Professional Documents
Culture Documents
Three-Dimensional Finite Element Analysis of Threaded Fastener Loosening Due To Dynamic Shear Load
Three-Dimensional Finite Element Analysis of Threaded Fastener Loosening Due To Dynamic Shear Load
www.elsevier.com/locate/engfailanal
Abstract
The two most widespread causes of failure of threaded fasteners subjected to dynamic loads are fatigue and vibration
induced loosening. This paper presents results of a study on failure of threaded fasteners by vibration induced loosening
caused due to dynamic shear loads. Previous experimental work has revealed that fastener loosening occurs as a result of
complete or localized slip at the thread and head contact surfaces. A three-dimensional nite element (FE) model is used to
study details of four dierent loosening processes that are characterized by either complete or localized slip at the head and
thread contacts. The FE model is found to be capable of adequately modeling factors that inuence slip and predicting
the dierent loosening processes. Primary factors that inuence slip at fastener contacts are discussed. The results show
that loosening can occur at relatively low shear loads due to the process of localized slip. # 2002 Elsevier Science Ltd.
All rights reserved.
Keywords: Threaded fasteners; Vibration induced loosening; Fastener failures; Joint failures
384
was attributed to reduction of circumferential holding friction as a result of slip at the fastener surfaces
caused by the applied shear load. There have been numerous attempts [713] to model the loosening phenomenon, however, all have had limited success in adequately predicting loosening. It was recently shown
[14] that a fastener could turn loose under dynamic shear loading as a result of accumulation of localized
slip in the form of strain at the fastener contacts surfaces. Adequate modeling of such loosening requires
inclusion of fastener geometry, stiness, as well as contact with friction. In this work a three-dimensional
nite element (FE) model that includes these features is used to study fastener loosening.
Most of the previous work relating to FE analysis [1520] of threaded fastener have been restricted to
axisymmetric models aimed primarily at stress analysis. Such models cannot be used to simulate loosening
because they do not include the helical thread geometry. Zadoks and Kokatam [21] have presented a threedimensional FE model of a screw that includes the thread helix. The screw in their model is not meshed as
a single continuous body; instead the screw body (a cylinder) and the threads (helix) are meshed separately
and then joined using xed contact elements. Although this approach makes the modeling process less
complex, it signicantly increases the solution computational cost due to the xed contacts elements used
to bond the threads to the body. In addition, for a given mesh density, the xed contact approach is less
accurate than using a continuous mesh for the entire screw. As a result of these considerations, the nite
element model in the present work utilizes a continuous mesh for the screw.
This work is aimed at improving the understanding of fastener loosening under dynamic shear loads as a
step towards development of quantitative guidelines for joint design based on fastener loosening. Previous
experimental work [14] has identied four dierent loosening processes that cause fasteners to turn loose
under shear loading. Details of the dierent loosening process are illustrated using results from the FE analysis. Factors that inuence slip, and the resulting loosening are discussed. The nite element model is used to
illustrate the eect of some signicant factors that inuence loosening.
385
Fig. 2. Typical nite element mesh: (a) joint model, (b) boundary conditions, and (c) contact regions on the screw.
consists of a screw, which fastens the top plate through a threaded insert. The geometry is simplied to
include only the essential features of the system. Since the base is assumed to be rigid, only a small region
around the threaded insert is modeled, and the nodes on the external surface of this region are constrained
(see Fig. 2b). Also, only a small region of the top plate around the screw is modeled and its end surfaces are
constrained to remain plane to model the behavior of a longer member (see Fig. 2b). Since the friction at the
interface between the top plate and the xed base is negligible due to the roller bearings, the bottom and side
contacts are simply modeled by nodal constraints at the bottom of the top plate in the z and y directions.
