Available online at www.sciencedirect.

com

ScienceDirect
Procedia CIRP 43 (2016) 292 – 296

14th CIRP Conference on Computer Aided Tolerancing (CAT)

Effect of hole-location error on the strength of fastened multi-

material joints
Ramzi Askria*, Christophe Boisa, Hervé Wargniera
aUniv. Bordeaux, I2M, UMR 5295, F-33400 Talence, France

* Corresponding author. Tel.: +33 5 56 84 79 85; fax: +33 5 56 84 58 43. E-mail address: [email protected]

Abstract

Fastened joint design is based essentially on good prediction of fastener-load distribution. The transmitted load depends on geometrical
specifications of coincident holes. Because of the low through-thickness compressive strength of laminated composite materials, a high clamp-
up cannot be applied to composite joints. So in single or double lap shear configurations, a sliding phase occurs during loading. Thus the final
transmitted load distribution is directly linked to bolt-hole clearance and location error. This work investigates the effect of hole-location error
on the strength of multi-material joints. A Monte Carlo and a Genetic Algorithm associated to Finite Element Method are used to estimate
maximal transmitted load evolution as a function of tolerance values of hole-location error.
© 2016 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license
Peer-review under responsibility of the organizing committee of the 14th CIRP Conference on Computer Aided Tolerancing.
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer-review under responsibility of the organizing committee of the 14th CIRP Conference on Computer Aided Tolerancing
Keywords: Fastened joint; Hole-location error; Composite; Finite Element Analysis

1. Introduction This phenomenon has a negative effect on joint strength and

material local damage [3–5]. In the case of a multi-fastened
The increase in the rate of aeronautical manufacturing has joint, load distribution between fasteners could be affected if,
led to a decentralization of production lines. In order to ensure in addition to the presence of clearances, hole pairs are not
interchangeability, different parts of an assembly must respect initially perfectly aligned [6,7]. Some fasteners could
tightened geometrical constraints, thus generating high costs. therefore be found to be in overload compared to nominal
These constraints are translated into tolerance value. An load, which could lead to fastener failure or material damage.
optimal tolerancing is generally a compromise between design Few studies have been performed on the effect of hole-
requirement and process capability and cost [1]. location error on multi-fastened joints. This may be due to the
A reduction in the process costing of a mechanical-fastened lack of reliable tools able to take into account existing
joint could be achieved by using loose tolerances on hole- variabilities in joints and to provide accurate information with
location errors. At the same time, larger hole diameters should reasonable calculation cost. The main origin of calculation
be considered in order to recover hole misalignment [2]. cost is the explicit modelling of contact between parts in order
However, in some cases, introducing larger pin-hole clearance to simulate the sliding of bolt heads and adherents.
may affect joint behaviour. It is for this purpose that a reduced finite element model of
In the case of metallic joints, bolt-hole clearance does not bolted joints was developed [8,9]. By using continuum shell
significantly affect joint strength. Indeed, due to their high elements for parts and rigid surfaces for fasteners it is possible
out-of-plane strength, a high clamp-up could be applied, with to take into account explicitly the contact between fastener
the load therefore transmitted by adherence. However, in the and adherents while controlling calculation time. The effect of
case of laminated composite joints, the material will be clearance, preload and friction is therefore considered in the
damaged during bolt torqueing if a high value of preload is model proposed here. The comparison with a solid 3-D model
considered. A loss of grip consequently occurs during shear showed a good estimate of global joint stiffness, bolt-load
load. The duration of the sliding phase of bolt heads and distribution and stress state around the hole. The saving in
adherents is directly related to clearance in each bolt-hole pair. calculation time is about 80%.

