Engineering Failure Analysis 18 (2011) 364–373

Contents lists available at ScienceDirect

Engineering Failure Analysis

journal homepage: www.elsevier.com/locate/engfailanal

Failure analysis of bolted joints: Effect of friction coefficients

in torque–preloading relationship
D. Croccolo ⇑, M. De Agostinis, N. Vincenzi
DIEM, University of Bologna, Viale Risorgimento, 2, 40136 Bologna, Italy

a r t i c l e i n f o a b s t r a c t

Article history: The aim of this study is to provide an experimental methodology useful to determine the
Received 14 June 2010 friction coefficients in bolted joints and, therefore, to relate precisely the tightening torque
Received in revised form 16 September to the preloading force. The components under investigation are clamped joints made of
2010
aluminium alloy and used in front motorbike suspensions to connect steering plates and
Accepted 17 September 2010
Available online 29 September 2010
legs, or legs and the wheel pin: static failures of clamps occurred during the tightening,
because of the bending stress introduced by the preloading forces. Some specific specimens
have been appropriately designed and realised with the same process of the actual compo-
Keywords:
Tightening test
nents. The bolt torque is given by a torque wrench whereas the preloading force has been
Friction coefficient evaluated by means of a strain gauge. The overall friction coefficient and the torque coef-
Bolted connections ficient (nut factor) have been calculated. Experimental tests have been carried out by
Preloading force applying the Design of Experiment method in order to obtain an accurate mathematical
model that involves the significant friction variables and their interactions. Then, results
of present study have been applied to actual components: the tightening torque has been
precisely related to the preloading force by means of the friction coefficients definition and
the tensile state of clamps have been evaluated both experimentally (strain gauges) and
numerically (FEA) in order to shed light on the failures which occurred during the
tightening.
Ó 2010 Elsevier Ltd. All rights reserved.

1. Introduction

Threaded fasteners are widely used in mechanical and structural applications because of their easiness to assemble and
disassemble for maintenance and repair purposes. Both the level and the stability of the clamp loads, which are created by
the tightening process, will govern the safety and reliability of bolted joints. In most of production applications, the fastener
tension (preloading force) is achieved by using a torque wrench applied to the head or to the nut. The tribological aspects of
the tightening phase are critical to define the actual torque–tension relationship [1,2]. The tightening torque is mostly con-
sumed during the process of overcoming two friction components: the underhead (or bearing) friction due to the sliding of
the fastener head on the flanges and the thread friction between the male and female thread. The residual torque component
produces the fastener tension by generating the joint clamping force. Inaccuracies in determining the friction components
may lead to an overestimation or underestimation of the bolted joint performances. The torque–preloading relationship is
often simplified by using a constant K (Eq. (1)), known as torque coefficient (or nut factor): some authors and Standards
[3–5] provide an approximate value of 0.20 for the nut factor, but warnings against using it for critical joints should be

Abbreviations: DOE, Design of Experiment; FEM, finite elements method; ANOVA, Analysis of Variance; F-tests, Fisher’s tests.
⇑ Corresponding author. Tel./fax: +39 0512093413.
E-mail address: dario.croccolo@unibo.it (D. Croccolo).

1350-6307/$ - see front matter Ó 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.engfailanal.2010.09.015
 D. Croccolo et al. / Engineering Failure Analysis 18 (2011) 364–373 365

Nomenclature

Aeq cylindrical equivalent section of the specimen (mm2)

At screw tensile stress area (mm2)
d2 pitch (mean) diameter of the thread, d2 = d  0.649  p (mm)
dt core (minor) diameter of the thread, dt = d  1.227  p (mm)
du underhead mean diameter (mm)
E Young’s modulus (MPa)
FV preloading force (N)
FC preloading force of the specimen (N)
K torque coefficient
Kt stress concentration factor
T total torque (N mm)
Tp torque due to sliding skew palne of the screw (N mm)
Tt torque due to the friction of the screw (N mm)
Tu torque due to the friction of the nut (N mm)
p thread pitch (mm)
SY,U yield, ultimate point (MPa)
eC external strain of the specimen
m Poisson’s ratio
lt friction coefficient of the screw
lu underhead friction coefficient
lm mean friction coefficient
rb_max maximum bending stress due to the preloading force FV (MPa)

considered. Bickford [6] provides some mean values of the torque coefficient for various combinations of joint materials and
surface conditions. However, the scatter is too great to provide a unique and reliable value, particularly in critical joints.

