You are on page 1of 6

MODELING THE DEFORMATION AND FRACTURE OF WELDBONDED

JOINTS

M.N. Cavalli1, M.D. Thouless1,2, Q.D. Yang3


1
Mechanical Engineering Department, 2Materials Science and Engineering Department
2250 GG Brown Building, University of Michigan
Ann Arbor, MI 48109-2125
3
Rockwell Science Center
Thousand Oaks, CA

ABSTRACT
Results of this study show that the deformation and fracture of weldbonded joints can be successfully predicted using the
cohesive zone theory of fracture. Cohesive fracture parameters were first determined separately for spot-welded and
adhesively bonded joints by comparing experimental results and finite element calculations. These parameter values were
then used without modification in numerical models of weldbonded coach peel and lap shear joints. Comparisons with
experimental data for weldbonded joints show good agreement with both peak loads and energy absorption during the fracture
process.

INTRODUCTION
An extremely important aspect of automotive design is how the various components of the structure are to be connected to
one another. Automotive structures are often spot-welded or adhesively bonded, depending on the materials being joined and
the expected loading in a given region. A technique called weldbonding has been developed that incorporates both adhesive
bonding and spot-welding in a single joint [1]. Weldbonded joints possess many of the most desirable aspects of both
adhesive and spot-welded joints [2-12]. The strength of weldbonded joints is typically higher than that of either spot-welded or
adhesively bonded joints [2-8] and the presence of a spot-weld can at least partly compensate for deterioration of the adhesive
over time or in hostile environments [9,11]. Fatigue behavior of weldbonded joints is more desirable than that of spot-welded
joints [9-12]. In addition, the stiffness of weldbonded joints is much greater than spot-welded joints due to the presence of the
continuous adhesive bondline [9,10]. A dramatic example of this is the fact that the overall stiffness of a car body can be
retained using weldbonded joints even when the number of spot-welds is reduced [9]. Weldbonding offers the potential for
safer, stronger and more reliable automotive structures.

At present, however, analytical tools to predict the behavior of weldbonded joints are very limited. Most of the current
knowledge of weldbonded joint behavior is based on experimentally observed trends in specific systems with little quantitative
predictive ability. Studies have investigated the effect of welding parameters on weld nugget size [2,5,8], the effect of nugget
size on joint strength [5], the effect of nugget expulsion on joint strength [10] and the effect of the number of welds in the joint
on joint strength [4,7]. Investigations into the effect of joint width on joint strength [3], the effect of environmental conditions on
strength and fatigue properties [6,9,11], and the effect of loading rate on joint strength and energy absorption [6] have also
been completed. The relative levels of stress in spot-welded, adhesively bonded and weldbonded joints have also been
calculated through the use of finite element modeling [13-15]. A general method for predicting the fracture behavior of
weldbonded joints has not yet been proposed, however.

In contrast, the literature contains several proposed failure criteria for spot-welded joints. The basis of each is that a single
parameter fully describes fracture behavior under a single mode of loading. Typically, the fracture parameter chosen is a
critical value of stress. In addition, failure of the weld nugget itself is ignored as a possible mode of fracture. This is likely due
to the industrial mandate that welds are acceptable only if they ‘pull a nugget’ (i.e. fail by pullout failure), a view reinforced by
at least one available standard [16].

Lee et al. [17] define a failure criterion based on the normal and shear strengths of the weld, S tu and S su. These values
represent the normal and shear forces acting on a weld at failure normalized by the circumferential area of the weld. The
values of S tu and S su are determined by subjecting the joint to pure Mode I or II loading until failure. Under mixed-mode
loading conditions (i.e. a combination of shear and tension) the proposed criterion has the form
Ê S ˆn Ê S ˆn
Á t ˜ +Á s ˜ = 1 (1a)
Ë S tu ¯ Ë S su ¯
where St and S s are the applied normal and shear forces and n is a fitting parameter. Lee et al. [17] showed that a value of n
could be found such that this equation described failure for one particular geometry over the full range of positive ratios of
S s / S t.

