ON ACHIEVING AN OPTIMAL RIVETED LAP JOINT DESIGN FOR FIBRE METAL LAMINATES

Transcription

1 ON ACHIEVING AN OPTIMAL RIVETED LAP JOINT DESIGN FOR FIBRE METAL LAMINATES A.M. Brown, C.D. Rans, and P.V. Straznicky Department of Mechanical and Aerospace Engineering, Carleton University, 1125 Colonel By Drive, K1S 5B6, Ottawa, Ontario, Canada ABSTRACT Fibre metal laminates are designed to provide a damage tolerant alternative to monolithic metallic materials in airframe structures without the need for new structural design solutions required by pure fibre reinforced composites. This advantage allows fibre metal laminates to be incorporated directly into existing airframe design practices with little effort. Although such a simple transition is possible, fibre metal laminates exhibit different properties and behaviour than monolithic metallic materials. Consequently, differences may exist in optimal design details for a given structural design solution. This paper overviews the results of a study examining this possibility for a common airframe joint design solution: the riveted lap joint. Focusing on fatigue performance, a series of studies were completed to assess how major differences between monolithic aluminum sheet and a glass fibre-aluminum fibre metal laminate variant known as GLARE could be treated and exploited to obtain an optimal GLARE joint. These studies include an examination of alternative countersinking methods for thin laminated sheets, the formation of residual stresses in GLARE due to fastener interference, and material stiffness effects on joint bending. Results from these studies will form the basis for a discussion on achieving an optimal riveted lap joint design for GLARE. INTRODUCTION Fibre metal laminates (FML s) are a family of hybrid laminates consisting of alternating layers of monolithic metallic sheet and fibre reinforced epoxy pre-impregnated fibre (prepreg) layers (Figure 1). Originating from the development of built-up structures pioneered by Fokker Aircraft in the late 1940 s, FML s were found to exhibit benefits from both their metallic and fibre reinforced composite constituents. Specifically, FML s possess many of the damage tolerant properties inherent in fibre reinforced composite materials while retaining the fabrication and repair capabilities of sheet metal. Within the aerospace field, these properties are attractive as they provide an opportunity to enhance the damage tolerance of airframe structures without the need for new structural design solutions required by pure fibre-reinforced composites. Although FML s can be incorporated directly into metallic airframe design practices with little effort, FML s do exhibit different properties and behaviour compared to monolithic metallic materials. Consequently, differences may exist in optimal design details for a given structural design solution. This paper provides an overview of studies examining this possibility for a common airframe joint design solution: the riveted lap joint. 1

2 Fibre Layers Metallic Layers Figure 1 Illustration of fibre metal laminate construction. Figure 2 Illustration of fibre bridging mechanism. GLARE LAMINATES Within the family of FML s, a material known as GLARE (GLAss REinforced aluminum) has been developed as a replacement for monolithic 2024-T3 aluminum in aircraft fuselage construction. This material consists of 2024-T3 aluminum for the metallic layers and S2-glass fibre epoxy for the prepreg layers. The focus of this paper is the GLARE3-2/1-0.3 laminate which would be a suitable replacement for 1 mm 2024-T3 fuselage skin. FML s and, more specifically, GLARE variants are known to have beneficial characteristics in the area of fatigue and damage tolerance with fatigue lives in excess of 10 times that of 2024-T3 [1]. The reason for the increased fatigue performance is the material s superior crack propagation life through a mechanism called fibre bridging. With fibre bridging, the load normally carried by the cracked aluminum plies gets redistributed into the adjacent intact fibre layers and the stress concentration at the crack tip is effectively reduced (Figure 2). A further benefit of the sandwich lay-up is that through thickness crack propagation does not occur. A crack growing through the thickness of an aluminum ply will stop once it hits a prepreg layer. It then remains for the other intact aluminum plies to nucleate cracks. This unique laminate construction contributes to several other desirable properties [2-4] for an aircraft material but will not be discussed here. Although the fatigue propagation behaviour of GLARE is superior to 2024-T3, crack nucleation actually occurs earlier than in monolithic sheet. This stems from the difference in flexibility of the material plies and from residual tensile stresses in the aluminum plies introduced during the curing cycle of the manufacturing stage. These tensile residual stresses contribute to any applied stress and helps lead to crack nucleation. In addition to the presence of tensile residual stresses, the lower stiffness of the prepreg plies in the sandwich lay-up means that the aluminum plies will experience greater stresses than the applied bulk stress. 2