T ¼ K  FV  d ð1Þ

T (N mm) is the input tightening torque applied to the fastener head or nut, FV (N) is the preloading force and d (mm) is the
nominal thread diameter.
Motosh [7] and VDI2230 [8] provided a more accurate T–FV relationship, reported in Eq. (2), as addition of three different
contributions: the pitch Tp, the thread Tt and the underhead Tu torque.

T ¼ T p ðF V ; pÞ þ T t ðF V ; lt ; d2 Þ þ T u ðF V ; lu ; du Þ
ð2Þ
T ¼ F V  ½0:16  p þ 0:58  lt  d2 þ 0:5  lu  du 

p (mm) is the thread pitch, lt is the friction coefficient between male and female threads, d2 (mm) is the mean thread diam-
eter (d2 = d  0.6495  p), lu is the underhead friction coefficient and du (mm) is the underhead mean diameter. The pitch
torque component Tp (N mm) creates the fastener tension and the joint clamp load FV. By considering a M8  1.25 bolt
(d2 = 7.19 mm, du = 11 mm) and a mean value for friction coefficients lt and lu equal to 0.15, it is easy to calculate that
Tp  0.12 T, Tt  0.38 T and Tu  0.50 T: the 88% of the total input torque T is consumed in overcoming friction (only 12%
is converted in preloading force).
This study aims to develop an experimental tool and procedure useful to define accurately the friction coefficients in
bolted joints and, therefore, at relating precisely the input bolt torque T to the bolt preloading force FV. The components
under investigation are clamped joints made of aluminium alloy used in front motorbike suspensions to connect steering
plates and legs or legs and wheel pin [9]: static failures occurred during the tightening phase of such components because
of the bending stress produced by preloading forces in the compliant part of the clamp (Fig. 1).
The aluminium alloy under investigation (G-AlSi 7, yield point SY = 196 MPa, ultimate point SU = 304 MPa) is realised by a
casting or a forging process with subsequent machining in order to obtain a surface roughness Ra = 3.2 lm, and finally anod-
ized or spray-painted (thickness 5–9 lm).
The same tightening torque could produce a high locking force if friction coefficients are low; conversely, high friction
coefficients could provide an insufficient locking force. Therefore an incorrect evaluation of friction parameters may lead
to dangerous failures [10], especially in applications where human safety is involved, such as front motorbike suspensions.
In previous works [9,11], Croccolo et al. investigated the tensile state of clamped joints in front motorbike suspensions,
via FEM: engineering design formulae are developed and proposed in order to estimate the maximum stress on the clamp
and the coupling pressure as a function of the bolt preloading force and of the clamp geometry. The present work deals with
the extension of the results presented in Refs. [9,11] by leveraging the accurate knowledge of friction coefficients and,
366 D. Croccolo et al. / Engineering Failure Analysis 18 (2011) 364–373

FAILURES

10mm

Fig. 1. Example of failure occurred during the tightening phase of a spray-painted clamp.

therefore by using, as loading input parameter for the calculation, the tightening torque (in the place of the preloading force),
which is actually controlled during the manufacturing process.

