- Email: [email protected]
Failure of adhesively bonded metallic T-joints subjected to quasi-static and crash loading
Accepted Manuscript Failure of adhesively bonded metallic T-joints subjected to quasi-static and crash loading Michael May, Olaf Hesebeck PII: DOI: Reference:
S1350-6307(14)00370-7 http://dx.doi.org/10.1016/j.engfailanal.2014.12.007 EFA 2470
To appear in:
Engineering Failure Analysis
Received Date: Revised Date: Accepted Date:
21 July 2014 11 December 2014 15 December 2014
Please cite this article as: May, M., Hesebeck, O., Failure of adhesively bonded metallic T-joints subjected to quasistatic and crash loading, Engineering Failure Analysis (2014), doi: http://dx.doi.org/10.1016/j.engfailanal. 2014.12.007
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Failure of adhesively bonded metallic T-joints subjected to quasi-static and crash loading Michael May1,*, Olaf Hesebeck2 1
Fraunhofer Institute for High-Speed Dynamics, Ernst-Mach-Institute, EMI, Eckerstrasse 4, 79104
Freiburg, Germany 2
Fraunhofer Institute for Manufacturing Technology and Advanced Materials, IFAM, Wiener Straße 12,
28359 Bremen, Germany
* Corresponding author, Email: [email protected], Tel: +49 (0)761 2714 – 337, Fax: +49 (0)761 2714 - 1337
Abstract
Experimental studies on adhesively bonded metallic joints are presented. The T-joints were mounted in a servo-hydraulic machine and loaded in three different directions (rear, front and side) and at three different loading velocities (0.5 mm/s, 50 mm/s and 5000 mm/s). The variation in loading direction allowed triggering of different failure mechanisms: metal plasticity, peel failure of the adhesive, shear failure of the adhesive and combined failure. Some scatter was seen in the joint performance which can be attributed to two major manufacturing parameters. The use of glass spheres for defining the bond line thickness has a positive effect on joint performance. Interestingly, T-joints manufactured from two different batches of adhesive showed large differences in behaviour although the manufacturing parameters were the exactly same.
Keywords: Decohesion, Failure mechanism, Mechanical testing, Plastic deformation, Strain rate
1
1.
Introduction
The limited availability of natural resources such as oil and the increasing environmental consciousness in the general public have driven federal organizations all over the world to impose strict regulations on the Automotive industry regarding sustainability. In recent years developments in the Automotive industry have therefore been driven by attempts to enhance sustainability by reducing fuel consumption and consequently CO2 emissions. On the one hand, developments have investigated novel engine concepts such as hybrid engines (see for example Toyota Prius) or electric engines (see for example BMW i3). On the other hand, developments have aimed at reducing vehicle weight. Structural adhesives offer significant potential for lightweight design as joining does not require bolts or rivets which cause a weight penalty as well as stress concentrations and therefore potential weak spots in the structure. However, debonding of adhesively bonded components is of major concern in the design process. At the same time passenger safety remains critical for the design process. Adhesively bonded T-joint structures are a popular validation case for material models for adhesive joints [1, 2]. Recently, May and co-workers [3, 4] proposed a novel geometry for adhesively bonded T-joint structures allowing quasi-static and crash loading under the same boundary conditions. In this paper, the same specimen geometry and test setup is used. In previous work it was shown that loading direction and loading speed have some influence on the failure behavior of this T-joint structure. In this paper the underlying processes leading to failure are investigated in more detail. Special attention is given to aspects of manufacturing and loading rate. Additional, previously unpublished, test data on a third loading direction (rear impact) is provided.
2
2. Manufacturing of T-joints
Manufacturing of adhesively bonded joints was carried out at the Fraunhofer Institute for Advanced Manufacturing Technology and Advanced Materials, IFAM, in Bremen. A total of 37 T-joints were manufactured using sheets of high strength steel DP-K 30/50, thickness 1.210± 0.005 mm and crash-optimized adhesive BETAMATE 1496 V. The joints consist of a rectangular profile and a top hat profile. Manufacturing was done in several steps. First, the flat steel sheets were bent into the require shapes (see Fig. 1).
Figure 1 about here Figure 1: Baseline material for profiles. Left: top hat profile, right: rectangular profile
Then, the sheets were cleaned using an ultrasonic bath and the solvents methyl ethyl ketone (MEK), isopropanol and acetone. The surface was then activated using low pressure plasma. This activation process was required as the steel used here does not contain any zinc. In a second step, the two individual profiles were manufactured. The adhesive was heated to 40°C and applied to the bent profiles. Glass spheres were used to ensure a bond line thickness of 0.3 mm. The flat sheets were then added and fixed with clamps. In a final step, the two profiles were adhesively bonded together (see Fig. 2). Again, the bond line thickness was adjusted using glass spheres of thickness 0.3 mm. The adhesive joints were then cured for 30 minutes using an oven heated to 180°C. The finished T-joint is shown in Fig. 3.
Figure 2 about here
Figure 2: Assembly of the T-joint from the two individual profiles.
It is noted that additional adhesive had to be ordered over the course of the project. Manufacturing of the T-joint structures was therefore performed with materials from two different deliveries. T-joints manufactured using the initially delivered adhesive are
3
subsequently called “Batch A”, T-joints manufactured using the material supplied at a later stage are subsequently called “Batch B”. It is noted that this is the only difference in the manufacturing process between “Batch A” and “Batch B” T-joints. Also, the use of glass spheres to adjust the thickness of the bond line was omitted for two T-joints.
Figure 3 about here
Figure 3: T-joint structure including approximate dimensions in mm.
3. Mechanical testing of T-joints
The T-joints were mounted in a servo-hydraulic testing machine, type Instron 8503, allowing test speeds of up to 5 m/s. The T-joints were clamped at the outer edges of the top hat profile and mounted onto a steel frame. The rig can be rotated by 90° allowing loading the T-joints in different directions. A semi-spherical Aluminium impactor (spherical segment of diameter 96 mm, height 24 mm, diameter of the full sphere 120 mm) was instrumented with a 60 kN load cell. All tests were recorded with PHOTRON APS high-speed cameras with a maximum frame rate of 100,000 fps. This allowed detailed analyses of the damage and failure sequences occurring during the tests. Three different loading directions were considered: Rear impact loading (impacting the rectangular profile on the back sheet, marked with an X in Fig. 3), front impact loading (loading the rectangular profile on the opposite side to the rear impact configuration), and side impact loading (loading the rectangular profile in the direction of the axis of the top hat profile). The experiments were performed at three different loading rates: 0.5 mm/s, 50 mm/s, and 4800 mm/s. At least four T-joints were tested for each loading rate and direction. Figure 4 exemplarily shows the side impact configuration.