3 FATIGUE OF RIVETED LAP JOINTS The basic riveted lap joint configuration consists of two sheets joined by one or more rows of rivets. The two sheets are differentiated by their location and accessibility for inspection along the joint overlap; in an airframe application, the outer sheet would be accessible from the outside of the aircraft while the inner sheet would be accessible from the inside. Additional variables and terminology used in defining a riveted lap joint design are illustrated in Figure 3. TOP ROW OUTER SHEET s (rivet pitch) Outer free surface t (sheet thickness) Outer faying surface BOTTOM ROW p (row pitch) D o (rivet diameter) INNER SHEET Inner faying surface Inner free surface Figure 3: 3-row riveted lap joint nomenclature. Developing a lap joint design based on fatigue performance requires the designer to manipulate the local stress state around fatigue prone rivet locations in order to prolong fatigue crack nucleation and reduce the rate of propagation of existing cracks. This local stress state can be broken down into components related to load transfer, and secondary loads such as secondary bending and rivet interference as illustrated in Figure 4. These components and the influence of lap joint design variables on them will be briefly discussed in order to establish a baseline for lap joint design which will be expanded on later in the context of GLARE specific lap joint design. Load Transfer Load transfer in riveted lap joints occurs by two mechanisms. The majority of load is transferred by means of bearing between the rivets and inner and outer sheets while the remainder is transferred through friction between the faying joint surfaces. The magnitude of the bearing stress at a rivet hole is related to the load transferred by a single rivet and the bearing area (rivet diameter multiplied by sheet thickness) over which that load is transmitted into the joined sheets. Large bearing stresses promote crack nucleation and reduce the crack free life of a joint. Although bearing stresses can be reduced easily in practice by increasing the total bearing area, either through increasing the rivet diameter, increasing the number of rivets, or both, these steps negatively impact crack growth. The reduced amount of sheet material along the net section of a rivet row that 3

4 would accompany such efforts to reduce bearing stress increase crack propagation rates and reduce critical crack lengths. Bearing Friction Load Transfer Bypass Secondary Loads Secondary Bending Interference Figure 4: Illustration of local stress field components in a riveted lap joint. Figure 5: Illustration of load carried by inner and outer sheets in a 3-row lap joint. Friction comprises a small portion of the overall load transfer in a riveted joint and is often neglected. Any load that is transferred through friction reduces bearing stress and generally results in an improvement in fatigue performance. More important than the amount of load transferred through friction is the amount of fretting damage that occurs as a result of friction. Fretting damage is a destructive process associated with clamped bodies undergoing oscillatory loading [5]. In riveted lap splices, the evidence of fretting damage is a black oxidized region surrounding the rivet hole on the sheet faying surface. This damaging process is known to generate microcracks in the surface that, under further loading, can propagate into the substrate as a fatigue crack. In multiple-row lap joints, each rivet row transfers a portion of the applied load from the outer to inner sheet while the remainder (referred to as the bypass load) remains in the outer sheet (Figure 5). The amount of load transferred by each rivet row is related to the relative flexibilities of the rivets and sheets being joined, neglecting the contribution from friction. For a 3-row lap joint, the top and bottom rivet rows transfer the same amount of load due to joint asymmetry while the load transferred by the middle rivet row can be determined by: 4

5 T f T rivet 2 1 frivet fsheet (1) where: T1, T 2 = load transferred by top and middle row respectively f, f = rivet and sheet flexibilities rivet sheet As negative flexibilities are not possible, this equation shows that the top and bottom rivet rows will always transfer (through bearing) a larger percentage of the load than the middle rivet rows. The combination of large bearing stresses and high bypass stresses along these rivet rows results in two fatigue critical locations: in the outer sheet along the top rivet row and in the inner sheet along the bottom rivet row (Figure 5). In practice, the inner sheet location is typically less critical due to improved hole filling within this sheet during riveting and due to the possible added stress concentration in the outer sheet due to countersinking. Although adding more rows of rivets beyond three can reduce the amount of load transferred through the top and bottom rivet rows, this reduction drops off exponentially with the number of rivet rows and is typically viewed as a weight and manufacturing penalty. A more in depth discussion on load transfer in riveted lap joints can be found in the work of Müller [6]. Secondary Bending Secondary bending refers to the out-of-plane deformation and resulting bending stresses that are inherent in the lap joint configuration due to load path eccentricities (Figure 6). As a result of secondary bending, large tensile stresses are created along the joint faying surfaces that contribute to the nucleation of fatigue cracks. The severity of this bending stress is often expressed as a ratio of the bending and applied stresses using the secondary bending factor, K b, given by: bending Kb (2) applied t1 t2 e1 e2 e3 Neutral Line Part 1 Part 2 Part 3 Part 4 Figure 6: Neutral line representation of a lap joint showing joint eccentricities. The role of secondary bending in crack nucleation makes it necessary to predict and assess the effects of secondary bending on a lap joint design. Schijve [7] developed a one-dimensional model to predict secondary bending stresses in a 2-row lap joint by representing it by its neutral line. Further extensions have subsequently been made, 5