2. Methodology and results

In order to evaluate accurately the friction coefficients and to gather their effective mathematical expression, the Design
of Experiment (DOE) method was applied. At first, a screening analysis was performed: the applied torque T (5–20 Nm), the
bolt type (hexagon flange bolts-DIN6921 or hex socket bolts-DIN912, according to Fig. 2) and the bolt diameters (M6  1 or
M8  1.25) are not significant in changing the friction coefficient values. The bolts used are zinc plated and realized with a
product grade B (according to ISO4799) while tightening was always performed at room temperature.
In light of the screening tests a full factorial plane, characterized by four variables with two levels each, was designed.
Three replicas were carried out, in order to reduce the influence of noise (experimental error) and any non-investigated fac-
tors [12]: a total of 3  24 = 48 experimental tests were carried out. In Table 1 the DOE parameters are summarized.
According to VDI2230 [8] and DIN946 [13], Eq. (2) was rewritten into Eq. (3), considering lt and lu equal to the same
value lm in order to describe the overall friction behaviour of the bolt/nut assemblies.

T ¼ F V  ½0:16  p þ 0:58  lm  d2 þ 0:5  lm  du 

T=F V  0:16  p ð3Þ
lm ¼
0:58  d2 þ 0:5  du
Each test planned by the DOE provided a value of the overall friction coefficient lm (Eq. (3)): in order to calculate lm a
specific specimen, reported in Fig. 3, was appropriately designed and realized with the same material, roughness, process
(cast or forged) and surface finishing (spray-painted or anodized) of the actual components (Fig. 1). The study considered
M8  1.25, SAE Standard 8.8, zinc-plated screws.

Fig. 2. Example of different M8 8.8 zinc-plated screws used for friction coefficient definition.
 D. Croccolo et al. / Engineering Failure Analysis 18 (2011) 364–373 367

Table 1
The Design of Experiment (DOE) parameters: variables and levels.

Variable Low level (0) High level (1)

A. Lubrication Unlubricated (A = 0) Lubricated–oil SAE10-EBH16 (A = 1)
B. Forming process Cast (B = 0) Forged (B = 1)
C. Surface finishing Spray-painted (C = 0) Anodized (C = 1)
D. Number of tightening First tightening–unspoiled surfaces (D = 0) Sixth tightening–spoiled surfaces (D = 1)

Specimen
Strain
gauge

Support

Fig. 3. The specimen used for the friction coefficient lm definition.

The total torque T was given by a calibrated torque wrench whereas the preloading force FV was evaluated by means of a
strain gauge, located on the external surface of the specimen, which is able to provide the axial compression strain eC. The
compression force FC (Eq. (4)) acting on the specimen is equal, in magnitude, to the preloading force acting on the screw,
since the system works like springs connected in series during the tightening phase. The innovative specimens designed
and manufactured by the authors are cheaper and easier to use than the most common friction machinery, based on dis-
placement transducers, load cells, rotary torque sensors/angle encoders and DC motors [2,13]. Furthermore, it is possible
to avoid the application of strain gauges on the bolt (difficult to perform in the case of nominal diameters of less than
12 mm) in order to measure the actual preloading force.