Fastener preload is simulated by an initial interference between the screw head and the upper surface of
the top plate. The transverse load in the x direction is applied to the clamped component through a spring
element (see Fig. 2a). The stiness of the spring element is based on the experimentally determined stiness
of the load transmitting members of the transverse test apparatus. Due to the presence of contact and
signicant rotation from the screw turn, the problem requires a non-linear solution that utilizes the New-
386
ton-Raphson method [22]. Loosening is modeled as a quasi-static process based on previous results that
show loosening in the transverse test apparatus to be independent of the loading frequency [6]. Each
loading cycle is divided into at least thirty two load increments with some cases utilizing higher number of
divisions to enhance convergence of the non-linear solution.
The nite element mesh used in this study utilizes a predominantly hexahedral mesh, which provides
good results within a reasonable runtime. A typical model with twelve screw threads and ve and a half
internal threads consist of 6126 nodes, and 4584 elements. The model mainly comprises of 8-node and 20node brick elements, with a small portion of the fastener meshed with 10 node tetrahedral elements. Higher
order elements (20-node brick) are used in the contact regions of the components (see Fig. 2c). These
contact regions are overlaid with high order general-purpose contact elements capable of modeling friction.
The model includes contact regions between the screw head and the clamped component (see Fig. 2c), the
clamped component hole and the screw body, and between the internal and external threads.
Preliminary studies were conducted to determine the eect of the mesh density on the loosening results.
It was found that a relatively coarse mesh (such as shown in Fig. 2) utilizing high order elements at the
contact regions provided good results in a reasonable amount of time. This is not surprising since loosening results are based on the displacements, which converge with a relatively coarse mesh compared to the
mesh required for accurate determination of stresses. Typical run time for a loading cycle is approximately
6 h on a 700 MHz Pentium III PC with 512MB RAM on the Windows NT platform.
3. Loosening processes
The earliest explanation for loosening under shear loading was provided by Junker [6] and recently
extended based on an experimental study by the authors [14]. Threaded fasteners have an inherent tendency to loosen due to the helical slope at the threads. Fig. 3 shows the thread reactions, RPn, n=1. . .4, to the
preload, FP, at four points around a thread. The loosening moment is developed from the circumferential
components RPLn,n=1. . .4. In absence of external loads, the fastener remains tight because of the circumferential friction that resists the loosening moment. However, as a result of an applied shear load, the
circumferential friction force reduces as the direction of contact forces changes from circumferential to the
direction of the applied shear force. Consequently the fastener turns loose as contacts at the head and
threads slip. For additional details of the loosening process, the reader is referred to Pai and Hess [14].
Slip at the contacts under the applied load can be classied as complete slip or localized slip. Complete slip
occurs when the entire contact surface (at the head or threads) slips, while localized slip occurs when only
parts of the contact regions slip. Complete slip requires that the loads acting on the fastener are suciently
large to overcome friction over the entire contact, while localized slip occurs only in parts of the contact
where the friction force has been overcome. Signicant loosening occurs only when the entire fastener
turns, which requires that the two fastener contact surfaces, i.e. at the head and threads, either undergo
complete slip, or localized slip that accumulates over loading cycles. Loosening processes can be divided
into four dierent types depending on the nature of slip at the head and thread contacts. These are characterized by (1) localized head slip with localized thread slip, (2) localized head slip with complete thread
slip, (3) complete head slip with localized thread slip, and (4) complete head slip with complete thread slip.
The inuence of the type of slip on loosening is illustrated in Fig. 4, which shows a typical preload versus
cycles plot of loosening obtained using the transverse vibration test apparatus. The loosening rate shows a
drastic increase as soon as head slip changes from localized to complete slip.
All the FE results presented in this paper model a 63.5 mm long, Grade 8, Class 2A/2B, 0.5-13 UNC
screw (12.7 mm diameter) screw subjected to 1 mm displacement through the spring element of stiness
1.2 kN/mm. The modulus of elasticity of the fastener materials used for the simulations is 200 GPa. The
coecient of friction is 0.261 for dry contact, 0.158 for contacts with machine oil lubrication, and 0.086 for
MoS2 grease lubricated contacts based on previously obtained experimental data [14].
