Bolt loosening at rotating joint interface and its influence on rotor dynamics

Zhaoye Qin, Qinkai Han, Fulei Chu

PII: S1350-6307(15)30133-3
DOI: doi: 10.1016/j.engfailanal.2015.11.002
Reference: EFA 2727

To appear in:

Received date: 25 June 2015

Revised date: 1 October 2015
Accepted date: 2 November 2015

Please cite this article as: Qin Zhaoye, Han Qinkai, Chu Fulei, Bolt loosening
at rotating joint interface and its influence on rotor dynamics, (2015), doi:
10.1016/j.engfailanal.2015.11.002

This is a PDF file of an unedited manuscript that has been accepted for publication.
As a service to our customers we are providing this early version of the manuscript.
The manuscript will undergo copyediting, typesetting, and review of the resulting proof
before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that
apply to the journal pertain.
 ACCEPTED MANUSCRIPT

Bolt Loosening at Rotating Joint Interface and Its Influence on Rotor Dynamics

Zhaoye Qin*, Qinkai Han, Fulei Chu

State Key Laboratory of Tribology, Department of Mechanical Engineering, Tsinghua University, P.R. China

T
Abstract. The rotors of large rotating machinery involve multiple stages of disks supported by drums

P
and shafts, where bolted joints are commonly employed to connect the adjacent disks and drums.
Those connecting bolts, subjected to numerous combinations of loads during normal operation, tend

RI
to get loosening, which will affect the rotor dynamics and even results in structural failures. However,

SC
little research has been done on the bolt loosening at the rotating joint interfaces of rotor systems.
Thus, the influence of the bolt loosening on the rotor dynamics is studied in this paper. First, the time-
varying stiffness at the joint interface with bolt loosening is investigated by means of three-

NU
dimensional (3D) nonlinear finite element (FE) models. Then, the motion equations for the rotor with
bolt loosening are deduced accounting for the local stiffness variation caused by the bolt loosening.
By taking a simple drum rotor with bolted joints as an example, the time-varying joint stiffness
MA
resulting from the bolt loosening and its influence on steady-state response of the rotor are calculated.
The studies in this paper provide the fundamental understanding about the influence of the bolt
loosening at the rotating joint interface on the rotor dynamics, and are helpful for the bolt loosening
ED

detection of rotating components in heavy-duty rotating machinery.

PT

1 Introduction
In large rotating machines, bolted joints are widely used to fasten the rotating components
CE

together at the mating surfaces. Fig. 1 shows a typical disk-drum joint interface in an aero-engine
rotor, where the adjacent disk and drums are connected by bolts distributed in the circumference of
joint interface. Those joint interfaces rotate along with the shafts during normal operation of the
AC

rotating machines. Though high levels of preload are applied to the bolts to prevent joint separation
and provide sufficient strength to maintain the integrity of the rotor systems, bolt loosening may
occur due to severe loading conditions. The bolt loosening will change the local stiffness at the joint
interfaces, which in turn affect the rotor dynamics and even result in structural failures. Investigation
of the bolt loosening and its influence on the rotor dynamics will assist in the early detection of the
bolt loosening before damage occurs. However, this issue has been generally disregarded up to now.
[Fig. 1]
Bolt loosening is the major failure form for bolted joints. Much research has been performed to
investigate the loosening mechanism and its influence on the joint stiffness and structural dynamics
[1-5]. However, those works focused on the bolted joints in the stationary, namely non-rotating,
structures. The studies of bolt loosening in rotor systems dealt with looseness of the pedestal, which

* Corresponding author. Tel.: +86 10 6278 8308; fax: +86 10 6278 8308.
E-mail address: qinzy@mail.tsinghua.edu.cn (Z.Y. Qin).

1
 ACCEPTED MANUSCRIPT

were also non-rotating [6-8]. To the authors’ best knowledge, the bolt loosening of the rotating joint
interfaces has not been yet studied.
Although no published literature has been found on the looseness of bolts connecting rotating
components, a so called bolt-removal method was utilized to investigate the dynamics of the cracked
rotors experimentally, where parts of the connecting bolts was removed from the joint interface to

T
simulate the breathing mechanism of a transverse crack [9-11]. Keiner, et al. compared the

P
experimental results using the bolt-removal method with the response of a real crack rotor, where

RI
good agreement was achieved [11]. The bolt removal can be treated as the limiting case of the bolt
loosening, where those bolts are loosened completely providing no constraint to the mating surfaces.

SC
Thus, the consistency of the experimental results indicates the similarity of the bolt loosening
signature to the breathing mechanism of the transverse crack to some extent. However, since the
purpose of those studies was to simulate the breathing mechanism of the transvers crack

NU
experimentally through the bolt-removal method, no theoretical studies were carried out, and the
experiments were limited to the complete loosening case without considering relatively lower degrees
MA
of loosening.
The aim of this work is to provide fundamental understanding of the dynamic characteristics of
rotors with bolt loosening at the rotating joint interface. Taking a simple drum rotor as an example,
quasi-static numerical simulations are performed on a 3D FE model accounting for the nonlinearity
ED

caused by the mating surface contact to investigate the variation of the local deformation and stiffness
in a full revolution of the rotor primarily. Three degrees of loosening, together with different numbers
PT

of the loosened bolts, are discussed. Then, the time-varying joint stiffness is incorporated into the
rotor dynamic model. The steady-state response of the rotor is calculated using harmonic balance
CE

method (HBM) to evaluate the influence of bolt loosening on the rotor dynamics.