jF C j ¼ F V ¼ jeC j  E  Aeq ð4Þ

2
E (MPa) is Young’s modulus of aluminium specimen (69,000 MPa) and Aeq (mm ) is the cylindrical equivalent cross section of
the compressed region. Since the specimen was designed to be very similar to a thin bush, the equivalent cross section Aeq
was considered equal to the actual cross area A, as reported in Eq. (5) and proposed in [14], in which DA (16 mm) is the exter-
nal and dh (8.5 mm) the internal diameters of the bush.
p 2
A¼  ðD2A  dh Þ ¼ 144:3 mm2 ð5Þ
4
According to [15], experimental tests were randomized: a total torque T equal to 15 Nm was applied to each specimen
(Fig. 4). The preloading force FV was evaluated and both the overall friction coefficient lm (Eq. (3)) and the torque coefficient
K (Eq. (1)) were calculated, as functions of the DOE variables and levels (Table 1). The DOE plan and results (48 different
tests) are reported in Table 2: the scatters were calculated by performing the three replicas under the same conditions
and evaluated by the DOE method in order to apply the Analysis of Variance (ANOVA) [15,16]. This technique is based on
a statistical approach and allows evaluation of the effect of each variable and of their interaction, in order to identify those
variables (or interactions) that are actually significant in changing the friction coefficients value.
By means of the software StagraphicsÒ Plus, Release 5.1, the ANOVA table was generated (an example is reported in
Table 3) and Fisher’s test (F-test) was executed. F0 value was determined subdividing the mean square of the single effect
(or interaction) with the mean square of the error [12,15,16]. The error probability a was set equal to 0.10 (10%) and Fisher’s
value Fm1;m2;a was obtained from Fisher’s table [15,16] (m1 are the degrees of freedom of the source of variation and m2 of the
source of error). The variables (or interactions) with F0 higher than the corresponding Fm1;m2;a (F1;33;10% = 2.87) are significant in
the analysis, whereas the others can be ignored in the response model (white cells in Table 3), with a confidential level of
(1  a) = 90%. ANOVA and F-test provide Eqs. (6) and (7) in which the Regression Coefficients and, therefore, the mathematical
models for lm and K are reported.
368 D. Croccolo et al. / Engineering Failure Analysis 18 (2011) 364–373

Fig. 4. Forged (left) vs cast (right) specimen.

lm ¼ 0:108  0:023  A þ 0:103  C þ 0:010  D  0:007  A  B  0:063  A  C þ 0:017  A  D þ 0:053  B  C

þ 0:117  C  D  0:153  A  C  D ð6Þ

K ¼ 0:157  0:028  A þ 0:123  C þ 0:013  D  0:008  A  B  0:075  A  C þ 0:015  A  D þ 0:065  B  C

þ 0:138  C  D  0:180  A  C  D ð7Þ
Surface finishing (specimen spray-painted or anodized), lubrication (specimen lubricated via oil SAE10-EBH16 or unlubri-
cated), their interaction, and finally the number of tightening are the most significant parameters (highest values of F0 are
written in bold-face in Table 3). In the bars diagrams of Figs. 5 and 6 results are compared on the basis of the aforementioned
parameters in the case of forged specimens.
It is possible to affirm that spray-painted specimens show the lower value of lm (the higher preloading forces FV). In the
presence of unspoiled surfaces, the overall friction coefficient lm decreases from a value of 0.26 in the case of forged, anod-
ized and unlubricated specimens to a value of 0.11 in the case of forged, spray-painted and unlubricated ones, so that the
preloading force doubles with the same tightening torque T = 15 Nm. Lubrication always increases the preloading forces:
in the presence of forged and unspoiled surfaces, the overall friction coefficient lm decreases from a value of 0.11 to a value
of 0.08 in the case of spray-painted specimens, and from a value of 0.26 to a value of 0.16 in the case of anodized ones.
Finally, considering the effect of the number of tightening and loosening (up to six maintenance operations in a standard
motorbike lifecycle) the preloading force is affected by the tightening replicas mainly in the case of unlubricated and
anodized surfaces so that the preloading forces progressively decrease with the same tightening torque. As a matter of fact
surfaces are subjected to wear and spoiling (Fig. 7) by increasing the number of tightening, while lubrication creates a sort of
protective film. In the presence of spoiled surfaces, the overall friction coefficient lm increases from a value of 0.26 to a value
of 0.39 in the case of forged, anodized and unlubricated surfaces, whereas a value of 0.16 to a value of 0.17 is observed in the
case of forged, anodized and lubricated surfaces.
In short, friction conditions are strongly affected by surface finishing (C variable), lubrication (A variable) and number of
tightening and loosening (D variable); conversely, the forming process (cast or forged aluminium alloy) seems to have no
significant influence on friction conditions. Spray-painted specimens (C = 0) present the lower values of lm and K (the higher
preloading forces FV). Lubrication (A = 1) always increases the preloading forces, while the tightening replicas (D = 1) pro-
gressively decrease the preloading forces, mainly in the case of dry surfaces, as demonstrated also in Ref. [17].

