Strength and Failure Mechanism of Composite-Steel Adhesive Bond Single Lap Joints

Carbon fiber-reinforced plastics(CFRP-) steel single lap joints with regard to tensile loading with two levels of adhesives and four levels of overlap lengths were experimentally analyzed and numerically simulated. Both joint strength and failure mechanismwere found to be highly dependent on adhesive type and overlap length. Joints with 7779 structural adhesive were more ductile and produced about 2-3 kN higher failure load than MA830 structural adhesive. Failure load with the two adhesives increased about 147N and 176N, respectively, with increasing 1mm of the overlap length. Cohesion failure was observed in both types of adhesive joints. As the overlap length increased, interface failure appeared solely on the edge of the overlap in 7779 adhesive joints. Finite element analysis (FEA) results revealed that peel and shear stress distributions were nonuniform, which were less severe as overlap length increased. Severe stress concentration was observed on the overlap edge, and shear failure of the adhesive was the main reason for the adhesive failure.


Introduction
With the aim to save energy and reduce emission, weight saving is of significant importance in the transportation industry.Composite, which exhibits high stiffness-to-weight and strength-to-weight ratios than traditional metal counterparts, has gained widespread usage for lightweight structures.In practical application, it is almost impossible to manufacture a structure as a whole body.Many structures are manufactured as single parts, and then connected through joints.e commonly used methods for joining composite parts are either through mechanical fastening or bonding.Mechanical fasteners including bolts, rivets, and pins have been commonly used for several decades [1][2][3].
e ease of disassembling components and allowing for reliable inspection has been a great benefit.However, the key problem is that high stress concentrations can develop around the fastener holes, and the joint can be brought to failure at far lower stress levels than expected [4].Due to its larger bond area to distribute loads and eliminate stress concentration as well as keeping structure integrity, adhesive bonding is more attractive as compared to mechanical fastening joining methods [5].
Extensive researches have been conducted to investigate the bonded joints through analytical, experimental, and numerical methods.Previous researches focused on different affecting factors on the joint strength and damage mechanism [6][7][8][9][10].
e failure load is found to increase with overlap length and adhesive thickness.Material properties and geometry size have been investigated to significantly affect the joint strength and failure modes.On account of the effect of factors mentioned above, researchers latterly focused on improving the strength of the joints.e joint strength increased by modifying the shape of the joint [11,12] and adding chamfer [13] and fillets [14].e quality of the bonded joints depends highly on the manufacturing process.Some researchers thus presented surface treatment [15,16] on the overlap region and curing conditions such as pressure and temperature.For the purpose of optimizing and designing a high-quality joint, stress distributions over the adhesive layer were obtained through numerical methods [17][18][19].
ese simulation works can also predict the joint strength compared to experimental results.Later the finite element method coupled with the cohesive zone model was performed for failure evolution analysis [20,21]. is can be used to model the failure initiation and further propagation.Although the mechanical behaviors of the adhesive joints have been investigated as mentioned above, the understanding the strength and failure mechanism of the joints is still local and rough due to the complexity of the mechanical behaviors, especially for joints with composite substrates [22].erefore, it is necessary to conduct a detailed research for the CFRP-tosteel adhesive joints.
e present study mainly focused on the mechanical properties and failure behavior of CFRP-to-steel adhesively bonded single lap joints.Different joints were fabricated and tested according to eight different variances, including two kinds of adhesives and four overlap lengths.Mechanical properties were firstly shown and compared with each other.Both the experimental and numerical results about joint strength were then displayed for further detailed analysis.Finite element analysis was then conducted to compare with the experimental results.A detailed stress distribution analysis for various overlap length values was then exhibited, followed by a stress distribution comparison at three typical moments during the tensile process.Failure propagation analysis was carried out for a detailed understanding of the joints' damage evolution.Finally, photographs of the failure joints were exhibited for failure mode analysis.