4
Figure 4 about here
Figure 4: Test setup for side impact loading.
3.1. Rear impact loading
The force-displacement curves recorded during rear impact loading are shown in Fig. 5.
Figure 5 about here
Figure 5: Force displacement curves recorded during rear impact loading. a) loading speed 0.5 mm/s. b) Loading speed 50 mm/s. c) Loading speed 4800 mm/s. One T-joint tested under high rates of loading was manufactured in batch B (shown in dark grey); all others were manufactured in batch A.
The force-displacement curves are characterized by three different regions. In the first part of the curve, up to approximately 10 mm displacement, the impactor is causing plastic deformation in the rectangular profile. Then, contact occurs between the rectangular profile and the top hat profile. This contact is caused by a small free edge sticking out of the rectangular profile on the opposite side of the loaded back sheet. If the rectangular profile is struck by the impactor, rotation starts. Once this free edge gets in contact with the radius on the top of the top hat profile, a stiffening effect is seen.
Figure 6 about here
Figure 6: Close-up view on the free edge causing contact between the rectangular profile and the top hat profile.
5
The last part of the force-displacement curve is governed by two mechanisms: Plastic deformation of the top hat profile and sliding of the impactor along the rectangular profile. For the case of rear impact loading no failure of adhesive bond lines was observed. A mild rateeffect was observed for the force-displacement curves recorded during quasi-static and medium rate loading: The peak loads recorded for medium rate loading are about 10% higher than the peak loads recorded for quasi-static loading. For high-rate loading the peak load increases even further, accompanied by some typical oscillations in the load signal. The typical damage sequence observed during rear impact loading is shown in Fig. 7.
Figure 7 about here
Figure 7: Damage sequence for a T-joint subjected to rear impact loading, loading speed 0.5 mm/s. Left: Beginning of test; middle: Plastic deformation of the rectangular profile, contact of the two profiles, beginning of plastic deformation of top hat profile; Right: end of test, large plastic deformation of the top hat profile.
3.2. Front impact loading
The force-displacement curves recorded during front impact loading are shown in Fig. 8.
Figure 8 about here
Figure 8: Force displacement curves recorded during front impact loading. a) loading speed 0.5 mm/s. b) Loading speed 50 mm/s. c) Loading speed 4800 mm/s. All T-joints tested at low rates were manufactured in batch A. One of these did not have glass spheres (shown in dashed line). One Tjoint tested at intermediate rate was manufactured in batch B (shown in dark grey); all T-joints tested at high rate were manufactured in batch B.
6
All T-joints tested under quasi-static loading conditions were manufactured in batch A. Here, the performance is governed by the peel properties of the adhesive. First, the load increases until peel failure starts, with increasing amount of debonding, the load is slowly reduced until catastrophic failure occurs, characterized by a sudden load drop. As highlighted in [4], the load does not drop to zero due to bending of the steel back face. High-speed video footage in Fig. 9 illustrates the typical failure sequence.
Figure 9 about here
Figure 9: Failure sequence observed during front impact loading, loading speed 50 mm/s. Left: beginning of test; middle: debonding; right: final failure.
Figure 8 shows that one T-joint showed weaker performance in terms of displacement at failure and peak load. In a first step it was investigated if this reduction in performance was due to differences in the area of the fracture surface. This analysis considered the “real” area of the fracture surface including contributions from partially fractured spew fillets. The mean area of the fracture surface was 1592 mm² with a coefficient of variation of 4.1 %. No clear correlations were found between the area of the fracture surfaces and the performance of the joint. In a second step, the fracture surfaces were examined. Here some differences were observed. The Tjoint with the weakest performance has failed in a mixed cohesive/adhesive failure mode whilst the remaining three T-joints failed cohesively as shown in Figure 10.
Figure 10 about here
Figure 10: Fractured surfaces in quasi-static front impact test. Top: Typical cohesive failure occurring in most cases; bottom: mixed cohesive adhesive failure occurring in absence of glass spheres.
7
It is also noted that the T-joint with the weakest performance was manufactured without the use of glass spheres. It is therefore assumed that the thickness of the adhesive bond line was not adjusted correctly, leading to a change of failure mode and joint performance.
A rate effect is seen for medium rate loading resulting in an increase of peak load with increasing loading rate, see Figure 8 middle.The experimental evidence for medium rate loading shows that one of the T-joints performed differently to the other joints. This particular joint is characterized by a less pronounced and substantially smaller peak load and smaller displacement at failure compared to the other T-joints tested at medium rate. The areas of the fractured surfaces were determined, but again did not show a clear correlation to the performance of the joints. The mean area of the fracture surface was 1587 mm² with a coefficient of variation of 8.9 %. It is noted that T-joints tested under medium rate loading were taken from two different batches of manufacturing. Three T-joints were manufactured in batch A, one T-joint was manufactured in batch B. The joint with the weakest performance was the joint originating from batch B. Each of the four T-joints failed cohesively. However, some difference was observed during the investigation of the fracture surface. The T-joint manufactured in batch B showed some darker blue color in one of the corners, see Figure 11. Optical microscopy revealed that failure in the area of dark blue color was a combination of cohesive and adhesive fracture; in the other areas failure was found to be purely cohesive.
Figure 11 about here
Figure 11: Comparison of fracture surfaces after front impact, v= 50 mm/s. Top: Typical T-joint manufactured in batch A; bottom: T-joint manufactured in batch B.
Due to dynamic effects, the load signal for high rate loading looks significantly different to the load signals recorded for quasi-static and medium rate loading. However, the scatter in terms of fractured surface (mean 1570 mm², coefficient of variation of 2.0%), peak load and displacement at failure is small. All T-joints tested under high rate loading were manufactured
8
in batch B. It is interesting to note that each fracture surface of T-joints tested under high rate loading showed the same dark blue color, see Figure 12.
