are signi cantly higher than the measurements of Ashby and Zhang because of the di erent test setup. Ciba or Hexcel use a short beam test, where shear strength and sti ness are out-of-plane and measured indirectly (see eq. (2) in ref. 67]). Zhang and Ashby has tested the honeycombs in in-plane shear with an appropriate testing rig. We believe that the latter source gives a more direct estimate of shear properties. The major di erence between equations (2.19a), (2.19b) and the experimental data of 27] for c= s > 0:1 are due to debonding of honeycomb specimens from the rig. We can use equations (2.19a), (2.19b) when c= s > 0:1, since in the material systems considered here no debonding occurs.

2.3 Construction of a Failure Mode Map

Sections 2.2.1 and 2.2.2 have described various mechanisms of failure which occur with honeycomb sandwich panels, and the honeycomb mechanics needed to evaluate the failure loads for each of these mechanisms. In this section we describe how a map can be constructed detailing which failure mechanism actually occurs for a given material combination and beam geometry. This follows the work of Trianta llou and Ashby 41] for foam-core sandwich panels. The failure mode map is illustrated by way of examples in section 2.3.1. The failure loads depend on the properties of the skin and honeycomb solid material, the relative density c= s of the core, the thicknesses t and c of the skin and the core, and the beam span L. Because we include indentation failure, failure also depends on the loading details. In the experiments we use rollers to apply the load, so that failure depends on the roller radius R. The failure line load Wo can then be expressed as a function of the material properties and the beam parameters 41] Wo = f (t=L t=R c= s). To evaluate this function the expressions for skin and core stresses equations (2.7), (2.13), (2.15)] are substituted into the various failure criteria equations (2.9), (2.10), (2.12) and (2.14), (2.16)] as described in section 2.2 to give the critical line loads as summarised in Table 2.1. The actual failure load and mode are
Stabilised compression means restriction of the cell walls from slipping between the specimen and the rig plates during the test
4

Chapter 2. Failure Mode Maps for Honeycomb Sandwich Beams

22

1.00

C IB A

R elativ e S ta b ilised C o m p ressiv e S tren g th , cc / sc

Ashby-Zhang

R elativ e Y o u n g 's M o d u lu s, E c /E s

0.10

mixture law

eq.(2.17)
0.10

eq.(2.18)

C IB A d a ta
0.01

H e xce l d a ta A sh by -Z h an g

0.03

0.10

0.03

0.10

R elativ e D e n sity , c / s

R e la tiv e D en sity , c / s

(a)
0.010

(b)
eq.(2.19a) eq.(2.19b)

R e lativ e S h e ar M o d u lu s, G c /G s

0.10

R elativ e S h ear S tren g th , c /E s

G 31 G32
C IB A HEXCEL A sh b y-Z h a n g

0.001

eq.(2.20a) eq.(2.20b)
0.01 0.03 0.10

31

32
C IB A HEXCEL A sh b y-Z h a n g

0.03

0.10

R e lativ e D e n sity , c / s

R elativ e D en sity , c / s

(c)

(d)

Figure 2.4: Comparison of theoretical and experimentally measured out-of-plane (a) com-

pressive modulus, (b) compressive strength, (c) shear modulus and (d) shear strength of Nomex honeycombs

Chapter 2. Failure Mode Maps for Honeycomb Sandwich Beams

23

given by the mode with the minimum failure load. Maps of the failure mode and failure load can then be drawn as a function of the beam geometry, for a given material system. The Matlab 68] programming language is used to evaluate the equations.

Table 2.1: Summary of failure criteria

Top skin yield Intracell buckling Face wrinkling Core Shear Indentation

Wo = 4
;

t fY L

t Ef L ;t Wo = 4B1Ef1=3 Es2=3 L

Wo = 1 8 f2

;t

c s

2=3

Wo = 2AEsd Wo = 3:25
sc

c s c s

3 5=3

Note: The quantities and A change according to the honeycomb ribbon orientation

2.3.1 A Failure Mode Map for Beams with a GFRP Skin and Nomex Core
The above section describes how a failure mode map can be constructed for a honeycomb sandwich panel. This is illustrated in this section using sandwich panels made of GFRP laminate skins and Nomex honeycomb cores of di erent densities. The core and skin thicknesses c and t are 9.4 mm and 0.38 mm respectively and the nominal honeycomb cell size is 3 mm. Experimental results for this sandwich panel type are presented in section 2.4. Further details of the panel construction, materials and material properties are given in that section. Fig. 2.5(b) shows the dependence of line load Wo at failure on core relative density c= s and face thickness to span t=L ratio, for a value of the radius of the roller to the skin thickness R=t of 26. Each surface corresponds to a di erent failure mode. The failure mode map, Fig. 2.5(a) is given by the projection of the intersections between failure surfaces on the c= s ; t=L plane. Slightly di erent failure maps are calculated depending on whether the honeycomb ribbon lies in the longitudinal direction along the beam axis, or transverse to the beam axis. Di erences arise from the anisotropy in the strength and sti ness of honeycomb in shear c.f. equations (2.8) and (2.19a,b)]. As noted in section 2.2.2, failure by indentation is estimated using an empirical approach and relies on experimental measurements of the contact area as described in section 2.4. Indentation is the only mechanism which depends on

Chapter 2. Failure Mode Maps for Honeycomb Sandwich Beams

24

the roller diameter. This will be a more widespread mechanism of failure with smaller roller diameters. For the honeycomb geometry chosen for these plots, intra-cell dimpling is not predicted. At the map boundaries, the failure loads for the mechanisms either side of the boundary are equal. In practice failure near a boundary may be due to a combination of the two mechanisms and coupling between the two mechanisms may reduce the load below that predicted for each of the modes independently. The failure mode map shown in Fig. 2.5(a) is useful where a designer has speci ed face and core thicknesses and wishes, for example, to select an appropriate span or relative density. More commonly, however, the span is xed, a standard skin construction and thickness is speci ed, and only the core thickness or density can be relatively easily changed. In this case it is more useful to plot a map of the failure modes and loads as a function of core relative density and core thickness to span ratio c=L, at a xed skin thickness to span ratio t=L and roller radius to face thickness ratio R=t. Figure 2.6 shows such maps for three typical values of t=L with R=t = 26. Note that for long spans (Fig. 2.6a) the core crushing failure mode vanishes, while for short spans face wrinkling is not predicted (Fig. 2.6c).

2.4 Experiments
Sandwich panels were made of Nomex honeycomb core and GFRP laminate skins and were supplied by Hexcel Composites. All panels had the same skin cross-ply laminate on either side of the core, as depicted in Fig. 2.7. Each laminate comprised two glass prepregs the outer with a resin content of 27% and the inner with 41%, giving a skin thickness t of 0.38 mm. The Nomex honeycomb cores used in the panels are designated by the manufacturer as Aerowebr type A1. Panels with core densities of 29, 48, 64 and 128 kg/m3 were used. For most of the tests the honeycomb had a nominal cell size of 3 mm, but for tests described in section 2.5.4 a cell size of 13 mm was used. The core thickness c was 9.4 mm. Mechanical properties of the skin and the honeycomb's constituent solid material (aramid paper + resin) are listed in Table 2.2. Since the compressive strength of the laminate has been inferred from bending tests of sandwich beams, this is not an independent measurement. Hexcel quote a value of 265 MPa, based on simple beam theory equation (2.4)] for long beams. Using the beam model equation (2.7)] with a correction for shear in the core, a revised estimate for the compressive strength of the laminate of 300 MPa is inferred from the data at long spans. Panels were cut into beams using a diamond wheel. A width b of 40mm was chosen

Chapter 2. Failure Mode Maps for Honeycomb Sandwich Beams

25

0.4

Relative density, c=

Face Yield

0.1

Core Crushing

Face Wrinkling

Core Shear

0.02 10
4

10

10

Face thickness / span, t=L

(a)

10

Failure line load, Wo

10 10 10 10 10 1

0.1 0.1 10 10
3 2

Relative density, c=

0.01

10

Face thickness / span, t=L

(b)

Figure 2.5: (a) Failure mode map. Solid lines refer to beams in which the honeycomb ribbon
lies along the beam axis dashed lines for when the ribbon lies transverse to the beam axis. The symbol identi es experimental measurements described in section 2.4, (b) Failure load surface for ribbon lying along the beam axis

Chapter 2. Failure Mode Maps for Honeycomb Sandwich Beams

26

Relative density, c=

Face Yield

0.1

Face Wrinkling

Core Shear 0.01 3 10

Core thickness, c (m) (a) t=L = 5 10;5

10

10

Relative density, c=

Face Yield

0.1

Face Wrinkling Core Crushing

Core Shear

0.01 3 10

Core thickness, c (m) (b) t=L = 5 10;4

10

10

Relative density, c=

Face Yield

0.1 Core Crushing

Core Shear

0.01 3 10

Core thickness, c (m) (c) t=L = 5 10;3

10

10

Figure 2.6: Failure mode map for three typical values of t=L and R=t = 26

Chapter 2. Failure Mode Maps for Honeycomb Sandwich Beams

Nomex 26] Laminate skin 69] Young's modulus, (GN/m ) Es = 0:9 Ef = 20:5 Shear modulus, (GN/m2 ) Gs = 0:32 Gf = 4:2 Compressive strength, (MN/m2 ) sc = 80 fY = 300 Poisson's ratio s = 0:4 f = 0:17 3) Density, (Mg/m { s = 0:724
2

27

Table 2.2: Material properties of Nomex and laminate skin

so that it was greater than twice the sandwich height and three times the cell size, as recommended in by ASTM standards 70]. Beams were cut with the ribbon direction either in the longitudinal direction along the beam axis or in the transverse direction, as depicted in Fig. 2.7. Beams of varying spans and core relative densities were made to probe the various regions of the failure mode maps (Fig. 2.5). Specimens were tested in 3-point bending, applying the load through rollers of diameter of 20 mm (c.f. ASTM standard C393-62 70]). The crosshead speed was kept constant through the test and was chosen so that the maximum load occurred between 3 and 6 min after the start of the test. Displacement of the central loading point relative to the end rollers was monitored using an LVDT and logged on a computer. The central section of the beam (where failure invariably occurred) was also recorded on a video recorder (Fig. 2.8) as the test progressed. To model failure due to local indentation (section 2.2.2), it is necessary to de ne the value the length of the contact area between the central roller and the top skin. This was estimated experimentally by putting carbon paper between the roller and the specimen. For all the specimens tested was between 2 and 3 mm,

Longitudinal Direction

Prepreg Fibre Direction Honeycomb Ribbon Direction

Transverse Direction

Figure 2.7: Layup details

Chapter 2. Failure Mode Maps for Honeycomb Sandwich Beams

28

with a typical value of 2.5 mm. This information was used to estimate the failure load due to indentation as described in section 2.2.2.

Figure 2.8: Test setup

2.4.1 Experimental Results

Figures 2.9(a-d) show photographs of the section under the central load just after failure, illustrating four failure modes observed during the tests of those beams with 3 mm honeycomb cell size. These were taken from videos recorded during the tests. Beam details for each of these gures are given in Table 2.3 to scale the photographs note that the cell size is 3 mm for all the specimens of Fig. 2.9, except Fig. 2.11(a,b) which has the larger cell width of 13 mm. For clarity the roller seen in the top half of the gures has been outlined in Fig. 2.9(a). Corresponding load{de ection curves are included in Fig. 2.9. With all the failure mechanisms except intracell buckling (Fig. 2.11a,b) and core shear (Fig. 2.9d), failure occurred in a brittle manner with little non-linearity in the load displacement curves before failure and a sharp drop in load at failure. For short span tests where indentation occurred (Fig. 2.9c), the beam has a signi cant post-failure strength as the core progressively crushes. Table 2.3 presents the average5 line loads Wo at failure and corresponding observed failure mechanisms. The mode is described as `Complex' when failure appeared to occur simultaneously in the core and in the face. Those mechanisms denoted as `mode 1 mode 2' (i.e. Intracell dimpling core crushing) indicates the appearance of mode 1, as elastic instability before the nal failure mode 2. The measured e ect of skin thickness to span ratio on the failure load is compared with theory in Figs. 2.10(a)-(d) for the di erent core densities. The theoretical graphs are sections through the load surface, Fig. 2.5(b), at constant core relative density. Experimental mechanisms of failure are indicated on these graphs to allow a comparison between the actual and predicted failure mechanisms.
5

Most of tests were repeated and the di erence between failure loads was less than 10%

Chapter 2. Failure Mode Maps for Honeycomb Sandwich Beams

Core Honeycomb Span Density ribbon length (kg/m3) direction (mm) 500 400 230 150 50 500 400 230 150 50 530 380 220 130 60 530 380 220 130 60 530 380 220 130 60 530 380 220 130 60 530 430 345 260 170 95 60 490 430 345 260 170 95 60 500 230 150 50 500 230 150 50 29 (3mm cell) 48 (3mm cell) 64 (3mm cell) 128 (3mm cell) 64 (13mm cell) Trans. Long. Trans. Long. Trans. Long. Trans. Long. Trans. Long.