387
Figs. 5 and 6 show loosening resulting from localized head slip and complete thread slip for a fastener
with head and thread contacts lubricated with machine oil and a preload of nearly 11 kN. This loosening
process is the same as the initial loosening process occurring in Fig. 4. Fig. 5a shows a hysteresis curve,
which is a plot of the shear force acting on the top plate versus its displacement. The slope of the hysteresis
388
Fig. 5. Loosening process characterized by localized head slip with complete thread slip: (a) hysteresis curve, and (b) contact status.
curve provides an indication of the joint stiness in the transverse direction. The reduction in slope of the
hysteresis curve is a sign of slip at the contacts. Fig. 5b shows the state of the fastener at three points during
the cycle. Black regions indicate contact regions that stick, while the gray regions indicate slip. Segments ab
and cd in Fig. 5a indicate parts of the cycle where the threads undergo complete slip while the head contact
undergoes localized slip (see Fig. 5bi and biii). Both thread and head undergo localized slip during parts of
the cycle indicated by the steeper segments bc and da (Fig. 5bii). Note that head contact regions that stick
during the rst half of cyclic loading, slip during the second half (see Fig. 5bi and biii). This enables the
entire head contact to slip over a complete cycle.
389
Fig. 6. Loosening process characterized by localized head slip with complete thread slip: (a) thread turn angle at four points: , at
0 ; , at 90 ; , at 180 ; . . ., at 270 ; , average, (b) head turn angle at four points: , at 0 ; , at 90 ; , at
180 ; . . ., at 270 ; , average, and (c) preload.
390
Fig. 6a and b shows the turning angle at four points around the thread and head respectively. The thread
turn angle shown is the average from all engaged threads. Thread loosening angles show a rapid increase
due to complete thread slip, while the head turning rate is lower due to localized slip. This results in the
screw body being under torsion. The plots indicate that points at 90 and 270 oscillate with the applied
load with a net loosening at the end of each half cycle as reected by the angles at 0 and 180 . Since the
screw body is in a state of torsion, the corresponding oscillating points at the head and thread move in
opposite directions. Note that the thread turn angle is far larger than the head angle, and that the preload
loss per cycle (Fig. 6c) is fairly low.
Figs. 7 and 8 illustrate loosening caused by complete head slip and complete thread slip for the thread
and head contacts lubricated with machine oil and at a preload of approximately 5.5 kN. The hysteresis
curve shown in Fig. 7a indicates regions with three distinct slopes. Region i reect parts of the loading
where complete thread slip occurs and the head undergoes localized slip (Fig. 7bi). At region ii the entire
head and thread slip as shown in Fig. 7bii. The initial stage of the unloading portion of the cycle is seen to
have a higher slope because of localized slip at both the head and the thread (see Fig. 7biii). Fig. 8a and b
shows the loosening at four points at the thread and head. The loosening angles at all four points are
nearly identical. The head loosening angle shows nearly a step increase at the point where complete slip
occurs, while the thread angle increases more gradually with a small step reecting the occurrence of
complete head slip. The loosening angles at the head and thread are comparable with the thread angle
leading the head angle. This is expected since the entire loosening moment is developed at the threads.
Also, the rate of preload drop (Fig. 8c) is signicantly larger than that found with localized head slip illustrated in Fig. 6c. This loosening process corresponds to the loosening occurring at a rapid rate in Fig. 4.
Hysteresis curves and loosening angles of the loosening process characterized by complete head slip and
localized thread slip are shown in Figs. 9 and 10 for a fastener with MoS2 grease at the head and dry
threads with a preload of about 11.9 kN. The hysteresis curve is characterized by two distinct slopes. The
stier segments indicate parts of loading where the head and thread undergo localized slip (see Fig. 9bi).
Regions ii and iii reect parts of the cycle where the entire head slips, while the threads continue to undergo
localized slip (see Fig. 9bii and biii). From Fig. 9bii and biii it is seen that dierent parts of the threads stick
during dierent parts of the cycle (note that orientation of the ii is opposite of that of iii). The thread angles
at four points (Fig. 10a) show that points at 90 and 270 oscillate with the applied input with a net increase
in the loosening angle per half a loading cycle. The head angles (Fig. 10b) display a step increase at the
point where the complete head slips, with the points at 90 and 270 reecting the oscillations at the threads.