Figure 12 about here
Figure 12: Typical fracture surface after front impact tests at high rate.
Peel failure is dominating in front impact loading. In [5] Tapered Double Cantilever Beam (TDCB) tests were presented indicating some influence of the loading rate on the location of failure. For high loading rates failure was always cohesive, for low loading rates the location of failure moved towards adherent-adhesive interface. This behavior was not seen in the T-joints subjected to front impact loading.
3.3 Side impact loading
The force-displacement curves recorded during side impact loading are shown in Fig. 13. Figure 13 about here
Figure 13: Force displacement curves recorded during side impact loading. a) loading speed 0.5 mm/s. b) Loading speed 50 mm/s. c) Loading speed 4800 mm/s. One T-joint tested under quasistatic loading was manufactured in batch A (shown in light grey); all other T-joints were manufactured in batch B. One T-joint tested under quasi-static loading did not have glass spheres (indicated by dashed line).
The typical sequence of events observed for side impact loading is more complex than for the previous load cases, see Fig. 14. At the beginning, plastic deformation of the rectangular profile is observed underneath the impactor. Then, damage initiates in the adhesive joint connecting the top surface of the top hat profile to the rectangular profile. In most cases, a fold is created in the metallic back sheet of the rectangular profile. After failure of this joint, folding is pronounced by rotation of the rectangular profile. At the same time, the back sheet starts debonding from the
9
bent section of the rectangular profile. Final failure is governed by shear failure of the adhesive joint connecting the back sheet of the rectangular profile to the side of the top hat profile.
Figure 14 about here
Figure 14: Typical failure sequence observed during medium rate and high rate side impact loading. From left to right: beginning of test; plastic deformation of rectangular profile and damage initiation at upper bond line and beginning of fold formation; failure of upper bond line, damage initiation in back face bond line; final failure.
In some cases, especially for low loading rates, the behaviour observed was slightly different. For these cases no or only limited fold formation was observed. As a consequence, catastrophic failure of the bond line connecting the back sheet to the top hat profile occurred. Looking at the force-displacement curves this manifests itself in a sudden load drop at a displacement of about 15 mm. For these cases no force plateau was observed after the initial load drop. The force plateau seen in most of the experiments is therefore attributed to the plasticity in the steel and formation of the fold shown in Fig. 14.
The T-joints tested under quasi-static loading conditions were taken from two different manufacturing batches. Three T-joints were taken from batch B, one T-joint was taken from batch A. This loading configuration is very interesting since the observed behavior varies significantly as seen in Figure 13 left. One T-joint failed catastrophically at a displacement of only about 11 mm. Also the maximum failure load was lower than for the remaining three Tjoints. Two T-joints can be clustered as both show the same type of behavior expressed by a maximum displacement at failure of approximately 18 mm. Finally, one T-joint formed a load plateau at about 3 kN starting at a displacement of about 18 mm. Final failure occurred at a displacement of 30 mm which is significantly larger than the displacements at failure observed for the other joints. Similar to the investigations for the front impact loading configuration, this scatter in joint performance cannot be correlated to different areas of the bond lines. The mean
10
area of the fracture surface on the side was 943 mm² with a coefficient of variation of 5.6 %. Recalling the analysis of the front impact there was a trend that the T-joints manufactured in batch A were stronger than the joints manufactured in batch B despite the fact that both batches were manufactured by the same technician under the same conditions. The same trend is seen here. The only joint forming the plateau was the also the only joint manufactured in batch A. The joint with the poor performance was manufactured without the use of glass spheres to adjust the thickness of the bond lines. Similarly to the front impact configuration, variations in thickness of the bond line have some significant influence on the mechanical performance of the joint. Numerical simulations showed a strong influence of the shear properties of the adhesive on the performance of the T-joint subjected to side impact loading [6]. A high shear strength of the adhesive was found to be critical for the fold formation in the back sheet of the adhesive joint and therefore for the formation of the plastic plateau. This is reflected in the experiments as the T-joints showed a more pronounced fold (height of the fold measured from the original flat position) with increasing joint performance. The characteristic size of the fold is the height of the fold measured from the original flat position. The weakest T-joint which was manufactured without glass spheres did not form a fold at all. The intermediate T-joints formed a mild fold of size less than 0.5 mm. The T-joint with the best performance formed a fold of size 5 mm. Figure 15 highlights the difference in the size of the fold for the two extreme cases:
Figure 15 about here
Figure 15: Comparison of the characteristic folds in metallic steel sheets. Left: strong fold formation was observed for the T-joint with the best performance; right: no fold was observed for the T-joint with the weakest performance.
All T-joints tested under medium rate loading conditions were manufactured in batch B. Here, the length of the plateau varies significantly resulting in a scatter for the displacement at failure ranging from 21 mm to 38 mm. Again, this cannot be correlated to the area of the fracture surfaces. The mean area of the fracture surface on the side was 949 mm² with a coefficient of
11
variation of 5.8 %. Also, all of the T-joints tested here were manufactured using glass spheres. Therefore the differences cannot be attributed to differences in the thickness of the bond line. Similarly to the quasi-static load case it was found that the length of the plateau can be correlated to the characteristic size of the fold in the metal. The characteristic size of the fold of the joint with the smallest displacement at failure is only 1.5 mm. For the joints with intermediate performance, the characteristic size of the fold is about 5 mm. For the joints with the strongest performance, the characteristic size of the fold is about 7 mm.
All T-joints tested under high rate loading conditions were manufactured in batch B. The results are very consistent with only one T-joint failing prematurely. Again, there was no correlation between the area of the fractured surface and the displacement at failure. The mean area of the fracture surface on the side was 905 mm² with a coefficient of variation of 2.3 %. However, the characteristic size of the fold shows a correlation to the displacement at failure. For the weakest T-joint, the size of the fold is about 4 mm, for the other cases the size of the fold is about 8 mm. For all cases failure occurred cohesively inside the adhesive. However, the fracture pattern at the bond line at the side of the T-joint with the weakest performance had a slightly different characteristic to the fracture pattern observed for the other cases as shown in Figure 16.
Figure 16 about here
Figure 16: Comparison of fracture patterns for the adhesive bond line failing in shear during highrate side impact loading. Left: Typical T-joint with good performance; right: T-joint with weak performance.