6 extending the capabilities to n-row lap joints [6, 8, 9]. In practice, the neutral line model represents each row of rivets as an infinitely thin and rigid bond line. This limits its ability to predict local influences of discrete fastener locations and variations along the joint width. Despite this shortcoming, the neutral line model is often used in design due to its speed and simplicity and is supplemented with more computationally expensive finite element analysis. Rivet Interference Expansion of the rivet shank during installation produces an interference that results in a beneficial residual stress field within the sheet material surrounding the rivet. This process is analogous to cold expansion processes where fastener holes are expanded to a desired level, typically using a tapered mandrel, before inserting a mechanical fastener. Although analogous, the expansion process resulting from riveting has added complexities compared to cold expansion process, namely: The rivet shank does not expand uniformly through the thickness of the joint, The manufactured and driven rivet heads exert a pressure through the thickness of the joint during riveting, resulting in a complex 3-dimensional stress state during expansion, The rivet remains in the fastener hole after riveting, preventing full elastic springback of the surrounding sheet. Developments in riveting technology as well as analysis software are providing more opportunity to exploit rivet installation during joint design. Müller [6] demonstrated that the use of force-controlled riveting techniques provided a repeatable residual stress distribution that could reduce fatigue life scatter in current lap joint designs. Furthermore, Müller found that the use of larger rivet squeeze forces could provide further improvements to fatigue performance. Currently, investigations using 2- and 3- dimensional finite element models [6, 10-19] are being performed to further our ability to predict and exploit residual stresses due to rivet installation in both FMLs and monolithic sheet materials. DESIGN CONSIDERATIONS FOR RIVETED GLARE LAP JOINTS Optimizing a lap joint design for GLARE requires careful consideration of how the material differs from its monolithic aluminum counterpart and how these differences affect the local stress state in fatigue critical locations. In the following sections, the potential problems with countersinking GLARE are explored as well as a discussion on laminate stiffness and the unique GLARE property of fibre layer elasticity. Laminate and Lamina Thickness One of the 2024-T3 design details being applied directly to GLARE is the method of countersinking rivets. The most common method of creating a flush joint surface is machine countersinking where a conical recess is machined into the sheet to accommodate a countersunk rivet. Another, less popular, method of creating a flush 6

7 Figure 7 Illustration of conventional dimpling set-up. surface is through dimpling where a recess is formed in the sheet by pressing the sheets between mated countersunk tools (Figure 7). This method was originally thought to be superior to machine countersinking but has since been relegated to use on secondary structures. When machine countersinking thin sheet materials, it is important to avoid the creation of a knife-edge condition. A knife-edge is created at a rivet hole when a countersinking tool penetrates through the entire thickness of the sheet creating a sharp edge at the faying surface (Figure 8 a). The presence of a knife-edge greatly increases the stress concentration at that point, reducing the crack nucleation life [20]. In thin monolithic sheet material, this condition is normally avoided by using rivets with a reduced height countersunk head. There are, however, obvious limits to how shallow a countersunk rivet head can be made while still maintaining its strength. (a) Figure 8 Schematic of a through-thickness knife-edge (a) and a lamina knife-edge (b). (b) When using GLARE, the superior overall fatigue performance provides the opportunity for weight savings through the application of thinner laminates in primary structures. The application of thinner laminates increases the risk for through-thickness knife-edge conditions in machine countersunk holes. Also, a condition unique to fibre metal laminates is the creation of lamina knife edges (Figure 8 b). The 0.3 mm aluminum layers in GLARE are too thin to avoid the formation of a knife-edge in at least one of the outer aluminum plies. While this may be less of a concern in thick GLARE variants where most of the aluminum plies are unaffected, variants such as GLARE3-2/1 have only two aluminum plies with at least one subject to a knife edge condition. Studies into the possibility of using dimpling methods on GLARE3-2/1 to avoid lamina knife edges was performed by the current authors with mixed results [21, 22]. The studies found that conventional dimpled single lap splice GLARE coupons obtained fatigue lives no better than their monolithic aluminum counterparts and modified versions of the dimple tools were only able to obtain a moderate increase in fatigue life in GLARE over the conventional tools. Results from the complete test series including machine countersunk coupons are presented in Figure 9. This test series was an instance where the 7