As in the earlier cases, the thread angle leads the head angle. The loosening rate (Fig. 10c) is not as severe
as in the case of complete thread and head slip (Fig. 8c).
Figs. 11 and 12 illustrate the loosening process resulting from localized head and localized thread slip
with dry threads and MoS2 grease at the head and at preload of nearly13.2 kN. The hysteresis curve shows
nearly the same stiness at all regions with a very slight reduction in the slope as the thread slip region
increases. The regions of localized slip (Fig. 11b) are seen to vary at dierent parts of the cycle, however the
lower thread contact stick throughout the entire cycle. Fig. 12a and 12b show the loosening angles at the
thread and head, which essentially display an oscillatory character at the 90 and 270 points with a slight
increase in the loosening angle per cycle. As with the other loosening processes, the thread loosening angle
leads the head angle. The rate of preload loss (Fig. 12c) is seen to be quite small. Although the lower
threads stick throughout the entire cycle in the results presented here, the preload loss resulting due to the
localized slip at the rst few threads can accumulate and cause progressively larger slip and eventually lead
to complete thread slip.
The above FE results illustrate the four possible loosening processes that can occur in a threaded fastener
subjected to dynamic shear loads. Of these four processes, only three are widespread in practice. In the
typical case of fastener loosening, the initial stage of loosening is characterized by localized slip at both the
head and the threads. This progresses to loosening by localized head slip with complete thread slip, and
391
Fig. 7. Loosening process characterized by complete head slip with complete thread slip: (a) hysteresis curve, and (b) contact status.
eventually to complete slip at the head and the threads. In some cases, the initial loosening starts with
localized head slip and complete thread slip, and progresses to complete slip at the head and the threads.
The loosening rate at the initial stages characterized by localized slip is fairly low and increases as the
loosening transitions from localized to complete slip (see Fig. 4). A contribution of the present work is the
identication of loosening caused by localized slip. Loosening at the initial stages resulting from localized
slip is critical since it can occur at a signicantly lower shear load than that required for complete slip. For
392
Fig. 8. Loosening process characterized by complete head slip with complete thread slip: (a) thread turn angle at four points: , at
0 ; , at 90 ; , at 180 ; . . ., at 270 ; , average, (b) head turn angle at four points: , at 0 ; , at 90 ; , at
180 ; . . ., at 270 ; , average, and (c) preload.
393
Fig. 9. Loosening process characterized by complete head slip with localized thread slip: (a) hysteresis curve, and (b) contact status.
example, in the data shown earlier, loosening by localized slip (Fig. 5a) occurs when the magnitude of the
shear force acting on the joint is approximately 9% of the preload, while loosening by complete slip
(Fig. 7a) occurs when the shear force is 16% of the preload. In this case, loosening by localized slip occurs
at about half of the load required to cause complete slip. Loosening by localized slip is therefore the critical
loosening process from a perspective of joint design. Loosening failure in a joint can be avoided by
ensuring that the dynamic loads acting on the fastener are lower than the loads required to cause loosening
by localized slip.
394
Fig. 10. Loosening process characterized by complete head slip with localized thread slip: (a) thread turn angle at four points: , at
0 ; , at 90 ; , at 180 ; . . ., at 270 ; , average, (b) head turn angle at four points: , at 0 ; , at 90 ; , at
180 ; . . ., at 270 ; , average, and (c) preload.
395
Fig. 11. Loosening process characterized by localized head slip with localized thread slip: (a) hysteresis curve, and (b) contact status.
396
Fig. 12. Loosening process characterized by localized head slip with localized thread slip: (a) thread turn angle at four points: , at
0 ; , at 90 ; , at 180 ; . . ., at 270 ; , average, (b) head turn angle at four points: , at 0 ; , at 90 ; , at
180 ; . . ., at 270 ; , average, and (c) preload.