In [7], Marzi presented some rate dependent Tapered End Notched Flexure (TENF) tests on the same adhesive. Marzi found that shear failure was cohesive for high rates of loading and close to the adhesive-adherent-interface for low loading velocities. This trend was not seen here, except for one T-joint which failed adhesively under quasi-static loading conditions. However,
12
this can be attributed to the lack of glass spheres for this particular joint and is not a result of loading rate.
4. Discussion
Adhesively bonded metallic T-joints were tested under different loading conditions. Both, the loading direction and the loading velocity were varied resulting in a broad database of 9 different configurations. Failure in T-joints subjected to front impact loading was dominated by peel failure. Failure in T-joints subjected to side impact loading was a combination of peel and shear failure of the adhesive and plastic deformation of the steel sheets. T-joints subjected to rear impact loading showed large plastic deformation without causing failure of any of the adhesive bond lines.
Strain rate effects were observed for all loading directions. In general there was a trend towards increasing peak loads and increasing displacements at failure with increasing loading rate.
Scatter in the experiments can be traced back to the manufacturing of the T-joints. Two T-joints showed significantly weaker performance than the remaining T-joints. During the manufacturing process of these two test pieces, glass spheres were not used for ensuring the correct thickness of the bond line. As a consequence, failure of the bond line shifted from the desired cohesive failure mode towards a mixed adhesive/cohesive failure mode. Consequently, the performance of these joints was weaker than for the remaining joints.
The T-joints were manufactured in two batches by the same technician using an identical process. The only difference is the batch of the adhesive used for manufacturing. Despite ensuring perfect repeatability of the process there is evidence that T-joints manufactured in the second batch (batch B) are weaker than T-joints manufactured in the first batch. The difference is most distinct for the load case “side impact, quasi-static”. Here, the T-joint manufactured in batch A showed a plateau in the force-displacement curve which was not seen for the T-joints
13
manufactured in batch B. The shear strength of the adhesive is governing the formation of the plateau as higher resistance in the bond line at the side of the joint causes the metallic back sheet to deform plastically and form a fold. The displacement at failure in side impact loading can be directly correlated to the characteristic size of the fold as shown in Figure 16. The characteristic size of the fold is the height of the fold measured from the original flat position. Measurements taken from joints tested at quasi-static loading rate are marked with a diamond; measurements taken from joints tested at a medium rate are marked with squares; measurements from high rate tests are marked in triangles. A linear relationship is found between the size of the fold and the displacement at failure, see Fig. 17. No additional rate effect is seen here.
Figure 17 about here
Figure 17: Correlation of the characteristic size of the fold in side impact configuration and the recorded displacement at failure.
For some cases, mixed adhesive/cohesive fracture was found which resulted in dark blue colors of the fractured surfaces. The reason for this behaviour remains unclear. No ductile-brittle transition was seen which could explain differences between quasi-static and high-rate tests. However, BETAMATE 1496V is a multi-phase system which may result in an even more complex behaviour in reality. Variation in bond-line thickness can be excluded as a cause for this behaviour as a sufficient amount of glass spheres was found in both – the mixed fracture zones as well as the purely cohesive fracture zones. Consequently there is a chance that this behaviour is also a result of differences in the manufacturing batch.
5. Conclusions
14
In summary it became clear that adhesively bonded metallic T-joints are very sensitive to manufacturing. The different behaviour observed for the two different batches is attributed to differences in shear strength of the adhesive bond lines. Similar batch-dependent properties have also previously been reported for other structural adhesives produced by other manufacturers [8]. We therefore recommend performing comparative studies on a typical shear test such as the single lap shear test [9] before using new batches of adhesive for manufacturing crash relevant large structures.
Acknowledgements The IGF project 338 ZN ”Robustheit und Zuverlässigkeit der Berechnungsmethoden von Klebverbindungen mit hochfesten Stahlblechen unter Crashbelastungen” of the research association Forschungsvereinigung Stahlanwendung e.V. - FOSTA, Sohnstrasse 65, 40237 Düsseldorf was funded by the AiF under the program for the promotion of joint industrial research and development (IGF) by the Federal Ministry of Economics and Technology based on a decision of the German Bundestag. The authors would like to thank Mr. Holger Voß for the experimental support.
References
[1] M. Wissling and O. Hahn, "Adhesive Joints under Crash Loading: Experiments and Modelling of the Failure Behaviour," Welding in the World, vol. 51, pp. 63-73, 2007.
[2] P. Jousset and M. Rachik, "Implementation, identification and validation of an elasto-plasticdamage model for the finite element simulation of structural bonded joints," Int J Adhes Adhes, vol. 50, pp. 107-116, 2014.
15
[3] M. May and H. Voß, "Experimental and Numerical Analyses of Adhesively Bonded T-joints under Crash Loading," in EPJ Web of Conferences 26, 2012.
[4] M. May, H. Voß and S. Hiermaier, "Predictive modeling of damage and failure in adhesively bonded metallic joints using cohesive interface elements," Int J Adhes Adhes, pp. 7-17, 2014.
[5] M. May, O. Hesebeck, S. Marzi, W. Böhme, J. Lienhard, S. Kilchert, M. Brede and S. Hiermaier, "Rate dependent behaviour of crash-optimized adhesives - Experimental characterization, model development, and simulation," Engng Fract Mech, vol. 133, pp.112137,2015.
[6] M. May and O. Hesebeck, "Assessment of experimental methods for calibrating ratedependent cohesive zone models for predicting failure in adhesively bonded metallic structures," submitted to Engineering Failure Analysis, 2014.
[7] S. Marzi, "Measuring the critical energy release rate in mode II of tough, structural adhesive joints using the tapered end-notched flexure (TENF) test," The European Physical Journal Special Topics, vol. 206, pp. 35-40, 2012.
[8] M. Schlimmer, "Methodenentwicklung zur Berechnung und Auslegung geklebter Stahlbauteile für den Fahrzeugbau," final report of AiF-Zutech project 593, Kassel, 2005.
[9] ASTM, ASTM D-1002-10:Standard Test Method for Apparent Shear Strength of Single-LapJoint Adhesively Bonded Metal Specimens by tension Loading (Metal-to-metal), ASTM, 2010.