2 Local stiffness at joint interface with bolt loosening

AC

2.1 Quasi-static finite element analysis

A typical disk-drum joint interface is shown in Fig. 2, where two drums with inner flanges are
fastened to one disk by a same set of connecting bolts distributed in the circumference of the joint
interface. A stationary local coordinate system is used for calculating the deformation at the joint
interface as shown in Fig. 2. In industrial applications, the spigot positioning is commonly adopted at
the disk-drum joint interface to resist the relative lateral slippage between the contact surfaces of the
disk and drum as shown in the zoom plot in Fig. 2. Owing to the constraint of the mating surfaces, the
joint interface can be assumed to be rigid in the lateral direction, experiencing only bending
deformations under external loads.

[Fig. 2]
To evaluate the local stiffness caused by the bolt loosening, a 3D FE model shown in Fig. 3 is
developed on the platform of ANSYS, on which nonlinear quasi-static simulations are performed. In
the FE model, the disk, drums and connecting bolts are all modelled using brick elements SOLID95,

2
 ACCEPTED MANUSCRIPT

and the frictional contact is simulated between the contact surfaces of the joint components using
contact elements CONTA173 and target segment elementsTARGE170, where a uniform friction
coefficient of 0.1 is assigned. To simulate the constraint of the mating surfaces in the lateral direction,
the radial degree of freedom (DOF) of the corresponding nodes of the disk and drums at the outer
edge of the contact areas are coupled together. The boundary conditions of the model are defined as

T
follows: all DOFs of the nodes at the left edge surface of the left-side drum are constrained; all DOFs

P
of the nodes at the right edge surface of the right-side drum are coupled together to form a rigid

RI
region with an auxiliary node defined at the region centre as the master node shown in Fig.3. The
auxiliary node has six DOFs, on which external loads are applied to realize loading of the joint

SC
interface during simulations. The displacements of the auxiliary node will be used to calculate the
local stiffness of the joint interface.
[Fig. 3]

NU
This work focuses on the influence of bolt loosening on the rotor dynamics, rather than the
mechanisms of bolt self-loosening. Therefore, with no concern of the loosening process, the
MA
loosening phenomenon is simulated by reducing the preload level of those loosened bolts in the FE
model. As the limiting case, the loosened bolts generate no constraint to the joint interface, and those
bolts are removed from the FE model.
According to the discussions of the bolt-removal method in section 1, it can be predicted that the
ED

local stiffness at the joint interface with bolt loosening varies periodically with respect to the rotation
angle when the static deflection of the rotor is dominant over the vibration, just like the breathing
PT

mechanism of the transverse crack. It is worth noting that both static and dynamic loads act on the
industrial rotors, and the onset of the periodically time-varying behaviour indicates a low level of the
dynamic loads, which have little contribution to the time-varying stiffness. Hence, those dynamic
CE

loads are neglected here, and a series of static simulations are performed to achieve the time-varying
stiffness. The static deflections of the joint interface are calculated as the FE model rotates step by
AC

step in different angular position, while the directions and magnitudes of the external loads remain
unchanged. The initial position of the joint interface is illustrated in Fig. 2, where the bolt loosening
region is on the top side of the joint interface.
Multiple load steps are defined for each angular position as follows: preloads are applied to the
bolts using the virtual thermal deformation method in the 1st load step, where a relatively low preload
level is assigned to the assumed loosened bolts; in the 2nd load step, a bending moment Mg about the
horizontal axis x is applied at the auxiliary node to simulate the gravity effect; finally, small bending
moments Mx and My about the two lateral axes, that is, x and y, are applied, respectively, in the
following load step. The above loading cycle is repeated for different angular positions until a full
revolution of the joint interface is achieved. The bending moments Mx and My applied in the
simulations are the resultants of the static loads mentioned above, which could be the gravity force,
the temperature-gradient-induced thermal force, and the bearing reaction force. In order to achieve
the time-varying stiffness of the joint interface, the values of Mx and My applied in the simulations

3
 ACCEPTED MANUSCRIPT

should be small enough to avoid separations of the joint components, which will lead to nonlinearity
at the joint interface.
Subjected to the bending moments Mx and My, the rotation angles x and  y with respect to the
J J

x and y axes at the load acting point, that is, the auxiliary node, are calculated. Since the bending
moment Mg applied in the 2nd load step is used only to account for the gravity effect, the rotation

T
angles x and  y caused by Mg should be subtracted from the total ones x and  y . It should also

P
g g J J

RI
be noted that the rotation angles at the auxiliary node are the combination of those of the joint
interface and the elastic deformation of the disk and drums. To obtain the deformation caused by the

SC
joint interface solely, the elastic deformation should also be subtracted from x and  y . An FE
J J

model with the disk and drums fixed together is developed, on which the rotation angles x and  y
F F

NU
due to the elastic deformation are calculated subjected to the bending moments of Mx and My,
respectively. Therefore, the rotation angles of the joint interface caused by the bolt loosening can be
MA
expressed as
x  xJ  xg  xF
 (1)
 y   y   y   y
J g F
ED

