You are on page 1of 15

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/242134924

Modelling the dynamic failure of riveted joints in aerospace fuselages

Article · January 2005

CITATIONS READS

4 10,734

4 authors, including:

Robert Birch Mark D White


University of Liverpool University of Liverpool
38 PUBLICATIONS   648 CITATIONS    110 PUBLICATIONS   918 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

A Novel Approach to Rotorcraft Simulation Fidelity Enhancement and Assessment View project

Ship Air Interface Modelling and Simulation View project

All content following this page was uploaded by Mark D White on 20 February 2015.

The user has requested enhancement of the downloaded file.


Modelling the dynamic failure of riveted
joints in aerospace fuselages
Robert S. Birch1 , Laura Vendrasco2 ,
Mark D. White1 & Michel Mahé2
1 Impact Research Centre, University of Liverpool, Department of
Engineering, UK
2 Airbus France, ESANT, France

Abstract

Crashworthiness of lightweight aerospace structures is a major research


topic in aeronautical engineering. Commonly, aircraft fuselages comprise
thin sheet materials that are jointed using fasteners such as rivets or bolts
that are arranged in various configurations depending on the required me-
chanical performance. Typically, a commercial aircraft has over one million
rivets and much effort is employed to optimise the design of structure prior
to an aircraft design progressing to prototype and then full production.
Nowadays, due to expense of experimental testing and greater pressure to
reduce product lead times, more use is made of numerical modelling using
finite element analysis (FEA) as a way of simulating and optimising a design
prior to prototyping.
This paper offers comment on the use of the explicit FEA technique for
modelling the dynamic behaviour of riveted joints of the type used in the
aerospace industry today. Comparison of an FEA model (up to failure) is
made with a series of tests on a series of riveted lap joints. Parameters, such
as edge distance and rate of deformation, are considered.

Keywords: riveted joints, dynamic failure, aerospace structures, finite element ana-
lysis

WIT Transactions on Engineering Sciences, Vol 49


Impact Loading of Lightweight Structures, M. Alves & N. Jones (Editors)
°c 2005 WIT Press, www.witpress.com, ISSN 1743-3533
122 Robert S. Birch, Laura Vendrasco, Mark D. White & Michel Mahé

1 Introduction

Ever-increasing safety standards and the desire by manufacturers to re-


duce the product design cycles means that greater reliance is placed on
finite element analysis (FEA) simulations and less on experimental studies.
Typically, the size of FEA models used to study the crashworthiness of a
commercial aircraft such as A320 is of order 2x105 shell elements. This large
number of elements is required to model the geometry of fuselage and also
describe non-linear behaviour and failure modes. Some hundred or so hours
of supercomputer time would be necessary to identify an ultimate failure of
the structure.
Much research has been carried on the failure of welded or fastened joints
for automotive structures [1–4]. Similarly, during a crash1 or collision with
an object2 , an aerospace structure is required to collapse in a controlled
stable manner to absorb energy effectively and to mitigate the effects of
the impact loading. However, under rapid collapse, the behaviour of a joint
element can be very different to that when the load is applied statically.
Modelling accurately the failure of these jointed interfaces is a major key
to predicting the behaviour of a real fuselage structure using FEA. Current
modelling methodologies appear to be inadequate to describe the complex
interaction between the structure and fasteners, particularly when consid-
ering the failure modes.
Despite the advances in computing technology it is not practical to model
every rivet connection in an aircraft structure with a 3 dimensional contin-
uum model. Therefore, lumped joint property models, comprising spring/
damper connection between two nodes of the FEA model, have been used to
describe the behaviour of the joint. Dependent on experimental data these
models also take into account properties such as bending and shear action
as well as providing a failure criterion. Unfortunately, several FE studies
[5, 6] showed that present models are not always satisfactory in dynamic
FE analysis, partly because of an inadequate methodology for calculating
model parameters. Further investigations based on both experimental and
numerical studies are therefore necessary to create a new FE joint model
and establish a methodology to fix all its parameters.