Experiments
2.1.Materials.Two different structural adhesives were used in this study.Type 1 is 7779 adhesive which belongs to twocomponent polyurethane structural adhesive produced by Ashland.Type 2 (MA830) is a kind of two-liquid acrylic structural adhesive provided by Plexus.Both structural adhesives were prepared and tested to compare the mechanical properties with each other.In this study, the uniaxial tensile test was conducted based on the ISO standard 527-2, which is the determination of tensile properties of plastics.T-peel test was employed with regard to the ISO standard 11339, which is mainly applied for flexible-to-flexible bonded assemblies.And the thick-adherend testing method based on the ISO standard 11003-2 is employed to determine the shear behavior for both structural adhesives.e mechanical properties are finally summarized in Table 1.
e adherends selected for this study were carbon fiberreinforced plastic (CFRP) laminates and DC04 steel commonly used for automobiles.
e CFRP laminates were prepared using the vacuum-assisted resin infusion molding process.Carbon fiber (CC-P400-12) was employed as the reinforcement, and weft and warp were balanced plain woven.Each ply has a thickness of 0.44 mm, and 4 plies were used with 0 °/90 °ply orientation.e epoxy (MA-8931A/B) was selected as the matrix.When the carbon fiber was soaked with resin under the impact of atmospheric pressure, the whole specimen was moved to a heating oven for curing at 120 °C for 6 minutes.e whole specimen was then released from the mould after being cured and cut into the required size.e mechanical properties of the CFRP are listed in Table 2. Another selected adherend is DC04 steel, which is a kind of deep drawing steel with low yield strength and high ductility.It is widely used in the complex parts of automobile.e mechanical properties of the DC04 steel are listed in Table 3.

Single Lap Joints.
e geometry size of the single lap joint for tensile testing according to ISO 4587 is presented in Figure 1. e various overlap length with regard to the experiment requirement between the CFRP laminate and DC04 substrate is denoted as L 0 .L represents the whole length of the specimen.
e compensating plate was prepared with the same thickness corresponding to specific adherend.
According to literature review [14], the abrasive paper of grit size up to 1000 was used to polish the substrates, and acetone was applied to eliminate impurity of the substrate surfaces.Compensating plates were firstly bonded to the corresponding adherends, and lines in both adherends were drawn to dominate the overlap length.Adhesive was then evenly distributed on the overlap region of the DC04 substrates.Four small steel wires were employed to control the adhesive thickness in this process.e two substrates were boned together using a clip, and the whole specimen was finally put into an oven to cure for 2.5 h under the temperature of 80 °C.
A full-factorial experimental design was employed with two levels of adhesive and four levels of overlap length giving total number of 8 tests.e detailed experiment with corresponding factors and levels is listed in Table 4.A quasistatic loading with the velocity of 2 mm/min was applied in tensile testing.For improving experimental accuracy, three replicates were conducted for each trial, and the normalized peak loads and failure displacements were recorded.