A similar approach was used by Wung [18] and Wung et al. [19] in which four modes of applied loading are assumed: (i) a
normal force, fn, (ii) a shear force, fs, (iii) a bending moment, m b, and (iv) a torque, m t. Failure parameters are specified for
each individual mode and are given by F n, F s, Mb, and M t, respectively. Under arbitrary combinations of the four modes, the
form of the failure criterion is taken to be
Ê f ˆa Ê f ˆb Ê m ˆg Ê m ˆd
Á n ˜ +Á s ˜ +Á b ˜ +Á t ˜ = 1 (1b)
Ë Fn ¯ Ë Fs ¯ Ë M b ¯ Ë M t ¯
where a, b, g, and d are empirical parameters that must be determined experimentally. Rather than fracture of the joint, failure
in this model is taken to be the point at which the load-displacement curve is no longer linear. Lin et al. [20] use a slightly
different approach to model the onset of yield at the circumference of a weld. Their analysis, based on a lower-bound limit-

load, results in a criterion similar to Eqn. 1a, with n = 2.

Most recently, a method has been developed by Cavalli et al. [21] to predict the behavior of spot-welded joints using a
combination of experimental testing and finite element calculations. A major difference in this approach from that of previous
work [17-20] is that two fracture parameters, peak stress and toughness, are used to describe fracture in each mode of
loading. The necessity of a two-parameter model to describe crack propagation in joints undergoing plastic deformation has
been well documented [22-25]. These parameters can be incorporated into numerical models of structural joints through the
concept of a cohesive zone [21-31]. Within the cohesive zone, under a single mode of loading, crack propagation occurs
when the applied energy is equal to the material toughness for that mode of loading. When mixed-mode loading is applied,
the contributions of each mode must be combined through a suitable failure criterion. Cavalli et al. [21] showed that the values
of the necessary parameters for both pullout failure and weld nugget fracture could be established by comparing experimental
data with finite element calculations for a range of test geometries. After the parameter values were established for a given
weld in a given material, the same parameters could be used to predict the fracture behavior of identical welds in new joint
geometries [21]. Additionally, the failure mode of the weld (nugget fracture or nugget pullout) did not need to be specified at
the outset of calculations for new geometries. The appropriate mode of fracture was predicted by the model based on the
material parameters and geometry/loading of the joint.

Cohesive zone models also have been successfully used to predict the deformation and fracture of adhesive joints [26-31].
The current study combines these approaches for adhesive joints [26-31] with the work of Cavalli et al. [21] on spot-welded
joints to predict the behavior of weldbonded joints. Results show that cohesive zone parameters established independently by
modeling spot-welded and adhesively bonded joints can be used to model weldbonded joints. Both joint strength and energy
absorption characteristics are accurately captured in the numerical calculations.

EXPERIMENTAL PROCEDURE
The base metal for all test specimens was 2mm thick 5754-O aluminum. Both coach peel and single lap shear weldbonded
joints were fabricated for the study. Schematics, with dimensions, for each joint are shown in Figure 1a (coach peel) and 1b
(lap shear). A waxy surface coat had been applied to the surface of the aluminum sheets by the supplier to facilitate adhesion.
No other preparation was performed on the surface of the metal prior to bonding. All specimens were fabricated using a
toughened epoxy adhesive, XD4601 (Ciba-Geigy). Glass beads, 0.25mm in diameter, were mixed into the uncured epoxy by
the manufacturer for bondline spacing. The boundaries of the bonded region were defined by strips of Teflon‚ tape. The
curing cycle for all specimens was 180°C for 45 minutes, as specified by the manufacturer.

Batches of 10-12 specimens were fabricated together from a single pair of aluminum sheets. Individual specimens were then
separated using a band saw. The first weld made on each batch was a shunt weld, with no adhesive between the two pieces
of aluminum. This was done because an initial current path is required to prevent arcing between the electrodes of the spot-
welder. Truncated cone electrodes with tip diameters of 8mm were used to apply a square waveform weld current of 23 kA for
17 cycles for each weld. The resulting nugget diameters were found to be 7 ± 1 mm for all specimens.
-1
A screw-driven test frame was used to load each of the samples. The rates of loading were 5 mmmin for coach peel
-1
samples and 0.5 mmmin for lap shear specimens. A C.C.D. camera was used to measure the displacements during each
test. Values of displacement were measured between the points of load application, (X-X in Figures 1a and 1b).
P/2 P/2
P