2212-8271 © 2016 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer-review under responsibility of the organizing committee of the 14th CIRP Conference on Computer Aided Tolerancing
doi:10.1016/j.procir.2016.02.040
 Ramzi Askri et al. / Procedia CIRP 43 (2016) 292 – 296 293

In this paper, the effect of hole-location error on joint

strength is studied using the reduced finite element model
quoted above. The long-term objective is to set up a robust
tool for aided-tolerancing. This tool should provide
geometrical specifications for the designer in order to ensure
non-exceedance of some essential dimensioning criteria, such
as maximal supported load by a fastener, relative displacement
between adherents or stress/strain-based failure criterion. The
dimensioning criterion used here is the force transmitted by
the most loaded fastener in a multi-bolted joint. This
information is considered as the primary indicator in joint
strength prediction. It could be compared with an admissible
bolt-load value or used to calculate the bearing failure
criterion [3,10].
Knowing the tolerance value to use for a given admissible
transmitted load depends firstly on the required reliability. In
some industrial fields, to reduce process costs a small Fig. 1. Geometry of the reference joint.
proportion of assemblies could be permitted. Statistical
tolerancing is therefore needed in order to establish the Table 1. Mechanical properties of materials used.
probability distribution of the dimensioning or tolerancing Properties of unidirectional E11 [GPa] E22 [GPa] E33 [GPa]
criterion [11–14]. The Monte Carlo method is used to generate ply for IMA/M21
143 7.89 7.89
random hole-location errors on the joint.
G12 [GPa] G13 [GPa] G23 [GPa]
With higher levels of reliability required, more than 99%, a
3.92 3.92 2.76
stochastic method such as Monte Carlo cannot estimate
extreme values. The problem of looking for a worst-case could ν12 ν13 ν23
then be solved by various optimization algorithms. Because of 0.33 0.33 0.43
the great number of possible configurations, the approach Properties of aluminium E [GPa] ν
chosen is the Genetic Algorithm. This is the second approach alloy 7075
70 0.3
presented in this paper.
Properties of titanium E [GPa] ν
2. Reference joint 110 0.3

2.1. Description 2.2. Finite element model

In order to evaluate the performance of proposed methods A finite element model of the reference joint described in
with multi-fastened and multi-material joints, a single-lap section 2.1 is created with Abaqus software. Fasteners are
aluminium/composite joint with 4 bolts is considered as a modelled by Multi-Connected Rigid Surfaces [8]. Adherents
reference joint, as shown in Fig. 1. With the purpose of are modelled with continuum shell elements [9] as shown in
expanding the domain of validity, the tensile load applied to Fig. 2. In this model, fasteners and holes are assumed to have
the joint is shifted by 2mm in the Y-direction from the (XZ) the same diameter and to be located in their nominal
symmetry plane so that load distribution between the fasteners positions.
becomes non-uniform. The composite material is a Boundary conditions are concentrated on two reference
unidirectional ply made of carbon fibre and thermoset matrix points which are rigidly linked to the nodes of each adherent
IMA/M21. The stacking sequence of the 17 plies of border. For these two reference points, rotation around Z is
composite adherent is allowed. The imposed tensile load of 30 kN is applied to the
>90 /  45 / 0 /  45 / 0 /  45 / 0 /  45 / 900.5 @s with ply thickness composite adherent. Concerning contact, a normal and
of 0.184 mm. The second joined adherent is an aluminium tangential contact with a friction coefficient of 0.2 is created
alloy 7075. The four titanium bolts with diameter of 6.35 mm between all fasteners and adherent surface pairs.
are fixed with an axial preload of 2700 N, generating under-
head pressure of 35 MPa. All material properties are listed in
Table 1.

Fig. 2. Finite element model of the reference joint.

294 Ramzi Askri et al. / Procedia CIRP 43 (2016) 292 – 296

3. Modelling hole-location error

3.1. Parametrization of hole-location error

All holes are assumed perfectly cylindrical and

perpendicular to the overlap plane. Nominal hole number i
C C
has a diameter Di and is centred on Ci X i ,Yi . The position
c
of the hole centre, noted C i , in a configuration
& with location
errors is obtained by a translation vector c
& ti Ci Ci . The
coordinates of ti are expressed in a polar coordinate system
(Eq. 1) as shown in Fig. 3. We note T the tolerance value.
& &
& ­° Di X , ti  S d Di d S ½°
ti ® & ¾ (1)
°̄ti ti 0 d ti d T °¿