29

t=L
7.6E-04 9.5E-04 1.7E-03 2.5E-03 7.6E-03 7.6E-04 9.5E-04 1.7E-03 2.5E-03 7.6E-03 7.2E-04 1.0E-03 1.7E-03 2.9E-03 6.4E-03 7.2E-04 1.0E-03 1.7E-03 2.9E-03 6.4E-03 7.2E-04 1.0E-03 1.7E-03 2.9E-03 6.4E-03 7.2E-04 1.0E-03 1.7E-03 2.9E-03 6.4E-03 7.2E-04 8.9E-04 1.1E-03 1.5E-03 2.2E-03 4.0E-03 6.4E-03 7.8E-04 8.9E-04 1.1E-03 1.5E-03 2.2E-03 4.0E-03 6.4E-03 7.6E-04 1.7E-03 2.5E-03 7.6E-03 7.6E-04 1.7E-03 2.5E-03 7.6E-03

Table 2.3: Experimental results. Photographs of failure given in Fig. 2.9 and 2.11 correspond to the entries in italics in the nal column of this table

Measured line fail.load (kN/m) 3.8 4.2 5.1 5.6 6.4 3.0 3.4 4.1 4.8 6.0 6.2 7.5 9.3 11.0 11.9 5.3 6.1 7.8 9.2 10.8 8.2 10.2 13.8 15.6 17.2 6.6 8.0 10.4 12.3 14.5 11.6 13.2 17.0 23.2 31.3 42.1 47.5 9.4 10.9 14.7 17.9 26.5 38.1 44.4 4.0 8.9 11.2 21.1 4.2 9.1 11.3 20.8

Observed failure modes Face wrinkling Face wrinkling Core shear face wrink. Core shear crushing Core shear crushing Face wrinkling Face wrinkling Core shear face wrink. Core shear crushing Core shear crushing Face wrinkling Face wrinkling Complex Core crushing Core crushing Face wrinkling Face wrinkling Complex Core crushing Core crushing Face yield Face yield+wrinkling Complex Core crushing Core crushing Face yield Face yield+wrinkling Complex Core crushing Core crushing Face yield Face yield Face yield Face yield Face yield Complex Core crushing Face yield Face yield Face yield Face yield Face yield Complex Core crushing Intracell dimpl. face peel. Intracell dimpl. face peel. Intracell dimpl. core crush. Intracell dimpl. core crush. Intracell dimpl. face peel. Intracell dimpl. face peel. Intracell dimpl. core crush. Intracell dimpl. core crush.

Chapter 2. Failure Mode Maps for Honeycomb Sandwich Beams

20

30

Core Density = 128 kg/m Span Length = 345 mm

L in e L oad, W (k N /m )

15

10

H oneycom b D irection

L o n g itu d in a l T ra n s ve rs e

0 0 10 20 30

M id sp an D eflection (m m )

(a) Face yield ( c = 128 kg/m3 , L = 340 mm)

8

Core Density = 48 kg/m Span Length = 380 mm L in e L oad , W (k N /m )

6

H o neyco m b D irectio n

L o n g itu d in a l T ra ns ve rse

0 0 5 10 15 20

M idsp an D eflection (m m )

(b) Face wrinkling ( c = 48 kg/m3 , L = 380 mm)

60
H oneycom b D irection

L o n g itu d in a l

L in e L oad, W (k N /m )

T ra n s ve rse

40

20

3 Core Density = 128 kg/m Span Length = 60 mm

0 0 2 4 6

M idspan D eflection (m m )

(c) Core crushing ( c = 128 kg/m3 , L = 60 mm)

8

3 Core Density = 29 kg/m Span Length = 50 mm

L in e L oa d , W (k N /m )

H on e yco m b D ire ctio n

L o ngitu d in al T ra nsverse

0 0 1 2 3

M id sp a n D eflection (m m )

Figure 2.9:

(d) Core crushing ( c = 128 kg/m3 , L = 340 mm) Photographs of the di erent failure modes (for transverse ribbon direction) and corresponding load de ection curves (for both ribbon directions)

Chapter 2. Failure Mode Maps for Honeycomb Sandwich Beams

31

100

Core Density = 128 kg/m

Long

Core Density = 64 kg/m

L in e L oa d , Wo (k N /m )

L in e L oad , Wo (k N /m )

Long

T rans

10
E xp e rim e n ta l D a ta Long T ra n s F a ce Y ie ld

10
E x p e rim e n ta l D a ta
L o n g T ra n s

C o m p le x
T h e o re tic a l C u rve s

T ran s
T h e o re tica l C u rv e s F a c e Y ie ld F a c e W rin klin g

F a c e Y ie ld C o m p le x F a c e W rin k lin g C o re C ru s h in g

C o re C ru s h in g

F a ce Y ie ld C ore C ru sh ing

C o re C ru sh in g

1 0 .0 0 1 0 .0 1 0

0 .0 0 1

0 .0 1 0

F ace th ick n ess / sp an , t/ L

F ace th ick n ess / sp an , t/ L

(a) c= s = 0:160
10

(b) c= s = 0:080

3 Core Density = 48 kg/m

10
Long

3 Core Density = 29 kg/m

Long

L in e L oad , Wo (k N /m )

E xp erim e nta l D ata

L in e L oad , Wo (k N /m )

L on g

T ran s
F ace W rinkling C om plex

T ra n s
T h e ore tica l C u rve s F ac e W rinklin g C o re S h ea r C o re C ru sh ing
Long

T ra n s

E x p e rim e n ta l D a ta
T ra n s

T h eo retica l C u rves

C ore C rushing

F a c e W rin k lin g C o re C ru s h in g

F a ce W rin kling C om ple x C ore S he ar + C ru shing

1 0 .0 1 0 0 .0 0 1

0 .0 0 1

0 .0 1 0

F ace th ick n ess / sp an , t/ L

F ace th ick n ess / sp an , t/ L

(c) c= s = 0:066

(d) c= s = 0:040

Figure 2.10: Variation of failure load Wo with skin thickness to span ratio t=L

Chapter 2. Failure Mode Maps for Honeycomb Sandwich Beams

32

2.5 Discussion
Figure 2.10 shows that, in general, the experimental failure loads agree satisfactorily with theoretical values and that the observed failure modes are generally the same as the predicted modes (Fig. 2.5). Where a transition from one failure mode to another occurs, the mechanism of failure is mixed, but the failure load is still adequately predicted. The main source of deviations arises from the errors in predicting the honeycomb material properties (section 2.2.3).

2.5.1 Skin Failure

The laminate face yielding strength has been chosen to t data at long spans, as discussed in section 2.4. At the longest spans with the 64 kg/m3 honeycomb, face yield occurs although wrinkling is predicted, but as shown in Fig. 2.5(a), these points are very close to the boundary between these two mechanisms.

2.5.2 Core Failure

Failure load predictions for core crushing are adequate for the higher density cores but predictions are poor for the lower density cores. The photograph in Figure 2.9(d) illustrates a beam with a low density core in transverse direction near the failure load, where core shear failure is predicted. Final failure is a mixture of core shear and crushing under the indenter. The di erence in the initial slopes of the load{de ection curves6 for the specimens with di erent ribbon directions re ects the anisotropy associated with the core shear sti ness. Non-linearity in the curve for the transverse direction indicates that there may be some ductile shear of the core prior to nal failure. The signi cant under-prediction of the failure load Fig. 2.10(d)] due to core shear is because nal failure does not occur when the core fails in shear. Predictions of failure based on the core crushing strength would give a relatively good agreement with measurements in this case. Again this highlights the need for more elaborate models of core failure near loading points.

2.5.3 E ect of Ribbon Direction

Specimens with ribbon running in the longitudinal direction fail at a signi cantly higher load than those with ribbon in the transverse direction in all cases except where failure
6

The kinks in the load de ection curves at a low load are an artefact of the measuring system.

Chapter 2. Failure Mode Maps for Honeycomb Sandwich Beams

33

is by intracell dimpling (section 2.5.4). For face failure modes, a slight di erence is predicted, associated with the e ect of core shear sti ness on the stress in the skins, although the measured di erences are somewhat greater than the prediction. The theoretical model described for core crushing does not include any anisotropic material properties. The substantial di erence between the strengths for the two ribbon orientations suggest that an improved model is necessary, looking in more detail at the stresses around an indenter and including the e ect of core shear.

2.5.4 Intra-cell Buckling

Section 2.2.1 de nes a critical honeycomb cell size above which failure occurs by intracell buckling instead of by top face yielding. For the materials used here, this transition is predicted at a cell size of approximately 9 mm. This hypothesis was tested by using specimens with a cell size 13 mm, a core density of 64 kg/m3 and a range of skin thickness to span ratios. All of the specimens with the larger cell size exhibited elastic intracell buckling. Final failure did not occur immediately, but at a higher load, either by delamination of the top-face or by core crushing for long or short spans respectively. Fig. 2.11(a) shows the section under the central roller at the onset of intra-cell buckling, while Fig. 2.11(b) shows the section when nal failure occurs by delamination. Figure 2.11(c) compares the failure load and failure mechanisms for this set of tests. The theory equation (2.10)] under-predicts the nal failure load because it relates to the initial buckling instability and not the nal delamination or core failure. Failure loads are independent of the honeycomb ribbon direction, because the single and double-walls paper walls make up a smaller proportion of the honeycomb for a given honeycomb density.

2.6 Conclusions
Previous research on honeycomb mechanics and the behaviour of sandwich beams in 3-point bending have been combined to model the behaviour of honeycomb sandwich panels. It is assumed that the skin and core materials behave in a brittle manner. The failure mechanisms considered were face yield, face wrinkling, intra-cell buckling, core shear and indentation at the load points, leading to core crushing. The latter is treated in an empirical way, using measurements of the bearing area at the load points. The failure loads for each region are estimated assuming that there is no coupling between failure mechanisms. Following the work of Trianta llou and Gibson 41], failure mode maps are derived with axes as the core relative density and the ratio of the face thickness

Chapter 2. Failure Mode Maps for Honeycomb Sandwich Beams

34

to span length. Each map is appropriate for a single value of the ratio of the radius of the roller to the skin thickness. Regions in which the various failure mechanisms occur are identi ed on the maps. Alternative axes for the map of core relative density and the core thickness are suggested, as a more useful set of parameters for a beam designer or manufacturer. Although the maps are generated for three-point bending, the method can straightforwardly be applied to other loading geometries, for example four-point bending. The method is illustrated using the widely used combination of cross-ply GFRP laminate skins with a Nomex/phenolic resin honeycomb core. Experimental tests showed that, in general, the maps predicted adequately failure modes and failure loads. The transition from face yielding to intra-cell buckling for long span beams was demonstrated by increasing the honeycomb cell size from 3 to 13 mm. Failure near the load points due either to core shear or core indentation was not modelled well. To model this behaviour more sophisticated models of the contact region and of the failure criteria in the core are needed. Chapters 3 and 4 show how the implementation of the high-order sandwich beam theory of Frostig 46] provides a better insight into the mechanism of localised e ects induced by concentrated loads and helps to produce an improved indentation failure analysis for honeycomb sandwich beams.

Chapter 2. Failure Mode Maps for Honeycomb Sandwich Beams

35

12
Ho neycom b D irection

L ong itudina l T ran sverse

L in e L o a d , W (k N /m )

(a) (b)

Core Density = 64 kg/m Cell size = 13 mm Span Length = 230 mm

0 0 4 8 12

M idspa n D eflection (m m )

(a) Elastic intra-cell buckling (b) Final failure due to top skin delamination
100
E xp e rim e nta l D a ta L on g T ran s Intra cell D im p l. + F a ce P ee ling Intra cell D im p l. + C o re C rush in g

L in e L oad , Wo (k N /m )

10

T he ore tical C urves In trac ell B u cklin g C ore C ru shing

1 0 .0 0 1

Core Density = 64 kg/m Cell Size = 13 mm

0 .0 1 0

F ace th ick n ess / sp an , t/ L

(c) Experimental results compared with theory

Figure 2.11: Photographs and results of specimens with 13 mm cell size

Chapter 3 Indentation Resistance of Sandwich Beams

3.1 Introduction
The experiments described in the previous chapter revealed the important role of the exural rigidity of the skins and the transverse shear sti ness of the core on the overall resistance of a sandwich beam to indentation. In the literature the importance of the core behaviour in a ecting indentation failure has been considered however, the in uence of the skin's exural rigidity is generally overlooked. This omission derives from the fact that in practical applications the skins are quite sti and the in uence of skin exural rigidity can be neglected. One of the main aims of this chapter is to focus on the role of skin rigidity in indentation failure. Before introducing any analytical methods, it is instructive to consider two extreme cases, as illustrated in Fig. 3.1. For very exible skins (Fig. 3.1a), there is large local deformation under the load which can easily lead to core failure. For very rigid skins (Fig. 3.1b), indentation failure will be relatively hard, as the skins spread the load. Skin exibility is a prerequisite for indentation failure and the core's transverse exibility enhances this susceptibility. In this chapter the high-order sandwich beam theory (HOSBT) is used to analyse the bending behaviour of sandwich beams under localised loads and extract conclusions about their indentation resistance by examining the exibility of the top skin to the external applied loads. The ability of HOSBT to cope with the localised e ects that appear in sandwich beams at the load points is experimentally veri ed by Surface Displacement Analysis in section 3.3. Section 3.4 introduces a length scale that characterises the susceptibility of a sandwich beam to localised e ects. 36

Chapter 3. Indentation Resistance of Sandwich Beams

37

In section 3.5 high-order sandwich beam theory is used to predict the contact stress distribution between a cylindrical indenter and the top skin of a sandwich beam.
(a) low flexural rigidity of skins

(b)

high flexural rigidity of skins

Figure 3.1: The behaviour of exible and rigid skins

3.2 High-Order Sandwich Beam Theory

The mathematical formulation of high-order sandwich beam theory (HOSBT) is introduced by Frostig and Baruch 46]. `High-order' refers to the non-linear way in which the in-plane and vertical displacements are allowed to vary through the height of the core. The core vertical displacement is assumed to have a quadratic variation with height z. Other core displacements also vary in a non-linear way, with the exact variation being expressed in terms of Fourier series. This contrasts with simple beam theory where the core in-plane displacements are assumed to vary in a linear way through the depth, and the out-of-plane displacements are assumed to be constant. These high-order variations allows modelling of the more complicated changes in core geometry which occur at loading points. Thus the basic assumptions of HOSBT are: The shear stresses in the core are uniform through the height of the core. The core vertical displacement variation is a quadratic polynomial in z see equation (3.11)], allowing the core to distort and its height to change. The core is considered as a 3D elastic medium, which has out-of-plane1 compressive and shear rigidity, whereas its resistance to in-plane2 normal shear stresses is negligible.
`Out-of-plane' refers to 3 direction as illustrated in Fig. 1.3 or the direction normal to the plane of the sandwich panel 2 `In-plane' refers to 1-2 plane as illustrated in Fig. 1.3 or the plane of the sandwich panel
1