Finite Element Analysis
A numerical model was implemented and developed using the commercial software ABAQUS.e objective of finite element analysis (FEA) is to develop a model that could accurately predict the experimental results and present a detailed  2(a).e end with CFRP substrate was xed thoroughly to clamp, while another end with DC04 substrate could only move in the loading direction.e loading was terminated when the displacement reached the set value with regard to experimental loading cases.e element type selected to mesh the adhesive was COH2D4, while the CFRP and DC04 adherends were meshed with C3H20.e mesh was re ned to have more concentration of elements in both the adherends near the adhesive for further stress analysis.e properties for the adherends were mainly based on the results obtained in experiments (Tables 2 and 3).
To reproduce the behaviors of the adhesive, the bilinear traction separate law was used to simulate the elastic behavior up to a peak and subsequent degradation of material properties up to failure.During tensile testing, the damage occurs under mixed-mode loading (Model I, Model II, and Model III). Figure 2(b) shows the bilinear traction separate law under a single loading mode.e curve associates stress with displacements connecting homologous nodes of the cohesive elements.e initial linear elastic corresponds to the rst section until the stress reaching the maximum, after the adhesive sti ness is degraded.
e cohesive failure mainly contains two stages, including damage initiation and crack propagation.In the rst stage, a quadratic nominal stress criterion is used to decide the damage initiation, as expressed below: where σ i and σ 0 i (i I, II, III) are the cohesion and interfacial strength under loading of Model I, Model II, and Model III, respectively.When the sum of the equation on the left is less than 1, there is no initial damage.Otherwise, initial damage will develop in the cohesive layer.B-K fracture criterion is applied to dominate the crack evolution in the second stage, as given below: where G I , G II , and G III represent fracture energy in three directions, respectively G iC (i I, II, III) is the critical strain energy release rate under the respective models, and G C is the total of the three.η is a constant that is related to the properties of the materials.When the left of the equation reaches the value of G C , the initial crack begins to propagate and nally leads to adhesive failure.In order to perform further failure evolution analysis, a parameter de ned as scalar sti ness degradation (SDEG) is used to represent the degradation degree of the adhesive.is parameter can be any value between 0 and 1.When adhesive is in the initial elastic part of mixed-mode loading, the adhesive elements have no damage to any degree and SDEG is set equal to 0. While, the adhesive elements failure completely, SDEG is set equal to 1, and the element is deleted.Advances in Materials Science and Engineering adhesives with various overlap lengths.For joints using 7779 adhesive, the tensile process could be divided into three stages as shown in Figure 3(a).e slopes of the curves are kept constant in the rst stage.When the load reached 2.5 kN, it entered the second stage, and the slope of the curves appeared a slight decrease.is mainly resulted from the strong toughness of the 7779 adhesive.As the adhesive was in force and began to soften, the sti ness of the joint decreased.When the load was between 6.5 and 7.5 kN in the third stage, the slope declined obviously, which was mainly due to that DC04 substrates achieved the yield strength, and plastic deformation occurs.As the loading continued, the joint nally failed when the load exceeded the strength.Since the failure load is the maximum force when the joint su ers failure, it is found that the failure load of the joint increased with the overlap length.For the joints using MA830 adhesive, the failure load also increased as overlap length increased, while the trend of the curves was di erent from those of using 7779 adhesive.When the overlap length was 12.5 and 20 mm, the slope of the curves was constant until complete failure of the joint.e maximum load was less than 6 kN, and thus there was no plastic deformation for the DC04 adherend.However, for the lap length of 30 and 40 mm, the maximum load was higher than 7 kN, and the slope of the curves appeared an obvious decrease.All the four curves showed no decrease when the load reaching 2.5 kN. is was mainly due to the fact that MA830 adhesive was more brittle, and there was no plastic deformation for MA830 adhesive.

Failure Load.
For detailed analysis of the joint, Figure 4 summarizes the failure load and the normalized value.
e error bars are also added.e deviation should be caused by deviation of experimental procedure and the dispersion of properties for the materials, especially the adhesive and the CFRP.Numerical results were found to agree well with the experimental results within 5% relative error.It was found that joints bonded with 7779 adhesive damaged at a higher load (2-3 kN) than that of the joints using MA830 adhesive with the same overlap length.is was mainly due to that the 7779 structural adhesive was much more ductile and  Advances in Materials Science and Engineering exible and could tolerate higher load than MA830 structural adhesive.For the failure load in Figure 4(a), both adhesives increased almost linearly with overlap length.As overlap length increased by 1 mm, the failure load of the 7779 structural adhesive joints increased by an average of about 147 N and that of the MA830 structural adhesive joint increased by an average of about 176 N.Moreover, normalized failure load, which is the failure load divided by the length of the overlap, is de ned and plotted in Figure 4(b).e normalized failure load decreased dramatically with the increase of the overlap length which indicated that increasing the overlap length could strengthen the joint with only a limit degree.