2.2 Local stiffness calculation

The rotation angles of the joint interface caused by the bolt loosening given by Eq. (1) indicate
the presence of the local flexibility at the joint interface. As stated in section 2.1, it is assumed that
PT

the bolt loosening at the joint interface provides no additional shear flexibility due to the constraints
at the mating surfaces. Therefore, a 2  2 local flexibility matrix of the joint interface with bolt
CE

loosening is considered here

 f xx f xy 
f  (2)
f yy 
AC

 f yx
where, the coefficient fij  i, j  x, y  denotes the rotation angles with respect to the i axis arisen by the
unit load with respect to the j axis, of which the value can be determined from the quasi-static
simulations in section 2.1 as
i
fij  (3)
Mj
The presence of the off-diagonal coefficients in the flexibility matrix are caused by the asymmetric
stiffness of the joint interface due to the bolt loosening, which indicates the coupling of the rotor
vibrations in the lateral directions. The local stiffness matrix at the joint interface can be obtained as
k  f 1 (4)
nd
It is worth noting that the bending moment Mg applied in the 2 load step is crucial for
calculating the local stiffness of the joint interface but tends to be disregarded. Without applying Mg,
that is, taking no account of the gravity effect, the values of fxx and fyy in the flexibility matrix would

4
 ACCEPTED MANUSCRIPT

be equal to each other with only a phase shift as they vary along with the revolution angle [12].
Actually, due to the gravity effect, the value and variation trend of fyy differ from fxx obviously as will
be shown in section 4. It should also be noted that the joint interface involves three joint members,
that is, the left- and right-side drums, and the middle disk, at the joint interface, which form two sets
of connection in both sides of the middle disk. The two sets of the connection, fastened by the same

T
group of the connecting bolts, are in series, and the coefficients in the stiffness matrix k are the

P
resultant stiffness of the joint interface.

RI
3 Dynamic model of rotor with bolt loosening

SC
In order to evaluate the influence of bolt loosening on the rotor dynamics, the local stiffness
matrix is incorporated into the rotor dynamic model in this section. A simple drum rotor consisting of
disks, drums and shafts supported by a rolling element bearing at each end as shown in Fig. 4 is

NU
studied here. In the rotor, the connections between the components are assumed to be rigid, except the
bolted joints at the disk-drum joint interface.
MA
[Fig. 4]
The FE model for the rotor involving N elements is shown in Fig. 5. To develop the rotor
dynamic model, a right-handed stationary coordinate system with its origin located at the left end of
the rotor is defined. In this model, the connecting bolts are represented by the local time-varying
ED

stiffness achieved from the static simulations in section 2. The disks are treated as lumped masses
with corresponding moments of inertia, the shafts are modelled using Timoshenko beam elements, of
PT

which the mass, stiffness and gyroscopic matrices are given in [13]. The drums are simplified as
hollow beams and also modelled using Timoshenko beam elements. The beam element has four
DOFs at each node as illustrated in Fig. 6, of which the displacement vector is
CE

δe  u1 w1 1 1 u2 w2 2 2 
T
(5)
[Fig. 5]
AC

[Fig. 6]
It should be noted that the drums are thin-walled cylindrical shells and vibrate in the beam mode,
that is, the circumferential wave order n=1, for low frequency vibrations of the rotor system. In this
case, reasonable results can be achieved by approximating the drum with a Timoshenko beam, which
allows for the effects of transverse shear deformation and rotary inertia. [14, 15]
As shown in Fig. 5, the joint interface separates the rotor into three parts, that is, the left and
right sub rotors and the middle disk, which are modelled separately and combined together by the
reactions transmitted through the joint interface. Detailed derivation of the rotor model with the joint
interface can be found in Ref. [16]. Here, the FE model for the rotor system subjected to the gravity
and unbalance forces are given directly as
Mδ   C  G  δ   K  K be  δ  Fg  F (6)
where, M, C, G and K are the mass, damping, gyroscopic and stiffness matrices of the rotor
neglecting the joint interface respectively, Kbe is the additional stiffness matrix due to the bolt

5
 ACCEPTED MANUSCRIPT

loosening at the joint interface, Fg is the vector of the gravity load, F is the vector of the mass
unbalance force. As stated in section 2, the small dynamic loads acting on the rotor, which have little
contribution to the time-varying stiffness, are neglected in the static simulations. However, as the
most common source of the dynamic loads, the mass unbalance caused by the manufacturing
imperfection of the rotating components is unavoidable and affects the rotor dynamics. Hence, a

T
small mass unbalance force is involved in the motion equations of the rotor system.

P
The expression of the additional stiffness matrix Kbe is

RI
1 i(u , w) i( L ,L ) i( M ,M ) i( R , R ) N 1

SC
  1
 
 
  i(u , w)

NU
 i (7)
kL k L
K be    ( L ,L )
 k L kL  kR k R  i( M ,M )
 
 k R  i( R ,R )
MA
kR
 
 
  N  1
where k L  k R  2k are the joint stiffness at the left and right sides of the middle disk, respectively.
ED

Rayleigh damping is adopted in the rotor model

PT

C  M   K (8)

where  and  are the Rayleigh damping parameters. The values of those two parameters are
CE

calculated by assuming that damping ratios of the 1st and 2nd modes of the rotor system at rest are
AC

equal to 0.001.