Fig. 4. Description of proposed method steps

4. Results and discussion

4.1. Statistical tolerancing approach

The principal interest in using the Monte Carlo method is

to estimate the percentage variation in the criterion given by a
configuration with defects compared to a criterion given by a
reference configuration (with neither clearance nor location
0
error) noted Fmax for a fixed tolerance value T. The total
number of configurations tested is 5,000.
The graphs shown in Fig. 5a and Fig. 5b represent the
'Fmax
probability density of 0 for two tolerance introduced
Fmax
Fig. 3. Parametrization of hole-location error.
values: T1 0.05 mm and T2 0.1mm .
3.2. Generation of FE models with hole-location error First, a comment on the presence of a significant number
0
of cases with maximal force Fmax lower than Fmax . Indeed,
In order to limit the number of configurations, clearances
due to the non-uniform load distribution because of eccentric
and location errors are generated in composite adherents only.
load, particular hole locations could contribute to a more even
In this study, we assume that there is no variability in hole-
distribution and consequently to a decrease in Fmax .
clearance. It is considered equal to tolerance value T. Finite
element models with defects are generated from the reference Secondly, comparing the two criterion distributions shows
model described in section 2.2. that the greater the tolerance value, the larger the mean
For each configuration tested, a programme coded in criterion value and standard deviation.
Python reads coordinates of nodes in each hole of the
composite adherent. Those nodes are then & displaced from
their original positions following vector ti as described in
section 3.1. A model of the joint with hole-location errors is
therefore created, executed and post-processed in order to
extract bolt-load values noted Fi . The maximal transmitted
force Fmax is defined by (Eq. 2). Both approaches, Monte
Carlo and Genetic Algorithms, use the same automated tasks
shown in Fig. 4.

Fmax sup Fi (2)

i 1, 4
 Ramzi Askri et al. / Procedia CIRP 43 (2016) 292 – 296 295

amplitudes and angles of translation vectors are attributed.

Different genes are generated randomly in bounded intervals.
Table 2 shows a comparison of transmitted forces and
percentages for variations in the criteria for two tested
configurations. The reference configuration is noted C0.
Forces given by worst-case and noted C1 are obtained by
Genetic Algorithm. In the present case, the values produced
by the Genetic Algorithm are quite close to some of the
configurations simulated by the reliability-based approach.
However, for more complex cases (more fasteners, more
parameters subjected to variability), the total number of
configurations tested with the Monte Carlo method should be
reduced and the space of interest scanned in a parsimonious
manner. The worst-case approach will thus become suitable.

Table 2. Load distribution for configurations C0 and C1.

Configu- T F1 F2 F3 F4 'Fmax [%]
ration [mm] [N] [N] [N] [N] 0
Fmax
C0 - 6973 6610 7995 7852 -
0.05 7375 5295 6450 10419 +30.3
C1
0.1 4979 7033 12689 4969 +58.7
Fig. 5a. Probability density of the criterion variation using tolerance value
T1=0.05 mm
4.3. Tolerance value and criterion relationship

Once the probability density of criterion and maximal

criterion given by the worst-case has been performed, a
relationship can be established between tolerance value and
the dimensioning criterion for a given reliability level. The
graph in Fig. 6 shows that, in addition to the proportional
relation between tolerance value and criterion, there was a
significant increase in the criterion for very high reliability
(more than 99%). In order to understand the usefulness of this
graph, a simple example is treated.
With a reliability level of 99%, a designer would know the
required tolerance value to apply to hole-location in order to
admissible
ensure non-exceedance of an admissible bolt-load Fmax of
0
10000 N. The reference maximal force Fmax (given by a
configuration with zero hole-location errors) is 7995'Fmax N, as
shown in Table 2. This means that the criterion 0 is
Fmax
about 25%. The designer should therefore choose a tolerance
value lower than 0.064 mm.