Stress Distribution.
e peel (S 33 ) and shear (S 13 ) stress distributions are compared at the adhesive layer in the middle width for various overlap lengths.All the stresses mentioned above were obtained under 6 kN load.Stress distributions in the adhesive layer of three typical moments during the whole tensile testing procedure were exhibited and compared with each other.Joints with MA830 adhesive were not considered in this section for the similar results.
e peel and shear stress distributions in the adhesive layer for di erent overlap lengths are presented in Figure 5.For the peel stress shown in Figure 5(a), the stress peaks located on the edge of the overlap region.And stress values on the left edge were slightly higher.is is mainly due to the Advances in Materials Science and Engineering two types of adherends with di erent sti ness.us, different degrees of adherend exure appeared at the overlap edges.On the left overlap edge, the higher degree of exure of the DC04 adherend produced higher peak peel stresses.
As the overlap length increased, peel stress peaks decreased on both edges, and stress distributions were relatively much more uniform.For the shear stress exhibited in Figure 5(b), the maximum stress values on the left were 2-6 MPa higher   Advances in Materials Science and Engineering than that on the right, which was mainly due to the fact that the two types of adherends had di erent tensile modulus.Furthermore, stress peaks of several types of joints located near the edge of the overlap rather than on the edge.As the overlap length increased, the position that peak stress appeared was close to both edges. is was mainly due to that under the 6 kN tensile loading, adhesive on both edges appeared degradation of varying degrees. is degeneration was slightly more severe on the left region.With increasing the overlap length, relatively more uniform peel distributions could be found.All the former analysis indicated that the increase of the lap length could reduce the stress concentration, thus leading to the failure load enhancement.However, both types of stress were distributed near the edges of the overlap region.e stresses in most of the middle regions were relatively small.e distribution characteristics indicated that increasing the overlap length can strengthen the joint with only a limit degree.
In order to investigate the stress distribution alongside the tensile loading process, di erent moments representing various stages were selected to analyze the stress state.e stress distributions are shown in Figure 6 with regard to three di erent moments, including t 6 s, t 49 s, and t 110 s, referring to no plastic deformation of the DC04 substrate, large plastic deformation of the DC04 substrate, and approaching to the failure load, respectively.For the moment of t 6 s, the tensile displacement was about 0.2 mm and in the stage of initial loading.e maximum peel stress of the adhesive was 14.92 MPa, and the maximum shear stress was about 13.95 MPa.While, the two stresses were far from their limits.For the moment of t 49 s, the displacement approached to be 1.64 mm, which mainly resulted from the deformation of the CFRP since the CFRP produced larger deformation than that of the DC04 adherend.e e ect of asymmetric rigidity of the adherends could be seen in stress distribution for its inhomogeneous distributions of both shear and peel stresses.On the right side of the adhesive layer, the shear stress dropped dramatically; however, the decline trend on the left was much slower.is was mainly due to that DC04 adherend produced a plastic deformation at this moment so that the tensile force was not transmitted to the adhesive on the right.e maximum peel stress reached 19.16 MPa, which was far from the stress limit (29.43 MPa).However, the highest shear stress was about 20.27 MPa and approached shear stress limits (20.4 MPa).When time is t 110 s, some of the cohesive elements became invalid and were deleted, which meant that complete failure appeared in the adhesive layer.
e whole joint assumed to be almost the maximum load at this moment.e peel stress was 12.55 MPa, still far from reaching the limit stress.
e maximum shear stress was 20.32 MPa and was very close to the stress limit, which implied that the cause of the joint failure was that the shear stress reached the shear limit of the adhesive.
4.4.Failure Propagation.Joint using 7779 structural adhesive with 30 mm overlap length was selected for the failure propagation analysis.As described in Section 3, SDEG is a parameter that represents the degradation degree of the adhesive.Figure 7(a) shows that the adhesive on the left was rst degenerated.is was related to the relatively higher peel and shear stresses which was analyzed in Section 4.2.As the tensile loading continued, adhesive on the right shown in the Figure 7(b) also began to degenerate and the degeneration was much more severe than that on the left.Furthermore, SDEG of adhesive on the right rstly reached the value of 1, and the failure elements were deleted as could be seen in Figure 7(c).is phenomenon was mainly due to that two di erent materials were used as the adherends.Adhesive was xed to the two substrates with various sti ness.For a speci c loading, CFRP produced a larger deformation than the DC04 adherend.us, adhesive on the right edge deformed much more severe, and failure appeared in this region at the early stage.In Figure 7 (d), all the adhesive elements were in failure and deleted, and the DC04 adherends produced permanent plastic deformations in the joint area.e asymmetry degeneration of the adhesive layer could be easily seen from the failure process and this asymmetry distribution mainly resulted from the two di erent adherends.