3. Failure analysis: a case study

The previous outcomes, summarized in Eq. (6), were applied to an unlubricated, spray-painted, cast aluminium wheel
clamp, realized by Paioli Meccanica of Bologna (IT), which produces front motorbike suspensions. Two M6  1 SAE Standard
8.8 zinc-plated screws (d = 6 mm, p = 1 mm, d2 = 5.35 mm and du = 8 mm) were tightened up to six times each. Strain gauges
 D. Croccolo et al. / Engineering Failure Analysis 18 (2011) 364–373 369

Table 2
DOE plan and results provided by experimental tests (T = 15 Nm, M8  1.25).

ID A – lubrication B – forming process C – surface finishing D – number of tightening FV (N) lm K

7 0 0 0 0 14,865 0.08 0.13
20 0 0 0 0 12,989 0.10 0.14
37 0 0 0 0 11,257 0.12 0.17
31 0 0 0 1 10,102 0.13 0.19
13 0 0 0 1 12,267 0.11 0.15
42 0 0 0 1 9381 0.14 0.20
4 0 0 1 0 7504 0.19 0.25
19 0 0 1 0 5484 0.26 0.34
34 0 0 1 0 6783 0.21 0.28
10 0 0 1 1 5051 0.29 0.37
30 0 0 1 1 4041 0.36 0.46
41 0 0 1 1 4330 0.34 0.43
5 0 1 0 0 10,391 0.13 0.18
17 0 1 0 0 14,432 0.09 0.13
35 0 1 0 0 11,978 0.11 0.16
12 0 1 0 1 7793 0.18 0.24
32 0 1 0 1 9814 0.14 0.19
45 0 1 0 1 8659 0.16 0.22
3 0 1 1 0 6783 0.21 0.28
24 0 1 1 0 6206 0.23 0.30
38 0 1 1 0 4762 0.31 0.39
14 0 1 1 1 4185 0.35 0.45
26 0 1 1 1 3608 0.41 0.52
43 0 1 1 1 3802 0.37 0.47
6 1 0 0 0 17,751 0.07 0.11
23 1 0 0 0 16,885 0.07 0.11
40 1 0 0 0 9381 0.14 0.20
15 1 0 0 1 16,019 0.08 0.12
27 1 0 0 1 15,298 0.08 0.12
47 1 0 0 1 9236 0.15 0.20
1 1 0 1 0 10,102 0.13 0.19
22 1 0 1 0 15,009 0.08 0.12
39 1 0 1 0 9669 0.14 0.19
16 1 0 1 1 10,824 0.13 0.18
29 1 0 1 1 15,153 0.08 0.12
44 1 0 1 1 8659 0.16 0.22
8 1 1 0 0 17,895 0.07 0.10
18 1 1 0 0 21,648 0.05 0.09
36 1 1 0 0 16,885 0.07 0.11
9 1 1 0 1 16,019 0.08 0.12
28 1 1 0 1 18,040 0.07 0.10
48 1 1 0 1 15,442 0.08 0.12
2 1 1 1 0 9092 0.15 0.21
21 1 1 1 0 8515 0.16 0.22
33 1 1 1 0 9525 0.14 0.20
11 1 1 1 1 10,679 0.12 0.18
25 1 1 1 1 10,102 0.13 0.19
46 1 1 1 1 9525 0.14 0.20

were placed on the critical section of the clamp (loaded by a bending stress, as amply demonstrated in [9]), where a stress
concentration factor Kt exists, mainly due to the presence of spot-facings and holes.
Referring to Fig. 8, the theoretical bending stress in the critical section was evaluated by applying Eq. (8) [9], in which FV is
the preloading force, a (mm) the lever arm of FV with respect to the critical section, n the number of bolts, b (mm) and h (mm)
are the width and the height of the rectangular cross section, respectively.