Chapter 3. Indentation Resistance of Sandwich Beams

38

This high-order analysis is based on variational principles the details of derivation of the governing equations and associated boundary conditions are presented in ref. 46]. In this study we examine the bending behaviour and the localised e ects of sandwich beams formulating a two dimensional model of a sandwich beam with unit width and span L, which consists of a core with thickness c, Young's and shear modulus Ec and Gc respectively, and two skins with the same thickness t = tt = tb , Young's modulus Ef and Poisson's ratio f , as depicted in Fig. 3.2(b). An external distributed load qt is applied on the top skin. The displacement and stress elds of the core are expressed in terms of the followings ve unknowns: the in-plane deformations ut and ub in the x-direction of the centroid of the top and bottom skin respectively their corresponding vertical displacements wt and wb and the shear stresses x in the core. The relevant geometric parameters and the notation of stresses and displacements appear in Fig. 3.2. The governing equations (see equations (37), (39) and (41)-(43) in ref 46]) are given below: At ut x + x = 0 Ab ub x ; x = 0 (3.1) Dt wt x ; Ecc (wb ; wt ) ; c+t x x = qt 2 Ec (w ; w ) ; c+t Dbwb x + c b t 2 x x = 0 3 ut ; ub ; c+t wt x ; c+t wb x ; Gcc x ; 12cEc x x = 0 2 2 2 2 where At = Ab = Ef t=(1 ; f ) and Dt = Db = Ef t3= 12(1 ; f )] are the in-plane and exural rigidity of the skins. The notation () x for example denotes the 4th partial derivative with respect to x. For a simply supported sandwich beam the solution can be expressed as a Fourier series expressing the variation of the relevant variables in the x direction as M X ut ut(x) = Cm cos mL x (3.2) m=1 M X ub ub(x) = Cm cos mL x (3.3) m=1 M X wt m x wt(x) = Cm sin L (3.4) m=1 M X wb m x wb(x) = Cm sin L (3.5) m=1 M X x m x Cm cos L (3.6) x (x) = m=1 ut ub wt wb where M = number of terms in the series and Cm Cm Cm Cm and Cmx are constants to be determined. We assume that the external loads are exerted only on the top skin.

Chapter 3. Indentation Resistance of Sandwich Beams

39

wt ut deformed top skin wc deformed cross section deformed bottom skin d

ub z x

wb

(a)
W
top skin

qt

txx
tt

zz(z=0)
core

x bxx

bottom skin

x L

tb
unit width

(b)
Figure 3.2: (a) Nonlinear displacement patterns, (b) Beam geometry and stresses. The
origin of the z-coordinate is always taken at the top of the beam element, either skin or core, which is being considered

Chapter 3. Indentation Resistance of Sandwich Beams

So in terms of the Fourier series we set

40

qt(x) =

M X m=1

qt Cm sin mL x

(3.7)

qt where Cm is a constant that depends on the distribution of the external load. After substituting every term of the Fourier series (3.2)-(3.6) in the governing equations (3.1), the problem can be expressed in matrix form as
D ] C ] = Q]

(3.8) 1 ;1

where

2 0 0 0 ;At ( m )2 6 0 L ;A ( m )2 6 0 0 b L 6 m )4 + Ec 6 0 D=6 0 ;Dt ( L ; Ec c c 6 0 Ec m )4 + Ec 4 0 ;c ;Db ( L c

1
;1 ; c+t m 2 L ; c+t m 2 L

c+t m 2 L c+t m 2 L c + c3 ( m Gc 12Ec L

)2

3 7 7 7 7 7 7 5

2 ut 3 C 6 Cm 7 ub 6 m7 6 wt 7 C = 6Cm 7 and 6 7 6C wb7 4 m5 Cmx

2 3 0 607 6 7 6 qt7 Q = 6 Cm 7 6 7 607 4 5

0

By solving (3.8) with respect to matrix C for every m = 1 2 : : : M the Fourier coe cients in equations (3.2)-(3.6) can be determined. These equations can then be used to calculate all the in-plane and vertical displacements and core shear stresses. The in-plane normal stresses txx in the top skin3 , and the out-of-plane normal stresses zz in the top skin-core interface are calculated indirectly as = Ef ut x c zz = x x + (wb ; wt ) 2 c
txx Ec

(3.9) (3.10)

3.3 Surface Displacement Analysis

An experimental veri cation of high-order sandwich beam theory (HOSBT) has been undertaken by Thomsen and Frostig 71]. They conducted a series of photoelastic
3

Only top skin compressive failure is considered

Chapter 3. Indentation Resistance of Sandwich Beams

41

experiments, compared the results with analytical calculations and demonstrated a good agreement between the photoelastic measurements and the predictions based on the HOSBT. In this section a similar comparative study is conducted with respect to the displacement eld in the core. The objective is to compare the analytical and experimental displacement distributions in sandwich beams with a `soft' core as, for example the Nomex cores used in Chapter 2. Special emphasis is given to the local out-of-plane compression of the core under the localised loads, as this will be important in determining failure of the core. For the analytical predictions, the equations described in section 3.2 are implemented in a Matlab code to calculate the vertical displacement eld within the core by the following equation (given in ref. 46])
2 + wc(x z) = ; x x ;z2E cz + (wb ; wt) z + wt c

(3.11)

The experimental determination of vertical displacements in the core is achieved by Surface Displacement Analysis. SDA is a Windows based software developed by Instron Corporation, which detects and maps horizontal and vertical displacement of points on the surface of a specimen or structure. There are two main parts to the procedure image capture and image analysis. SDA captures images of the surface of a material or structure using a video camera and a frame grabber board. The surface requires a random speckle pattern, either natural or applied, which has to be illuminated to ensure good contrast. Image frames of the surface of the specimen are captured at intervals while it deforms under load. Then SDA compares and analyses pairs of these image frames using a designated area of interest, as depicted in Fig. 3.3. Each area of interest contains one or more analysis cells. During analysis, the pattern in each cell of the current frame is compared with the pattern in the corresponding cell of the reference frame. The software maps the deformation of the surface as a series of vectors displayed on the image, one vector for each cell. The software also calculates the strain for each cell. SDA contains a graphics module that plots displacement and strain. If the software is calibrated to the dimensions of the image, absolute displacement values can be obtained. In this study we tested sandwich beams cut out from the panels described in section 2.4. Testing was under 3-point bending at a constant displacement rate of 0.5 mm/min. The beams' width was 30mm, the span 60 mm and the rollers' diameter 20 mm. The experimental setup is shown in Fig. 3.4. The clip gauge measures the midspan core compression during loading, while the video camera captures images of the other cross section of the beam. To apply a random speckle pattern of that cross

Chapter 3. Indentation Resistance of Sandwich Beams

42

Figure 3.3: Surface Displacement Analysis software window, showing the reference image

frame of a side cross section (painted to have a random speckle pattern) and the `area of interest' where the analysis takes place

clip gauge beam camera

measured core compression

Figure 3.4: Experimental setup

Chapter 3. Indentation Resistance of Sandwich Beams

43

section, at rst it was painted black and then ne-sprayed with white paint to produce a pattern of random white dots on a black background (see the image in Fig. 3.3). The three dimensional surface of the cross section, arising from the honeycomb structure and the subsequent buckling of the cell walls, causes errors in the image processing and mapping of displacements using SDA. However in the initial stage of loading (with little buckling) the results are satisfactory and illustrative of the localised e ects that occur under the central load. For a sandwich beam with Nomex honeycomb core with density of 48 kg/m3 the load de ection curve is shown in Fig. 3.5. The same gure also shows the clip gauge measurements of midspan core compression against top skin de ection. For Surface Displacement Analysis two frames are chosen the rst reference frame, just before loading (point A in Fig. 3.5), and the second when the applied line load reaches the arbitrary value of 6 kN/m, labelled B in Fig. 3.5. The SDA graphics module, after calibration, plots contours of equal vertical displacements within the core in mm, as depicted in Fig. 3.6(a). It should be noted that the camera was xed relative to the central roller so that the specimen appears to move upwards with zero displacements at the roller. The SDA measurements demonstrate the occurrence of localised de ections under the applied load due to the core local compression. The contour of 0.045 mm close to the bottom skin indicates the total compression of the core in the midspan of the sandwich beam. This value agrees well with that measured by the clip gauge (see the dashed line in Fig. 3.5). Fig. 3.6(b) shows a theoretical plot at the corresponding load of 6 kN/m, using equation (3.11). These calculations use the material properties of the skin and the honeycomb and the geometric attributes of the sandwich beam (Table 2.2). The total applied load is taken as 6 kN/m and applied uniformly over an area of 2 mm long4. A comparison between the theory and experiments shown in Fig. 3.6 show that HOSBT provides an analytical tool that is capable of predicting the localised e ects due to concentrated loads both qualitatively and quantitatively.

3.4 E ect of Spreading Stresses

After demonstrating the applicability of high-order sandwich beam theory, in this section we use it to investigate the mechanism that controls the behaviour of a sandwich beam under localised loads. One of the key factors that governs the behaviour of a sandwich beam under localised loads is the shape of the normal stress distribution
4

The reason for this is explained in section 3.5

Chapter 3. Indentation Resistance of Sandwich Beams

44

12

L in e loa d , W (k N /m )

4
S p eed = 0 .5 m m /m in C ore D en sity = 4 8 k g /m S p an L en gth = 6 0 m m 3

0 0 .3

M id -sp an core com p ression (m m )

0 .2

0.1

B
0 .0 0 .0

A
0 .5 1 .0 1 .5 2 .0

T op sk in d eflection (m m )

Figure 3.5: Load-de ection and core compression curves. A and B correspond to the frames
used for processing by SDA. The dashed line indicates the total midspan core compression at a line load of 6 kN/m

Chapter 3. Indentation Resistance of Sandwich Beams

45

(a)

Vertical coordinate, z (mm)

7 0.035

0.15

0.045

0.04

0.06

4 0.1 0.15

0.1

24

26

Horizontal coordinate, x (mm)

(b)

28

30

32

34

36

Figure 3.6: Vertical displacement eld in the core produced by (a) SDA and (b) HOSBT
model. All contour values are in mm and the scaling in both plots is the same

Chapter 3. Indentation Resistance of Sandwich Beams

46

along the beam in the top skin-core interface (see Fig. 3.2). Two extreme cases can be considered a very rigid top skin that spreads the external stress in the interface, or a `membrane' skin allows the applied stresses to pass intact through the interface into the core. In this section we extract a dimensionless quantity which characterises the behaviour of a sandwich beam under 3-point bending, independently of the applied central load. We do this by de ning an index which measures the `transparency' of the top skin - i.e. to what extent the top skin-core interface `feels' the stresses exerted on the external surface of the top skin. For this purpose we rewrite the matrices D and Q introduced in equation (3.8) as

2 2 3 0 A 607 6 0f 6 7 6 6 qt 7 6 Q = 6Cm 7 and D = 6 0 6 7 6 607 60 4 5 4

0

0 Df Ec Ct 7 where >Ec = ; Ecc 7 > > > 5 0 Ec Df Ct 7 > > > >Ct = c+t m 1 ;1 ;Ct ;Ct Gc > > 2 L > >

Af

0 0

0 0

1 7 ;1 7 7

8 >Af = ;At ( m )2 > > L > > > > > > >D = ;D ( m )4 + Ec > f > t L > c > > > > <

3 + 12cEc ( m )2 L Using the Symbolic Toolbox of Matlab equation (3.8) is solved with respect to the matrix C 2 3 2 ut 3 ; R1CtR2 C + 6 7 6 Cm 7 Ct ub 7 6 7 6 m7 R1+R2 7 6 wt 7 qt 6 7 6 C ] = Q] D] 1 =) 6 Cm 7 = Cm 6 (Df EcR1 1 +R2 ) 7 (3.12) 6 7 )(R 6 7 6C wb5 6; (D E R2R +R ) 7 4 m 4 f c)( 1 2 5 Af Ct x Cm R1+R2
; ; ;

> > > > >G = : c

c Gc

where R1 = Df (Af Gc ; 2) + Af (Ct)2 and R2 = Ec(Af Gc ; 2) + Af (Ct)2 Substituting equations (3.4)-(3.6) and the above Fourier coe cients into (3.10) the normal stresses in the top skin-core interface are given as
zz =

cm 2 L m=1 M X cm = ; 2L m=1
;

M X

wb wt Cmx + Ec (Cm ; Cm ) sin mL x c Af Ct + Ec C qt sin m x R1 + R2 Df ; Ec m L

(3.13)

Chapter 3. Indentation Resistance of Sandwich Beams

47

Equation (3.13) shows that the Fourier coe cients for the stress zz can be separated qt into two parts one part containing the coe cient Cm (units of stress), which depends on the distribution of the load, and a second part represented by the dimensionless `transmission coe cient' Cmzz , which is entirely dependent on the geometric attributes and material properties of the sandwich beam, where Cmzz is given by m f Ct Cmzz = ; c2L RA+ R + D Ec E 1 2 f; c (3.14) Af (Ct)2 + Ec ; R1 + R2 Df ; Ec if t c For any geometry and material combination this transmission coe cient is a discrete function of the parameter L=m, which is the semi-wavelength of every term in the Fourier series. The value of Cmzz varies from 0 (small L=m) to 1 (large L=m). 0 implies no transmission of this Fourier component of applied stress while 1 implies total transmission through the top skin. Thus we conduct a parametric study, in which we calculate the variation of Cmzz with L=m for the range of material and geometric parameters given in Table 3.1. Fig. 3.7 plots this variation. Each sub-plot shows curves for variations in the skin thickness t and core depth c. Changing from one sub-plot to the next represents a change in either skin sti ness Ef or core density c, as indicated by the arrows to the left and top of the gure.

Parameter Low Value t, (mm) 0.381 Ef , (GPa) 1 48 c , (kg/m3) c, (mm) 2.5

Note: The comb with
c

! ! ! ! !

