Mechanical Systems and Signal Processing 218 (2024) 111564

Contents lists available at ScienceDirect

Mechanical Systems and Signal Processing

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

Dynamic modeling of gear compound faults and stiffness

sensitivity analysis against arbitrary spatial configuration of defect
Xiao Yang , Xiang Li , Yaguo Lei *, Bin Yang , Huan Liu , Xuanyu Gao
Key Laboratory of Education Ministry for Modern Design and Rotor-Bearing System, Xi’an Jiaotong University, Xi’an 710049, China

A R T I C L E I N F O A B S T R A C T

Keywords: Gear systems play a crucial role in various machines to transmit power and motion. One common
Dynamic modeling issue in the gear systems is the frequent gear collision, which results in increased wear and tear,
Gear transmission system and ultimately, damage to the gears. Consequently, the gear faults tend to develop in a compound
Compound faults
manner rather than in isolation, i.e., a faulty gear often encompasses multiple defects, such as
Time-varying meshing stiffness (TVMS)
Stiffness sensitivity analysis
pitting, cracking, and so forth. Nevertheless, the existing literature has primarily centered on the
gear model with a single fault, leaving a research gap in the modeling of compound faults.
Further, to elucidate the root causes behind the degradation of gear systems, it is necessary to
figure out the relationship between the intrinsic characteristics of gear systems and the pro­
gressive development of compound faults of gears. To address these issues, this paper proposes a
compound fault gear modeling method with a generic framework to handle arbitrary spatial
configuration of compound faults. The compound fault gear model is then verified by finite
element model (FEM). Further, the relationship between the progressive development of gear
compound faults and the evolution of time-varying meshing stiffness (TVMS) of gears is inves­
tigated, presenting a detailed comparison of the sensitivity of TVMS to different spatial config­
uration of gear faults. The results offer a promising avenue for modeling machinery degradation
with compound faults, paving the way for a more comprehensive understanding of the intricate
dynamics involved in the degeneration of gear systems.

1. Introduction

Gear transmission system plays a crucial role to machinery safely and reliably operation [1,2]. Under external excitation as complex
working condition and intrinsic excitation like meshing stiffness impact, gear tooth faults occur during transmission system operation.
Typical types of faults include the tooth surface defects, such as pitting and spalling, and gear tooth crack, whose modeling method is
rather distinct from surface defects [3]. The gear crack fault has no certain parts dropped out from gear tooth, but the existing of crack
separate the failure part of gear and tooth itself. Due to this, most works modeled the different types of faults based on independent
framework. However, the gear faults tend to develop in a compound manner rather than in isolation, i.e., a faulty gear often en­
compasses multiple defects [4,5]. Various types of faults are coupled with each other, which brings difficulties to the modeling of
compound faults.
The time-varying meshing stiffness (TVMS) of gear pair is the main intrinsic excitation during transmission. To study the dynamic
behaviors of fault gear, many works regarding to the stiffness model are presented. Most researchers focus on the study of modeling of

* Corresponding author.
E-mail address: [email protected] (Y. Lei).

https://doi.org/10.1016/j.ymssp.2024.111564
Received 3 January 2024; Received in revised form 31 March 2024; Accepted 23 May 2024
Available online 30 May 2024
0888-3270/© 2024 Elsevier Ltd. All rights are reserved, including those for text and data mining, AI training, and similar technologies.
X. Yang et al. Mechanical Systems and Signal Processing 218 (2024) 111564

Fig. 1. Stress analysis of healthy gear during meshing.