Mb n  F V  a
rb th ¼ ¼ ð8Þ
W b b  h2 =6

8 nF V a
< rb max ¼ rb th  K t ¼ bh 2
=6
 Kt
ð9Þ
: Kt M6 ¼ 2:438 þ 0:548 
hsf
 1:131  d a  0:393  d v
h sf sf

The maximum bending stress on the clamp was calculated by applying Eq. (9) in which the stress concentration factor Kt
takes into account the perturbation produced by the spot-facings located close to the coupling zone. The geometrical dimen-
sions of clamps under investigation are reported in Table 4: the actual Kt value is equal to 1.37 [9,11].
370 D. Croccolo et al. / Engineering Failure Analysis 18 (2011) 364–373

Table 3
Example of ANOVA table results for lm.

Effect (interaction) Sum of square DoF Mean square F0

A – lubrication 0.1610 1 0.1610 53.31
B – forming process 0.0085 1 0.0085 2.83
C – surface finishing 0.1776 1 0.1776 58.81
D – number of tightening 0.0352 1 0.0352 11.66
AB 0.0114 1 0.0114 3.78
AC 0.0690 1 0.0690 22.85
AD 0.0320 1 0.0320 10.61
BC 0.0108 1 0.0108 3.58
BD 0.0030 1 0.0030 1.00
CD 0.0102 1 0.0102 3.38
ABC 0.0004 1 0.0004 0.14
ABD 0.0056 1 0.0056 1.87
ACD 0.0176 1 0.0176 5.84
BCD 0.0010 1 0.0010 0.33
Error 0.0997 33 0.0030 1.00

Fig. 5. Bar diagrams of the friction coefficient lm for forged specimens ((a) spray-painted and (b) anodized) in case of tightening torque T = 15 Nm (different
series for lubricated and unlubricated specimens).

Fig. 6. Bar diagrams of the preloading force FV for forged specimens ((a) spray-painted and (b) anodized) in case of tightening torque T = 15 Nm (different
series for lubricated and unlubricated specimens).

A tightening torque T = 9.5 Nm was applied to the wheel clamp: the preloading forces produced by the tightening were
evaluated by means of Eq. (10), in which the overall friction coefficient lm was equal to 0.108 during the first tightening
(A = B = C = D = 0, according to Table 1 and Eq. (6)) and equal to 0.118 during the sixth tightening (A = B = C = 0, D = 1, accord-
ing to Table 1 and Eq. (6)).

T
FV ¼ ð10Þ
½0:16  p þ 0:58  lm  d2 þ 0:5  lm  du 
By applying the Von Mises equivalent stress criterion, the maximum preloading forces, with respect to the bolt yielding or
failure, was evaluated by means of Eq. (11) (see also the sketch reported in Fig. 9), in which SY (MPa) is the yielding point and
SU (MPa) is the ultimate point of the screw (SAE Standard 8.8: SY = 640 MPa, SU = 800 MPa), while At (mm2) is the screw ten-
sile stress area and dt its diameter (At = 20.1 mm2, dt = 5.06 mm, in case of bolt M6  1) [8].
 D. Croccolo et al. / Engineering Failure Analysis 18 (2011) 364–373 371

(a) (b)

Spoiling

Fig. 7. Spray-painted specimens (a) unspoiled surfaces (D = 0) and (b) spoiled surfaces (D = 1).

Critical
section

Fig. 8. Example of wheel clamp: geometrical dimensions.

Table 4
Wheel clamp geometrical dimensions useful for the calculation according to Fig. 8.