2 Joint systems

Riveted lap joints are used extensively in the construction of lightweight


aircraft fuselage. Typically, the single lap or doubler joint designs (see Fig-
ure 1) are used to interconnect structural components such as skin panels

1 Considering survivable crash scenarios such as hard landings or ditching.


2 Typically, bird strike or foreign object impact such as runway debris.

WIT Transactions on Engineering Sciences, Vol 49


Impact Loading of Lightweight Structures, M. Alves & N. Jones (Editors)
°c 2005 WIT Press, www.witpress.com, ISSN 1743-3533
Modelling the dynamic failure of riveted joints 123

Figure 1: (a) Simple lap joint, (b) Lap joint with internal doubler plate, (c)
Lap joint with internal and external doubler plate

onto the airframe with single and multiple rows of rivets. These joints are
subjected to combined loading from the fuselage pressurisation and bending
due to flight loads. Clearly, the load is transferred from one panel to another
through the rivets. In the case of multiple rivet rows, then the highest load
is carried in the first rivet row while the second highest load is carried in
the second row and so on. A complete analysis needs to consider the super-
position of both the far field stress (i.e., pressure loads) and the local rivet
loading on the skin (i.e., bearing loads).
The stress state in the region of the joint is complicated by the presence of
rivets. These introduce residual stresses around the hole that are dependent
on the amount of fixing force. Countersunk holes introduce additional stress
concentrations around the holes and in the region of the countersinks. Such
a complex stress state requires detailed analysis to determine accurately the
behaviour of these joints under sever loading.
Since riveted joints in structural components often fail under combined
loads, a general failure criterion for riveted joints is useful for crashworthi-
ness analysis especially if it can be implemented easily into a finite element
code. But, experimental work reveals unexplained differences between the
static and dynamic failure modes that make the accurate simulation of riv-
eted structural components a useful way to improve the understanding of
their mechanical behaviour.

WIT Transactions on Engineering Sciences, Vol 49


Impact Loading of Lightweight Structures, M. Alves & N. Jones (Editors)
°c 2005 WIT Press, www.witpress.com, ISSN 1743-3533
124 Robert S. Birch, Laura Vendrasco, Mark D. White & Michel Mahé

3 Failure mechanisms

Significant differences between quasi-static and dynamic failure modes have


been observed [1] in metallic lap joints when subjected to a simple in-plane
loading. But, not all joint configurations exhibit this behaviour and the work
in this area is by no means extensive.
Work that was carried out by a consortium of European aerospace indus-
tries and research institutes under a European Framework III programme
in the mid 1990s attempted to improve FEA technology through a step-
wise programme of benchmark tests and FEA modelling. As part of this
programme a series of dynamic and static tests were carried out on various
riveted lap joint configurations common to the structure of a commercial
aircraft [2, 3].
Generally, when subject to an axial in-plane loading, common modes of
failure were identified as the well-known rivet-shear, tensile failure in the
parent material, shear-out, pull-out (bearing) and cleavage (see Figure 2).
Joint configuration and the relationship between rivet and parent material
strengths were identified as being two significant parameters that controlled
the failure mode of a joint.

Figure 2: Common failure modes in riveted lap joints. (a) Tensile failure in
the parent material, (b) shear out, (c) cleavage, (d) rivet-shear,
(e) pull-out (bearing)

Rivet-shear and tensile are relatively simple modes of failure that depend
largely on the whether low or high strength rivets are used in a rigid joint
configuration. Under these circumstances these simple modes do not appear
to be influenced greatly by deformation rate. For example, lap joints that
are formed using low strength aerospace rivets such as LN9198 (universal

WIT Transactions on Engineering Sciences, Vol 49


Impact Loading of Lightweight Structures, M. Alves & N. Jones (Editors)
°c 2005 WIT Press, www.witpress.com, ISSN 1743-3533
Modelling the dynamic failure of riveted joints 125

head) or NAS1097 (countersunk head) always failed by rivet-shear and did


not exhibit major differences in failure mode over the deformation rates
considered (namely, quasi-static to 7 m/s approximately, see Figure 3).