Fig. 5b. Probability density of the criterion variation using tolerance value
T2=0.1 mm

4.2. Worst-case approach

Statistical methods such as Monte Carlo are recommended

for estimating tolerance values dependent on reliability,
however, due to the large number of combinations and non-
linear responses to the problem, worst-cases could not been
found with this kind of approach. It is therefore more relevant
to use an efficient method, able to maximize Fmax and give the
worst-case. The method selected is a Genetic Algorithm.
Concerning Genetic Algorithm parameters, the initial
population count is 200 individuals. For each individual, 8
genes (t1, t2 , t3 , t4 ,D1 ,D2 ,D3 ,D4 ) representing location-error
296 Ramzi Askri et al. / Procedia CIRP 43 (2016) 292 – 296

References

[1] K.W. Chase, A.R. Parkinson, A survey of research in the application

of tolerance analysis to the design of mechanical assemblies,
Research in Engineering Design. 3 (1991) 23–37.
[2] F. Scholz, Hole Alignment Tolerance Stacking Issues, Mathematics
and Computing Technology, Boeing Phantom Works. (1999).
[3] M.A. McCarthy, V.P. Lawlor, W.F. Stanley, C.T. McCarthy, Bolt-
hole clearance effects and strength criteria in single-bolt, single-lap,
composite bolted joints, Compos. Sci. Technol. 62 (2002) 1415–1431.
[4] V.P. Lawlor, M.A. McCarthy, W.F. Stanley, An experimental study
of bolt-hole clearance effects in double-lap, multi-bolt composite
joints, Compos. Struct. 71 (2005) 176–190.
[5] M.A. McCarthy, C.T. McCarthy, G.S. Padhi, A simple method for
determining the effects of bolt–hole clearance on load distribution in
single-column multi-bolt composite joints, Composite Structures. 73
(2006) 78–87.
[6] J. Lecomte, C. Bois, H. Wargnier, J.-C. Wahl, An analytical model
for the prediction of load distribution in multi-bolt composite joints
including hole-location errors, Compos Struct. 117 (2014) 354–361.
[7] G.L. Kulak, J.W. Fisher, J.H.A. Struik, Guide to Design Criteria for
Bolted and Riveted Joints, American Institute of Steel Construction,
Chicago, IL, 1987.
[8] R. Askri, C. Bois, H. Wargnier, J. Lecomte, A reduced fastener model
using Multi-Connected Rigid Surfaces for the prediction of both local
stress field and load distribution between fasteners, Finite Elements in
Analysis and Design. 110 (2016) 32–42.
Fig. 6. Evolution of the criterion with tolerance value [9] R. Askri, C. Bois, H. Wargnier, J. Lecomte, Reduced bolted joint
model using multi-connected rigid surfaces and continuum shell
5. Conclusion elements, in: The 20th International Conference on Composite
Materials, Copenhagen (Denmark), 2015.
[10] ASTM standard D 5961/D 5961M- 96, Standard test method for
A study of the effect of hole-location error on the strength bearing response of polymer matrix composite laminates, (1996).
of a multi-material multi-fastened joint is described in this [11] J.-Y. Dantan, A.-J. Qureshi, Worst-case and statistical tolerance
analysis based on quantified constraint satisfaction problems and
paper. Tolerancing analysis based on statistical tolerance and
Monte Carlo simulation, Computer-Aided Design. 41 (2009) 1–12.
worst-case approaches was performed. The statistical [12] J. Camelio, S.J. Hu, D. Ceglarek, Modeling Variation Propagation of
tolerancing method computes the criterion with a given Multi-Station Assembly Systems with Compliant Parts, Journal of
reliability requirement. The worst-case approach was used to Mechanical Design, Transactions of the ASME. 125 (2003) 673–681.
[13] P. Franciosa, S. Gerbino, S. Patalano, Simulation of variational
estimate the highest criterion value that could be attempted by compliant assemblies with shape errors based on morphing mesh
the joint. approach, International Journal of Advanced Manufacturing
In future research, by combining the two methods, Technology. 53 (2011) 47–61.
[14] R. Söderberg, K. Wärmefjord, L. Lindkvist, Variation simulation of
statistical and worst-case, we will be able to establish a stress during assembly of composite parts, CIRP Annals -
continuous probability density distribution function and select Manufacturing Technology. 64 (2015) 17–20.
the tolerance value for a given criterion value and reliability
level.