397
Fig. 13. Comparison of experimental hysteresis curves and FE results: (a) , experiment 11 kN preload with oil lubrication; , FE
11 kN; , experiment 5.5 kN preload with oil lubrication; &, FE 5.5 kN preload. (b) , experiment 11.9 kN preload with dry
threads and MoS2 grease at head; , FE, and (c) , experiment 13.2 kN preload with dry threads and MoS2 grease at head; , FE.
(63.5 mm long, Grade 8, 0.5-UNC 13 screw, 1 mm eccentric setting).
398
capture the trends, however the net slip occurring during complete head slip is smaller in the FE results than in
the experiments. The amount of slip reected in the hysteresis curve is a function of the fastener dimensions, and
is largely inuenced by the clearance between the internal and external threads. The dierence in results is most
likely because the thread dimensions utilized for the FE simulations were dened in the middle of the allowable
range for Class 2A/2B threads, and the thread dimensions in the test specimens may be slightly dierent.
The hysteresis curves shown in Fig. 13b show loosening by complete head slip and localized thread slip
for a screw with MoS2 grease at head and dry threads at 11.9 kN preload. The FE result is seen to capture
the experimental data quite well, and as with the previous case the slip is slightly smaller. A similar trend is
observed for loosening caused by localized thread and head slip shown in Fig. 13c for the same lubrication
condition at 13.2 kN preload.
In all four cases, especially in Fig. 13b and c, the FE results are seen to be slightly stier than the
experimental data. This is because the FE model includes six threads before the rst engaged thread instead
of 12 threads found in the test specimens. The additional threads in the screws used in the experiments contribute to the lower bending stiness observed in the experimental data. A FE model with 12 exposed threads
could not be used because of the signicant additional computational cost. The inuence of this omission is
clearly not very signicant as indicated by the reasonably good comparisons between the results.
5. Fastener slip
The four dierent loosening processes described earlier are a result of the slip occurring at the fastener
contact surfaces. Slip at the fastener surfaces is a function of the distribution of loads acting tangential to
the contact surface (i.e. the forces that cause slip), and the contact normal force distribution (since it
inuences the friction). Fig. 14 summarizes the main factors that inuence slip, S, and normal reactions,
R, in a joint subjected to shear load [14]. The fastener has an initial loosening moment L due to the thread
reaction to preload (see Fig. 3), which is balanced by the friction moments at the threads and the head.
During the loading, the clamped component is subjected to a shear load FS, which is transferred to the screw
through the head friction. The shear force can also be transferred to the screw through side contact between
the clamped component and the screw body, which is caused by head slip or fastener bending. The component of the shear force acting tangential to the thread ank contributes to thread slip, while the remaining
part changes the thread normal reaction distribution. The bending moment developed from the shear force
changes the head and thread reaction distributions. The bending moment also contributes to slip along the
thread ank due to the turning action about the moment axis. As a result of the shear load and the bending
moment, the fastener undergoes elastic deformation at the head and the threads, which also contributes to
slip. The resulting slip from all of the above factors is also inuenced by the state of side contact at the
head and the thread (not shown).
From the above description it can be seen that several factors inuence the loosening process. The main
factors have been separated for the purpose of illustration. However, all these factors are coupled non-linearly and together determine the nal state and process of loosening. Some important aspects of these factors
that inuence loosening are discussed below.
6. Fastener bending
It is seen in Fig. 14 that the bending moment contributes to slip and loosening in several ways. The
bending moment contributes to localized slip by changing the reaction distribution which causes slip in
regions with lower reaction force (for example see slip regions at the head in Figs. 5b and 11b). The
bending moment also causes slip at the thread anks, and contributes to slip by elastic deformation.
399
Fig. 14. Summary of factors inuencing slip and reaction force distribution in a joint subjected to shear load. (FSApplied shear force,
Lloosening moment, Stangential slip force, Rnormal reaction).