(a) (b)

Figure 3: Simple rivet shear failure mode in lap joints made using low
strength rivets: (a) single row of universal rivets, (b) multiple rows
of countersunk rivets

This appeared to be true for both single and multiple rivet row lap joint
configurations. Equally, rigid lap joints that are formed with a doubler plate
and multiple rows of high strength rivets such as the HI-LOK HL10 and
HL11 failed by parent material tension along the first row and were also
insensitive to deformation rate (see Figure 4).
Conversely, the failure modes of shear-out, pull-out and cleavage involve
a more complex mechanical process and were associated with the failure of
less rigid joint designs such as a simple lap fixed by a single row of high
strength rivets. These modes appeared to be sensitive to deformation rate.
For example, Figure 5 shows how the mode of failure switched from tensile
to cleavage and then to pull-out with respect to increased deformation rate.

(a) (b)

Figure 4: Simple tensile failure mode of lap joints made using high strength
rivets: multiple rows of (a) countersunk and (b) universal HI-LOK
rivets

WIT Transactions on Engineering Sciences, Vol 49


Impact Loading of Lightweight Structures, M. Alves & N. Jones (Editors)
°c 2005 WIT Press, www.witpress.com, ISSN 1743-3533
126 Robert S. Birch, Laura Vendrasco, Mark D. White & Michel Mahé

4 Material behaviour

The yield (0.2%) and Ultimate Tensile Stress of aerospace alloys such as
2024-T351 are not considered to be overly sensitive to strain rate. However,
more significant is the influence of strain rate on the fracture strain. For
example, Figure 6 shows the influence of strain rate on the uni-axial tensile
fracture strain of a 2024-T351 aluminium alloy sheet material for a range of
rates up to 100s−1 , approximately. Although, not entirely conclusive, these
test results indicate an increase in fracture strain with respect to strain rate.
Reference [7] confirms this trend, giving a true failure strain value of over
30% for a strain rate of 8000s−1 .
Strain rates of order 100s−1 are typical of the averaged global strain rates
that a fuselage structure might experience during a crash. But, observation
of the failure modes in the test joints showed highly localised regions of
deformation where the strain rate is likely to be much higher, and conse-
quently the failure strain too. Moreover, for impact analysis (bird strike,
tyre or hard debris impact) the strain rate can reach an order of magnitude
of 1000s−1 or more.
The fracture is modelled with the Johnson-Cook criterion [8]:
· µ ¶¸
σH
εf ail = D1 + D2 exp D3 (1 + D4 ln ε∗ )
σvm

where:

ε̇
ε∗ = ε̇0 Normalized effective plastic strain rate (the reference
strain rate is usually equal to 1.0 s−1 )
D1, D2, D3, D4 Failure constants
σH Hydrodynamic stress (first invariant of stress tensor)
σvm Von Mises stress (second invariant of stress tensor)

Fracture occurs when the damage parameter D exceeds 1:


X δ εpl calc
D= ≥ 1 ⇒ f ailure
εf ail

During the calculation, element stresses are all set to zero when the failure
criterion is fulfilled, and remain equal to zero for the rest of the calculation.
The fist bracket of the criterion represents the observation that for duc-
tile fracture the failure strain decreases with increasing stress triaxiality
σH /σvm . At a micro-scale, in fact, the ductile fracture occurs due to nucle-
ation, growth and coalescence of micro-voids, and the rate of void growth

WIT Transactions on Engineering Sciences, Vol 49


Impact Loading of Lightweight Structures, M. Alves & N. Jones (Editors)
°c 2005 WIT Press, www.witpress.com, ISSN 1743-3533
Modelling the dynamic failure of riveted joints 127