High Value 4 0.381 100 1 3 48 4 2.5

refers to the density of Nomex honey3 s = 724 kg/m

Table 3.1: The range of the geometric and material parameters used in the parametric study
To characterise the transparency of the skin, it is helpful to identify a semi-wavelength L=m below which the values of the coe cient Cmzz are close to zero so that these wavelengths cannot be transmitted. Examining the curves in Fig. 3.7 it can be seen that the curves all have similar shapes, and that the in ection point in each curve can be used to characterise where Cmzz is small. The semi-wavelength at the in ection point characterises the susceptibility of sandwich beams to localised e ects, and this semi-wavelength is de ned as the spreading length . In general the larger is, the broader is the distribution of the normal out-of-plane stresses in the top skin-core interface, and vice versa.

Chapter 3. Indentation Resistance of Sandwich Beams

48

Ef = 1 GPa
1 1

Ef = 100 1 GPa

Coe cient, Cmzz

0.8 0.6 0.4 0.2 0

= 48 kg/m3

t = 0:381 mm t = 4 0:381 mm c = 2:5 mm c = 4 2:5 mm

0 5

Coe cient, Cmzz

0.8 0.6 0.4 0.2 0

c = 2:5 mm c = 4 2:5 mm t = 0:381 mm t = 4 0:381 mm

L=m (mm)
t = 0:381 mm

10

15

20

L=m (mm)

10

15

20

Coe cient, Cmzz

0.8 0.6 0.4 0.2 0

Coe cient, Cmzz

= 3 48 kg/m3

0.8 0.6 0.4 0.2 0

c = 2:5 mm c = 4 2:5 mm t = 0:381 mm t = 4 0:381 mm

t = 4 0:381 mm c = 2:5 mm c = 4 2:5 mm

0 5

L=m (mm)

10

15

20

L=m (mm)

10

15

20

Figure 3.7: Parametric Study change of transmission coe cient Cmzz with wavelength L=m
for variations in the parameters t, Ef , c and c . The boxed labels show each time the parameters and the arrows indicate the direction of each parameter's increase. The position of in ection is marked by a symbol

Chapter 3. Indentation Resistance of Sandwich Beams

49

Spreading length is a function of the skin thickness t, skin Young's modulus Ef , core density c and core thickness c, but it is independent of span length L. Table 3.2 shows the in uence of each parameter to spreading length . This table shows the in uence of changing the various parameters from the low values given in the left hand column of Table 3.1. Table 3.2 shows that the skin thickness t is the most signi cant factor as expected the larger the skin thickness the stronger the e ect of `spreading stresses'. A similar e ect is shown with increasing sti ness Ef although, given the big change in sti ness of 100, this e ect is less important. Core density also a ects the sti er is the core, the less the external stresses spread. Core thickness c seems to play a small role.

Table 3.2: Change of spreading length

Low values from Ta- t ! 4t c ! 4c Ef ble 3.1 for t c Ef c
! 100Ef

c!3 c

1.6

4.4

2.1

5.0

1.2

At this point, it is important to note that is not directly related to the strength of the sandwich beam. Instead it measures how the top skin is able to spread the external load over the core. is property of the beam material and geometry. In the following sections we show how it can be normalised and provide a dimensionless index to characterise the spreading e ect in sandwich beams under indentation loads.

3.5 Contact Pressure Distribution

Having used the high-order sandwich beam theory to characterise the way in which contact loads are transmitted through the top skin, in this section we investigate how this methodology can be applied to the contact between an indenter and a sandwich beam. In the literature there are only a few references relating to the contact stresses exerted between the top skin of a simply supported sandwich beam and an indenter which applies 3-point bending on the beam. Frostig & Baruch 45] examine analytically the e ect of four types of localised load distributions a point load, uniform, sinusoidal distributions or two concentrated loads. These represent very exible, intermediate and very sti indenters respectively. Johnson 72] (p.143) investigates thin plates in contact with a rigid indenter and concludes that the contact stress distribution changes from having a maximum in the centre to one in which the pressure is concentrated at the edges, when the thickness of the plate increases. A di erent problem is examined by Keer and Ballarini 73], who investigate the problem of a rigid punch in smooth

Chapter 3. Indentation Resistance of Sandwich Beams

50

contact with an initially stressed transversely isotropic elastic beam. In this study we investigate how sandwich beams behave under contact with a rigid central roller, calculating the pressure distribution under the indenter and through the beam. Fig. 3.8 shows a rigid cylinder of radius R which is pressed into contact with a sandwich beam of unit width and length L, such that the contact width is . Equation (3.4) can be used to calculate the top skin displacements for a given distribution of the external load. Following standard contact methods 72] we can divide the contact into a number of discrete sections, calculate the displacements resulting from a unit load on each of these sections, and superimpose the results to nd de ections for any contact load. In particular, for a given pressure distribution qt (x) it is possible to nd the corresponding distribution of heights Pi (Fig. 3.8a,b) along the top skin. Conversely, we shall show that for a given roller geometry, we can evaluate the contact width and contact pressure by inverting this approach. The shape of discrete pressure elements can be step functions (Fig. 3.8a) or overlapping triangles (Fig. 3.8b). The second choice is better because the total distributed load function is smoother along the x direction. In addition the Fourier series converges more quickly for a triangle shape rather than for a rectangular one. The pressure elements on the top skin can be described using the series

qi = Piqi = Pi ^

M X

where Pi is the height of the triangular distribution and q Cmi = ; m8 L2v sin( mL xi ) ;1 + cos( m Lv ) 2 2 are the Fourier coe cients for an isosceles triangle of unit magnitude. Each unit length has coordinate xi and base width v. qi is the distribution of pressure qi with ^ unit amplitude. For this and subsequent symbols, the `hat' refers to variables relating to this unit pressure distribution. Using the procedure described in section 3.2, the vertical displacements wti(x) of the top skin associated with each of the triangular pressure elements are given by equation (3.4)

m=1

q Cmi sin mL x

(3.15)

wti(x) = Pi

M X

If N is the number of pairs of symmetrical triangular pressure elements covering a contact width , then the total overall top skin de ection is given by

m=1

w (Cm t )i sin mL x = Piwti ^

(3.16)

wt(x) =

N X n=1

wti(x) =

N X n=1

Piwti ^

(3.17)

Chapter 3. Indentation Resistance of Sandwich Beams

51

(a)

Pi v/2 (b) top skin surface

Pi
Contact width

xi
End of beam

x=L/2

(c)

Figure 3.8: Discrete contact pressure elements (a) uniform (piecewise constant), (b) overlapping triangles (piecewise linear) and (c) Geometric de nitions

Chapter 3. Indentation Resistance of Sandwich Beams

52

In the case where load is applied by a rigid indenter, which is in perfect contact with the top skin, the top skin de ection follows the shape of the indenter within the contact area. In this section we examine the case of a cylindrical indenter, for comparison with our experiments, but the calculations can be applied to any indenter shape. Having a cylinder of R radius as a central indenter and assuming perfect contact with the top skin within contact of width , then the top skin de ection for all x L ; 2 L + 2 ] is 2 2

s wtR(x) = R2 ; x ; L 2 s = R2 ; x ; L 2 s = R2 ; x ; L 2

; R + wt(L=2)

N M X X wt ;R+ Pi (Cm )i sin m 2

;R+

n=1 N X n=1

m=1

(3.18)

Piwti(L=2) ^

where the superscript R denotes the de ection at the roller. We can also write the equality, only for x L ; 2 L + 2 ], as 2 2

wt(x) = wtR(x) s N X i giving Pi(wt(x) ; wti(L=2) ) = R2 ; x ; L ^ ^ 2 n=1

;R

(3.19)

If we write the above equality for N di erent values of xj L ; 2 L ], which represent 2 2 di erent values of wt (xj ) = wtR(xj ) along the left half5 of the contact width , then we get a system of N equations with N unknowns which are the distributions of pressure Pi. This system can be written in a matrix form
P ] W ] = R] P

where

(3.20)

= P1 P2

PN

2 1 3 wt(x1 ) ; wt1(L=2) wt1(x2 ) ; wt1(L=2) ^ ^ ^ ^ wt1(xN ) ; wt1(L=2) ^ ^ 6w2 ; w2 6 ^t(x1 ) ^t(L=2) wt2(x2 ) ; wt2(L=2) ^ ^ wt2(xN ) ; wt2(L=2) 7 ^ ^ 7 6 7 W=6 ... ... ... 7 ... 5 4 wtNx1 ) ; wtNL=2) wtNx2 ) ; wtNL=2) ^( ^( ^( ^( wtNxN ) ; wtNL=2) ^( ^(
R=
5

hq q 2 ; (x ; L )2 ; R R R2 ; (x2 ; L )2 ; R 1 2 2

; (xN ; L )2 ; R 2

Or right half, because of symmetry

Chapter 3. Indentation Resistance of Sandwich Beams

53

Substitution of Pi into equation (3.15) and summation for all N triangular pressure elements gives us the contact pressure distribution

qt (x) =

N X i=1

qi

(3.21)
zz .

The same concept of superposition can be used to calculate the other stresses, i.e. The total load is derived by integration along the x direction

W=

N M X X qi Z L m x qt (x) = Pi Cm sin L ) 0 i=1 m=1 N M X X q W = 2L Pi mCmi for m = 1 3 : : : M i=1 m=1

(3.22)

In principle this approach can be used to determine the variation in load distribution and the contact arc as the total load W increases. In practice a problem arises at small . As it is necessary to span the whole beam length with elements, the number of elements needed to deal with sharp variations in pressure becomes prohibitively large. This method is used in the following section to answer the following questions. 1. How accurate is it to assume that the load is distributed uniformly over the contact length ? 2. How `transparent' is the top skin to the external loads. In other words, how does the distribution of normal stress zz (x) in the top skin-core interface compare with the contact stress distribution qt (x) ?

3.5.1 Case Study

In this section the results from the aforementioned methodology for two extreme cases of skin's exural rigidity are presented to illustrate the e ect of indenter's size on the contact stress distribution and the top skin's transparency to external loads. We consider one sandwich beam (denoted as beam A) with very exible skins, i.e. 1-ply 90] GFRP laminates with t = 0:191 mm and Ef = 1 GPa, which gives = 0:8 mm, and another (beam B) with very rigid skins, i.e. 2-ply 0,90] laminates with t = 0:381 mm and Ef = 20 GPa, which corresponds to = 3:7 mm. Both beams have unit depth, are 60 mm long, have Nomex honeycomb with core density of 128 kg/m3, are simply supported and loaded in 3-point bending by a central roller. For each beam two radii for the central roller are considered R = 5 mm and R = 10 mm. The methodology described in the previous section requires the contact width as an input data. To

Chapter 3. Indentation Resistance of Sandwich Beams

54

compare results for each roller, is adjusted manually in a simple numerical algorithm so that, for each beam, the maximum normal stresses in the top skin-core interface are approximately equal. Fig. 3.9 and Fig. 3.10 show the distributions of contact stresses qt = cc and the corresponding normal stresses in the top skin-core interface zz = cc (normalised by the core's out-of-plane compressive strength) for beams A and B, respectively. For beam B the input value of is chosen to give max zz = cc 1. However, for beam A this is not feasible. The much greater exibility of the beam means that it is not possible to generate stresses in the core greater than half the compressive strength of the core without the beam bending excessively. Further loading, by increasing the value of , induces large skin deformations, which also make HOSBT incapable to model the behaviour of the sandwich beam. Thus for beam A the input value of is chosen to give max zz = cc 0:4. However, this di erence in peak stress between the two beams is immaterial, as the calculations are elastic and the stress distributions are independent of the applied load. A particular feature of the contact pressure distribution for the exible beam A, with R = 5 mm, and for the sti er beam B with R = 10 mm is the sharp peaks in pressure at the edges of the contact. This feature is also noted by Johnson 72] for similar contacts. These peaks, however, are not transmitted through the skins. To characterise the degree to which the skins are able to spread the load, we need to compare the width of the contact patch and the width over which there are signi cant pressures transmitted across the interface. This latter width can be characterised by a length , given when the pressures have positive non-zero values. Results show that, for the beam B with a rigid skin, Fig. 3.10, the pressure width of the pressure distribution in the interface is considerably more than the contact width , and does not change signi cantly with roller diameter or contact width. The skin is su ciently rigid to spread the local pressure distribution out. For beam A with a more exible skin, Fig. 3.9, the width of the pressure distribution in the interface corresponds to that of the contact. The ratio = can be used to measure how well the contact load has been spread out. For a very exible skin this will equal 1, while for a very rigid skin this will tend to a large value which will depend on the roller radius. By considering a range of skin sti nesses changing the thickness t or/and Young's modulus Ef , this method can be used to nd the variation of = for a range of beam spreading lengths , for the two roller diameters 5 and 10 mm. This is illustrated in Fig. 3.11. This gure shows that, for very exible skins with small , the width of the contact equals the width of the region in which there are signi cant pressures across

Chapter 3. Indentation Resistance of Sandwich Beams

55

1 Load distribution 0.9 0.8 0.7

R = 5mm R = 10mm

Normalised, qt =

cc
0.6 0.5 0.4 0.3 0.2 0.1 0 1 Normal stresses in the interface

Normalised,

zz = cc

0.8 0.6 0.4 0.2 0 0.1 0.05

0 0.05 0.1

(x ; L=2)=L

Figure 3.9: Distributions of contact stresses qt= cc and the corresponding normal stresses

in the top skin-core interface zz = cc (normalised by the core's out-of-plane compressive strength) for beam A with = 0:8 mm

Chapter 3. Indentation Resistance of Sandwich Beams

56

1 0.9 0.8 0.7

Load distribution

Normalised, qt =

cc
0.6 0.5 0.4 0.3 0.2 0.1 0 1 Normal stresses in the interface

R = 5mm R = 10mm

Normalised,

zz = cc

0.8 0.6 0.4 0.2 0 0.1 0.05

0 0.05 0.1

(x ; L=2)=L

Figure 3.10: Distributions of contact stresses qt= cc and the corresponding normal stresses
in the top skin-core interface zz = cc (normalised by the core's out-of-plane compressive strength) for beam B with = 3:7 mm

Chapter 3. Indentation Resistance of Sandwich Beams

57

the interface, with = = 1 as expected. As the spreading length increases, the distribution of pressure in the interface broadens with respect to the contact width, with larger = . This change depends on roller diameter. If we take the transition from exible to rigid behaviour when = equals 2, then this occurs for =R equal to 0.24 and 0.25 for the 5 and 10 mm rollers respectively. This case study demonstrates the usefulness of the spreading length in characterising the beam exibility. It also allows us to answer the questions posed in section 3.5, relating to the assumptions commonly used to model the contact. Firstly, for practical beams with large , the distribution of pressure over the contact is not important, and the assumption that the external load is applied uniformly across the beam will be reasonable. However, this contact distribution will not be transmitted to the interface and the core unchanged. By contrast, Fig. 3.10 shows that it is essential to model beams with sti skins accurately in order to get a reasonable estimate of the stress distribution in the core. The commonly used assumption that the failure can be predicted using (Failure line load) = (Core's out-of-plane compressive strength, cc) (Contact width, ) will be poor, unless a judicous choice of is made, based on experimental measurements. In the nal section 3.6, we summarise the conclusions from this chapter, including guidelines of a more accurate failure analysis.

