294 COMPOSITE MATERIALS FOR AIRCRAFT STRUCTURES quantifying minimum strength;even zero is a possibility.Environmental degradation of the interface in service is much more likely if the adherends are not given the correct surface treatment before bonding.Suitable methods will be discussed later. Fatigue damage or creep in the adhesive layer can be avoided,or at least minimized,by maintaining the adhesive in an elastic state for most of its service life.Ideally,significant plastic deformation of the adhesive should be permitted only when the joint is stressed to limit load.Limit load is the highest load expected during the service life of the aircraft.Even at ultimate load(1.5x limit), the strain in the adhesive should not approach the failure strain. The design aim is to maintain the adhesive in a state of shear or compression.Structural adhesive joints (and composites)have relatively poor resistance to through-thickness(peel)stresses and,where possible,this type of loading is avoided.The classical joint types,suitable for joining composites to either composites or metals2(Fig.9.2),are 1)the double lap,2)the single lap,3)the single scarf,4)double scarf,5)the single-step lap,and 6)the double-step lap. Figure 9.3,by Hart-Smith,3 illustrates schematically the load-carrying capacities of these joints and some simple design improvements. The single-lap joint is generally the cheapest of all joints to manufacture. However,because the loads are offset (eccentric),a large secondary bending moment develops that results in the adhesive being subjected to severe peel stresses.This type of joint is therefore only used for lightly loaded structure or is supported by underlying structure such as an internal frame or stiffener. The double-lap joint has no primary bending moment because the resultant load is collinear.However,peel stresses arise due to the moment produced by the unbalanced shear stresses acting at the ends of the outer adherends.The resulting stresses,although relatively much smaller in magnitude than in the single-lap joint,produce peel stresses limiting the thickness of material that can be joined. Peel(and shear)stresses in this region are reduced by tapering the ends of the joint.As shown in Figure 9.3,this markedly increases the load capacity of this joint. The scarf and step-lap joints,when correctly designed,develop negligible peel stresses and may be used (at least in principle)to join composite components of any thickness. To explore the feasibility of using primary lap joints that use only adhesive bonding,the USAF funded the Primary Adhesively Bonded Structure Technology (PABST)program which,although concerned with the bonding of aluminum alloy airframe components,must be mentioned as a landmark in the development of bonded joints for aeronautical applications;many of its conclusions are relevant to bonded composite construction.The Douglas Aircraft Company was the major contractor.The program (based on a full-scale section of fuselage for a military transport aircraft)demonstrated that significant improvements could be obtained in integrity,durability,weight,and cost in an
294 COMPOSITE MATERIALS FOR AIRCRAFT STRUCTURES quantifying minimum strength; even zero is a possibility. Environmental degradation of the interface in service is much more likely if the adherends are not given the correct surface treatment before bonding. Suitable methods will be discussed later. Fatigue damage or creep in the adhesive layer can be avoided, or at least minimized, by maintaining the adhesive in an elastic state for most of its service life. Ideally, significant plastic deformation of the adhesive should be permitted only when the joint is stressed to limit load. Limit load is the highest load expected during the service life of the aircraft. Even at ultimate load (1.5x limit), the strain in the adhesive should not approach the failure strain. The design aim is to maintain the adhesive in a state of shear or compression. Structural adhesive joints (and composites) have relatively poor resistance to through-thickness (peel) stresses and, where possible, this type of loading is avoided. The classical joint types, suitable for joining composites to either composites or metals 2 (Fig. 9.2), are 1) the double lap, 2) the single lap, 3) the single scarf, 4) double scarf, 5) the single-step lap, and 6) the double-step lap. Figure 9.3, by Hart-Smith, 3 illustrates schematically the load-carrying capacities of these joints and some simple design improvements. The single-lap joint is generally the cheapest of all joints to manufacture. However, because the loads are offset (eccentric), a large secondary bending moment develops that results in the adhesive being subjected to severe peel stresses. This type of joint is therefore only used for lightly loaded structure or is supported by underlying structure such as an internal frame or stiffener. The double-lap joint has no primary bending moment because the resultant load is collinear. However, peel stresses arise due to the moment produced by the unbalanced shear stresses acting at the ends of the outer adherends. The resulting stresses, although