8 effects of a direct substitution of GLARE for monolithic 2024-T3 could be observed for a specific splice geometry and loading condition. The results were surprising in that GLARE performed poorer than the aluminum coupons in all cases. Laminate thickness also plays a role in the degree of secondary bending in a single lap splice. It may be thought that as the sheet thickness decreases in size, the secondary bending stresses will also decrease due to the smaller moment arm between the neutral lines of the mating sheets. In reality, the difference in Figure 9 Results of dimple coupon fatigue tests. material stiffness of a GLARE single lap splice constructed of 0.86 mm thick sheets will show greater secondary bending than a monolithic splice constructed with 1 mm sheets. The effects of the differences in material stiffness are discussed further in the next section. Laminate Stiffness The difference in stiffness between monolithic aluminum and GLARE poses a challenge for designers. The difference can have detrimental effects on component design as observed in the case of the riveted single lap splice results shown earlier. A lower material stiffness allows for a splice to show increased secondary bending, greater slip and, possibly, greater fretting damage around rivet holes in comparison to a stiffer material. The serious effects of increased secondary bending deflection due to low laminate stiffness were shown in the machine countersunk riveted coupon results from the dimple study mentioned above. Where sheet failure occurred in the traditional machine countersunk 2024-T3 coupons, the machine countersunk GLARE3 coupons failed due to fatigue of the NAS1097AD-4-4 rivet heads with eventual pull through (Figure 10). As the sheet deflected out-of-plane due to secondary bending, the top portion of the rivet head was Figure 10 Rivet head fatigue failure. forced to carry a greater portion of the load. Continual cyclic loading resulted in failure of the top portion of the rivet head before crack nucleation could occur in the sheet. Even though the fatigue behaviour of the sheet 8

9 material was superior, the reduced material stiffness created a joint where the rivets became fatigue critical and the overall fatigue life was lower. Figure 11 Typical relationship between wear rate, slip amplitude, and fretting fatigue life. [5] In a loaded splice, a lower laminate stiffness can also result in an increase in relative motion between the sheets. This can contribute to an increase in fretting damage around the rivet heads and a reduced crack nucleation life. Figure 11 shows a typical relationship between slip amplitude, wear, and fretting fatigue [5]. As the amount of slip around a rivet increases from a low slip amplitude to a higher one, Figure 11 shows that the life of the material drops due to an increase in fretting damage and the formation of micro-cracks. The life curve only begins to rise again once sliding motion takes over and wear rates become high. High wear, associated with sliding, will wear away micro-cracks before they have a chance to propagate into the material substrate. Fibre Layer Elasticity Effect Relative to its aluminum layers, the fibre layers in GLARE can be considered perfectly elastic. This insensitivity of the fibre layer to plastic deformation influences the manner in which the entire laminate yields. When a portion of an aluminum layer begins to yield, its stiffness begins to drop, causing load to be redistributed into the elastic fibre layer and bridged around the yielded portion. This mechanism is analogous to how load is redistributed around a crack within an aluminum layer as discussed earlier. Plastic strains are thus distributed over a larger region of the GLARE laminate compared to a monolithic aluminum sheet. This influence of the fibre layer on yielding affects the formation of residual stresses in riveted GLARE lap joints. Finite element simulations of the cold expansion process [23] and the rivet installation process [14, 16] in GLARE laminates confirmed that the redistribution of load by the fibre layer results in larger zones of plastic deformation around the fastener holes (Figure 12). Consequently, larger zones of beneficial compressive residual stress are formed which contribute further to the damage tolerant advantage of GLARE. 9