single type gear fault. The first sort of modeling work address surface defects. The characteristics of surface defects is that the faults
present spot distribution and the failure part break away from gear tooth. For pitting fault, a gear single tooth pit model based on the
magnitude-phase changes stiffness analyzing was proposed by Chaari et al. [6]. Liang et al. [7] established a model of multiple gear pit
fault on a gear tooth and investigated the impact of different pitting severity levels on the gear mesh stiffness. Lei et al. [8] applied
uniform distribution to model the pitting on the surface of gear tooth as a series of pits and the TVMS of pitting gears is evaluated.
Above studies regards pitting as cylinder defects. Rincon et al. [9] introduced a finite element model to assess the consequences of
pitting on the meshing stiffness, where a single pit was modeled in an elliptical shape. Meng et al. [10] proposed a method where the
pitting shape is considered as approximately a part of sphere. In addition, for spalling fault, Chaari et al. [11] built a spalling model that
regard the fault as a rectangular defect and studied effects of the spalling damage on the gear mesh stiffness. Similarly, Ma et al. [12]
also shaped the spalling fault a rectangular defect and calculated the TVMS. Those similar work was carried out by Abouel-seoud et al
[13] and Cheng et al [14] as well. Chen et al. [15] proposed a finite element method to model different spalling defects utilizing
spalling variables, including the spalling lengths, widths, axial position. And a calculation model of gear spalling fault was presented by
Liang et al. [16] to analyze the effect of fault on the TVMS. Above studies consider spalling as certain geometrical shape, most of which
is rectangular shape. Luo et al. [17] put forward a calculation method of gear mesh stiffness in the presence of spalling fault, which was
suitable for different shapes of spalling faults.
The second sort of modeling works addressed gear crack faults. The characteristics of crack defects is that the faults present line
distribution and the failure part remain on gear tooth. Chen et al. [18] proposed a gear tooth root crack model of a spur gear pair and
obtained TVMS under different crack length and depth. Based on Chen’s work [18,19], Mohammed et al. [20] modified the limiting
line as a parabola line, and the meshing stiffness under different crack grades is compared with the finite element analysis results. Han
et al. [21] extended the studies in [22,23] and illustrated the stress distribution of a rectangular cantilever beam with cracks, which
inspired the proposition of exponential limiting line crack model [24]. Above works addressed two-dimensional crack problem. For
spatial crack, Yu et al. [25] considered the spatial distribution of root cracks and the load distribution of teeth, the two-dimensional
crack model is extended to three-dimensional crack model. Similar works such as Chen [26] and Wang [27]. Duan et al. [28,29]
developed a three-dimensional numerical model of gear crack propagation considering the initial crack location, ring gear backup ratio
and constraint mode based on the linear elastic fracture mechanics. The above literature mainly focused on study of gear tooth root
crack. However, gear crack occurs not only on tooth root, but also on arbitrary position along meshing line. According to the above
literature review, it can be concluded that the works addressed gear single type fault modeling mostly aimed at giving precise define to
the shape of defects, which also increases the restriction of model framework, and causes more difficult to model compound faults.
Hence, research on gear tooth compound fault dynamic modeling is rather limited. Overall, a universal gear compound faults dynamic
modeling framework is not yet established.
In addition, the propagation of compound gear faults will lead to degeneration of gear transmission system, and further causes
failure of mechanical equipment. In order to study the degradation mechanism of machinery system, it is of significant to illustrate the
develop path of defects and attempt to clarify the propagation mechanism of gear faults, i.e., the modeling framework is required to
reflect the system condition not only on certain time, but also on entire degeneration process. To tackle this problem, it is feasible to
study the relation between stiffness reduction magnitude and defects propagation path along different direction, such as gear tooth
width, gear tooth thickness and meshing position. To be specific, it is common that the decline of stiffness is caused by the development
of defects. Correspondingly, the decrease of stiffness also leads to extension of gear tooth defects. Hence, clarifying the stiffness
sensitivity to different defects extension path would be rather helpful to the conduction of gear degeneration mechanism. Moreover, it
is able to provide theory foundation to degradation model of machinery system.

2
X. Yang et al. Mechanical Systems and Signal Processing 218 (2024) 111564

To address the above shortcomings, a gear compound faults TVMS model based on improve potential energy method is proposed. In
the proposed model, the defects are not defined to certain shape, i.e., the proposed method is adapted to arbitrary defect spatial
configuration. The simulation results are compared with the results of finite element model (FEM) and further validate the effec­
tiveness and accuracy of proposed model. Based on which, the relation of stiffness sensitivity to defects extension path is conducted.
The sensitivity of TVMS to the defect extension path along gear tooth width, thickness and meshing path are analyzed finally.
The remainder of this paper is organized as follows. Section 2 details modeling of TVMS against arbitrary defect spatial configu­
ration based on improved potential energy method. Section 3 presents the overall framework of gear compound faults model. In
Section 4, the simulation results are obtained, and FEM models are conducted for effectiveness demonstration. Section 5 analyzes the
stiffness sensitivity to different defects extension path. Finally, conclusions are drawn in Section 6.