Assuming that a mass unbalance mi is attached to the disk located at node i, the unbalance force

vector at this node is given by

 mi 2 cos  t    
 i 2 
m  sin  t    
Fi   (9)
 0 
 
 0 

where γ is the angle of the mass unbalance with respect to the x axis at t = 0. Eq. (9) can be rewritten

in the complex form

Fi  Fi e jt  c.c. (10)

6
 ACCEPTED MANUSCRIPT

where, c.c. denotes the complex conjugate of the preceding term, Fi  Fxi 2  j Fyi 2 , in which

Fxi  mi2 cos  sin  0 0 and Fxi  mi2   sin  cos  0 0 are the vectors of the
T T

unbalance force amplitude at node i in the x and y directions. Then, the total unbalance force vectors

T
of the rotor system are

P
RI
F  Fe jt  c.c. (11)

SC

where F   F1   F2   Fi   FN 1   . For the nodes without unbalance mass, the
   

NU
corresponding force amplitude vectors in F are zero.

4 Steady-state Response of Rotor with Bolt Loosening

MA
Direct numerical integration can be employed to calculate the steady-state response of the rotor
system given by Eq (6). However, due to the time-varying property, the stiffness matrix in the motion
equations needs to be modified in each iteration, which is time consuming. HBM, which is a more
ED

efficient approach for solving the time-varying problems [17, 18], is used to calculate the steady-state
T
response of the rotor system here. By denoting x  δ δ  as the state vector, Eq. (6) can be
PT

rewritten in the state form as

x  Ax  Fs  Fu e jt  c.c. (12)
CE

1 1 1
 0 M  K  K be C  G   0 M  Fg   0 M  FE 
where, A       , Fs      , Fu     .
M 0   0 M  M 0   0  M 0   0 
AC

As mentioned in section 2, the local stiffness matrix k varies periodically along with the shaft
revolution. Therefore, the coefficient matrix A is periodical and can be expressed as a Fourier series

A(t )   Ae
k 
k
jk t
(13)

where, A-k is the complex conjugate of Ak. Then, the rotor response can also be represented as

x(t )  xe
s 
s
jst
(14)

Substituting Eqs. (13) and (14) into Eq. (12) and balancing the harmonic terms yields the following
infinite-dimensional linear algebraic equations

7
 ACCEPTED MANUSCRIPT

    
 Γ3 A 1 A 2 A 3 A 4 A 5 A 6  x   0 
   3   
 A1 Γ2 A 1 A 2 A 3 A 4 A 5   x 2   0 
    
 A2 A1 Γ1 A 1 A 2 A 3 A 4   x 1   Fu _ 
 A3 A2 A1 Γ0 A 1 A 2 A 3   x0   -  Fs  (15)

T
    
 A4 A3 A2 A1 Γ 1 A 1 A 2   x1   Fu 

P
 A5 A4 A3 A2 A1 Γ 2 A 1 x   0 
  2   

RI
 A6 A5 A4 A3 A2 A1 Γ 3   x3   0 
    
    

SC
where, Γs  A0  isI  s  ,  3,  2, 1, 0, 1, 2, 3,  , I is the unit matrix of

8  N  2   8  N  2  . Eq. (15) is solve for the Fourier coefficient matrices x s . Then, substituting

NU
those coefficient matrices into Eq. (14), the rotor response x  t  can be obtained.
The equations given above will be truncated to finite-dimensional ones in order to obtain an
MA
approximate numerical solution for the steady-state response of the rotor system.

4 Simulations and Discussion

ED

4.1 Local stiffness of joint interface

The physical parameter values of the drum rotor are listed in Table 1. According to the previously
described procedure, a 3D FE model for its joint interface is developed as shown in Fig.3, where a
PT

preload of 0.5 kN is applied to the normal bolts. As for those loosened bolts, three different loosening
degrees, that is, the loosened bolts carrying half of the normal preload, carrying ten per cent of the
CE

normal preload, and being completely removed from the FE model, are investigated here.
[Table 1]
Figs.7-10 shows the local flexibilities at the joint interface invovling 1, 3, 5, and 10 loosened bolts,
AC

respectively. Due to the bending moment Mg applied in the 2nd load step simulating the gravity effect,
the upper portion of the joint interface is always in the compression state, and the lower one is
subjected to tension. Then, as the bending moment Mx is applied, the region involving the loosened
bolts is in compression at the initial stage. In this case, the bolt loosening has little influence on the
local flexibility, which is equal to the one without loosening. As the shaft rotates, the bolt loosening
region moves into the tensile stress region, where the disk and drums tend to separate and the local
flexibility with respect to the horizontal axis increases. The flexibility will reach its maximum as the
loosening region reaches the bottom of the interface. Further rotation decreases the flexibility again.
Thus, along with the shaft revolution, the local flexibility fxx varies periodically. Similarly, when the
bending moment My acts, a time-varying flexibility fyy is observed, of which the maximum value has
a phase lag of 90° approximately from that of fxx. In addition to the phase lag, the variation trends of
fxx and fyy show obvious divergence due to the gravity effect. If the gravity load is neglected, the two
curves of fxx and fyy will turn to be identical with only a phase lag of 90° left. The bolt loosening also
leads to time-varying cross-coupling flexibilities fxy and fyx, of which the values are equal to each