Description Symbol (Fig. 8) (mm) Value

Wheel pivot diameter D 20
Distance between the wheel pivot centre and the bolt axis L 15.5
Lever arm of the preloading force (L–D/2) a 5.5
Distance between the wheel pivot and the clamp side c 16
Spot-facing diameter ds–f 11
Clamp total height htot 45
Clamp height h 18.5
Thread height hth 24.5
Spot-facing height hs–f 4.3
Clamp width b 36
Distance between bolt axes v 17.5
Mill radius r 2

SY
FV yielding ¼ sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 2ffi
 2
1 0:16pþ0:58lm d2
At
þ3 pðd Þ3
t
16
ð11Þ
SU
FV failure ¼ sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 2ffi
 2
1 0:16pþ0:58lm d2
At
þ3 pðd Þ3
t
16

Numerical values of the preloading forces are reported in Table 5, as a function of the overall friction coefficient lm (Eq.
(6)).
372 D. Croccolo et al. / Engineering Failure Analysis 18 (2011) 364–373

Fig. 9. Sketch of the load diagrams (axial and torsional) acting on the bolt during tightening.

Table 5
Preloading force values as a function of the overall friction coefficient (bolt: M6  1, 8.8).

Clamp parameters (Table 1) T (Nm) lm FV (kN) (Eq. (10)) FV_yielding (kN) (Eq. (11)) FV_failure (kN) (Eq. (11))
A=B=C=D=0 9.5 0.108 10.220 10.640 13.300
A = B = C = 0, D = 1 9.5 0.118 9.500 10.430 13.040

During the first tightening (fresh and unspoiled surfaces) the strain gauge applied on the clamp critical section (Fig. 10),
gave a strain value that was equivalent to a stress of 79 MPa. A numerical (FEM) nonlinear analysis, with contact elements
between the shaft and the hub [9], was performed on the same joint by imposing the preloading force accurately calculated
by the proposed formulae (FV = 10.220 kN of Table 5). The stress evaluated via FEM was equal to 82 MPa so that the differ-
ence between the experimental test and the FEM values is lower than 4%. During the sixth tightening (spoiled surfaces) the
test provided a strain value equivalent to a stress of 73 MPa: the ratio between the sixth and the first tightening bending
stresses (73/79 = 0.92) was very close to the ratio between the first and the sixth preloading force calculated by the formulae
(9.500/10.220 = 0.93).
The engineering design formulae of Ref. [9], reported in Eqs. (8) and (9), were able to provide, during the first tightening, a
maximum bending stress rb_max equal to 75 MPa (theoretical bending stress rb_th equal to 55 MPa and stress concentration
Kt equal to 1.37): the discrepancy with respect to the experimental test value was equal to 5%. Findings are summarized in
Table 6.
Another clamp, with the same geometry, was studied during the tightening phase: in this case, the bolt was tightened until
its failure, in correspondence to the sixth tightening. Since the yielding of the screw was overcome, the torque–preloading
relationship (Eq. (10)) is no longer effective. However, according to Eq. (11), it was possible to calculate the preloading force
in correspondence to the bolt failure, which is equal to 13.040 kN (Table 5). When the bolt failure occurred (Fig. 11) the strain
gauge applied on the clamp gave a strain value equivalent to a stress of 107 MPa.

Kt

Fig. 10. Experimental tightening test (T = 9.5 Nm) vs FEM analysis (FV = 10.220 kN).
 D. Croccolo et al. / Engineering Failure Analysis 18 (2011) 364–373 373

Table 6
Comparison between the values provided by the experimental (strain gauge), the numerical (FEM) and the design formulae applied to an unlubricated, spray-
painted, cast aluminium clamp.

Clamp parameters (Table 1) T (Nm) FV (kN) rstrain_gauge (MPa) rFEM (MPa) (e%) rdesign_formulae (MPa) (e%)
A=B=C=D=0 9.5 10.220 79 82 (+4%) 75 (5%)
A = B = C = 0, D = 1 9.5 9.500 73 76 (+4%) 70 (4%)

Fig. 11. Experimental failure of a bolt, during the tightening phase.