10 In addition to influencing the how the laminate yields, the elasticity of the fibre layer also influences elastic recovery. During rivet installation, the fibre layer is expanded along with the aluminum layers, resulting in a build-up of strain energy. In the aluminum layers, a portion of this strain energy is dissipated through yielding. Despite the lower stiffness of the fibre layers relative to the aluminum layers, the lack of plasticity allows the fibre layers to store more elastic strain energy. After rivet installation, elastic recovery of the sheet compresses the potion of the sheet which yielded during riveting. This compression introduces the desired residual compressive hoop stresses around the rivet hole. The added stored strain energy in the fibre layer increases the springback response of GLARE, resulting in added compression of the yielded aluminum layers. This effect increases the magnitude of the residual compressive hoop stresses in GLARE as illustrated in Figure 12. DISCUSSION AND CONCLUSIONS rivet installation force, F Sq (kn) Decades of research and development of GLARE has shown the material to have superior properties over monolithic aluminum especially in the area of fatigue propagation and damage tolerance. Much of the development and promotion of the material was that it was a direct replacement for monolithic sheets and traditional design practices could be applied to the new material. In many situations it may be that direct substitution of GLARE will result in a more fatigue resistant design; however, optimizing the design specifically for GLARE is often overlooked. This new material differs greatly from monolithic material and there are negative aspects that must be considered and positive aspects that should be exploited. The key negative effect that must be treated when designing with GLARE is the reduced stiffness of the material. More than all other factors, this should be the driving element when designing for an optimum splice. The difference in stiffness has become such a difficulty that designers are now looking at sizing the GLARE sheets based on aluminum ply thickness alone and considering the prepreg layers as supporting material for damage tolerance only. This will help to deal with many of the problems associated with a reduced laminate stiffness. The presence of residual tensile stresses will, however, remain and thicker sheets will result giving a larger secondary bending moment arm than the monolithic sheet it replaces. radius of plastic zone (mm) Results plotted for inner sheet faying surface only GLARE3-2/ T Figure 12: Radii of plastic zone and peek residual compressive hoop stress resulting from rivet installation as determined by finite element analysis. peek residual compressive hoop stress (MPa) 10

11 The poor machine countersunk GLARE coupon fatigue results from the dimpling study mentioned above (Figure 9) showed that splice designs and loading levels that are easily accommodated by the stiffer monolithic aluminum sheet may prove disastrous when applied directly to GLARE. Although the GLARE sheet itself did not fail, the response of the material for that splice design was to transfer load to the rivets in a damaging way and this type of failure is not necessarily easy to predict. Rivets with larger heads may be a solution as long as the material does not lead to a laminate knife edge condition. The dimpled coupon tests also showed that directly applying a design principle from monolithic material directly to an FML will not necessarily result in improved fatigue performance. With the dimpled GLARE coupons, failure did occur in the sheet but with fatigue lives lower than the monolithic aluminum counterparts. The difference in material stiffness, again, was to blame. In this situation, the reduced stiffness of the sheet acted in conjunction with the reduced stiffness of the dimple cone allowing for rivet rotation and greater secondary bending deflection around the critical rivet rows. The bending became more pronounced in the dimpled coupons that had a solid lubricant applied to the faying surface. For the 2024-T3, the lubricant was able to stop the nucleation of fretting induced micro-cracks into the substrate and resulted in a substantially increased fatigue life. In the GLARE splice, however, the lubricant exacerbated bending in the sheet by allowing the outer dimple cone to slide up the inner cone and slightly separate the sheets. The result of the reduced material stiffness in combination with faying surface lubrication was a severely reduced fatigue life. Despite the negative fatigue results discussed above, in many cases the substitution of GLARE for monolithic aluminum will provide greater fatigue life and damage tolerance. This paper has shown that there are other characteristics that can be exploited when designing with GLARE to work towards an optimal splice design. The use of high rivet forming loads and load-controlled riveting can induce a beneficial compressive residual stress distribution in the GLARE sheet that is enhanced by the fibre elasticity effect. The superior damage tolerance behaviour also allows for expansion of previous design principles such as minimum allowables on rivet size and rivet pitch. Closer rivet spacing could improve load transfer while the crack propagation behaviour of GLARE would provide enough damage tolerance to detect cracks before reaching a critical length. Although the focus of this paper was the thin laminate GLARE3-2/1-0.3, many of the aspects discussed above, such as the fibre layer elasticity effect, can apply to thicker GLARE laminates and to other FMLs. What must be recognized by the designer is that there are negative aspects of FMLs that should be avoided as well as the positive aspects that could be exploited. In both cases, the behaviour of the material is different from monolithic 2024-T3 and further research is required to find an optimal joint design principles for GLARE. 11