8
 ACCEPTED MANUSCRIPT

other. Compared with fxx and fyy, the cross-coupling flexibilities fxy and fyx are small. Along with the
increase of the loosening degree and loosened bolt number, the amplitudes of fxy and fyx are enlarged,
which indicates that the coupling of the lateral vibrations becomes significant.
[Fig. 7]
[Fig. 8]

T
Beyond the common signatures mentioned above, the local flexibility curves vary with the

P
loosening degree and loosened bolt number. For the limiting case where the loosened bolts are
removed, no constraint is applied on the bolt loosening region, and the variation of the local

RI
flexibility coefficients quite resembles that of the breathing transverse crack [11]. As the number of

SC
the loosened bolt increases, the amplitudes of the flexibility curves increase significantly, whereas,
the curve shape changes little as shown in Figs. 7-10 (c). For lower loosening degrees, the fluctuation
of the flexibility coefficients is also amplified as the loosening degree and loosened bolt number

NU
increase. At the meantime, the constraint of the loosened bolts on the loosening region in the tensile
state varies. Consequently, the shape of those flexibility curves shows obvious divergence for
different loosening cases.
MA
[Fig. 9]
[Fig. 10]
4.2 Steady-state response of drum rotor
ED

In this section, the influence of the bolt loosening on the rotor dynamics is evaluated. Without
loss of generality, two loosening cases, that is, half preload left for 3 loosened bolts and complete
loosening of 5 bolts are considered here. The coefficients of the joint stiffness matrices for the two
PT

cases are expressed using the truncated Fourier expansions with the first 4 harmonic components, as
shown in Fig 11, where it can be seen that the truncated Fourier expansion agrees with the FE
CE

simulation results well.

[Fig. 11]
A small mass unbalance of m  106 kg m is attached to the middle disk with a zero phase angle
AC

at t = 0. Subjected to both the gravity load and the unbalance force, the steady-state response of the
rotor at the middle disk with bolt loosening is illustrated in Fig. 12. In the frequency-response curves,
peaks at the critical backward whirling speeds appear. This is caused by the asymmetric property of
the rotor stiffness due to the bolt loosening and the gravity force. In addition, more peaks are
observed at the subcritical speeds resulting from the periodically varying stiffness of the loosened
joint interface. This phenomenon is consistent with the response of the rotor containing a breathing
transverse crack [18, 19], and can be adopted to detect the occurrence of the bolt loosening. By
comparing the two frequency-response curves in Fig. 12, it is revealed that the variations of the
loosening degree and loosened bolt number have little influence on the critical and subcritical speeds.
For the case of lower loosening degree, the resonant amplitudes are small, and some peaks at the
subcritical speeds even disappear as shown in Fig. 12 (a). As the loosening gets severe, the peaks at
the critical and subcritical speeds increase significantly, which is caused by the increment of the
stiffness fluctuation.

9
 ACCEPTED MANUSCRIPT

[Fig. 12]
Figs. 13 and 14 show the whirling orbits of the middle disk in the vicinity of the 1st critical and
subcritical speeds. Multiple loops can be observed in the whirling orbits around the subcritical speeds,
which indicates the presence of the harmonics caused by the bolt loosening. Hence, the multiple loops
in the whirling orbits can also be considered as a feature for the detection of the bolt loosening of the

T
rotating joint interface.

P
[Fig. 13]
[Fig. 14]

RI
5 Conclusions

SC
In this paper, the bolt loosening at the rotating joint interface and its influence on the rotor
dynamics was investigated. Nonlinear FE simulations were performed to calculate the time-varying

NU
stiffness at the joint interface with bolt loosening. The joint stiffness was then incorporated into the
motion equations of the rotor system, on which the steady-state response of the rotor system with bolt
loosening at the rotating joint interface was calculated using HBM. Taking a simple drum rotor as an
MA
example, the rotor behaviour with different loosening degrees and loosened bolt numbers was
investigated.
The nonlinear FE analyses revealed that the local flexibility of the joint interface with bolt
ED

loosening varied periodically in a full revolution of the rotor. Along with the increase of the loosening
degree and number of the loosened bolts, the flexibility fluctuation became significant. For the
limiting case of the bolt loosening, that is, those bolts were loosened completely providing no
PT

constraint to the joint interface, the local flexibility at the joint interface with bolt loosening was quite
similar with that of the transverse crack in the shaft.
CE

An analysis of the steady-state response of the rotor system with bolt loosening showed that
resonant peaks took place at both the forward and backward whirling critical and subcritical speeds
AC

due to the time-varying property of the joint interface with bolt loosening. As the bolt loosening
degree got severe, the peaks at the critical and subcritical speeds were amplified, which indicated the
significance of the stiffness fluctuation.
The studies in this paper help understanding about the influence of the bolt loosening at the
rotating joint interface on the rotor dynamics, and those signatures of the rotor response can be
adopted to detect the bolt loosening in rotating machines.

Acknowledgements
This research is supported by National Natural Science Foundation of China (Grant nos.
11472148 and 51335006) and Beijing Natural Science Foundation (Grant no. 3143024).