can be determined as a function of triaxiality applying the theory of plas-


ticity [9, 10]. The second bracket represents the effect of the strain rate on
the material ductility.
The Johnson-Cook failure model assumes the level of damage being pro-
portional to the plastic strain level. That is not always true, because it has
been shown that the damage can start at a strain level higher than 0, and
intervenes mostly at the end of the plastic stress flow. That happens when
the triaxiality of the stress favours the nucleation and the growth of fragile
intergranular microcracks, lead by the transverse stress. The mechanism of
void growth intervenes only in the ultimate phase, during the final break of
the ductile grains separated by the microcracks. The model also supposes
that the value of the critical damage is equal to 1. Several studies on highly
ductile alloys showed that the critical value varies between 0.5 and 0.8. In
the case of less ductile materials it can reach values of 0.3-0.4. Therefore,
Johnson-Cook failure model gives good results for very ductile materials
damaged by the voids growth. For less ductile materials, the evaluation of
damage evolution is incorrect, but is nevertheless possible to use this model
as a failure criterion.
Modelling of alloys fracture is complex because of the statistical nature
of the void nucleation and growth processes that are fundamental to ductile
failure. More fundamental ductile fracture models exist, which allow for the
statistical nature of void initiation and growth and the effect of voids on the
yield surface, but it is difficult to fill them in with available experimental
data.

5 FEA modelling

The large amount of rivets in aeronautical structures requires the employ-


ment of a simplified model able to predict the behaviour of the joint with
an acceptable computational cost. In actual project applications, joints are
defined by 1D elements (beam type spring) between two nodes that are
fixed onto the plates with tied interfaces. This solution allows defining a
non-linear behaviour in tension/compression, shear (two modes), bending
(two modes) and torsion, and a multidirectional failure criterion.
Experience shows that these models are not always adequate in dynamic
FE analysis in terms of failure mode and strength, due to limitations bound
to the models themselves (coupling failure criteria in different direction,
interaction between plates and fasteners) or to the identification of the re-
quired numerical parameters. Therefore, further investigations are still on-
going to understand better the joint behaviour in order to create a new FE
joint model and establish a methodology to fix all its parameters.

WIT Transactions on Engineering Sciences, Vol 49


Impact Loading of Lightweight Structures, M. Alves & N. Jones (Editors)
°c 2005 WIT Press, www.witpress.com, ISSN 1743-3533
128 Robert S. Birch, Laura Vendrasco, Mark D. White & Michel Mahé

(a)

(b)

(c)

Figure 5: Simple rivet shear failure mode of lap joints made using high
strength HI-LOK rivets: axial pull velocities of (a) quasi-static,
(b) 2.6 m/s and (c) 6.6 m/s

To develop such a model it is necessary to perform a considerable num-


ber of tests. Given the large number of rivet types, joint geometries, loading
conditions, and accrued complex behaviour of junctions, it would be im-
practical to perform tests for every possible configuration. Therefore, the
best choice is to develop a virtual test bed for numerical simulation of joint
failure, reducing the experimental tests to those few necessary for corre-
lation and verification of the tool. With such a tool, it would be possible
to perform simulations for every planned configuration and pursue the de-
velopment of the simplified model, identifying the leading parameters that
govern the junction behaviour up to failure.
An example of this procedure of improvement and validation of the virtual
test bed is the study of the behaviour up to failure of the simple lap joint
family shown in Figure 7, where the geometric variable is the width of plate,
w. The specimens were configured as a single lap, fastened with one type
ASNA 2026V3 countersunk fastener with four values of w, namely 12 mm,

WIT Transactions on Engineering Sciences, Vol 49


Impact Loading of Lightweight Structures, M. Alves & N. Jones (Editors)
°c 2005 WIT Press, www.witpress.com, ISSN 1743-3533
Modelling the dynamic failure of riveted joints 129

20 mm, 30 mm and 40 mm. Tensile loading was applied in the in-plane axial
direction with a displacement velocity of up to 1 m/s. Using FEA and tests,
the aim of the study was to examine the changes in behaviour up to failure
of the joint given by the geometric change in plate width, w.
Figure 8 shows the axial load vs. cross head displacement curves obtained
for the different width, and Figure 9a to 9d show the failure modes. The
comparison of the FEA predicted failure modes with those obtained after-
wards from the experiments (see Figure 9) shows a good correlation. For
lower width (12 and 20 mm) we obtain the same tensile failure in the plates;
for the 30 mm and 40 mm width specimen we obtain a similar mixed failure
(tensile and shear-out), even if the model does not show the change in frac-
ture direction. This limitation could be partly due to the mesh geometry
influencing the fracture pattern after initial failure.