X
Weld Nugget X
Transition Zone
Adhesive Layer
5mm 106mm Weld Nugget
Transition Zone
Adhesive Layer

45mm
3.625mm
26mm 3.5mm
3.5mm 3.625mm
2mm
X
2mm
X
P/2 P/2
23mm P

Width (out of plane): 25mm Width (out of plane): 25mm

Figure 1a: Schematic, with representative dimensions, of Figure 1b: Schematic, with representative dimensions, of
the weldbonded coach peel joints tested in the study. the weldbonded lap shear joints tested in the study.

NUMERICAL MODELING
A commercial finite element package (ABAQUS, version 5.8.16) was used for all numerical calculations. Models were fully
three-dimensional and the majority of the elements were 8-noded continuum block elements. The properties of these
elements were determined via tensile tests on samples of 5754-O aluminum [32]. True stress-true strain data from these tests
was fitted to the customary relationship
s = Ae n (2)
where A = 494 ± 6 MPa and n = 0.3 ± 0.01. The material was found to have a Young’s modulus, E, of 69 ± 5 GPa, and a yield
stress, s Y , of 113 ± 3 MPa. Poisson’s ratio was taken to be 0.3. For plastic deformation, isotropic hardening and a Von
mises yield criterion
† were assumed.
th
Based on the planes of symmetry in the coach peel and lap shear geometries, only 1/8 and 1/2 of the joints, respectively,
†were required to be modeled in the finite element calculations. Suitable boundary conditions enforced the symmetry. Fracture
was allowed to occur in these models by replacing continuum elements in some regions with 8-noded user-defined elements.
Under a single mode of loading, these elements deform according to a traction-separation law of the form shown in Figure 2.

Figure 2: Schematic representation of the traction-separation law governing behavior of the user-defined elements under a
single mode of loading.

The material toughness, G , is defined as the area under the traction-separation curve. Material failure (crack advance) occurs
when the applied mechanical energy to a potential plane of cracking is equal to the toughness of the material. During each
numerical increment, the relative displacements between nodes 1-5, 2-6, 3-7 and 4-8 are evaluated independently. If the
toughness has been reached at any of the four node pairs, all tractions acting on that pair are set to zero and the calculations

proceed. For mixed-mode deformation, the contribution from each mode must be considered. In this paper, a simple mixed-
mode failure criterion of the form [29,33,34]
G I G II G III
+ + =1 (3)
GI GII GIII


is assumed, where GI , GII , GIII represent the energy-release rates for each mode (the area under the curve up to the applied
displacement), and GI , GII , GIII represent the total area under the traction-separation curve for each mode. The values of the
two shear toughnesses, GII and GIII are always assumed to be identical in the current work. In general, however, GII and
GIII will be†different.
† †
† † †
User-defined elements were placed in three regions within each model. The first region was adhesive layer. Whether
† † †
cracking in this region is adhesive or cohesive does not explicitly affect the placement of the elements, but could affect the
† properties assigned to them. The second region of user elements was at the midline of the weld nugget to allow for failure by
nugget fracture. Finally, a ring of elements was placed through the thickness of the adherend just outside the weld nugget to
allow for pullout failure.