relatively much smaller in magnitude than in the single-lap joint, produce peel stresses limiting the thickness of material that can be joined. Peel (and shear) stresses in this region are reduced by tapering the ends of the joint. As shown in Figure 9.3, this markedly increases the load capacity of this joint. The scarf and step-lap joints, when correctly designed, develop negligible peel stresses and may be used (at least in principle) to join composite components of any thickness. To explore the feasibility of using primary lap joints that use only adhesive bonding, the USAF funded the Primary Adhesively Bonded Structure Technology (PABST) program 4 which, although concerned with the bonding of aluminum alloy airframe components, must be mentioned as a landmark in the development of bonded joints for aeronautical applications; many of its conclusions are relevant to bonded composite construction. The Douglas Aircraft Company was the major contractor. The program (based on a full-scale section of fuselage for a military transport aircraft) demonstrated that significant improvements could be obtained in integrity, durability, weight, and cost in an
JOINING OF COMPOSITE STRUCTURES 295 Double overlap Single overlap Scarf Double scarf Stepped lap Double-stepped lap Fig.9.2 Schematic illustration of several types of bonded joint. aluminum alloy fuselage component by the extensive use of bonded construction. The demonstrated weight-saving was about 15%,with a 20%saving in cost. Lap joints relying solely on adhesive bonding,although structurally very attractive,are not generally used by major aircraft manufacturers in primary structural applications(such as fuselage splice joints)because of concerns with long-term environmental durability.These concerns stem from some early poor service experience with the environmental durability of adhesive bonds,resulting from the use of inadequate pre-bonding surface treatments and ambient-curing adhesives
JOINING OF COMPOSITE STRUCTURES 295 Double overlap Single overlap Scarf Double scarf ~===---~====~::::.~ ..... ,/ Stepped lap ZZZZZZZZ-'--ZZZZ_ZZZZZ__'_=_~ Double-stepped lap Fig. 9.2 Schematic illustration of several types of bonded joint. aluminum alloy fuselage component by the extensive use of bonded construction. The demonstrated weight-saving was about 15%, with a 20% saving in cost. Lap joints relying solely on adhesive bonding, although structurally very attractive, are not generally used by major aircraft manufacturers in primary structural applications (such as fuselage splice joints) because of concerns with long-term environmental durability. These concerns stem from some early poor service experience with the environmental durability of adhesive bonds, resulting from the use of inadequate pre-bonding surface treatments and ambient-curing adhesives
296 COMPOSITE MATERIALS FOR AIRCRAFT STRUCTURES SCARF-AND STEPPED-LAP JOINTS FAILURES OUTSIDE SHEAR FAILURES FAILURES SHOWN REPRESENT THE BEST POSSIBLE FROM TAPERED-LAP JOINT EFFICIENT DESIGN FOR EACH GEOMETRY LAMINATE STRENGTH OUTSIDE JOINT DOUBLE-LAP JOINT PEEL FAILURES SINGLE-LAP JOINT BENDING OF AOHFRENDS DUE TO ECCENTRIC LOAD PATH ADHEREND THICKNESS Fig.9.3 Load-carrying capacity of adhesive joints.Taken from Ref.3. 9.3.2 Design/Analysis of Bonded Lap Joints Reviews of analytical procedures for joints involving composites are provided in Refs.5 and 6.Hart-Smith undertook comprehensive analytical studies7-9 on adhesive joints,particularly advanced fiber composite to composite and composite to metal joints.His studies,based on the earlier approaches,cover the important aspect of non-linear (elastic/plastic)deformation in the adhesive. The stress level for joint (adhesive)failure is determined by shear strain to failure of the adhesive(y+yp)in the bondline;the design aim being that this stress level should well exceed adherend strength.Peel stresses are avoided by careful design rather than considered as a potential failure mode. Several earlier attempts were made to represent non-linear behavior in the adhesive assuming realistic shear stress/shear strain behavior,but they were too
296 COMPOSITE MATERIALS FOR AIRCRAFT STRUCTURES O Z o SCAPJ:- AND STEPPED-LAP JOINTS i , , J" FAILURES SHOWN REPRESENT THE BEST POSSIBLE FROM EFFICIENT DI~SION FOR EACH GEOMETRY TAPERED-!~P JOINT ' DOUBLE-LAP JOINT II . , '! PEEL FAILURES SINGLE-LAP JOINT i BENDING OF AOHFRENDS DUE TO ECCENTRIC LOAD PATH ADHEREND THICKNESS t Fig. 9.3 Load-carrying capacity of adhesive joints. Taken from Ref. 3. 9.3.2 Design~Analysis of Bonded Lap Joints Reviews of analytical procedures for joints involving composites are provided in Refs. 5 and 6. Hart-Smith undertook comprehensive analytical studies 7-9 on adhesive joints, particularly advanced fiber composite to composite and composite to metal joints. His studies, based on the earlier approaches, cover the important aspect of non-linear (elastic/plastic) deformation in the adhesive. The stress level for joint (adhesive) failure is determined by shear strain to failure of the adhesive (% + "yp) in the bondline; the design aim being that this stress level should well exceed adherend strength. Peel stresses are avoided by careful design rather than considered as a potential failure mode. Several earlier attempts were made to represent non-linear behavior in the adhesive assuming realistic shear stress/shear strain behavior, but they were too