(a) (b)

Figure 6: Rupture strain of aerospace sheet alloy materials (a) 2024-T351


and (b) LN9073 with respect to strain rate

Concerning axial force vs. cross-head displacement, Figures 10a to 10d


show the comparison between numerical results (in grey) and those ob-
tained experimentally at 1 m/s and quasi-static pull velocities (in colours).
As often happens in FEA predictive studies, the global shape of the curve
and the force level are well predicted, but the displacement at failure is
wrong. This is partly due to the localisation of the deformation and to the
pre-stress induced by riveting or interference fitting, which induce changes
in the failure strain in some zones, and that can be hardly reproduced in
a FE model without using a very fine and time-consuming mesh. The size
of the mesh itself can have an influence on the failure strain. Moreover, re-
ducing stress tensor to operate with one or two values (Von Mises’ stress

WIT Transactions on Engineering Sciences, Vol 49


Impact Loading of Lightweight Structures, M. Alves & N. Jones (Editors)
°c 2005 WIT Press, www.witpress.com, ISSN 1743-3533
130 Robert S. Birch, Laura Vendrasco, Mark D. White & Michel Mahé

and hydrostatic pressure) could be too limiting in case of strong 3D defor-


mation patterns such the one around the fastener. Another potential source
of error is an incomplete knowledge of the material behaviour or a lacking
modelling concerning damage, failure hardening and fracture propagation,
which is often crucial to obtain realistic predictions.

Figure 7: Geometry of single lap joint specimens

6 Other considerations

The influence of other parameters on dynamic failure of joints has yet to


be explored. For example, the thin layer of adhesive (sealant) that is nor-
mally applied to lap joints in order to provide a pressure seal, restrict the
entry of moisture and retard corrosion adds to the performance of the con-
nection. High and low modulus adhesives have been widely used to bond
metallic-to-metallic and metallic-to-composite structures to minimize the
stress concentrations and seal the joints. Acting like a compliant adhesive
and a viscous lubricant, the sealant may also have beneficial mechanical
effect by reducing fretting and fatigue damage. Adhesive stiffness and ad-
hesive layer thickness are two parameters that control the performance of
riveted lap joints. Stress concentration factors can be reduced by increasing
stiffness of the adhesive layer and or decreasing its thickness.
A damage tolerance analysis may require several analytical methods to
firstly characterise the complex stress state in the joint prior to applying
fracture mechanics methods to determine the stress distributions and stress
intensity factors within the joint. These methods are part of an overall ana-
lysis approach to do a complete assessment of the structural integrity of a
fuselage. Ultimately, a detailed assessment is important for the determina-
tion of critical areas and safety limits.

WIT Transactions on Engineering Sciences, Vol 49


Impact Loading of Lightweight Structures, M. Alves & N. Jones (Editors)
°c 2005 WIT Press, www.witpress.com, ISSN 1743-3533
Modelling the dynamic failure of riveted joints 131

Figure 8: Quasi-static axial load vs. displacement for a single lap joint for
w = 12, 20, 30 and 40 mm

7 Conclusions

An understanding of the dynamic failure of joint interfaces is crucial in


the study of crashworthiness of thin-walled structures such as aircraft fuse-
lages. Currently, considering the large number of riveted joints in a typical
aerospace structure, cost and limitations in computing power, it is impracti-
cal to use a high resolution FEA model for every individual joint connection.
Therefore, the use of more practical FEA approaches using validated global
joint models are necessary, but the exact parameters and influence of each
has yet to be examined thoroughly.
Previous research on typical aerospace joints has shown that there are
significant differences between the quasi-static and dynamic modes of failure
for certain configurations of joint interface. The sensitivity of the rupture
strain in the parent material to strain rate may be a significant parameter
in this behaviour.
The FEA modelling of rivet joints shows that the global failure mode and
the force level at failure are usually well predicted, but the displacement at
failure is often inaccurate. This could be due to basic faults in the model
(i.e., the mesh not fine enough to catch localised deformations, missing or