The fracture properties given to the user-elements in the weld nugget and pullout region were those established previously
through modeling of spot-welded joints [21]. By assuming the weld properties in the spot-welded and weldbonded joints are
identical, the assumption has been made that these properties are not affect by the presence of the adhesive. In general, it is
possible that the adhesive layer may affect the thermal properties of the joint [2]. In the present case, however, the base metal
alloy is non-heat treatable [35-37]. As a result, any thermal effects of the adhesive were neglected. Values of the parameters
) ) -2 -2
used for the nugget region were s N = 290 ± 20 MPa, t N = 185 ± 20 MPa, GIN = 15 ± 3 kJm , and GIIN = 23 ± 4kJm , where
sˆ indicates the Mode I peak stress and tˆ indicates the Mode II and Mode III peak stress. In the pullout region, the parameter
-2 -2
values were sˆ P = 298 ± 10 MPa, tˆ P = 190 ± 10 MPa, GIP = 10 ± 2 kJm , and GIIP = 15 ± 3 kJm .
† † † †
† Mode I cohesive properties of the
† XD4601 adhesive were determined according to the method described by Yang et al. [26].
Adhesive failure in all specimens was seen to be primarily interfacial. The Mode I peak stress and toughness were found to be
-2
sˆ A †
= 63 ± 8 MPa and G†IA = 1.1 ± 0.3 kJm †. These values are very†close to those found by Yang et al. [29] for the XD4600
adhesive (also from Ciba-Geigy). In addition, the tensile and shear data provided by the manufacturer from tests on bulk
specimens of cured XD4600 and XD4601 is very similar. Based on these comparisons, the Mode II cohesive properties of the
two adhesives were also assumed to be identical. The Mode II parameters for the adhesive were thus taken to be tˆ A = 35
† † -2
MPa and GIIA = 5.4 kJm . Details on how these values were calculated can be found in Yang et al. [28].

RESULTS

Figure 3 shows a comparison between the experimentally measured load-displacement response of the weldbonded coach
† joints and the numerically predicted behavior. Agreement with both the first peak, corresponding to the initial cracking of
peel
the adhesive layer, and the second peak, corresponding to pullout of the nugget, is good. The final failure mode, pullout
failure, also agrees well between experimental results and numerical calculations. During the finite element modeling, some
crack propagation is seen in the cohesive elements at the midline of the nugget. This crack arrests before the maximum load
is reached, however, and final failure is by pullout. Cavalli et al. [21] offer a more detailed examination of this combination of
failure modes in the numerical model for the coach peel geometry.

3000

Experimental Range

2500
Finite Element Predictions

2000
Load (N)

1500

1000

500

0
0 5 10 15 20 25 30

Displacement (mm)
Figure 3: Comparison of experimental data from weldbonded coach peel joints with finite element predictions using the
cohesive parameters for the weld region determined in Cavalli et al. [21] and parameters for the adhesive determined as
outlined by Yang et al. [28,29].
A comparison between experimental results and numerical predictions for the load-displacement behavior of weldbonded lap
shear joints is seen in Figure 4. The numerical predictions agree well with the experimental measurements from initial loading
to beyond the point of maximum load. Experimentally, the flattening of the load-displacement trace near the peak load
corresponds to the onset of cracking in the adhesive. Final failure of the adhesive corresponds to the sharp drop near the end
of the curves. Numerical instabilities caused non-convergence in the model prior to complete failure of the adhesive layer.
Nugget fracture was seen in both the experimental samples and the numerical calculations.

12000

10000

8000
Load (N)

6000

4000

Experimental Range
2000 Finite Element Predictions

0
0 1 2 3 4 5 6 7 8

Displacement (mm)
Figure 4: Comparison of experimental data from weldbonded lap shear joints with finite element predictions using the cohesive
parameters for the weld region determined in Cavalli et al. [21] and parameters for the adhesive determined as outlined by
Yang et al. [28,29].

CONCLUSIONS
Results of the current study show that the deformation and fracture of weldbonded joints can be captured using cohesive zone
modeling within finite element calculations. This method requires that two cohesive fracture parameters, peak stress and
toughness, be determined for each mode of deformation in each potential region of fracture. If thermal effects of the adhesive
layer on weld nugget formation can be neglected, the required properties can be established by separate tests on spot-welded
and adhesively bonded joints. If not, it may be necessary to modify the methodology outlined in Cavalli et al. [21] to determine
the cohesive parameters by comparing experimental data from several weldbonded joint geometries and fracture modes to
numerical calculations. Once this has been done, the deformation and fracture of any new joint geometry containing the
identical spot-weld and adhesive combination can then be predicted using the same cohesive parameters.

ACKNOWLEDGEMENTS
The authors would like to thank Mr. Ron Cooper of Ford Motor Company Scientific Research Laboratory, Dearborn, MI for his
help with the resistance spot-welder. In addition, thanks are due Dr. John Hill, Dr. Susan Ward and Ms. Kimberly Lazarz, also
of Ford SRL, for providing the aluminum and adhesive used in this study.