2. TVMS modeling against arbitrary defect spatial configuration.

As mentioned above, there are two typical forms of defects occur on gear tooth, i.e., the surface defects and gear crack fault. For
crack fault, the failure part remains on gear tooth but separates from the tooth effective part, we define the gear crack fault as cracking
faults. In this section, we model the gear faults based on above two main types of spatial configuration firstly. The defect parts are
divided to minute integrate elements for the derivation of proposed model. Below we first simply introduce the theory of potential
energy method and present the TVMS model of healthy gear tooth, which provide foundation to proposed method.
According to [30], considering each stress component comprehensively, the bending potential energy, shearing potential energy
and axial compressive potential energy of the element showed in Fig. 1 are,
∫ ∫ ∫
M2 d
[Fb (d − x) − Fa h]2 Rb − Rr
[Fb (d + x1 ) − Fa h]2
Ub = dx = dx + dx1 (1)
l 2EIx 0 2EIx 0 2EIx1

∫ ∫ ∫
1.2Fb2 d
1.2(Fcosα1 )2 Rb − Rr
1.2(Fcosα1 )2
Us = dx = dx + dx1 (2)
l 2GAx 0 2GAx 0 2GAx1

∫ ∫ ∫
Fa2 d
(Fsinα1 )2 Rb − Rr
(Fsinα1 )2
Ua = dx = dx + dx1 (3)
l 2EA x 0 2EAx 0 2EAx1

where E represents Young’s modulus, G represents elasticity modulus, Rb is radius of gear base circle, Rr is radius of gear tooth root
circle, I represents the cross sectional moment of inertia, A represents the area of the cross sectional, Ub , Us , Ua represent the bending
potential energy, shearing potential energy, axial compressive potential energy, respectively.
Based on geometrical analyze, it can be obtained that,
d = Rb [(α1 + α2 )sinα1 + cosα1 − cosα2 ] (4)

h = Rb [(α1 + α2 )cosα1 − sinα1 ] (5)

x = Rb [(α2 − α)sin(− α) + cos(− α) − cosα2 ] (6)

hx = Rb [(α2 − α)cos(− α) − sin(− α)] (7)

and the cross sectional moment of inertia and the area of the cross sectional on the position x from the tooth base circle are derived as,
1 2
Ix = (2hx )3 L = R3b [(α2 − α)cos(α) + sin(α)]3 L (8)
12 3

Ax = 2hx L = 2Rb L[(α2 − α)cos(α) + sin(α)] (9)

In addition, the potential energy between the tooth root circle and tooth base circle is also considered in proposed method, as
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
shown in Fig. 1. It can be easily obtained from geometrical relationship that RR = (R2b − R2r )/(2Rr ) and hx1 = hb + RR − RR2 − x21 .
And the cross sectional moment of inertia and the area of the cross sectional on the position x1 from the tooth base circle are derived as,
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ 3
1 2
Ix = (2hx1 )3 L = (Rb sinα2 + RR − RR2 − x21 ) L (10)
12 3
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
Ax = 2hx L = 2(Rb sinα2 + RR − RR2 − x21 )L (11)

Through geometry analysis and coordinates transformation between x and α, and combine the theory of Hooke’s law, we can obtain
the bending stiffness, shearing stiffness, axial compressive stiffness, the Hertz contact stiffness and foundation deflection stiffness of a
healthy gear tooth, respectively.

3
X. Yang et al. Mechanical Systems and Signal Processing 218 (2024) 111564

Fig. 2. Stress analysis of gear surface defects during meshing.