3.6 Concluding Remarks

The bene ts of using the high-order sandwich beam theory (HOSBT) to analyse the behaviour of sandwich beams under indentation are presented. The capabilities of HOBST are veri ed experimentally by Surface Displacement Analysis. The symbolic manipulation of the governing equations of HOSBT is used to extract a spreading length . This is a property of a sandwich beam, depending mainly on the skin's exural sti ness and characterising the susceptibility of the sandwich beam to indentation loads. A further manipulation of HOSBT gives an insight into the contact mechanics for loading of a beam by a cylindrical indenter. The way in which the contact pressure is transmitted through the core is examined. A case study shows the fact that sandwich beams used as standard in industry have skins which are rigid enough to spread the external loads. The most important conclusion from this section is that the maximum stresses zz , which are responsible for indentation failure, cannot be predicted in a straightforward way for sandwich beams with rigid skins, even if the contact width is known. For

Chapter 3. Indentation Resistance of Sandwich Beams

58

20

R = 5 mm R = 10 mm
15

10

Flexible skins

Rigid skins

(mm)
Figure 3.11: Dependence of spreading e ect on roller's radius R and determination on how
exible or rigid are the skins with respect to the indentation resistance of the sandwich beam

Chapter 3. Indentation Resistance of Sandwich Beams

59

sandwich beams with very exible skins, the approximation of dividing the total line load W by can give us reliable estimation of the failure stresses in the skin-core interface. However for commercially applied sandwich con gurations the former case is the dominant. Thus, in practice one should follow the following steps for indentation failure analysis of sandwich beams: 1. For a particular indenter's size and shape do one test to measure the just before failure (since develops with the increase of W ). This value for can be assumed for other beam geometries. 2. Use the approximation that the external stresses are uniform within this measured length. For sandwich beams with small this approximation is good, while for sandwich beams with large the exact distribution of contact load does not a ect the core stresses signi cantly. 3. Use high order sandwich beam theory and apply the equations (3.8), (3.9) and (3.10) to calculate the stress eld in the core. Compare them with the maximum allowables to predict failure. This approach is adopted in the next chapter to undertake an indentation failure analysis.

Chapter 4 Indentation Failure Analysis

4.1 Introduction
If we assume that the behaviour of a sandwich beam is elastic up to failure, the highorder sandwich beam theory can be used to calculate the stresses in the core due to the indentation loading. This can then be compared with the core's out-of-plane compressive strength. However, a more accurate failure criterion for the core is determined from biaxial tests on Nomex honeycomb cores and used in this analysis. Finally short beam 3-point bending tests with three di erent central roller diameters verify the theoretical predictions and show the importance of the skin exural rigidity on the beam's overall strength.

4.2 Failure Envelope for Nomex Honeycombs

A straightforward way to calculate the failure load by indentation for honeycomb cores is by equating the maximum of the out-of-plane normal stresses zz with the allowable stresses cc. However the experiments in Chapter 2 showed that the interaction of core shear and out-of-plane stresses was important. The concept of the indentation failure due to these combined stress components is depicted in Fig 4.1. Therefore, before performing an indentation analysis, we investigate both theoretically and experimentally in this section the two-dimensional failure envelope of Nomex honeycombs under shear and out-of-plane compression loading and we de ne a new failure criterion. References to biaxial loading tests for cellular materials are not common in the literature. Gibson et al. 74] have performed a series of tests on a variety of foams under biaxial, axisymmetric and hydrostatic loading conditions. Zhang and Ashby 28] have investigated the in-plane biaxial buckling behaviour of Nomex honeycombs. More 60

Chapter 4. Indentation Failure Analysis

9 > > > > > > > > > > > = > = > > > > > > > > > >

61

Figure 4.1: Combined indentation failure mechanism

recently Stronge and Klintworth 75] have studied the yield surfaces for honeycombs under biaxial macroscopic stresses. In this chapter Nomex honeycombs are tested under combined shear and compressive out-of-plane loading. For this purpose, we used the Arcan test rig as illustrated in Fig. 4.2. This geometry was adapted for shear testing of composite materials by Arcan et al. 76] and extended to biaxial testing of butter y-shaped specimens by Voloshin and Arcan 77]. The rig consists of two pairs of plane circular S-shaped parts with antisymmetric cutouts. To accommodate honeycomb specimens we added two loading platens. The bottom pair of semicircular xtures are gripped to the frame of a servohydraulic testing machine. The upper pair are mounted to the load cell and actuator transferring the load to the specimen, which is attached between the two rectangular plates. To ensure a uniform stress through the specimen, the grips are clamped rather than being pivoted, since Marlo 's nite element analysis 78] shows that stress uniformity under biaxial conditions is not good. The perimetric holes in the semicircular plates allow a change of the angle ' between the plane of the specimen and the loading direction and therefore the ratio (tan ') of compression to shear applied to the specimen between the central plates. The specimens are rectangular plates of no more than 45 mm in width or length. The success of such tests depends on being able to attach the specimen to the loading platens e ectively. In this case the at external surface of the skins can be easily attached with standard acrylic adhesive. Specimens were prepared by cutting rectangular pieces from each of the available sandwich panels of 4 di erent densities as described in Chapter 2. Although the aim of this work is to examine the behaviour of the honeycomb core, the presence of the skins does not pose any di culties on the contrary they restrict movement of the honeycomb walls relative to the grips and help attachment of the specimen to the rig. For each core density, tests were conducted for

Chapter 4. Indentation Failure Analysis

62

Section A-A A Fitted to crosshead

P
200mm 10mm 15mm

Grip Pin Loading platen Specimen

50mm

Loading platens

Specimen

Fitted to frame

P
A

Figure 4.2: The Arcan-type rig

Chapter 4. Indentation Failure Analysis

63

both orientations of the honeycomb ribbon (longitudinal and transverse). Loading was monotonic up to failure, with a constant displacement rate of 0.3 mm/min. In order to cover a wide range from nearly pure compression to nearly pure shear, tests were made at three angles of ' = 21:5o 51:5o and 81:5o. Fig. 4.3 shows the corresponding load paths. Failure for each of the shear angles ' de nes one point on the failure surface, with corresponding values of the core normal and shear stresses zz and x at failure. In Fig. 4.4 photos of the three angle setups and typical load-de ection curves are presented. The de ection is measured from the crosshead displacement of the machine.
Pure compressive strength, cc Load path

Compressive stresses, zz

Failure surface

zz x

= 81.5

= 51.5 = 21.5
Pure shear strength, cs

Shear stresses, x

Figure 4.3: Load paths and determination of failure envelope

The failure peak loads give the combination of normal compressive load (P sin ') and shear load (P cos ') which leads to failure. Assuming that these loads are distributed uniformly over the external surfaces (of area b L) of both skins, the compressive and shear stresses in the core are given by P sin '=(bL) and P cos '=(bL) respectively. The calculated failure stresses zz and x are normalised by the core's out-of-plane compressive strength cc and shear strength cs respectively, as given by equations (2.18), (2.19a) and (2.19b). Normalised stresses zz = cc are plotted against x= cs for every

Chapter 4. Indentation Failure Analysis

64

12

' = 81:5o

L oad (k N )

0 0.0

0.5

1.0

D eflection (m m )
8

' = 51:5o

L oad (k N )

0 0 .0

0 .5

1.0

D eflection (m m )
6

' = 21:5o

L oad (k N )
2 0 0 .0

0 .5

1.0

D eflection (m m )

Figure 4.4: The three angle setups and the corresponding load-de ection curves (here for
specimens with 128 kg/m3 core density)

Chapter 4. Indentation Failure Analysis

65

angle, core density and both honeycomb ribbon directions in Fig. 4.5. Except for the honeycombs with density 128 kg/m3 the failure envelopes are well approximated by a linear failure criterion given by
zz cc

x cs

=1

(4.1)

as illustrated in Fig. 4.5 by the dashed line. The inconsistency observed in the 128 kg/m3 honeycombs is due to an inconsistency in the measured shear stresses, and does not a ect the accuracy of our calculations. Thus equation (4.1) allows us to de ne a failure criterion for combined loading cases, as required for an indentation failure analysis.

4.3 Failure Analysis with HOSBT

In section 3.2 the basics of high-order sandwich beam theory's implementation to sandwich beam bending behaviour analysis were presented. Here the equations needed for a failure analysis for honeycomb sandwich beams loaded under 3-point bending are assembled. To simplify our calculations we assume that the external loads (applied by the central roller) are uniform within a given width in the midspan of the beam. This assumption is discussed in section 3.6. So for a uniformly distributed load applied in qt the midspan of the beam the Fourier coe cient Cm c.f. eqn. (3.7)] reads 4 qt qt ^ qt Cm = mW sin( m ) sin( mL ) or Cm = W Cm (4.2) 2 2 where W is the total line load. For the failure analysis, we need to know the stress eld in the core-top skin interface i.e. the out-of-plane normal stresses zz and x qt (see Fig. 3.2b). Substitution of the above coe cients Cm into equation (3.12) gives the wt wb Fourier coe cients Cm , Cm , and Cmx for the top and bottom skin de ections and for core shear stresses

R1 f ; Ec )(R1 + R2 ) R2 wb qt Cm = Cm (D ; E;)(R + R ) f c 1 2 f Ct qt Cmx = Cm RA+ R 1 2

wt qt Cm = Cm (D

Chapter 4. Indentation Failure Analysis

66

1.0

1.0

R el. C om p ressiv e S tren gth , z z / cc

0.8

R el. C om p ressiv e S tren gth , z z / cc

0.8

0.6

0.6

0.4
3

0.4

C ore de nsity = 29 kg/m

0.2
Long Tran s

0.2

C o re density = 48 kg/m Lon g T rans

0.0 0.0

0 .2

0.4

0.6

0.8

1 .0

0.0 0.0

0.2

0.4

0.6

0.8

1.0

1 .2

R el. S h e a r S tren gth , x / cs

1.0
1.0

R el. S h e a r S tren gth , x / cs

R el. C om p ressiv e S tren gth , z z / cc

R el. C om p ressive S tren g th , z z / cc

C ore density = 128 kg/m

0.8

0.8

L on g T rans

0.6

0.6

0.4
3

0.4

C ore de nsity = 64 kg/m

0.2
Long Trans

0.2

0.0 0.0

0 .2

0.4

0.6

0.8

1 .0

0.0 0.2

0.4

0 .6

0.8

1.0

1.2

1 .4

R el. S h e a r S tren gth , x / cs

R el. S h ea r S tren g th , x / cs

Figure 4.5: Failure envelopes for Nomex honeycombs. The dashed line corresponds to the
linear failure criterion given by equation (4.1)

Chapter 4. Indentation Failure Analysis

M X ^ qt m Cm D Ec E ; c2L zz (x) = W f; c m=1 M X ^ qt (x) = W Cm Af Ct cos m x x R +R L m=1

67 Af Ct sin m x W ^ (x) zz R1 + R2 L
W ^x(x)

Then from equations (3.4)-(3.6) and (3.10), the required stresses are given as follows (4.3) (4.4)

The variation of the stresses zz (x) and x (x) with position x along the beam and with load W can be represented in - space by lines such as that shown dashed in Fig. 4.6. For every W the lowest part of the curve corresponds to the edge of the beam (x = 0), where shear stresses are dominant, but still there are some out-of-plane normal stresses (hence the curve does not touch the x-axis). Towards the midspan of the beam, (i.e. the upper section of the curve) the out-of-plane normal stresses increase, while the shear stresses decrease (falling to zero at x = L=2). As W increases, the stresses increase until the failure envelope, shown by a solid straight line, is reached at a failure load Wo. This failure load will di er signi cantly from that based on pure compressive loading.

cc
x = L/2 (at centre of beam) Location of failure, xf [0, L/2]

Normal stresses, zz

Failure surface
Increasing x Increasing W

Locus of stresses in beam x = 0 (at end of beam)

cs Shear stresses, x

Figure 4.6: Combined failure criterion

Chapter 4. Indentation Failure Analysis

68

Wo can be calculated after substituting equations (4.4) and (4.3) into equation (4.1) and nding the minimum value of the following expression W = ^ (x) +cc cc ^ (x) zz cs x
(4.5)