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
• • •
Testing of adhesive structures Rate- and loading direction dependent behaviour Influence of manufacturing
34
S1350-6307(14)00370-7 http://dx.doi.org/10.1016/j.engfailanal.2014.12.007 EFA 2470
To appear in:
Engineering Failure Analysis
Received Date: Revised Date: Accepted Date:
21 July 2014 11 December 2014 15 December 2014
Please cite this article as: May, M., Hesebeck, O., Failure of adhesively bonded metallic T-joints subjected to quasistatic and crash loading, Engineering Failure Analysis (2014), doi: http://dx.doi.org/10.1016/j.engfailanal. 2014.12.007
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Failure of adhesively bonded metallic T-joints subjected to quasi-static and crash loading Michael May1,*, Olaf Hesebeck2 1
Fraunhofer Institute for High-Speed Dynamics, Ernst-Mach-Institute, EMI, Eckerstrasse 4, 79104
Freiburg, Germany 2
Fraunhofer Institute for Manufacturing Technology and Advanced Materials, IFAM, Wiener Straße 12,
28359 Bremen, Germany
* Corresponding author, Email: [email protected], Tel: +49 (0)761 2714 – 337, Fax: +49 (0)761 2714 - 1337
Abstract
Experimental studies on adhesively bonded metallic joints are presented. The T-joints were mounted in a servo-hydraulic machine and loaded in three different directions (rear, front and side) and at three different loading velocities (0.5 mm/s, 50 mm/s and 5000 mm/s). The variation in loading direction allowed triggering of different failure mechanisms: metal plasticity, peel failure of the adhesive, shear failure of the adhesive and combined failure. Some scatter was seen in the joint performance which can be attributed to two major manufacturing parameters. The use of glass spheres for defining the bond line thickness has a positive effect on joint performance. Interestingly, T-joints manufactured from two different batches of adhesive showed large differences in behaviour although the manufacturing parameters were the exactly same.
Keywords: Decohesion, Failure mechanism, Mechanical testing, Plastic deformation, Strain rate
1
1.
Introduction
The limited availability of natural resources such as oil and the increasing environmental consciousness in the general public have driven federal organizations all over the world to impose strict regulations on the Automotive industry regarding sustainability. In recent years developments in the Automotive industry have therefore been driven by attempts to enhance sustainability by reducing fuel consumption and consequently CO2 emissions. On the one hand, developments have investigated novel engine concepts such as hybrid engines (see for example Toyota Prius) or electric engines (see for example BMW i3). On the other hand, developments have aimed at reducing vehicle weight. Structural adhesives offer significant potential for lightweight design as joining does not require bolts or rivets which cause a weight penalty as well as stress concentrations and therefore potential weak spots in the structure. However, debonding of adhesively bonded components is of major concern in the design process. At the same time passenger safety remains critical for the design process. Adhesively bonded T-joint structures are a popular validation case for material models for adhesive joints [1, 2]. Recently, May and co-workers [3, 4] proposed a novel geometry for adhesively bonded T-joint structures allowing quasi-static and crash loading under the same boundary conditions. In this paper, the same specimen geometry and test setup is used. In previous work it was shown that loading direction and loading speed have some influence on the failure behavior of this T-joint structure. In this paper the underlying processes leading to failure are investigated in more detail. Special attention is given to aspects of manufacturing and loading rate. Additional, previously unpublished, test data on a third loading direction (rear impact) is provided.
2
2. Manufacturing of T-joints
Manufacturing of adhesively bonded joints was carried out at the Fraunhofer Institute for Advanced Manufacturing Technology and Advanced Materials, IFAM, in Bremen. A total of 37 T-joints were manufactured using sheets of high strength steel DP-K 30/50, thickness 1.210± 0.005 mm and crash-optimized adhesive BETAMATE 1496 V. The joints consist of a rectangular profile and a top hat profile. Manufacturing was done in several steps. First, the flat steel sheets were bent into the require shapes (see Fig. 1).
Figure 1 about here Figure 1: Baseline material for profiles. Left: top hat profile, right: rectangular profile
Then, the sheets were cleaned using an ultrasonic bath and the solvents methyl ethyl ketone (MEK), isopropanol and acetone. The surface was then activated using low pressure plasma. This activation process was required as the steel used here does not contain any zinc. In a second step, the two individual profiles were manufactured. The adhesive was heated to 40°C and applied to the bent profiles. Glass spheres were used to ensure a bond line thickness of 0.3 mm. The flat sheets were then added and fixed with clamps. In a final step, the two profiles were adhesively bonded together (see Fig. 2). Again, the bond line thickness was adjusted using glass spheres of thickness 0.3 mm. The adhesive joints were then cured for 30 minutes using an oven heated to 180°C. The finished T-joint is shown in Fig. 3.
Figure 2 about here
Figure 2: Assembly of the T-joint from the two individual profiles.
It is noted that additional adhesive had to be ordered over the course of the project. Manufacturing of the T-joint structures was therefore performed with materials from two different deliveries. T-joints manufactured using the initially delivered adhesive are
3
subsequently called “Batch A”, T-joints manufactured using the material supplied at a later stage are subsequently called “Batch B”. It is noted that this is the only difference in the manufacturing process between “Batch A” and “Batch B” T-joints. Also, the use of glass spheres to adjust the thickness of the bond line was omitted for two T-joints.
Figure 3 about here
Figure 3: T-joint structure including approximate dimensions in mm.
3. Mechanical testing of T-joints
The T-joints were mounted in a servo-hydraulic testing machine, type Instron 8503, allowing test speeds of up to 5 m/s. The T-joints were clamped at the outer edges of the top hat profile and mounted onto a steel frame. The rig can be rotated by 90° allowing loading the T-joints in different directions. A semi-spherical Aluminium impactor (spherical segment of diameter 96 mm, height 24 mm, diameter of the full sphere 120 mm) was instrumented with a 60 kN load cell. All tests were recorded with PHOTRON APS high-speed cameras with a maximum frame rate of 100,000 fps. This allowed detailed analyses of the damage and failure sequences occurring during the tests. Three different loading directions were considered: Rear impact loading (impacting the rectangular profile on the back sheet, marked with an X in Fig. 3), front impact loading (loading the rectangular profile on the opposite side to the rear impact configuration), and side impact loading (loading the rectangular profile in the direction of the axis of the top hat profile). The experiments were performed at three different loading rates: 0.5 mm/s, 50 mm/s, and 4800 mm/s. At least four T-joints were tested for each loading rate and direction. Figure 4 exemplarily shows the side impact configuration.