REFERENCES
[1] Irving, Bob, “Weld Bonding Finds New Applications on Automotive Assembly Lines,” Welding J., 73, pp. 41-42, 1994.
[2] Ghosh, P. K., N. Sambasivarao, “Weldbonding of Thin Sheet Aluminum,” Int. J. for the Joining of Materials, 10, pp. 45-53,
1998.
[3] Hills, D. R., J. D. Parker, N. T. Williams, “The Effect of Increasing Sample Width on the Strength of Weldbonded Lap
Joints,” Key Engineering Materials, 99-100, pp. 119-126, 1995.
[4] Hills, D. R., J. D. Parker, N. T. Williams, “Effect of Number of Welds in Spot Welded and Weldbonded Arrays on Static
Failure Load of Individual Spot Welds and Adhesive Bond,” Ironmaking and Steelmaking, 23, pp. 150-156, 1996.
[5] Moniatis, G., L. Dorn, A. ElSawy, “A Comparison Between Adhesive-Bonding and Weldbonding of HSLA Steel,” Proc. of
an Int. Conf. on Trends in Welding Research, Gatlinburg, TN, USA, May 18-22, 1986.
[6] ElSawy, A. H., A. S. Salama, Z. B. Nan, “A Comparison Between the Spot Welding and Weldbonding of Aluminum
Alloys,” Proc. of an Int. Conf. on Trends in Welding Research, Gatlinburg, TN, USA, May 18-22, 1986.
[7] Renshaw, T., K. Wongwiwat, A. Sarrantonio, “Comparison of Properties of Joints Prepared by Ultrasonic Welding and
Other Means,” J. of Aircraft, 20, pp. 552-556, 1983.
[8] Jones, T. B., “Weldbonding: The Mechanisms and Properties of Weldbonded Joints,” Sheet Metal Industries, 72, pp. 29-
31, 1995.
[9] Bassler, H. I., I. El-Sebakhy, D. J. Malik, “Weldbonding of Structural Adhesives for Body Stiffening,” Sae Technical
Papers, #922118, pp. 39-55, 1992.
[10] Ring-Groth, M., C. Magnusson, “The Fatigue Properties of Weldbonded Stainless Steel Joints,” J. of Advanced Materials,
32, pp. 21-27, 2000.
[11] Melander, A., M. Larsson, H. Stenslo, A. Gustavsson, J. Linder, “Fatigue Performance of Weldbonded High Strength
Sheet Steels Tested in Artic, Room Temperature and Tropical Environments,” Int. J. of Adhesion and Adhesives, 20, pp.
415-425, 2000.
[12] Wang, P. C., S. K. Chisholm, G. Banas, F. V. Lawrence, Jr., “The Role of Failure Mode, Resistance Spot Weld and
Adhesive on the Fatigue Behavior of Weld-Bonded Aluminum,” Welding J., 74, pp. 41s-47s, 1995.
[13] Chang, B. H., Y. W. Shi and S. J. Dong, “A Study on the Role of Adhesives in Weldbonded Joints,” Welding J., 78, pp.
275s-279s, 1999.
[14] Chang, B. H., Y. W. Shi and S. J Dong, “Comparative Studies on Stresses in Weld-Bonded, Spot-Welded and Adhesive-
Bonded Joints,” J. of Materials Processing Technology, 87, pp. 230-236, 1999.
[15] Al-Samhan, A., S. M. H. Darwish, “Strength Prediction of Weld-Bonded Joints,” Int. J. of Adhesion and Adhesives, 23, pp.
23-28, 2003.
[16] Standard AWS C1.4M, “Specification for Resistance Welding of Carbon and Low-Alloy Steels,” AWS, 1999.
[17] Lee, Y. L., T. J. Wehner, M. W. Lu, T. W. Morrissett, E. Pakalnins, “Ultimate Strength of Resistance Spot-welds Subjected
to Combined Tension and Shear,” J. of Testing and Evaluation, 26, pp. 213-219, 1998.