4.4 Experimental Work

To validate the theoretical predictions for the indentation failure loads, 3-point bending tests were performed on short sandwich beams. Beams 70 mm long and 30 mm wide were cut in both honeycomb ribbon directions from Nomex core sandwich panels. Panels had nominal core densities 29, 48, 64 and 128 kg/m3 with a 3 mm cell size and 32 kg/m3 with a 5 mm cell size. To observe in detail the vertical de ections during loading we painted one of the specimens' long cross sections so that the surface of the honeycomb walls became black and the edges of both skins and of honeycomb walls became white. The experimental setup is depicted in Fig. 4.7. It consists of a video camera and recorder, which keep track of the high contrast black and white cross sections of the specimens during loading. The specimens are simply supported by two rollers of 20 mm diameter. The distance between them is 60 mm. 3-point bending tests were performed with 3 di erent diameters (6, 10 and 20 mm) of the central roller in a servohydraulic testing machine. The loading rate was 0.5 mm/min. As well as the midspan top skin de ection and the load measurements, which are given by the machine's transducers, we attached in the midspan of every specimen a clip gauge to measure the relative displacement between the skins.

clip gauge beam camera

measured core compression

Figure 4.7: Experimental setup

The experimental measurements of midspan core compression from the clip gauge, of load from the load cell and of top skin de ection from the crosshead displacement

Chapter 4. Indentation Failure Analysis

69

transducer are plotted for every central roller size and every core density in Figs. A.1, A.2 and A.3 in Appendix A. A representative combined plot is shown here in Fig. 4.8. The bottom skin de ection vs top skin de ection curves show clearly the di erent response of the two skins the top skins exhibit larger absolute de ections than the bottom skins. This is due to the compression of the core as shown clearly by the core compression vs top skin de ection curves. Core compression is almost linear up to failure. For all core densities the core compression is more severe for the beams with longitudinal honeycomb ribbon direction, due to the higher shear sti ness of the Nomex honeycomb in the longitudinal direction than in the transverse direction. This di erence con rms the importance of the core shear sti ness in indentation failure. The di erence between longitudinal and transverse ribbon directions is also re ected in the line load-de ection curves in the former case the beam has a higher overall exural rigidity. The letters A, B, C, and D in the upper plot of Fig. 4.8 indicate critical stages of the loading procedure (i.e. before loading, just before failure, just after failure and well after failure). At these points video snapshots were captured. These are presented for every central roller size and core density in Figs. A.4, A.5 and A.6 in Appendix A. Here we present some key photos for discussion. Fig. 4.9 shows two video images captured just before failure (position B in Fig. 4.8) during 3-point bending loading with a 6mm diameter central roller. It is apparent how the low density core is deformed with signi cant shear, causing the honeycomb cell walls to buckle elastically. Fig. 4.10 shows the e ect of core density after failure, when the total de ection is approximately 2 mm. The high out-of-plane sti ness of the high density core does not allow the damage to propagate within the core and the skins fail eventually by compressive macrobuckling. For the same reason the indented top skin follows tightly the curvature of the roller for high density cores, while for low density cores the top skin is indented over a wider area without following the indenter's curvature (see Fig. 4.11). After comparing the corresponding curves amongst Figs. A.1, A.2 and A.3 we realise that there are no signi cant di erences due to the central roller's size. This is in accordance with the prediction made in Chapter 3. This insensitivity to the indenter's size is con rmed by comparing the corresponding failure peak loads in Fig. 4.12. The same gure also shows the satisfactory theoretical predictions using the high-order sandwich beam theory in conjunction with the mixed failure criterion. The values of out-ofplane core compressive strength and core shear strength calculated in section 2.2.3 for every core density and ribbon direction are substituted in equation (4.5) and the minimum value of this provides the theoretical predictions of failure loads, which are represented by the lines in Fig. 4.12. The advantage of using the mixed failure crite-

Chapter 4. Indentation Failure Analysis

8

70

B
L in e loa d, W (k N /m )
6

C D

A
0 0.3

C ore com pression (m m ) B ottom sk in d eflection (m m )

0.2

0.1

0.0 2.0

1.5

1.0

0.5

0.0 0.0

0 .5

1 .0

1 .5

2 .0

T op sk in deflection (m m )

Figure 4.8: Typical experimental results for sandwich beam with 29 kg/m3 core density

loaded by a roller with diameter of 6 mm. Midspan bottom skin de ection, midspan core compression, and line load are plotted against midspan top skin de ection. Lines legend: (|) longitudinal and ({ {) transverse honeycomb ribbon direction

Chapter 4. Indentation Failure Analysis

29 kg/m3 128 kg/m3

71

Figure 4.9: The pre-failure in uence of core density. Video images captured just before
failure during 3-point bending loading with a 6mm diameter central roller

29 kg/m3

128 kg/m3

Figure 4.10: The post-failure in uence of core density. Video images captured after failure
(2 mm total de ection) during 3-point bending loading with a 10mm diameter central roller

29 kg/m3

64 kg/m3

Figure 4.11: The post-failure in uence of core density on extent of damage. Video images

captured after failure (2 mm total de ection) during 3-point bending loading with a 10mm diameter central roller

Chapter 4. Indentation Failure Analysis

72

rion instead of simply checking when the maximum out-of-plane stresses in the top skin-core interface reach the out-of-plane compressive strength of the honeycomb core is illustrated in Fig. 4.12. It is shown that, without the correction of the mixed criterion, the theoretical predictions are larger than the experimental data measurements of the failure loads, especially as the core density increases. Also the use of the mixed failure criterion `catches' the di erent indentation resistance of sandwich beams due to the di erent honeycomb ribbon direction. The sandwich beams with longitudinal honeycomb ribbon direction have slightly higher strengths than those with a transverse ribbon direction. The di erence increases with the core density.

4.5 Concluding Remarks

A systematic approach has been developed to determine the failure load of sandwich honeycomb structures under indentation loading. Firstly the failure envelope for Nomex honeycombs under simultaneous out-of-plane compression and shear has been determined by biaxial tests using an Arcan rig. A linear dependence on pure compression and shear proves to be a good approximation for the failure envelope of Nomex honeycombs. The implementation of high-order sandwich beam theory allows a more accurate failure analysis for sandwich beams subjected to localised loads. Using the failure envelope determined by the biaxial tests, a mixed failure criterion has been introduced to predict the indentation failure, caused by the simultaneous action of out-of-plane compressive and shear stresses exerted in the vicinity of a localised load. This criterion can predict failure that ranges from pure core crushing to pure core shear. The short beam bending tests validated the theoretical predictions of high order beam model. Video captures of the deformed side cross section illustrate the involvement of core shear in the elastic behaviour of sandwich beams with low density cores. Also these tests showed the di erent post-failure damage extent between sandwich beams with di erent core densities. They showed that the mixed failure criterion offers a signi cant improvement in predicting indentation strength, as compared with models which do not include combined loading. This model is particularly needed in high core densities which induce more severe stress elds in the core. Moreover this approach explains observed di erences in behaviour for longitudinal and transverse ribbon directions.

Chapter 4. Indentation Failure Analysis

73

60
L on g T ra n s In d en ter d ia m eter 20 m m

L in e loa d , W o (k N /m )

10 m m

40

6 mm

L on gitu d in al

T ran sverse

20
F ailu re C riteria
O n ly co m p ression W ith m ixe d sh e a r a n d com p ression

0 0 40 80 120

C ore d en sity, (k g/m 3 )

Figure 4.12: Theory (lines) vs experimental results (symbols) for failure line load Wo

Chapter 5 Failure Maps using HOSBT

5.1 Introduction
In Chapter 2 failure mode maps for Nomex honeycomb sandwich beams with GFRP laminate skins are produced using simpli ed beam models and honeycomb mechanics. However, the simpli ed model for indentation failure prediction used in chapter 2 can not be used with great con dence, especially where core crushing and shear interact. In the previous chapter a theoretical model based on high-order sandwich beam theory (HOSBT) and a mixed failure criterion for Nomex honeycombs has been shown to give accurate predictions of the peak loads when sandwich beams made of GFRP skins and Nomex honeycomb fail by indentation. Here the use of high-order beam theory is extended as a compact computational tool, which determines the maximum stresses in the skins and the core. Thus by comparing the calculated maximum stresses with the relevant allowable stresses, we nd the range of design parameters where each failure mode is dominant. Frostig and Shenhar 40] have conducted an analogous failure analysis for foam core sandwich panels by using the HOSBT, but for failure patterns and criteria di erent than those we present here. In particular, the failure mode of skin wrinkling is not examined. Furthermore, speci cally for the failure mode of indentation, the failure criterion for the core was based on the maximum principal stresses. Although this approach is appropriate for foams that are isotropic, in case of honeycombs, which are orthotropic, a more suitable failure criterion is that based on the normal and shear stresses exerted in the top skin-core interface. Thus in this chapter we make use of the mixed failure criterion, introduced in the previous chapter. Based on the above approach, new failure mode maps for Nomex honeycomb sandwich beams are constructed and provide the basis for design optimisation carpet plots.

74

Chapter 5. Failure Maps using HOSBT

75

5.2 Reconstruction of Failure Maps

In chapter 3 we presented the basics of high order beam theory's implementation to sandwich beam bending behaviour analysis. By assuming a linear elastic response for the sandwich beam up to failure, it is feasible to use the appropriate equations derived by high order beam theory to calculate the maximum stresses in the skins and the core under 3-point bending. The advantages of such approach are: (1) use of a compact computational tool to calculate all the maximum stresses in the sandwich beam, and (2) ability to use mixed failure criteria for the core to model more accurately the indentation failure, as shown in the previous chapter. The experimental results presented in Chapter 2 show that the short beams nally failed by core crushing under the central load, but with an increased interaction of core shear as the core density gets smaller. Thus in this chapter the core shear is not considered as a separate failure mode. The failure in the honeycomb core is treated as one mixed failure mode and denoted as core indentation. The theoretical determination of the peak load for this mode is presented thoroughly in the previous chapter and gives

Wo = min ^ (x) +cc cc ^ (x) zz cs x

(5.1)

These calculations are based on the assumption that the external loads (applied by the central roller) are uniform within a measured width in the midspan of the beam. As explained in Chapter 3, the rigidity of the laminate skins used in the examined sandwich beams is high enough that the calculations are insensitive to the contact area between the indenter and the top-skin. It is assumed in this chapter that this remains true. For the skin failure we need to know the maximum normal stresses txx exerted on the cross-section of the top skin. Equation (3.9), after substitutions, gives
M X ^ qt Ct m Cm R + R L sin mL x W ^txx (x) (5.2) txx (x) = W 1 2 m=1 This is a function of the x coordinate and the maximum value is reached (as expected) in the midspan of the beam, i.e. for x = L=2

max

txx

= W ^txx(x=L=2)

(5.3)

Now the failure loads for every failure mode are calculated by equating these maximum stresses with the allowable stresses fY , fw for skin compressive yield and skin wrinkling respectively.

Chapter 5. Failure Maps using HOSBT

76

As in Chapter 2, a failure map for this type of sandwich beam is constructed by nding the minimum failure loads for each of the failure modes in Table 5.1 for each value of the design parameters (i.e. t=L and c= s). Then we plot the minimum load

Table 5.1: Expressions for peak failure loads

Top skin yield Face wrinkling Indentation

Wo = Wo =

fY = ^txx fw = ^txx

Wo = min ^zz (x)+cccs ^x (x) cc

Wo against the skin thickness over span ratio t=L and the core relative density c= s. The projection of the lines where two modes intersect provides the boundaries for the corresponding failure mode map, as depicted in Fig. 5.1. By comparing this map with those given in Fig. 2.5, one can see that the failure modes of pure core crushing or shear are now replaced by a core indentation region. The upper part of the indentation `area' (higher core densities) corresponds to core crushing under the vicinity of the localised loads, whereas the lower part (lower densities) is characterised mainly by core shear. A comparison between the failure loads calculated with HOSBT and the experimental results presented in Chapter 2 is depicted in Fig. 5.2 for all four examined core densities. The theoretical predictions for core crushing are more accurate for the new failure prediction, especially for the lower core densities of 29 or 48 kg/m3 (c.f. Fig. 2.10). Also this new approach predicts the observed di erences in behaviour for longitudinal and transverse honeycomb ribbon direction. Although the high-order sandwich beam theory provides a robust and compact computational tool for all the modes, there are two drawbacks rstly, when the skin fails, the model calculates the same failure load for the longitudinal and transverse direction of the honeycomb core and secondly the predictions for the mode of skin wrinkling are overpredicted, especially closer to the transition from skin wrinkling mode to indentation mode. The latter is due to fact that HOSBT allows for local bending of the top skin before failure and therefore the calculated in-plane normal stresses txx in the midspan are lower than those calculated by simpli ed beam models. Indeed the distribution of in-plane normal stresses txx along the beam is triangular when predicted by simple beam theory, while HOSBT gives a curve tangential to this triangular distribution close to midspan. This discrepancy causes the di erences in prediction.