4
Figure 4 about here
Figure 4: Test setup for side impact loading.
3.1. Rear impact loading
The force-displacement curves recorded during rear impact loading are shown in Fig. 5.
Figure 5 about here
Figure 5: Force displacement curves recorded during rear impact loading. a) loading speed 0.5 mm/s. b) Loading speed 50 mm/s. c) Loading speed 4800 mm/s. One T-joint tested under high rates of loading was manufactured in batch B (shown in dark grey); all others were manufactured in batch A.
The force-displacement curves are characterized by three different regions. In the first part of the curve, up to approximately 10 mm displacement, the impactor is causing plastic deformation in the rectangular profile. Then, contact occurs between the rectangular profile and the top hat profile. This contact is caused by a small free edge sticking out of the rectangular profile on the opposite side of the loaded back sheet. If the rectangular profile is struck by the impactor, rotation starts. Once this free edge gets in contact with the radius on the top of the top hat profile, a stiffening effect is seen.
Figure 6 about here
Figure 6: Close-up view on the free edge causing contact between the rectangular profile and the top hat profile.
5
The last part of the force-displacement curve is governed by two mechanisms: Plastic deformation of the top hat profile and sliding of the impactor along the rectangular profile. For the case of rear impact loading no failure of adhesive bond lines was observed. A mild rateeffect was observed for the force-displacement curves recorded during quasi-static and medium rate loading: The peak loads recorded for medium rate loading are about 10% higher than the peak loads recorded for quasi-static loading. For high-rate loading the peak load increases even further, accompanied by some typical oscillations in the load signal. The typical damage sequence observed during rear impact loading is shown in Fig. 7.
Figure 7 about here
Figure 7: Damage sequence for a T-joint subjected to rear impact loading, loading speed 0.5 mm/s. Left: Beginning of test; middle: Plastic deformation of the rectangular profile, contact of the two profiles, beginning of plastic deformation of top hat profile; Right: end of test, large plastic deformation of the top hat profile.
3.2. Front impact loading
The force-displacement curves recorded during front impact loading are shown in Fig. 8.
Figure 8 about here
Figure 8: Force displacement curves recorded during front impact loading. a) loading speed 0.5 mm/s. b) Loading speed 50 mm/s. c) Loading speed 4800 mm/s. All T-joints tested at low rates were manufactured in batch A. One of these did not have glass spheres (shown in dashed line). One Tjoint tested at intermediate rate was manufactured in batch B (shown in dark grey); all T-joints tested at high rate were manufactured in batch B.
6
All T-joints tested under quasi-static loading conditions were manufactured in batch A. Here, the performance is governed by the peel properties of the adhesive. First, the load increases until peel failure starts, with increasing amount of debonding, the load is slowly reduced until catastrophic failure occurs, characterized by a sudden load drop. As highlighted in [4], the load does not drop to zero due to bending of the steel back face. High-speed video footage in Fig. 9 illustrates the typical failure sequence.
Figure 9 about here
Figure 9: Failure sequence observed during front impact loading, loading speed 50 mm/s. Left: beginning of test; middle: debonding; right: final failure.
Figure 8 shows that one T-joint showed weaker performance in terms of displacement at failure and peak load. In a first step it was investigated if this reduction in performance was due to differences in the area of the fracture surface. This analysis considered the “real” area of the fracture surface including contributions from partially fractured spew fillets. The mean area of the fracture surface was 1592 mm² with a coefficient of variation of 4.1 %. No clear correlations were found between the area of the fracture surfaces and the performance of the joint. In a second step, the fracture surfaces were examined. Here some differences were observed. The Tjoint with the weakest performance has failed in a mixed cohesive/adhesive failure mode whilst the remaining three T-joints failed cohesively as shown in Figure 10.
Figure 10 about here
Figure 10: Fractured surfaces in quasi-static front impact test. Top: Typical cohesive failure occurring in most cases; bottom: mixed cohesive adhesive failure occurring in absence of glass spheres.
7
It is also noted that the T-joint with the weakest performance was manufactured without the use of glass spheres. It is therefore assumed that the thickness of the adhesive bond line was not adjusted correctly, leading to a change of failure mode and joint performance.
A rate effect is seen for medium rate loading resulting in an increase of peak load with increasing loading rate, see Figure 8 middle.The experimental evidence for medium rate loading shows that one of the T-joints performed differently to the other joints. This particular joint is characterized by a less pronounced and substantially smaller peak load and smaller displacement at failure compared to the other T-joints tested at medium rate. The areas of the fractured surfaces were determined, but again did not show a clear correlation to the performance of the joints. The mean area of the fracture surface was 1587 mm² with a coefficient of variation of 8.9 %. It is noted that T-joints tested under medium rate loading were taken from two different batches of manufacturing. Three T-joints were manufactured in batch A, one T-joint was manufactured in batch B. The joint with the weakest performance was the joint originating from batch B. Each of the four T-joints failed cohesively. However, some difference was observed during the investigation of the fracture surface. The T-joint manufactured in batch B showed some darker blue color in one of the corners, see Figure 11. Optical microscopy revealed that failure in the area of dark blue color was a combination of cohesive and adhesive fracture; in the other areas failure was found to be purely cohesive.
Figure 11 about here
Figure 11: Comparison of fracture surfaces after front impact, v= 50 mm/s. Top: Typical T-joint manufactured in batch A; bottom: T-joint manufactured in batch B.
Due to dynamic effects, the load signal for high rate loading looks significantly different to the load signals recorded for quasi-static and medium rate loading. However, the scatter in terms of fractured surface (mean 1570 mm², coefficient of variation of 2.0%), peak load and displacement at failure is small. All T-joints tested under high rate loading were manufactured
8
in batch B. It is interesting to note that each fracture surface of T-joints tested under high rate loading showed the same dark blue color, see Figure 12.
Figure 12 about here
Figure 12: Typical fracture surface after front impact tests at high rate.