∫
1 2Ub α2
3{1 + [(α2 − α)sinα − cosα]cosα1 }2 (α2 − α)cosα
= 2 = dα
kb F − α1 2EL[(α2 − α)cosα + sinα]3
∫ Rb − Rr
(12)
3(Rb − Rb cosα1 cosα2 + x1 cosα1 )
+ √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ 3 dx1
0 2EL(Rb sinα2 + RR − RR2 − x21 )
∫ α2
1 2Us 1.2(1 + v)cos2 α1 (α2 − α)cosα
= 2 = dα
ks F − α1 EL[(α2 − α)cosα + sinα]
∫ Rb − Rr (13)
1.2(1 + v)cos2 α1
+ √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ dx1
0 EL(Rb sinα2 + RR − RR2 − x21 )
∫ α2
1 2Ua (α2 − α)cosαsin2 α1
= 2 = dα
ka F − α1 2EL[(α2 − α)cosα + sinα]
∫ Rb − Rr (14)
sin2 α1
+ √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ dx1
0 2EL(Rb sinα2 + RR − RR2 − x21 )

1 πLE
= (15)
kh 4(1 − v2 )
{ }
1 cos2 αm uf uf
= L* ( )2 + M* ( ) + P* (1 + Q* tanαm ) (16)
kf EL Sf Sf

where αm is the gear pressure angle, L is the tooth width, ν is Poisson’s ratio, and more details about the coefficients uf ,Sf ,L* , M* , P* , and
Q* can be found in [32]. In addition, considering coupling effect due to the elastic deformation of tooth fillet foundation, we calculate
the tooth fillet foundation stiffness mainly referring to method proposed in [3] and [39]. The tooth fillet foundation stiffness was
derived analytically as below,
( )2
1 cosα1 cosα2 u1 u2 u1
=[ L1 + (tanα2 M1 + P1 )( )+
kf21 WE S2 S
(17)
u2
(tanα1 Q1 + R1 )( ) + (tanα1 S1 + T1 )tanα2 + U1 tanα1 + V1 )]
S

4
X. Yang et al. Mechanical Systems and Signal Processing 218 (2024) 111564

( )2
1 cosα1 cosα2 u1 u2 u2
= [L2 + (tanα1 M2 + P2 )( )+
kf12 WE S2 S
(18)
u1
(tanα2 Q2 + R2 )( ) + (tanα2 S2 + T2 )tanα1 + U2 tanα2 + V2 )]
S

where the symbols ui and αi are defined the same as Eq. (16), the coefficients Mi , Qi , Ti , Ui , Li , Pi , Ri , Si and Vi can be obtained through
curve-fitted with polynomial functions based on the numerical calculation method proposed in [39].
Finally, the comprehensive TVMS of gear is calculated by,
2
∑ 1
k= (19)
i=1
1/kh,i + 1/kb1,i + 1/ka1,i + 1/ks1,i + 1/kf12,i + 1/kb2,i + 1/ka2,i + 1/ks2,i + 1/kf21,i

and i = 1, 2 represents the first and the second pair of meshing teeth, respectively.

2.1. TVMS model of surface defects with arbitrary spatial configuration

As shown in Fig. 2, for the minute integrate element, assuming that the initial and terminal position corresponding rotation angles
of driving gear are αis1 and αis2 , the defect width is Bi , the defect depth is tj , where i = 1, 2, ..., m and j = 1,2,...,k, m and k are decided by
integration step.Bmc is the current width length of gear defect on the surface of tooth. It noted that only external gear mesh is
considered here.
It is obvious that the bending stiffness, shearing stiffness, axial compressive stiffness and Hertz contact stiffness are modified due to
surface defects. The foundation deflection stiffness does not change. The cross sectional moment of inertia and the area of the cross
sectional on the position that defect occurs are derived as,
2
Ixsufd = R3b [(α2 − α)cos(α) + sin(α)]3 (L − Bmc ) (20)
3

Asuf
x d = 2Rb L[(α2 − α)cos(α) + sin(α)](L − Bmc ) (21)