Chapter 5. Failure Maps using HOSBT

77

10

Failure line load, Wo (N/m)

10

10

10

10 10

0 2

10 10
1

10

Relative density, c=

10

10

Face thickness / span, t=L

10

newmap2.m 54600

Skin yielding
26400

relative density Relative density, c=

s
10
1

6160 12700

Skin wrinkling
2980

Core indentation
695 336
4

1440

10

10

10 skin thickness / span

10

Face thickness / span, t=L

Figure 5.1: Improved failure map for Nomex honeycomb sandwich beams. Each contour
represents sandwich beams of equal strength in N/m

Chapter 5. Failure Maps using HOSBT

78

100

C o re D e nsity = 1 28 kg/m

C ore D e nsity = 6 4 kg /m

P eak L oad, W o (k N /m )

C o re In de nta tio n

P eak L oa d , W o (k N /m )

10

10

C ore In d en tatio n S kin Y ie ld or W rin kling

S kin Y ie ld

0.001

0.010

0.001

0.010

F ace th ick n ess / sp an , t/ L

F ace th ick n ess / sp an , t/ L

(a) c /s = 0.160
10

(b) c /s = 0.080

C ore D ensity = 48 kg/m

P ea k L oa d, W o (kN /m )

10

P eak L oad , W o (k N /m )

C ore Indentation

C ore Indentation S kin W rinkling

0 .00 1 0.010

S kin W rinkling C ore D ensity = 29 kg/m

1 0.00 1 0 .010

F ace th ickn ess / sp an , t/ L

F ace th ick n ess / span , t/ L

(c) c /s = 0.066

(d) c /s = 0.040

Figure 5.2: Comparison with the experimental results from Chapter 2. Lines show the pre-

dictions of HOSBT failure analysis and the symbols represent the experimental data (c.f. Fig. 2.10). Solid lines and symbols correspond to the longitudinal ribbon direction, while dashed lines and hollow symbols correspond to the transverse ribbon direction

Chapter 5. Failure Maps using HOSBT

79

5.3 Optimisation Carpet Plots

When designing a load-bearing structure the most important functional requirements are sti ness, strength and weight. In general these are attributes that we want to maximise or minimise. Minimum weight design is considered by Trianta llou and Gibson 33] for foam core sandwich beams and extended in Gibson's book 25]. In both cases, indentation is not examined as a possible core failure mode only core shear failure is considered. They conclude that, in minimum weight design of a foam core sandwich beam of a given strength, the top skin and core must fail simultaneously. In this work we use the failure maps to optimise the design of honeycomb sandwich beams with a graphical methodology. Here it is assumed that the failure in the core is by indentation and is a result of the simultaneous e ect of core shear and out-of-plane compression (Fig. 4.1). Fig. 2.6 shows a useful practical form of a failure map. In this section we consider a typical case study, seeking an optimum design for a sandwich beam which is loaded under 3-point bending and for which the material properties and skin thickness are given. The ratio c=L and the relative core density c= s are variables to be optimised. Hence it is most useful to present failure maps in terms of c=L and c= s. Fig. 5.3 presents the data for Fig. 5.1 in this form, showing contours of equal failure line load. By adding to this map contours of equal sti ness and mass, carpet plots are produced which can be used to nd the optimum design con guration. Such a carpet plot is shown in Fig. 5.5, where for clarity the mass contours and only one strength and one sti ness contour are shown. The full set of sti ness contours are plotted in Fig. 5.4. Two optimisation cases are considered strength limited design or sti ness limited design. These constraints are represented as hatched contours in Fig. 5.5. In addition a geometric constraint is included. In this case we suppose that the ratio of core thickness to span cannot exceed 5 10 2. Where a minimum strength constraint is applied the potential choices are restricted to the upper right area of the map. By considering the mass contours, it is clear that the design which minimises the mass lies at point A on the intersection of the skin wrinkling - core indentation region on the strength contour. However, if other considerations impose a value of c=L less than 2:4 10 2 (the value of c=L which corresponds to point A) this conclusion would no longer hold. It may also be desirable to choose a panel which fails by indentation because this is a safer failure mode than the alternatives. For the material combination considered, the optimum design point lies on the border between core indentation and skin failure for a range of strength constraints. This accords with the relevant conclusion of ref. 33]
; ;

Chapter 5. Failure Maps using HOSBT

10
0

80

strength contours

skin yielding
Relative density, c /s

10

-1

increasing strength

skin wrinkling core indentation

10

-2

10

-2

10

-1

Core thickness / span, c/L

Figure 5.3: Strength contour plot corresponding to Fig. 5.1

10
0

stiffness contours

skin yielding

Relative density, c /s

10

-1

increasing stiffness

skin wrinkling core indentation

10

-2

10

-2

10

-1

Core thickness / span, c/L

Figure 5.4: Sti ness contour plot

Chapter 5. Failure Maps using HOSBT

81

and 25] that the minimum design is such that the skin and core fail at the same peak load. The above illustration assumes that sti ness criteria are always satis ed. Consider instead a sti ness limited design with a sti ness constraint shown by the hatched contour in Fig. 5.5. Now, the contours of mass suggest that the beam with minimun mass would be that with the minimum available relative density of the core. In practice some other constraint, for example a geometric constraint, a strength constraint, or a manufacturing constraint, will limit the extent to which the core density can be reduced. Indeed this `sti ness-limited' case study illustrates the need to model strength of these panels, as this will be necessary to determine the optimum design.
10
0

skin yield

Relative density, c /s

decreasing mass

10

-1

skin wrinkling
mass contours 10
-2

core indentation

10

-2

geometric constraint 10
-1

stiffness constraint

Core thickness / span, c/L

Figure 5.5: Optimisation carpet plot for t=L = 5 10 4 with c= s and c=L as design
;

parameters

strength constraint

Chapter 5. Failure Maps using HOSBT

82

5.4 Concluding Remarks

The high-order sandwich beam theory can be used, not only to predict indentation failure loads, but also the failure modes of skin failure, enabling the construction of failure mode maps. This approach o ers the advantage of using only one computational tool to calculate the failure loads for every failure mode. The predictions follow satisfactorily the experimental data. However, the disadvantages of using the HOSBT for failure maps are overprediction of the failure loads for skin wrinkling and inability to predict the di erent behaviour between the two honeycomb ribbon directions when the top skin fails, either by yield or wrinkling. Finally it is shown that failure maps can help with the preliminary design of sandwich beams under bending, by superimposing contours of mass, sti ness and strength.

Chapter 6 Conclusions and Future Work

This thesis addresses theoretical and experimental aspects of the design of honeycomb sandwich panels. The research focuses on sandwich beams made of cross-ply GFRP laminate skins with a Nomex/phenolic resin honeycomb core, a widely used combination for aircraft ooring panels and freight loading pallets. Indentation resistance of such sandwich panels with transversely exible cores is an important factor in their design. An accurate model of the elastic deformation of such sandwich beams under three point bending has been derived. In particular, this model includes the local behaviour under the central load, allowing indentation failure to be examined.

6.1 Conclusions
Firstly the behaviour of sandwich beams of di ering lengths and core densities is considered, to examine the importance of the possible failure modes, including failure by skin yielding, skin wrinkling, intra-cell buckling, core shear and indentation. At this initial treatment indentation is treated in an empirical way, using measurements of the bearing area at the load points. Previous research on honeycomb mechanics and simple beam models has been combined to derive failure mode maps for 3-point bending with axes as the core relative density and the ratio of the skin thickness to span length. These maps are based on those of Trianta llou and Gibson 41], who focused on sandwich beams with ductile aluminium skin and isotropic foam cores. This thesis appears to be the rst attempt to construct maps for sandwich beams with laminate skins and honeycomb core. Since commercial panels are generally provided with standard skin thicknesses, but with di ering core thickness and density, alternative maps with these two core variables are presented, since these will be more useful for a beam designer or manufacturer. Although the maps are generated for three-point bending, the method can straightforwardly be applied to other loading geometries, for example four-point 83

Chapter 6. Conclusions and Future Work

84

bending. Experimental results for failure under three point bending are summarised as follows. A transition from face yielding to intra-cell buckling for long span beams was observed for a honeycomb cell size above a critical value. The measured peak loads were dependent on the direction of the honeycomb ribbon. This di erence is due to the shear anisotropy of the honeycomb, illustrating the important role that core shear plays in the bending behaviour of these sandwich beams. Experimental results veri ed satisfactorily the predicted failure loads. The boundary between skin and core failure on the failure mode maps was also predicted with good accuracy. However, failure near the load points due either to core shear or core indentation was not modelled well. A high-order sandwich beam theory (HOSBT) was implemented to provide a better deformation model of localised e ects under concentrated loads and so to produce an improved indentation failure analysis for honeycomb sandwich beams. The bene ts of using the high-order sandwich beam theory to analyse the behaviour of sandwich beams under indentation are presented and veri ed experimentally by measurements of the core deformation close to an indenter. HOSBT is used to extract a characteristic spreading length . This is a property of a sandwich beam, depending mainly on the skin's exural sti ness and characterising the susceptibility of the sandwich beam to indentation loads. Small values of correspond to sandwich beam with very exible skins which are transparent to the external loads. Large values of indicate rigid skins that restrict the transmission of localised loads to the core. HOSBT is further used to give an insight into the contact mechanics for a beam loaded by a cylindrical indenter. The way in which the contact pressure is transmitted through the core is examined. A case study shows the fact that sandwich beams used as standard in industry have skins which are rigid enough to spread the external loads. This spreading e ect allows to predict the indentation behaviour of a sandwich beam with rigid skins without having to model accurately the contact between the indenter and the top skin. The assumption of a uniform distribution over a roughly estimated width (which depends on the size of the indenter) and the use of the high-order sandwich beam theory can provide reliable predictions of the stress eld in the core. The most important conclusion from this analysis is that the maximum normal stresses in the top skin-core interface, which are mainly responsible for indentation failure, cannot be predicted in a straightforward way for sandwich beams with rigid skins, even if the contact width is known. For sandwich beams with very exible skins, the approximation of dividing the total line load W by can give us reliable estimation of the failure stresses in the skin-core interface. However for commercially applied

Chapter 6. Conclusions and Future Work

85

sandwich con gurations the skin is rather rigid and the exible skin approximation cannot be used. A systematic approach has been developed to determine the failure load of sandwich honeycomb structures under indentation loading. Biaxial tests of Nomex honeycombs using an Arcan rig show that a linear dependence on pure compression and shear proves to be a good approximation for the failure envelope of these honeycombs. Using this failure envelope and the core stress eld determined by HOSBT, a mixed failure criterion has been introduced to predict the indentation failure, caused by the simultaneous action of out-of-plane compressive and shear stresses exerted in the vicinity of a localised load. This criterion can predict failure that ranges from pure core crushing to pure core shear. Short beam bending tests validated the theoretical predictions of high order beam model and showed that the mixed failure criterion o ers a signi cant improvement in predicting indentation strength, as compared with models which do not include combined loading. This new approach is particularly needed in high core densities which induce more severe stress elds in the core. It also predicts observed di erences in behaviour for longitudinal and transverse ribbon directions. Video captures of the deformed side cross section of the tested sandwich beams illustrate the involvement of core shear in the elastic behaviour of sandwich beams with low density cores. Also they showed the di erent post-failure damage extent between sandwich beams with di erent core densities the higher the core density the less damage propagates into the core and skins can su er eventually macrobuckling failure. The high-order sandwich beam theory can also be used to predict the failure modes of skin failure, enabling the construction of failure mode maps. This approach o ers the advantage of using only one computational tool to calculate the failure loads for every failure mode. The predictions follow satisfactorily the experimental data. However, the disadvantages of using the HOSBT for failure maps are overprediction of the failure loads for skin wrinkling and inability to predict the di erent behaviour between the two honeycomb ribbon directions when the top skin fails, either by yield or wrinkling. Finally it is shown that failure maps can help with the preliminary design of sandwich beams under bending, by superimposing contours of mass, sti ness and strength.

6.2 Future Work

With this study a framework of failure analysis of Nomex honeycomb panels has been established and provides the basis for further investigation in the following areas.

Chapter 6. Conclusions and Future Work

86

Bending tests should be done with sandwich beams with di erent skin thicknesses and core densities than those tested in this research work, in order to validate further the applicability of the proposed methodologies for failure analysis and for failure map construction. The failure envelope for biaxial loading of Nomex honeycombs determined in section 4.2 should also be investigated for other core materials. In section 3.4 a parametric study is conducted to de ne the dependence of the spreading length on the material and geometric properties of a sandwich beam. The in uence of each parameter has been determined qualitatively. Further work on this area could attempt to derive a single normalised curve Cmzz with respect to L=m (instead of all the curves in Fig. 3.7) that accurately captures the indentation resistance behaviour of a wide range of sandwich beam designs. This approach would lead to simple formulae relating to the material and geometric properties of a sandwich beam. Miller 79] adopts such an approach to model the indentation behaviour of foamed metals. A nite element analysis of a model (see Fig 6.1) of a rigid indenter applying loads to a simply supported beam would be useful to determine the contact pressure between the indenter and the top skin, and the corresponding normal stresses transmitted to the core. These stresses can be compared with those calculated by the high-order beam theory model in section 3.5. In Chapter 3 we showed the ability of the spreading length parameter to characterise the static indentation resistance of sandwich beams. It should be possible to compare values with experimental results on dynamic indentation resistance and investigate the applicability of as a quality factor in sandwich panel manufacturing. For this purpose, the software code used in section 3.4 to calculate could be provided to industry for assessment of its practical usefulness. Finally the methodology for failure analysis proposed in this study should be used to explore the bene ts of unsymmetrical sandwich beams with bottom skins thinner than the top ones. This option would exploit rstly the fact that composite laminates have approximately tensile strength twice as much their compressive and secondly the higher indentation resistance that a thicker top skin can provide.