Peel failure is dominating in front impact loading. In [5] Tapered Double Cantilever Beam (TDCB) tests were presented indicating some influence of the loading rate on the location of failure. For high loading rates failure was always cohesive, for low loading rates the location of failure moved towards adherent-adhesive interface. This behavior was not seen in the T-joints subjected to front impact loading.
3.3 Side impact loading
The force-displacement curves recorded during side impact loading are shown in Fig. 13. Figure 13 about here
Figure 13: Force displacement curves recorded during side impact loading. a) loading speed 0.5 mm/s. b) Loading speed 50 mm/s. c) Loading speed 4800 mm/s. One T-joint tested under quasistatic loading was manufactured in batch A (shown in light grey); all other T-joints were manufactured in batch B. One T-joint tested under quasi-static loading did not have glass spheres (indicated by dashed line).
The typical sequence of events observed for side impact loading is more complex than for the previous load cases, see Fig. 14. At the beginning, plastic deformation of the rectangular profile is observed underneath the impactor. Then, damage initiates in the adhesive joint connecting the top surface of the top hat profile to the rectangular profile. In most cases, a fold is created in the metallic back sheet of the rectangular profile. After failure of this joint, folding is pronounced by rotation of the rectangular profile. At the same time, the back sheet starts debonding from the
9
bent section of the rectangular profile. Final failure is governed by shear failure of the adhesive joint connecting the back sheet of the rectangular profile to the side of the top hat profile.
Figure 14 about here
Figure 14: Typical failure sequence observed during medium rate and high rate side impact loading. From left to right: beginning of test; plastic deformation of rectangular profile and damage initiation at upper bond line and beginning of fold formation; failure of upper bond line, damage initiation in back face bond line; final failure.
In some cases, especially for low loading rates, the behaviour observed was slightly different. For these cases no or only limited fold formation was observed. As a consequence, catastrophic failure of the bond line connecting the back sheet to the top hat profile occurred. Looking at the force-displacement curves this manifests itself in a sudden load drop at a displacement of about 15 mm. For these cases no force plateau was observed after the initial load drop. The force plateau seen in most of the experiments is therefore attributed to the plasticity in the steel and formation of the fold shown in Fig. 14.
The T-joints tested under quasi-static loading conditions were taken from two different manufacturing batches. Three T-joints were taken from batch B, one T-joint was taken from batch A. This loading configuration is very interesting since the observed behavior varies significantly as seen in Figure 13 left. One T-joint failed catastrophically at a displacement of only about 11 mm. Also the maximum failure load was lower than for the remaining three Tjoints. Two T-joints can be clustered as both show the same type of behavior expressed by a maximum displacement at failure of approximately 18 mm. Finally, one T-joint formed a load plateau at about 3 kN starting at a displacement of about 18 mm. Final failure occurred at a displacement of 30 mm which is significantly larger than the displacements at failure observed for the other joints. Similar to the investigations for the front impact loading configuration, this scatter in joint performance cannot be correlated to different areas of the bond lines. The mean
10
area of the fracture surface on the side was 943 mm² with a coefficient of variation of 5.6 %. Recalling the analysis of the front impact there was a trend that the T-joints manufactured in batch A were stronger than the joints manufactured in batch B despite the fact that both batches were manufactured by the same technician under the same conditions. The same trend is seen here. The only joint forming the plateau was the also the only joint manufactured in batch A. The joint with the poor performance was manufactured without the use of glass spheres to adjust the thickness of the bond lines. Similarly to the front impact configuration, variations in thickness of the bond line have some significant influence on the mechanical performance of the joint. Numerical simulations showed a strong influence of the shear properties of the adhesive on the performance of the T-joint subjected to side impact loading [6]. A high shear strength of the adhesive was found to be critical for the fold formation in the back sheet of the adhesive joint and therefore for the formation of the plastic plateau. This is reflected in the experiments as the T-joints showed a more pronounced fold (height of the fold measured from the original flat position) with increasing joint performance. The characteristic size of the fold is the height of the fold measured from the original flat position. The weakest T-joint which was manufactured without glass spheres did not form a fold at all. The intermediate T-joints formed a mild fold of size less than 0.5 mm. The T-joint with the best performance formed a fold of size 5 mm. Figure 15 highlights the difference in the size of the fold for the two extreme cases:
Figure 15 about here
Figure 15: Comparison of the characteristic folds in metallic steel sheets. Left: strong fold formation was observed for the T-joint with the best performance; right: no fold was observed for the T-joint with the weakest performance.
All T-joints tested under medium rate loading conditions were manufactured in batch B. Here, the length of the plateau varies significantly resulting in a scatter for the displacement at failure ranging from 21 mm to 38 mm. Again, this cannot be correlated to the area of the fracture surfaces. The mean area of the fracture surface on the side was 949 mm² with a coefficient of
11
variation of 5.8 %. Also, all of the T-joints tested here were manufactured using glass spheres. Therefore the differences cannot be attributed to differences in the thickness of the bond line. Similarly to the quasi-static load case it was found that the length of the plateau can be correlated to the characteristic size of the fold in the metal. The characteristic size of the fold of the joint with the smallest displacement at failure is only 1.5 mm. For the joints with intermediate performance, the characteristic size of the fold is about 5 mm. For the joints with the strongest performance, the characteristic size of the fold is about 7 mm.
All T-joints tested under high rate loading conditions were manufactured in batch B. The results are very consistent with only one T-joint failing prematurely. Again, there was no correlation between the area of the fractured surface and the displacement at failure. The mean area of the fracture surface on the side was 905 mm² with a coefficient of variation of 2.3 %. However, the characteristic size of the fold shows a correlation to the displacement at failure. For the weakest T-joint, the size of the fold is about 4 mm, for the other cases the size of the fold is about 8 mm. For all cases failure occurred cohesively inside the adhesive. However, the fracture pattern at the bond line at the side of the T-joint with the weakest performance had a slightly different characteristic to the fracture pattern observed for the other cases as shown in Figure 16.
Figure 16 about here
Figure 16: Comparison of fracture patterns for the adhesive bond line failing in shear during highrate side impact loading. Left: Typical T-joint with good performance; right: T-joint with weak performance.