The inuence of bending moment on slip is illustrated in Fig. 15. It shows the shear force required to
cause complete slip at the threads for a 0.513 UNC screw of dierent lengths at a preload of 13.3 kN. The
shear force required to cause complete thread slip in a 63.5 mm screw is nearly half that for a 31.8 mm
screw. This indicates that the additional bending moment developed in the longer screw is a signicant
factor responsible for thread slip. The bending moment also contributes to head slip by elastic deformation
of the head. This is because the bending moment causes a change in the head normal reaction distribution,
which changes the head surfaces deformation and contributes to slip as shown by the illustration at the
bottom right corner in Fig. 14.
The above data indicate that long fasteners require lower shear force to loosen. However, this is applicable only to force loaded systems, such as fasteners used to mount a pillow block bearing subjected to
dynamic loads from an unbalanced rotating mass. In the case of displacement-loaded systems such as the
transverse vibration test apparatus (Fig. 1) longer fasteners can reduce loosening. This is because the shear
force acting at the contacts in such systems is a function of the applied displacement and the fastener
stiness. Since longer fasteners have a lower bending stiness, they are subjected to lower shear force, and
therefore are less vulnerable to loosening. Experience with displacement loaded systems support the notion
that longer bolts inhibit vibration induced loosening [2,23]. However, the opposite is true (i.e. longer bolts
are more prone to loosening) for force-loaded systems, and this is not widely appreciated.
400
Fig. 15. Shear force required to cause complete thread slip for dierent screw length. (0.5-13 UNC screw with dry contacts).
7. Other factors
Another aspect of the loosening process is the inuence of the thread load distribution on thread slip. In
the results presented earlier, it is found that the rst few threads slip before the last ones (see Fig. 11b for
example). This is mainly because of the normal load distribution at the threads as shown in Fig. 16. The
data show that the rst three threads in a FE model with ve and a half engaged threads carry about 73%
of the preload. This means that the rst three threads have signicantly higher frictional resistance, and
only once these are overcome can the lower threads begin to slip. This is reected in Fig. 17, which shows the
turn angle for nodes at the 90 location on the rst ve threads for the loosening process caused by localized
head and thread slip (Figs. 11 and 12). The magnitude of the turn angle is largest at the rst thread and reduces
in subsequent threads. The turn angles of the rst few threads reect slip as well as elastic deformation, while
those at the last few threads are mainly a result of elastic deformation. These results indicate that the load
distribution at the threads inuence the loosening process.
Fig. 16. FE results of load distribution at the engaged threads for 0.5-13 UNC screw.
401
Fig. 17. Thread turn angle at 90 node for loosening process characterized by localized head slip with localized thread slip: , rst
engaged thread; , second thread; , third thread; . . , fourth thread; , fth thread.
In all the cases presented in this work, the fastener preload has been fairly low so has to minimize the
inuence of localized thread yielding, which is likely to inuence the thread load distribution. Inclusion of
thread yielding in the FE model requires a denser mesh to ensure accurate determination of stresses. This
was not attempted at this stage of the study due to the extremely high computational cost. The eect of
thread yielding on loosening will be addressed in future work.
In addition to fastener length, and thread load distribution, there are various other parameters that inuence
loosening, including the fastener material, and dimensional tolerances. Since all these factors are coupled nonlinearly, it is important to study their individual eect as well as the inuence of their interaction on screw
loosening. Results from a parameter study investigating these eects will be reported in a future paper.
8. Conclusion
The nite element model presented is capable of predicting the four dierent loosening processes
observed experimentally. The FE results capture the essential features displayed by the experimental data.
Several factors inuence loosening at dierent stages of loading, and the nal outcome is a result of their
non-linear interactions. The FE model includes the primary factors that cause loosening, and provides a
powerful tool for evaluation of the details of fastener loosening. The loosening process caused by localized
slip can occur at signicantly lower shear force than loosening caused by complete slip, and therefore is
critical in joint design.
Acknowledgements
The authors gratefully acknowledge the support from the National Science Foundation under Grant No.
CMS-9629217.
References
[1] Hess DP. Vibration- and shock- induced loosening. In: Bickford JH, Nasser S, editors. Handbook of bolts and bolted joints. New
York: Marcel Dekker, 1998. p. 757824.
402
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
[22]
[23]