Reference
[1] Zadoks RI, Yu X. An investigation of the self-loosening behavior of bolts under transverse vibration. J Sound
Vib 1997; 208(2): 189-209.

10
 ACCEPTED MANUSCRIPT

[2] Izumi S, Yokoyama T, Iwasaki A, Sakai S. Three-dimensional finite element analysis of tightening and
loosening mechanism of threaded fastener. Eng Fail Anal 2005; 12(4): 604-615.
[3] Zhang M, Jiang YY, Lee CH. Finite element modeling of self-loosening of bolted joints. ASME J Mech Des
2007; 129(2): 218-226.
[4] Yokoyama T, Olsson M, Izumi S, Sakai S. Investigation into the self-loosening behavior of bolted joint
subjected to rotational loading. Eng Fail Anal 2012; 23: 35-43.

T
[5] Jiang XJ, Zhu YS, Hong J, Chen X, Zhang YY.Investigation into the loosening mechanism of bolt in curvic

P
coupling subjected to transverse loading. Eng Fail Anal 2013; 32: 360-373.

RI
[6] Muszynska A, Goldman P. Chaotic responses of unbalanced rotor/bearing/stator systems with looseness or
rubs. Chaos Soliton Fract 1995; 5(9): 1683–1704.

SC
[7] Chu F, Tang Y. Stability and non-linear responses of a rotor-bearing system with pedestal looseness. J Sound
Vib 2001; 241(5): 879-893
[8] Ma H, Huang J, Zhang SY, Niu HQ. Nonlinear vibration characteristics of a rotor system with pedestal

NU
looseness fault under different loading conditions J Vibroeng 2013; 15(1): 406-418.
[9] Tamura A, Iwata Y, Sato H. Unstable vibration of a rotor with a transverse crack. In: IMechE Conference of
Vibrations in Rotating Machinery; 1988. p. 647-653.
MA
[10] Liao M, Gasch R. Crack detection in rotating shafts – an experimental study. In: IMechE Conference of
Vibrations in Rotating Machinery; 1992. p. 289 -295.
[11] Keiner H, Gadala S, Tang W. Validity of bolt-removal and gap-insertion techniques to experimentally simulate
the vibration of a cracked rotor. Fatigue Fract Engng Mater Struct 2004;27: 449-458.
ED

[12] Georgantzinos SK, Anifantis NK. An insight into the breathing mechanism of a crack in a rotating shaft. J
Sound Vib 2008; 318(1-2): 279-295.
[13] Friswell MI, Penny JET, Garvey SD, Lees AW. Dynamics of rotating machines. New York: Cambridge
PT

university press; 2010.

[14] Saito T, Endo M. Vibration analysis of rotating cylindrical shells based on the Timoshenko beam theory. B
CE

JSME 1986; 29(250): 1239-1245.

[15] Trotsenko YV. Frequencies and modes of vibration of a cylindrical shell with attached rigid body. J Sound Vib
2006; 292(3-5): 535-551.
AC

[16] Qin ZY, Han QK, Chu FL. Analytical model of bolted disk-drum joints and its application to dynamic analysis
of jointed rotor. P I Mech Eng C - J Mec 2014; 228(4): 646-663.
[17] Han QK, Chu FL. Dynamic instability and steady-state response of an elliptical cracked shaft. Arch Appl
Mech 2012; 82(5): 709-722.
[18] Al-Shudeifat, MA, Butcher EA. New breathing functions for the transverse breathing crack of the cracked rotor
system: Approach for critical and subcritical harmonic analysis. J Sound Vib 2011; 330(3): 526-544.
[19] Sinou JJ, Lees AW. The influence of cracks in rotating shafts. J Sound Vib 2005; 285(5): 1015-1037.

11
 ACCEPTED MANUSCRIPT

Figure Caption

Figure 1. Sketch of fan rotor of a certain type of aero-engine.

Figure 2. Sketch of the disk-drum joint interface.
Figure 3. Finite element model for joint components.

T
Figure 4. Sketch of the simple drum rotor.

P
Figure 5. Finite element model for the drum rotor.

RI
Figure 6. Timoshenko beam element for rotor shaft and drum.
Figure 7. Flexibility coefficients of joint interface involving 1 loosened bolt with (a) half preload left, (b) ten

SC
percent preload left, and (c) bolt removed.
Figure 8. Flexibility coefficients of joint interface involving 3 loosened bolts with (a) half preload left, (b) ten
percent preload left, and (c) bolt removed.

NU
Figure 9. Flexibility coefficients of joint interface involving 5 loosened bolts with (a) half preload left, (b) ten
percent preload left, and (c) bolt removed.
Figure 10. Flexibility coefficients of joint interface involving 10 loosened bolts with (a) half preload left, (b)
MA
ten percent preload left, and (c) bolt removed.
Figure 11. Joint stiffness variation for (a) 3 loosened bolts with half preload left and (b) complete loosening
of 5 bolts in a full revolution.
Figure 12. Frequency-response curves of the rotor with (a) 3 loosened bolt with half preload left and (b)
ED

complete loosening of 5 bolts

Figure 13. Whirling orbits for 3 loosened bolts with half preload left at (a) 1010, (b) 1400, (c) 2100, and (d)
4150 rev/min.
PT

Figure 14. Whirling orbits for complete loosening of 5 bolts at (a) 1010, (b) 1400, (c) 2100, and (d) 4150
rev/min.
CE

Table 1 Values of physical parameters of the rotor system.