12 REFERENCES 1. Alderliesten, R.C., Fatigue, in Fibre Metal Laminates an Introduction, A.V.a.J.W. Gunnink, Editor., Kluwer Academic Publishers: Dordrecht. p Borgonie, B., M.S. IJpma and W.G.J. 't Hart, Corrosion, in Fibre Metal Laminates an Introduction, A.V.a.J.W. Gunnink, Editor., Kluwer Academic Publishers: Dordrecht. p Hagenbeek, M., Impact properties, in Fibre Metal Laminates an Introduction, A.V.a.J.W. Gunnink, Editor., Kluwer Academic Publishers: Dordrecht. p Hooijmeijer, P.A., Burn-through and lightning strike, in Fibre Metal Laminates an Introduction, A.V.a.J.W. Gunnink, Editor., Kluwer Academic Publishers: Dordrecht. p Hills, D.A., and D. Nowell, Mechanics of Fretting Fatigue. Solid Mechanics and Its Applications. Vol. 30: Kluwer Academic Publishers. (1994). 6. Müller, R.P.G., An experimental and analytical investigation on the fatigue behaviour of fuselage riveted lap joints: the significance of the rivet squeeze force and a comparison of 2024-T3 and Glare , Delft University of Technology: Delft, The Netherlands. 7. Schijve, J. and A. Hartman, The effect of secondary bending on the fatigue strength of T3 Alclad riveted joints. 1969, National Aerospace Laboratory NLR. 8. de Rijck, R., Stress Analysis of Fatigue Cracks in Mechanically Fastened Joints, in Aerospace Engineering. 2005, Delft University of Technology: Delft, the Netherlands. p Homan, J.J., et al., Fatigue of riveted joints, in Fibre Metal Laminates an Introduction, A. Vlot and J.W. Gunnink, Editors., Kluwer Academic Publishers: Dordrecht. p Deng, X. and J.W. Hutchinson, The clamping stress in a cold-driven rivet. International Journal of Mechanical Science. 40(7): p (1998). 11. Langrand, B., et al., Riveted joint modeling for numerical analysis of airframe crashworthiness. Finite Elements in Analysis and Design. 38(1): p (2001). 12. Li, G. and G. Shi, Effect of the riveting process on the residual stress in fuselage lap joints. CASI. 50(2): p (2004). 13. Li, G., G. Shi, and N.C. Bellinger, Study of the residual strain in lap joints. Journal of Aircraft. 43(4): p (2006). 14. Rans, C.D., R.C. Alderliesten, and P.V. Straznicky. Modelling of the rivet forming process in aluminum and Glare for design against fatigue. in ICAF Hamburg, Germany Rans, C.D., R.C. Alderliesten, and P.V. Straznicky. Effects of rivet installation on residual stresses and secondary bending stresses in a riveted lap joint. in AIAA SDM Honolulu, Hawaii Rans, C.D., R.C. Alderliesten, and P.V. Straznicky, Riveting process induced residual stresses around solid rivets in mechanical joints. Journal of Aircraft. 44(1): p (2007). 17. Szolwinski, M.P., The mechanics and tribology of fretting fatigue with application to riveted lap joints. 1998, Purdue University: West Lafayette, IN, U.S.A. 18. Szolwinski, M.P. and T.N. Farris, Linking riveting process parameters to the fatigue performance of riveted aircraft structures. Journal of Aircraft. 37(1): p (2000). 19. Urban, M.R., Analysis of the fatigue life of riveted sheet metal helicopter airframe joints. International Journal of Fatigue. 25: p (2003). 20. Rans, C.D., An experimental investigation into the fatigue behaviour of dimple countersunk Glare riveted lap joints. 2003, Carleton University: Ottawa, Canada. 21. Brown, A.M., and P.V. Straznicky, Dimple tool modification for press-countersinking GLARE, in COM : Montreal, Canada. 22. Rans, C.D.a.P.V.S., Avoiding knife-edge countersinks in GLARE through dimpling. Fatigue and Fracture of Engineering Materials and Structures. 28(7): p (2005). 23. Rans, C.D., R.C. Alderliesten, and P.V. Straznicky. Residual stresses in GLARE laminates due to the cold expansion process. in CANCOM Vancouver, Canada