By performing again the numerical nonlinear analysis on the joint and imposing the failure preloading force
FV_failure = 13.040 kN a bending stress equal to 104 MPa was reached, as shown in Fig. 11: the discrepancy with respect to
the experimental result was equal to 3%. The engineering design formulae (Eq. (10)) provide a maximum bending stress
rb_max equal to 96 MPa: the discrepancy with respect to the experimental result was equal to 10%.

4. Conclusions

Friction and torque coefficients have been calculated for aluminium bolted joints: specific and innovative specimens have
been designed and realised in order to evaluate different process and product parameters shared by different industrial
applications (i.e. cast vs forged aluminium alloy, spray-painted vs anodized surfaces, lubricated vs unlubricated screws
and, finally, unspoiled vs spoiled surfaces). Friction conditions are significantly affected by surface finishing, lubrication
and the number of tightening and loosening, whereas the forming process (forged or cast aluminium alloy) seems to have
no significant influence. The range of variability of the friction coefficient is between 0.08 and 0.39, while for the nut factor it
falls between 0.12 and 0.50: with the same applied torque, preloading force can change up to 320%. Spray-painted surfaces
have the lowest friction coefficients: in the presence of such finishing it is necessary to apply an appropriate tightening
torque in order to avoid clamp failure (see Fig. 1). Every maintenance operation (loosening and tightening) provides a loss
in preloading force, which is particularly evident in the presence of dry conditions: wear pattern indicates that coating is
progressively peeled off. Tightening torque should be increased with the number of tightening in order to guarantee the
same preloading force and coupling pressure between the parts assembled by means of the clamp.

References

[1] Nassar SA, El-Khiami H, Barber GC, Zou Q, Sun TS. J Tribol 2005;127:263–72.
[2] Nassar SA, Ganeshmurthy S, Ranganathan RM, Barber GC. ASME J Pres Ves Technol 2007;129:426–40.
[3] Juvinall RC, Marshek KM. Fundamental of machine component design. 3rd ed. New York: Wiley; 2000.
[4] Nassar SA, Matin PH, Barber GC. Thread friction torque in bolted joints. ASME J Pres Ves Technol 2004;127:387–93.
[5] CNR-UNI10011. Costruzioni in acciaio: istruzioni per il calcolo, l’esecuzione, il collaudo e la manutenzione; 1988.
[6] Bickford JH. An introduction to the design and analysis of bolted joints. 3rd ed. New York: Marcel Dekker; 1997.
[7] Motosh N. Development of design charts for bolts preload up to the plastic range. ASME J Eng Ind 1976;98:849–51.
[8] VDI Handbuch Konstruktion 2230. Systematic calculation of high duty bolted joints – joints with one cylindrical bolt; 2001.
[9] Croccolo D, Cuppini R, Vincenzi N. Finite Elem Anal Des 2009;45:406–14.
[10] Eliaz N, Gheorghiu G, Sheinkopf H, Levi O, Shemesh G, Ben-Mordechai A, et al. Eng Fail Anal 2003;10:443–51.
[11] Croccolo D, De Agostinis M, Vincenzi N. Eng Fail Anal 2010;17:1173–87.
[12] Croccolo D, Cuppini R, Vincenzi N. Strain 2008;44:170–9.
[13] DIN946. Determination of coefficient of friction of bolt/nut assemblies under specified conditions; 1991.
[14] Niemann G, Winter H, Hohn BR. Maschineelemente B.d. I. Berlin: Springer-Verlag; 2005.
[15] Montgomery DC. Design and analysis of experiment. 5th ed. New York: Wiley; 2001.
[16] De Vor R, Chang TH, Sutherland JW. Statistical quality design and control. New York: Maxwell Macmillan International Edition; 1992.
[17] Eccles W, Sherrington I, Arnell RD. Tribol Int 2010;43:700–7.