AC

12
 ACCEPTED MANUSCRIPT

T
R IP
SC
NU
MA
Blade

Drum
Disc
D

Connecting bolt
TE

Bearing
Bearing Turbine shaft
P
CE

Fig. 1. Sketch of fan rotor of a certain type of aero-engine.

AC

13
 ACCEPTED MANUSCRIPT

T
R IP
SC
Spigot positioning
configuration

NU
Loosened bolts MA
y y
D

O x z O
P TE
CE

Bolts without loosening

Fig. 2. Sketch of the disk-drum joint interface.

AC

14
 ACCEPTED MANUSCRIPT

T
R IP
SC
NU
MA
D
P TE
CE

Fig. 3. The (a) whole and (b) 1/4 sector of finite element model for joint components.
AC

15
 ACCEPTED MANUSCRIPT

T
R IP
SC
NU
MA
D
TE

Fig. 4. Sketch of the simple drum rotor.

P
CE
AC

16
 ACCEPTED MANUSCRIPT

T
R IP
SC
z
NU
MA
kL kR y
1 i-1 iL iM iR i+1 j N+1
D

x
TE

Fig. 5. Finite element model for the drum rotor.

P
CE
AC

17
 ACCEPTED MANUSCRIPT

T
R IP
SC
z
NU
MA
w1 w2

1 2
y
D

u1 1 u2 2
TE

x
Fig. 6. Timoshenko beam element for rotor shaft and drum.
P
CE
AC

18
 ACCEPTED MANUSCRIPT

T
R IP
SC
x 10
-5
x 10
-5

NU x 10
-5
Flexibility coefficients (1/kNm)

Flexibility coefficients (1/kNm)

Flexibility coefficients (1/kNm)

12 12 14
f xx f yx f xy f yy f xx f yx f xy f yy f xx f yx f xy f yy
10 10 12
MA
8 8 10
8
6 6
6
4 4
4
2 2 2
D

0 0 0
-2 -2 -2
TE

0 90 180 270 360 0 90 180 270 360 0 90 180 270 360

Rotation angle (deg) Rotation angle (deg) Rotation angle (deg)

(a) (b) (c)

P

Fig. 7. Flexibility coefficients of joint interface involving 1 loosened bolt with (a) half preload left, (b) ten
CE

percent preload left, and (c) bolt removed.

AC

19
 ACCEPTED MANUSCRIPT

T
R IP
SC
-5 -5 -4

NU
x 10 x 10 x 10

Flexibility coefficients (1/kNm)

Flexibility coefficients (1/kNm)

Flexibility coefficients (1/kNm)

12 14 2.5
f xx f yx f xy f yy f xx f yx f xy f yy f xx f yx f xy f yy
10 12 2
8 10
1.5
8
6
MA
6 1
4
4
0.5
2 2
0 0
0
-2 -2 -0.5
D

0 90 180 270 360 0 90 180 270 360 0 90 180 270 360

Rotation angle (deg) Rotation angle (deg) Rotation angle (deg)
TE

(a) (b) (c)

Fig. 8. Flexibility coefficients of joint interface involving 3 loosened bolts with (a) half preload left, (b) ten
P

percent preload left, and (c) bolt removed.

CE
AC

20
 ACCEPTED MANUSCRIPT

T
R IP
SC
NU
-5 -5 -4
x 10 x 10 x 10
Flexibility coefficients (1/kNm)

Flexibility coefficients (1/kNm)

Flexibility coefficients (1/kNm)

12 20 5
f xx f yx f xy f yy f xx f yx f xy f yy f xx f yx f xy f yy
10 16 4
8 3
12
MA
6 2
8
4 1
4
2 0
0 0 -1
D

-2 -4 -2
0 90 180 270 360 0 90 180 270 360 0 90 180 270 360
Rotation angle (deg) Rotation angle (deg) Rotation angle (deg)
TE

(a) (b) (c)

Fig. 9. Flexibility coefficients of joint interface involving 5 loosened bolts with (a) half preload left, (b) ten
P

percent preload left, and (c) bolt removed.

CE
AC

21
 ACCEPTED MANUSCRIPT

T
R IP
SC
NU
-5 -4 -3
x 10 x 10 x 10

Flexibility coefficients (1/kNm)

Flexibility coefficients (1/kNm)

Flexibility coefficients (1/kNm)

12 2.5 4
f xx f yx f xy f yy f xx f yx f xy f yy f xx f yx f xy f yy
10 2 3
8
MA
1.5 2
6
1 1
4
0.5 0
2
0 0 -1
D

-2 -0.5 -2
0 90 180 270 360 0 90 180 270 360 0 90 180 270 360
TE

Rotation angle (deg) Rotation angle (deg) Rotation angle (deg)

(a) (b) (c)

Fig. 10. Flexibility coefficients of joint interface involving 10 loosened bolts with (a) half preload left, (b) ten
P

percent preload left, and (c) bolt removed.