WIT Transactions on Engineering Sciences, Vol 49


Impact Loading of Lightweight Structures, M. Alves & N. Jones (Editors)
°c 2005 WIT Press, www.witpress.com, ISSN 1743-3533
132 Robert S. Birch, Laura Vendrasco, Mark D. White & Michel Mahé

inexact pre-stresses caused by fastening, limiting material models, etc.) or


to an incomplete knowledge of the material behaviour and its sensitivity
to mesh dimension and strain rate. An improvement in FE simulation of
joints requires a better understanding and modelling of material behaviour
and damage, strain localisation, residual stress and strain due to fastening
procedures, strain rate influence and mesh sensitivity.

Figure 9: Failure modes for a single lap joint where w = (a) 12, (b) 20, (c)
30 and (d) 40 mm; finite element predictions and test results

WIT Transactions on Engineering Sciences, Vol 49


Impact Loading of Lightweight Structures, M. Alves & N. Jones (Editors)
°c 2005 WIT Press, www.witpress.com, ISSN 1743-3533
Modelling the dynamic failure of riveted joints 133

(a) (b)

(c) (d)

Figure 10: Comparison of axial load vs. displacement for quasi-static (three
curves) and 1 m/s pull velocities for a single lap joint where w =
(a) 12, (b) 20, (c) 30 and (d) 40 mm; finite element predictions
and test results

References

[1] Birch, R.S. & Alves, M., Dynamic failure of structural joint systems.
Journal of Thin-Walled Structures, 36, pp. 137–154, 2000.
[2] Birch, R.S. & White, M., European framework iii project crashworthi-
ness for commercial aircraft. Contract AERO-CT92-0030, University
of Liverpool: 18 month report, July 1994.
[3] Birch, R.S. & White, M., European framework iii project crashworthi-
ness for commercial aircraft. Contract AERO-CT92-0030, University
of Liverpool: 24 month report, January 1995.
[4] Langrand, B. & Combescure, A., Non-linear and failure behaviour of
spotwelds: a ”global” finite element and experiments in pure and mixed
modes I/II. International Journal of Solids and Structures, 41(24-25),
pp. 6631–6646, December 2004.
[5] Langrand, B., Bayart, A.S., Deletombe, E. & Chauveau, Y., Assessment
of multi-physics FE methods for bird strike modelling. application to
a metallic riveted airframe. ICRASH 2002, Melbourne, 2002.
[6] Hanssen, A.G., Porcaro, R., Alberg, A. & Langseth, M., Identification
and modelling of self-pierce rivet failure for crash analyses. Technical

WIT Transactions on Engineering Sciences, Vol 49


Impact Loading of Lightweight Structures, M. Alves & N. Jones (Editors)
°c 2005 WIT Press, www.witpress.com, ISSN 1743-3533
View publication stats

134 Robert S. Birch, Laura Vendrasco, Mark D. White & Michel Mahé

report, Structural Impact Laboratory; Norwegian University of Science


and Technology, Trondheim.
[7] Lesuer, D.R., Experimental investigation of material models for Ti-6Al-
4V titanium and 2024-T3 aluminium. The Federal Aviation Adminis-
tration, 2000.
[8] Johnson, G.R. & Cook, W.H., Fracture characteristic of three metals
subjected to various strains, strain rates, temperatures and pressures.
Engineering Fracture Mechanics, 21, pp. 31–48, 1985.
[9] Hancock, J.W. & Mackenzie, A.C., On the mechanisms of ductile failure
in high-strength steels subjected to multi-axial stress-states. J Mech
Phys Solids, 24, pp. 147–169, 1976.
[10] El-Magd, E. & Brodmann, M., Ductility of aluminium alloy AA7075
at high strain rates. Journal de Physique IV, 10, pp. 335–340, 2000.

WIT Transactions on Engineering Sciences, Vol 49


Impact Loading of Lightweight Structures, M. Alves & N. Jones (Editors)
°c 2005 WIT Press, www.witpress.com, ISSN 1743-3533

You might also like