JOINING OF COMPOSITE STRUCTURES 297 Table 9.2 Computer Programs Developed by Dr John Hart-Smith for Stress Analysis of Bonded Joints Joint to be analyzed Program Joint to be analyzed Program Single-lap joint: A4EA Double-lap joint:Elastic A4EB Joint strengths and efficiencies adherend and elastic/plastic in non-dimensional form. adhesive. Deals only with identical Can deal with unbalanced adherends.Three failure cases joints and allows for thermal are considered:a)adherend mismatch between adherends. bending,b)adhesive shear, Provides ratio of maximum to and c)adhesive peel. average shear strength and non-dimensionalized joint strength. Scarf Joint: A4EE Step-lap joint: A4EG Elastic adherend and elastic/ Elastic adherend and elastic/ plastic adhesive. plastic adhesive. Provides a)shear stress Provides a)shear stress distribution along the joint b) distribution along the joint, displacement of inner and b)displacement of inner and outer adherends,and c) outer adherends,and c) potential joint strength. potential joint strength. Step-lap joint: A4EI Elastic adherend and elastic/ plastic adhesive. Similar to A4EG but more comprehensive;allows for variations in adhesive thickness and adhesive defects.Bond width can also be varied. complex for most analytical approaches.However,as discussed later,Hart-Smith shows that a simple elastic/ideally plastic formulation gives similar results to more realistic representations of adhesive behavior,providing the strain energy density in shear in the adhesive (area under the stress-strain curve)is comparable to that expected for the real curve. As a major part of these studies,software programs were developed for the analysis of double-overlap and the other types of joint discussed here;these are listed in Table 9.2.Similar programs are available through the Engineering Sciences Data Unit (ESDU)9 and proprietary programs have been developed by manufacturers. Inevitably,many of the complications in real joints are neglected or inadequately dealt with in these relatively simple studies.These include:
Table 9.2 JOINING OF COMPOSITE STRUCTURES Computer Programs Developed by Dr John Hart-Smith for Stress Analysis of Bonded Joints 297 Joint to be analyzed Program Joint to be analyzed Program Single-lap joint: A4EA Double-lap joint: Elastic A4EB Joint strengths and efficiencies adherend and elastic/plastic in non-dimensional form. adhesive. Deals only with identical Can deal with unbalanced adherends. Three failure cases joints and allows for thermal are considered: a) adherend mismatch between adherends. bending, b) adhesive shear, Provides ratio of maximum to and c) adhesive peel. average shear strength and non-dimensionalized joint strength. A4EE Step-lap joint: A4EG Elastic adherend and elastic/ plastic adhesive. Provides a) shear stress distribution along the joint, b) displacement of inner and outer adherends, and c) potential joint strength. Scarf Joint: Elastic adherend and elastic/ plastic adhesive. Provides a) shear stress distribution along the joint b) displacement of inner and outer adherends, and c) potential joint strength. Step-lap joint: Elastic adherend and elastic/ plastic adhesive. Similar to A4EG but more comprehensive; allows for variations in adhesive thickness and adhesive defects. Bond width can also be varied. A4EI complex for most analytical approaches. However, as discussed later, Hart-Smith shows that a simple elastic/ideally plastic formulation gives similar results to more realistic representations of adhesive behavior, providing the strain energy density in shear in the adhesive (area under the stress-strain curve) is comparable to that expected for the real curve. As a major part of these studies, software programs were developed for the analysis of double-overlap and the other types of joint discussed here; these are listed in Table 9.2. Similar programs are available through the Engineering Sciences Data Unit (ESDU)9 and proprietary programs have been developed by manufacturers. Inevitably, many of the complications in real joints are neglected or inadequately dealt with in these relatively simple studies. These include:
298 COMPOSITE MATERIALS FOR AIRCRAFT STRUCTURES Influence of flaws in the form of local porosity,local disbonds,etc. Adhesive thickness variations .Through-thickness variation of shear stresses Through-thickness stresses .Stress-free state at the ends of the adhesive .Highly beneficial effect of adhesive spew,excess adhesive that forms a fillet at the edges of the joint True shear stress/shear strain behavior Most of these complexities are best modelled using finite element procedures.For example,the simple analytical procedures for lap joints mentioned here predict that the maximum shear stress occurs at the free ends of the overlap.However, because the end of the overlap is a free surface,the principle of complimentary shears is violated because the horizontal shear force at the ends cannot be balanced by a vertical shear force.In reality,therefore,the stress along the bond line right at the edge must fall to zero.More realistic stress analysis-using the finite element approach shows that this is the case-shear stress falls rapidly to zero over a distance of the order of the adherend thickness;these observations are confirmed by direct experimental observations.However,the shear stress distribution along the bond line and magnitude of the maximum stress predicted by the simple analytical procedures turns out to be approximately correct.Similar observations have been made concerning normal or peel stresses. A further considerable complication,difficult to handle even with finite element methods,is the time dependency or viscoelastic (and viscoplastic) behavior of adhesives. 9.3.3 Models for Adhesive Stress/Strain Behavior For analysis of stress distribution in the joint,a model for the shear stress/ strain behavior of the adhesive is required.The simplest model assumes that the adhesive is strained only within its linear elastic range.This model may be adequate if fatigue is a major concern and the primary aim is to avoid plastic cycling of the adhesive;then the stresses must not be allowed to exceed Tp. However,use of the elastic model is overly conservative for assessing the static strength of a joint,particularly if it is bonded with a highly ductile adhesive. To account for plastic deformation,the actual stress/strain behavior must be modelled.In computer-based approaches such as the finite element method,the stress/strain curve can be closely modelled using the actual constitutive rela- tionship.However,for analytical approaches,much simpler models are needed. Figure 9.4 shows stress/strain behavior for a typical ductile adhesive and the models of this behavior used for joint analysis by Hart-Smith.The intuitive simple non-linear model is the bilinear characteristic because this most closely approximates to the real curve.However,even use of this simple model is mathematically complex,greatly limiting the cases that can be analyzed to produce closed-form solutions
298 COMPOSITE MATERIALS FOR AIRCRAFT STRUCTURES • Influence of flaws in the form of local porosity, local disbonds, etc. • Adhesive thickness variations • Through-thickness variation of shear stresses • Through-thickness stresses • Stress-free state at the ends of the adhesive • Highly beneficial effect of adhesive spew, excess adhesive that forms a fillet at the edges of the joint • True shear stress/shear strain behavior Most of these complexities are best modelled using finite element procedures. For example, the simple analytical procedures for lap joints mentioned here predict that the maximum shear stress occurs at the free ends of the overlap. However, because the end of the overlap is a free surface, the principle of complimentary shears is violated because the horizontal shear force at the ends cannot be balanced by a vertical shear force. In reality, therefore, the stress along the bond line right at the edge must fall to zero. More realistic stress analysis--using the finite element approach 1° shows that this is the case--shear stress falls rapidly to zero over a distance of the order of the adherend thickness; these observations are confirmed by direct experimental observations. However, the shear stress distribution along the bond line and magnitude of the maximum stress predicted by the simple analytical procedures turns out to be approximately correct. Similar observations have been made concerning normal or peel stresses. A further considerable complication, difficult to handle even with finite element methods, is the time dependency or viscoelastic (and viscoplastic) behavior of adhesives. 9.3.3 Models for Adhesive Stress/Strain Behavior For analysis of stress distribution in the joint, a model for the shear stress/ strain behavior of the adhesive is required. The simplest model assumes that the adhesive is strained only within its linear elastic range. This model may be adequate if fatigue is a major concem and the primary aim is to avoid plastic cycling of the adhesive; then the stresses must not be allowed to exceed "rp. However, use of the elastic model is overly conservative for assessing the static strength of a joint, particularly if it is bonded with a highly ductile adhesive. To account for plastic deformation, the actual stress/strain behavior must be modelled. In computer-based approaches such as the finite element method, the stress/strain curve can be closely modelled using the actual constitutive relationship. However, for analytical approaches, much simpler models are needed. Figure 9.4 shows stress/strain behavior for a typical ductile adhesive and the models of this behavior used for joint analysis by Hart-Smith. 11 The intuitive simple non-linear model is the bilinear characteristic because this most closely approximates to the real curve. However, even use of this simple model is mathematically complex, greatly limiting the cases that can be analyzed to produce closed-form solutions