Failure Modes.
e failure modes of the two adhesives with various overlap lengths are shown in Figure 8.For the 7779 structural adhesive, there were two main failure modes including cohesion failure and interface failure.When the overlap length was 12.5 mm, it merely presented cohesion failure as shown in Figure 8(a).As the overlap length increased, the bending moment caused by load eccentricity became more severe.
e peel stress on the edge of the overlap produced higher values and reached the peel stress limit of the adhesive.us, interface failure occurred and   Advances in Materials Science and Engineering became one of the failure modes as seen in Figure 8. Furthermore, the failure of bonding interface only occurred on the bonding surface between adhesive layer and DC04 substrate.is indicated that the 7779 adhesive had a better connection with CFRP adherends than DC04 adherends.As for joints of MA830 structural adhesive as shown in Figure 8, there was only cohesion failure for all di erent overlap lengths. is is mainly owing to the fact that joints of MA830 adhesive fail at a relative low load.us, peel stresses on the edge of the overlap were far from the stress limit of the adhesive, and there was no interface failure.e detailed observations of the morphology for the facture surfaces were conducted through a digital microscope (KEYENCE VHX-5000).
e morphologies of the facture surfaces are presented in Figure 9.As shown in Figure 9(a), for the 7779 adhesive in the joint, the adhesive is completely peeled o from the CFRP, and the CFRP is obviously visible in this specimen.us, this observation rmly con rms the interface failure model.On the contrary, as shown in Figure 9(b), residual MA830 adhesive can be clearly found on the CFRP, demonstrating the cohesion failure occurred within the adhesive.

Conclusions
e present work investigated the in uences of the adhesive type and overlap length on the mechanical behavior and failure modes of bonded single lap joints via experimental and numerical studies.Main conclusions can be drawn as follows: (1) e joint strength was found to be highly dependent on adhesive type and overlap length.Joints using 7779 structural adhesive provided 2-3 kN failure load higher than that using MA830 structural adhesive.And failure load with two adhesives increased about 147∼176 N with increasing 1 mm of the overlap length.(2) A nite element analysis model was established, and numerical results were within 5% relative error on predicting the failure load and joint strength compared with the experimental results.ere was severe stress concentration on the edge of the joint.And the stress distributions in the overlap area were nonuniform.e peel stress was responsible for the interface failure, and the cohesion failure is mainly caused by large shear stress.e shear stress reaching the limit stress of the adhesive was the main reason for the failure of the adhesive layer.(3) An asymmetry degeneration of the adhesive layer could be easily seen from the failure process analysis, which was mainly due to the asymmetric rigidity of the adherends.

Figure 4 :Figure 5 :
Figure 4: Experiment and simulation results: (a) failure load and (b) normalized failure load (failure load divided by overlap length).

Figure 6 :
Figure 6: Peel and shear stress distributions at three di erent moments of the joint using 7779 adhesive with 30 mm overlap length.(a) t 6 s; (b) t 49 s; (c) t 110 s.

Figure 9 :
Figure 9: e morphology of the facture surfaces con rms the (a) interface failure and (b) cohesion failure (observed through a digital microscope KEYENCE VHX-5000).

Table 1 :
Mechanical properties of the two adhesives.

Table 2 :
Mechanical properties of the CFRP adherend.
2Advances in Materials Science and Engineering stress distribution and failure evolution analysis of the joints.e single lap joint tests were numerically built in a threedimensional model with the geometry and boundary conditions exhibited in Figure

Table 4 :
Full-factorial experimental design with corresponding factors and levels.

Table 3 :
Mechanical properties of the DC04 adherend.