CE
AC

22
 ACCEPTED MANUSCRIPT

T
R IP
SC
4
x 10
14000 3
kxx from FE analysis

NU
12000 2.5 kxy from FE analysis

Stiffness coefficients (kNm)

Stiffness coefficients (kNm)

10000 kyy from FE analysis

2
kxx from FE analysis Truncated Fourier expansion
8000
1.5
MA
kxy from FE analysis
6000
kyy from FE analysis 1
4000 Truncated Fourier expansion
0.5
2000
0
D

-2000 -0.5
TE

0 90 180 270 360 0 90 180 270 360

Rotation angle (deg) Rotation angle (deg)

(a) (b)
P

Fig. 11. Joint stiffness variation for (a) 3 loosened bolts with half preload left and (b) complete loosening of 5
CE

bolts in a full revolution.

AC

23
 ACCEPTED MANUSCRIPT

T
IP
-1
10
f1
Response amplitude at middle disk (mm)

b1 x direction
y direction

 f 1 b1 f 1
-2

R
10
f 2
b1 2 2 f 2
3
 f 13

SC
10
-3 3
4 b 2
b1
-4
10 4

NU
-5
10

-6
MA
10
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Rotating speed (rev/min) x 10
4

(a)
D

0
10
b1  f 1
Response amplitude at middle disk (mm)

b1  f 1 x direction
TE

2 2 y direction
10
-1
b1  f 1
3 3
f1
-2 4 b 2  f 2 f 2
10 b1 f 2
P

3 3 b 2 b 2
4 2
2
CE

-3
10

-4
10
AC

-5
10
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Rotating speed (rev/min) x 10
4

(b)
Fig. 12. Frequency-response curves of the rotor with (a) 3 loosened bolt with half preload left and (b) complete
loosening of 5 bolts

24
 ACCEPTED MANUSCRIPT

T
R IP
SC
NU
Disp. in y direction (mm)
Disp. in y direction (mm)

Disp. in y direction (mm)

Disp. in y direction (mm)

-0.0406 -0.0406 -0.0406 -0.02
MA
-0.0406 -0.03
-0.0406 -0.0406
-0.0407 -0.04
-0.0407 -0.0407
-0.0407 -0.05
-0.0407 -0.0408 -0.0407 -0.06
-5 5 -5 0 5 -5 0 5 -0.02 0 0.02
D

x 10-5 x 10-5 x 10-5

Disp. in x direction (mm) Disp. in x direction (mm) Disp. in x direction (mm) Disp. in x direction (mm)
TE

(a) (b) (c) (d)

Fig. 13. Whirling orbits for 3 loosened bolts with half preload left at (a) 1010, (b) 1400, (c) 2100, and (d) 4150
P

rev/min.
CE
AC

25
 ACCEPTED MANUSCRIPT

T
R IP
SC
NU

Disp. in y direction (mm)

Disp. in y direction (mm)
Disp. in y direction (mm)
Disp. in y direction (mm)

-0.039 -0.038 -0.036 0.05

MA
-0.04 -0.039 -0.038 0

-0.041 -0.04 -0.04 -0.05

-0.042 -0.041 -0.042 -0.1

-0.043 -0.042 -0.044 -0.15

D

-0.001 0 0.001 -0.002 0.002 -0.005 0 0.005 -0.1 0 0.1

Disp. in x direction (mm) Disp. in x direction (mm) Disp. in x direction (mm) Disp. in x direction (mm)
TE

(a) (b) (c) (d)

Fig. 14. Whirling orbits for complete loosening of 5 bolts at (a) 1010, (b) 1400, (c) 2100, and (d) 4150
P

rev/min.
CE
AC

26
 ACCEPTED MANUSCRIPT

Table 1 Values of physical parameters of the rotor system.

T
Description Value
Length of drum, Lm 80 mm

P
Outer radius of drum, Rmo 101.2 mm

RI
Inner radius of drum, Rmi 100 mm
Length of left shaft, Lfl 200 mm

SC
Length of middle shaft, Lfm 140 mm
Length of right shaft, Lfr 195 mm
Radius of shaft, Rf 15 mm

NU
Outer radius of middle disk, Rko 120 mm
Inner radius of middle disk, Rki 40 mm
Thick of middle disk, tkm 10 mm
MA
Radius of side disks, Rks 101.2 mm
Thick of side disks, tks 8 mm
Radius of right disk, Rkr 80 mm
ED

Thick of right disk, tkr 20 mm

Elastic modulus of rotor, E 210 GPa
Poisson’s ratio of rotor, υ 0.3
PT

Density of rotor, ρ 7800 kg/m3

Bearing stiffness, ( kxx, kyy ) 1.3×107 N/m
7×102 Ns/m
CE

Bearing damping, ( cxx, cyy )

Number of bolts at joint interface, nb 20
AC

27
 ACCEPTED MANUSCRIPT

Highlights:

• Time varying stiffness due to rotating bolt loosening is studied by 3D FE analysis.

• Time varying stiffness matrix is incorporated into the rotor dynamic model.

T
• HBM is used to calculate the rotor response with bolt loosening.

P
RI
• Stiffness variation trend of bolt loosening resembles breathing transverse crack.

Peaks at forward and backward whirling critical and subcritical speeds are excited.

SC
•

NU
MA
ED
PT
CE
AC

28