Chapter 6. Conclusions and Future Work

87

ABAQUS

2
DISPLACEMENT MAGNIFICATION FACTOR = 6.77 ORIGINAL MESH DISPLACED MESH

RESTART FILE = indent1

STEP 1

INCREMENT 1 .000E+00

TIME COMPLETED IN THIS STEP ABAQUS VERSION: 5.6-1

2.220E-16 TOTAL ACCUMULATED TIME TIME: 03:53:10

DATE: 17-APR-98

Figure 6.1: Vertical displacements calculated by Abaqus nite element analysis

Appendix A Experimental Results

88

Appendix A. Experimental Results

89

Figure A.1: For 6 mm diameter roller: midspan bottom skin de ection, midspan core com-

pression, and line load variation curves with respect to midspan top skin de ection. Lines legend: (|) longitudinal and ({ {) transverse honeycomb ribbon direction

Appendix A. Experimental Results

90

Figure A.2: For 10 mm diameter roller: midspan bottom skin de ection, midspan core

compression, and line load variation curves with respect to midspan top skin de ection. Lines legend: (|) longitudinal and ({ {) transverse honeycomb ribbon direction

Appendix A. Experimental Results

91

Figure A.3: For 20 mm diameter roller: midspan bottom skin de ection, midspan core

compression, and line load variation curves with respect to midspan top skin de ection. Lines legend: (|) longitudinal and ({ {) transverse honeycomb ribbon direction

Appendix A. Experimental Results

Figure A.4: Video snapshots during loading with a 6 mm diameter roller

92

Appendix A. Experimental Results

Figure A.5: Video snapshots during loading with a 10 mm diameter roller

93

Appendix A. Experimental Results

Figure A.6: Video snapshots during loading with a 20 mm diameter roller

94

Bibliography
1] J. E. Gordon, Structures: or, why things don't fall down, A Pelican Original, Penguin, Harmondsworth, 1978. 2] M. F. Ashby, Materials Selection in Mechanical Design, Pergamon Press, Oxford, 1983. 3] R. W. Birmingham and J. A. D. Wilcox, Charting the Links Between Material Selection and Elemental Form in Structural Design, Journal of Engineering Design, 4(2), 1993, pp. 127{140. 4] S. W. Tsai, Theory of Composite Design, Think Composites, 1992. 5] R. T. Haftka, Z. Gurdal and M. P. Kamat, Elements of Structural Optimization, Kluwer Academic Publishers, Boston, USA, 1990. 6] D. Hull, An Introduction to Composite Materials, Cambridge University Press, 1981. 7] X. J. Xin, M. P. F. Sutcliffe, N. A. Fleck and P. T. Curtis, Cambridge Composite Designer - User's Manual, Cambridge, 1995. 8] M. G. Bader, Materials Selection, Preliminary Design and Sizing for Composite Laminates, Composites Part A, 27A(1), 1996, pp. 65{70. 9] M. Miki, A Graphical Method for Designing Fibrous Laminated Composites with Required In-plane Sti ness, Trans. JSCM, 9(2), 1983, pp. 51{55. 10] S. W. Tsai and J. M. Patterson, Design Rules and Techniques for Composite Materials, Longman, 1990. 11] J. A. Quinn, Composites - Design Manual, James Quinn Associates Ltd, Liverpool, 1995. 12] J. A. Quinn, Composites Selection Methods for Engineers, in FRC 84, Liverpool. 13] J. A. Quinn, Cost and Weight Optimisation Decisions for GRP, in ICCM III, Paris, 1980. 95

Bibliography

96

14] R. Martin, Composite Structures - A Dual Approach to Design, Materials World, 3(7), 1995, pp. 320{322. 15] D. J. Farrar, The Design of Compression Structures for Minimum Weight, Journal of Royal Aeronautical Society, 1949, pp. 1041{1052. 16] A. Zahorski, E ect of Material Distribution on Strength of Panels, Journal of Aeronautical Sciences, 1944, pp. 247{253. 17] M. S. Anderson and W. J. Stroud, A General Panel Sizing Computer Code and its Applications to Composite Structural Panels, AIAA Journal, 17(8), 1979, pp. 892{897. 18] D. C. Jegley, Study of Structurally E cient Graphite - Thermoplastic Trapezoidal - Corrugation Sandwich and Semisandwich Panels, Journal of Aircraft, 31(2), 1994, pp. 411{418. 19] D. L. Graessar, Z. B. Zabinsky, M. E. Tuttle and G. I. Kim, Optimal Design of a Composite Structure, Composite Structures, 24, 1993, pp. 273{281. 20] M. B. Snell and P. Bartholomew, The Application of Geometric Programming to the Structural Design of Aircraft Wings, Aeronautical Journal, 86(857), 1982, pp. 259{268. 21] M. B. Snell and P. Bartholomew, Initial Design of Stringer Sti ened Bend Boxes Using Geometric Programming, Aeronautical Journal, 87(861), 1983, pp. 21{25. 22] M. B. Snell and P. Bartholomew, The Engineering Optimisation of Hybrid Composite/Metallic Wing Boxes for Buckling and Strength Constraints, Composite Structures, 7(1), 1987, pp. 21{58. 23] R. J. Duffin, E. L. Peterson and C. M. Zener, Geometric Programming, John Wiley, 1967. 24] K. F. Karlsson and B. T. Astrom, Manufacturing and Applications of Structural Sandwich Components, Composites Part A, 28A, 1997, pp. 97{111. 25] L. J. Gibson and M. F. Ashby, Cellular Solids: Structure and Properties, Pergamon Press, Oxford, 1988. 26] J. Zhang, The Mechanics of Foams and Honeycombs, Ph.D. thesis, Cambridge University Engineering Department, Cambridge, U.K., 1989. 27] J. Zhang and M. F. Ashby, The Out-of-plane Properties of Honeycombs, International Journal of Mechanical Science, 34(5), 1992, pp. 475{489.

Bibliography

97

28] J. Zhang and M. F. Ashby, Buckling of Honeycombs under In-plane Biaxial Stresses, International Journal of Mechanical Science, 34(6), 1992, pp. 491{509. 29] G. Shi and P. Tong, Equivalent Transverse Shear Sti ness of Honeycomb Cores, Int. Journal of Solids Structures, 32(10), 1995, pp. 1383{1393. 30] M. Grediac, A Finite Element Study of the Transverse Shear in Honeycomb Cores, Int. Journal of Solids Structures, 30(13), 1993, pp. 1777{1788. 31] H. G. Allen, Analysis and Design of Structural Sandwich Panels, Pergamon Press, London, 1969. 32] F. J. Plantema, Sandwich Construction, John Wiley & Sons, New York, 1966. 33] T. C. Triantafillou and L. J. Gibson, Minimum Weight Design of Foam Core Sandwich Panels for a Given Strength, Materials Science and Engineering, 95, 1987, pp. 55{62. 34] Y. W. Kwon, M. C. Murphy and V. Castelli, Buckling of Unbalanced Sandwich Panels with Titanium and GRP Skins, Journal of Pressure Vessel Technology, 117, 1995, pp. 40{44. 35] T. R. A. Pearce, The Stability of Simply{supported Sandwich Panels with Fiber Reinforced Face Plates, Ph.D. thesis, University of Bristol, Bristol, U.K., 1973. 36] D. Zankert, An Introduction to Sandwich Construction, Engineering Materials Advisory Services, 1997. 37] P. J. Holt and J. P. H. Webber, Exact Solutions to Some Honeycomb Beam, Plate, and Shell Problems, Journal of Strain Analysis, 17(1), 1982, pp. 1{8. 38] A. K. Mukhopadhyay and R. L. Sierakowski, On Sandwich Beams with Laminated Facings and Honeycomb Cores Subjected to Hygrothermal Loads: Part I { Analysis, Journal of Composite Materials, 24(4), 1990, pp. 382{400. 39] E. W. Kuenzi, Sandwich Sandwich Design Criteria, Tech. 2161, Forest Products Laboratory, Madison, WI, Oct. 1959. 40] Y. Shenhar, Y. Frostig and E. Altus, Stresses and Failure Patterns in the Bending of Sandwich Beams with Transversely Flexible Cores and Laminated Composite Skins, Composite Structures, 35, 1996, pp. 143{152. 41] T. C. Triantafillou and L. J. Gibson, Failure Mode Maps for Foam Core Sandwich Beams, Materials Science and Engineering, 95, 1987, pp. 37{53. 42] A. P. S. Selvadurai, Elastic Analysis of Soil - foundation Interaction, Developments in Geotechnical Engineering, 17, 1979.

Bibliography

98

43] O. T. Thomsen, Analysis of Local Bending E ects in Sandwich Plates with Orthotropic Face Layers Subjected to Localised Loading, Composite Structures, 25, 1993, pp. 511{520. 44] O. T. Thomsen, Theoretical and Experimental Investigation of Local Bending E ects in Sandwich Plates, Composite Structures, 30, 1995, pp. 85{101. 45] Y. Frostig and M. Baruch, Bending of Sandwich Panels with Transversely Flexible Core, AIAA Journal, 28(3), 1990, pp. 523{531. 46] Y. Frostig and M. Baruch, Localized Load E ects in High-order Bending of Sandwich Panels with Flexible Core, Journal of Engineering Mechanics, 122(11), 1996, pp. 1069{1076. 47] Y. Frostig, M. Baruch, O. Vilnay and I. Sheinman, High-Order Theory for Sandwich-Beam Behaviour with Transversely Flexible Core, Journal of Engineering Mechanics, ASCE, 118(5), 1992, pp. 1026{1043. 48] Y. Frostig, On Stress Concentration in the Bending of Sandwich Beams with Transversely Flexible Core, Composite Structures, 24, 1993, pp. 161{169. 49] E. Bozhevolnaya and Y. Frostig, Nonlinear Closed-Form High-Order Analysis of Curved Sandwich Panels, Composite Structures, 38(1-4), 1997, pp. 383{394. 50] Y. Frostig, Hygothermal (environmental) E ects in High-Order Bending of Sandwich Beams with a Flexible Core and a Discontinuous Skin, Composite Structures, 37, 1997, pp. 205{221. 51] Y. Frostig, Inaccuracies and Validity of Simpli ed Models in the Theory of Sandwich Structures, in Proc. of the 4th Int. Conf. on Sandwich Construction, vol. 1, Stockholm, 1998, (pp. 167{189). 52] K. B. Armstrong, Cost-E ective Design of Indentation Resistant Sandwich Panels for Aircraft Floors and Other Purposes, Tech. EEA.S.4.8644, BOAC, Dec. 1969. 53] Ciba Composites, Duxford, England, Honeycomb Sandwich Design Technology, August 1995. 54] G. Lubin, Handbook of Composites, Van Nostrand Reinhold Company, New York, 1982. 55] H. Razi, B. Sergeev, S. Shkarayev and E. Madenci, Analysis of Sandwich Panels with Multiple-Site Damage, in Proc. of the 4th Int. Conf. on Sandwich Construction, vol. 1, Stockholm, 1998, (pp. 213{226).

Bibliography

99

56] P. A. Lagace, J. E. Williamson, P. H. W. Tsang, E. Wolf and S. Thomas, A Preliminary Proposition for a Test Method to Measure (Impact) Damage Resistance, Journal of Reinforced Plastics and Composites, 12, 1993, pp. 584{601. 57] R. A. W. Mines, C. M. Worrall and A. G. Gibson, The Static and Impact Behaviour of Polymer Composite Sandwich Beams, Composites, 25(2), 1994, pp. 95{110. 58] R. A. W. Mines and N. Jones, Approximate Elastic-Plastic Analysis of the Static and Impact Behaviour of Polymer Composite Sandwich Beams, Composites, 26(12), 1995, pp. 803{814. 59] C. L. Wu and C. T. Sun, Low Velocity Impact Damage in Composite Sandwich Beams, Composite Structures, 34, 1996, pp. 21{27. 60] E. J. Herup and A. N. Palazotto, Low-Velocity Impact Damage Initiation in Graphite/Epoxy/Nomex Honeycomb-Sandwich Plates, Composites Science and Technology, 57(12), 1997, pp. 1581{1598. 61] P. D. Soden, Indentation of Composite Sandwich Beams, Journal of Strain Analysis for Engineering Design, 31(5), 1996, pp. 353{360. 62] F. M. Shuaeib and P. D. Soden, Indentation Failure of Composite Sandwich Beams, Composites Science and Technology, 57(9-10), 1997, pp. 1249{1259. 63] R. Olsson and H. L. McManus, Improved Theory for Contact Indentation of Sandwich Panels, AIAA Journal, 34(6), 1996, pp. 1238{1244. 64] E. W. Kuenzi, Edgewise Compression Strength of Panels and Flatwise Flexural Strength of Strips of Sandwich Construction, Tech. 1827, Forest Products Laboratory, Nov. 1951. 65] C. B. Norris and W. J. Kommers, Short-column Compressive Strength of Sandwich Constructions as A ected by the Size of the Cells of Honeycomb-core Materials, Tech. 1817, FPL, Aug. 1950. 66] T. Wierzbicki, Crushing Analysis of Metal Honeycombs, International Journal of Impact Engineering, 1(2), 1983, pp. 157. 67] K. Lingaiah and B. G. Suryanarayana, Strength and Sti ness of Sandwich Beams in Bending, Experimental Mechanics, 31(1), 1991, pp. 1{7. 68] THE MATHWORKS INC., MATLAB, The Language of Technical Computing, Using Matlab, 1996. 69] 20 Nov 1996, Personal communication from Hexcel Composites.

Bibliography

100

70] Anonymous, Method of Flexure Tests of Flat Sandwich Structures, ASTM, 15.03(C393-62), 1989. 71] O. T. Thomsen and Y. Frostig, Localised Bending E ects in Sandwich Panels: Photoelastic Investigation versus High-Order Sandwich Theory Results, Composite Structures, 37(1), 1997, pp. 97{108. 72] K. L. Johnson, Contact Mechanics, Cambridge University Press, Cambridge, 1985. 73] L. M. Keer and R. Ballarini, Smooth Contact Between a Rigid Indenter and an Initially Stressed Orthotropic Beam, AIAA Journal, 21(7), 1983, pp. 1035{ 1042. 74] L. J. Gibson, M. F. Ashby, J. Zhang and T. C. Triantafillou, Failure Surfaces for Cellular Materials under Multiaxial Loads | 2. Comparison of Models with Experiment, International Journal of Mechanical Science, 31(9), 1989, pp. 665{678. 75] W. J. Stronge and J. W. Klintworth, Biaxial Testing of Honeycomb in the Transverse Direction, Tech. Report No. CUED/C-Mechanics/TR70, Cambridge University Engineering Department, March 1996. 76] M. Arcan, Z. Hashin and A. Voloshin, A Method to Produce Uniform Planestress States with Applications to Fiber-reinforced Materials, Experimental Mechanics, 18(4), 1978, pp. 141{146. 77] A. Voloshin and M. Arcan, Failure of Unidirectional Fiber-reinforced Materials | New Methodology and Results, Experimental Mechanics, 20(3), 1984, pp. 280{284. 78] R. H. Marloff, Finite Element Analysis of Biaxial Stress Test Specimen for Adhesive Characterisation, in I. M. Daniel (editor), ASTM STP, vol. 787, 1988, (pp. 34{49). 79] R. E. Miller, A Continuum Plasticity Model for the Constitutive and Indentation Behaviour of Foamed Metals, To be submitted to Intl. J. Mech. Sci.