In [7], Marzi presented some rate dependent Tapered End Notched Flexure (TENF) tests on the same adhesive. Marzi found that shear failure was cohesive for high rates of loading and close to the adhesive-adherent-interface for low loading velocities. This trend was not seen here, except for one T-joint which failed adhesively under quasi-static loading conditions. However,
12
this can be attributed to the lack of glass spheres for this particular joint and is not a result of loading rate.
4. Discussion
Adhesively bonded metallic T-joints were tested under different loading conditions. Both, the loading direction and the loading velocity were varied resulting in a broad database of 9 different configurations. Failure in T-joints subjected to front impact loading was dominated by peel failure. Failure in T-joints subjected to side impact loading was a combination of peel and shear failure of the adhesive and plastic deformation of the steel sheets. T-joints subjected to rear impact loading showed large plastic deformation without causing failure of any of the adhesive bond lines.
Strain rate effects were observed for all loading directions. In general there was a trend towards increasing peak loads and increasing displacements at failure with increasing loading rate.
Scatter in the experiments can be traced back to the manufacturing of the T-joints. Two T-joints showed significantly weaker performance than the remaining T-joints. During the manufacturing process of these two test pieces, glass spheres were not used for ensuring the correct thickness of the bond line. As a consequence, failure of the bond line shifted from the desired cohesive failure mode towards a mixed adhesive/cohesive failure mode. Consequently, the performance of these joints was weaker than for the remaining joints.
The T-joints were manufactured in two batches by the same technician using an identical process. The only difference is the batch of the adhesive used for manufacturing. Despite ensuring perfect repeatability of the process there is evidence that T-joints manufactured in the second batch (batch B) are weaker than T-joints manufactured in the first batch. The difference is most distinct for the load case “side impact, quasi-static”. Here, the T-joint manufactured in batch A showed a plateau in the force-displacement curve which was not seen for the T-joints
13
manufactured in batch B. The shear strength of the adhesive is governing the formation of the plateau as higher resistance in the bond line at the side of the joint causes the metallic back sheet to deform plastically and form a fold. The displacement at failure in side impact loading can be directly correlated to the characteristic size of the fold as shown in Figure 16. The characteristic size of the fold is the height of the fold measured from the original flat position. Measurements taken from joints tested at quasi-static loading rate are marked with a diamond; measurements taken from joints tested at a medium rate are marked with squares; measurements from high rate tests are marked in triangles. A linear relationship is found between the size of the fold and the displacement at failure, see Fig. 17. No additional rate effect is seen here.
Figure 17 about here
Figure 17: Correlation of the characteristic size of the fold in side impact configuration and the recorded displacement at failure.
For some cases, mixed adhesive/cohesive fracture was found which resulted in dark blue colors of the fractured surfaces. The reason for this behaviour remains unclear. No ductile-brittle transition was seen which could explain differences between quasi-static and high-rate tests. However, BETAMATE 1496V is a multi-phase system which may result in an even more complex behaviour in reality. Variation in bond-line thickness can be excluded as a cause for this behaviour as a sufficient amount of glass spheres was found in both – the mixed fracture zones as well as the purely cohesive fracture zones. Consequently there is a chance that this behaviour is also a result of differences in the manufacturing batch.
5. Conclusions
14
In summary it became clear that adhesively bonded metallic T-joints are very sensitive to manufacturing. The different behaviour observed for the two different batches is attributed to differences in shear strength of the adhesive bond lines. Similar batch-dependent properties have also previously been reported for other structural adhesives produced by other manufacturers [8]. We therefore recommend performing comparative studies on a typical shear test such as the single lap shear test [9] before using new batches of adhesive for manufacturing crash relevant large structures.
Acknowledgements The IGF project 338 ZN ”Robustheit und Zuverlässigkeit der Berechnungsmethoden von Klebverbindungen mit hochfesten Stahlblechen unter Crashbelastungen” of the research association Forschungsvereinigung Stahlanwendung e.V. - FOSTA, Sohnstrasse 65, 40237 Düsseldorf was funded by the AiF under the program for the promotion of joint industrial research and development (IGF) by the Federal Ministry of Economics and Technology based on a decision of the German Bundestag. The authors would like to thank Mr. Holger Voß for the experimental support.
References
[1] M. Wissling and O. Hahn, "Adhesive Joints under Crash Loading: Experiments and Modelling of the Failure Behaviour," Welding in the World, vol. 51, pp. 63-73, 2007.
[2] P. Jousset and M. Rachik, "Implementation, identification and validation of an elasto-plasticdamage model for the finite element simulation of structural bonded joints," Int J Adhes Adhes, vol. 50, pp. 107-116, 2014.
15
[3] M. May and H. Voß, "Experimental and Numerical Analyses of Adhesively Bonded T-joints under Crash Loading," in EPJ Web of Conferences 26, 2012.
[4] M. May, H. Voß and S. Hiermaier, "Predictive modeling of damage and failure in adhesively bonded metallic joints using cohesive interface elements," Int J Adhes Adhes, pp. 7-17, 2014.
[5] M. May, O. Hesebeck, S. Marzi, W. Böhme, J. Lienhard, S. Kilchert, M. Brede and S. Hiermaier, "Rate dependent behaviour of crash-optimized adhesives - Experimental characterization, model development, and simulation," Engng Fract Mech, vol. 133, pp.112137,2015.
[6] M. May and O. Hesebeck, "Assessment of experimental methods for calibrating ratedependent cohesive zone models for predicting failure in adhesively bonded metallic structures," submitted to Engineering Failure Analysis, 2014.
[7] S. Marzi, "Measuring the critical energy release rate in mode II of tough, structural adhesive joints using the tapered end-notched flexure (TENF) test," The European Physical Journal Special Topics, vol. 206, pp. 35-40, 2012.
[8] M. Schlimmer, "Methodenentwicklung zur Berechnung und Auslegung geklebter Stahlbauteile für den Fahrzeugbau," final report of AiF-Zutech project 593, Kassel, 2005.
[9] ASTM, ASTM D-1002-10:Standard Test Method for Apparent Shear Strength of Single-LapJoint Adhesively Bonded Metal Specimens by tension Loading (Metal-to-metal), ASTM, 2010.
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
• • •
Testing of adhesive structures Rate- and loading direction dependent behaviour Influence of manufacturing
34