[18] Wung, P., “A Force-Based Failure Criterion for Spot-weld Design,” Exp. Mech., 41, pp. 107-113, 2001.
[19] Wung, P., T. Walsh, A. Ourchane, W. Stewart, M. Jie, “Failure of Spot Welds Under In-Plane Static Loading,” Exp. Mech.,
41, pp. 100-106, 2001.
[20] Lin, S. H., J. Pan, S. R. Wu, T. Tyan, P. Wung, “Failure Loads of Spot-welds Under Combined Opening and Shear Static
Loading Conditions,” Int. J. Solids and Structures, 39, pp. 19-39, 2001.
[21] Cavalli, M. N., Q. D. Yang, M. D. Thouless, “Cohesive-Zone Modeling of the Deformation and Fracture of Spot-Welded
Joints,” manuscript in preparation.
[22] A. Needleman, “A Continuum Model for Void Nucleation by Inclusion Debonding,” J. Appl. Mech., 54, pp. 525-531, 1987.
[23] V. Tvergaard, J. W. Hutchinson, “The Relation Between Crack Growth Resistance and Fracture Process Parameters in
Elastic-Plastic Solids,” J. Mech. and Phys. Solids, 40, pp. 1377-1397, 1992.
[24] V. Tvergaard, J. W. Hutchinson, “On the Toughness of Ductile Adhesive Joints,” J. Mech. and Phys. of Solids, 44, pp.789-
800, 1996.
[25] S. Roychowdhury, Y. D. Arun Roy, R. H. Dodds, Jr., “Ductile Tearing in Thin Aluminum Panels: Experiments and
Analyses Using Large-Displacement, 3-D surface Cohesive Elements,” Eng. Fract. Mech., 69, pp. 983-1002, 2002.
[26] Yang, Q. D., M. D. Thouless, S. M. Ward, “Numerical Simulation of Adhesively Bonded Beams Failing with Extensive
Plastic Deformation,” J. Mech. and Phys. Solids, 47, pp. 1337-1353, 1999.
[27] Yang, Q. D., M. D. Thouless, S. M. Ward, “Analysis of the Symmetrical 90°–Peel Test with Extensive Plastic
Deformation,” J. Adhesion, 72, pp. 115-132, 2000.
[28] Yang, Q. D., M. D. Thouless, S. M. Ward, “Elastic-Plastic Mode-II Fracture of Adhesive Joints,” Int. J. Solids and Struct.,
38, pp. 3251-3262, 2001.
[29] Yang, Q. D., M. D. Thouless, “Mixed-Mode Fracture Analysis of Plastically Deforming Adhesive Joints,” Int. J. Fracture,
110, pp. 175-187, 2001.
[30] Thouless, M. D., J. L. Adams, M. S. Kafkalidis, S. M. Ward, R. A. Dickie, “Determining the Toughness of Plastically
Deforming Joints,” J. Mat. Sci., 33, pp. 189-197, 1998.
[31] Kafkalidis, M. S., M. D. Thouless, Q. D. Yang, S. M. Ward, “Deformation and Fracture of Adhesive Joints Constrained by
Plastically-Deforming Adherends,” J. Adhesion Sci. and Tech., 14, pp. 1593-1607, 2000.
[32] Standard E8M-99, “Standard Test Method for Tension Testing of Metallic Materials [Metric],” ASTM, 1999.
[33] Wang, J. S., Z. Suo, “Experimental Determination of Interfacial Toughness Using Brazil-nut-Sandwich,” Acta Metall., 38,
pp. 1279-1290, 1990.
[34] Hutchinson, J. W., Z. Suo, “Mixed-Mode Cracking in Layered Materials,” Adv. in App. Mech., 29, pp. 63-191, 1992.
nd
[35] Bringas, H. E., CASTI Metals Red Book, Nonferrous Metals, 2 ed., Edmonton, 1999.
[36] Davis, J. R., ed., ASM Specialty Handbook: Aluminum and Aluminum Alloys, ASM Int., Material Park, OH, pp. 680, 1994.
[37] Holt, J. M., C. Y. Ho, eds., Structural Alloys Handbook, CINDAS/Purdue University, West Lafayette, IN, 1994.

You might also like