The bending energy, shearing energy and axial compressive energy of the defect are derived as,
mc ∑ki ∫ dis 2 ∫ dis 2
∑ Bi 2 [Fb (d − x) − Fa h] 2 [Fb (d − x) − Fa h]
Ub d = ( dx − dx) (22)
i=1 j=1
L i
ds 2EIx d i
ds 2EI sx d
1 1

mc ∑ki ∫ dis 2 ∫ dis 2

∑ Bi 2 1.2(Fcosα1 ) 2 1.2(Fcosα1 )
Us d = ( dx − dx) (23)
i=1 j=1
L i
ds 2GA x d i
ds 2GA sx d
1 1

mc ∑ki ∫ dis 2 ∫ dis 2

∑ Bi 2 (Fsinα1 ) 2 (Fsinα1 )
Ua d = ( dx − dx) (24)
i=1 j=1
L ds i 2EAx d i
ds 2EAsx d
1 1

where
2
Ix d = R3b [(α2 − α)cos(α) + sin(α)]3 L (25)
3

Ax d = 2Rb L[(α2 − α)cos(α) + sin(α)] (26)

1 1
Isx d = (2hx − tj )3 L = {2Rb [(α2 − α)cos(α) + sin(α)] − tj }3 L (27)
12 12

Asx d = 2(2hx − tj )L = 2(2Rb [(α2 − α)cos(α) + sin(α)] − tj )L (28)

ds1 = Rb (α2 + αs1 )sinαs1 + Rb cosαs1 − Rb cosα2 (29)

ds2 = Rb (α2 + αs2 )sinαs2 + Rb cosαs2 − Rb cosα2 (30)

According to meshing position, there are three stages during the whole engagement. The models are conducted as below address
three stages:

(1) No defect enters into meshing: the TVMS model perform the same as healthy model.
(2) Defect enters into meshing and located on meshing position: the modification of bending stiffness, shearing stiffness, axial
compressive stiffness and Hertz contact stiffness are derived as below,

5
X. Yang et al. Mechanical Systems and Signal Processing 218 (2024) 111564

∫
1 2Ub L α2
3{1 + [(α2 − α)sinα − cosα]cosα1 }2 (α2 − α)cosα
= 2 = dα
kb F L − Bmc − α1 2EL[(α2 − α)cosα + sinα]3
∫ Rb − Rr
3(Rb − Rb cosα1 cosα2 + x1 cosα1 )
+ √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ 3 dx1
0 2EL(Rb sinα2 + RR − RR2 − x21 )
(31)
∑mc ∑ki Bi ∫ αis2 3{1 + [(α2 − α)sinα − cosα]cosα1 }2 (α2 − α)cosα
− j=1 L
( dα
i=1
− αis
1
2EL[(α2 − α)cosα + sinα]3
∫ αis
2 3{1 + [(α2 − α)sinα − cosα]cosα1 }2 (α2 − α)cosα
− tj 3 dα )
− αis
1
2EL[(α2 − α)cosα + sinα − 2Rb
]

∫
1 2Us L α2
1.2(1 + v)cos2 α1 (α2 − α)cosα
= 2 = dα
ks F L − B mc − α1 EL[(α2 − α)cosα + sinα]
∫ Rb − Rr
1.2(1 + v)cos2 α1
+ √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ dx1
0 EL(Rb sinα2 + RR − RR2 − x21 )
∑mc ∑ki Bi ∫ − αis2 1.2(1 + v)cos2 α1 (α2 − α)cosα (32)
− j=1 L
( dα
i=1
− αis EL[(α2 − α)cosα + sinα]
1

∫ − αis
1 1.2(1 + v)cos2 α1 (α2 − α)cosα
− t j dα )
− αis
2 EL[(α2 − α)cosα + sinα − ]
2Rb
∫ α2
1 2Ua L (α2 − α)cosαsin2 α1
= 2 = dα
ka F L − Bmc − α1 2EL[(α2 − α)cosα + sinα]
∫ Rb − Rr
sin2 α1
+ √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ dx1
0 2EL(Rb sinα2 + RR − RR2 − x21 )

∑mc ∑ki Bi ∫ − αis2 (33)

(α2 − α)cosαsin2 α1
− ( dα
− αis1 2EL[(α2 − α)cosα + sinα]
i=1 j=1 L

∫ − αis
2 (α2 − α)cosαsin2 α1
− tj dα
− αis1 2EL[(α2 − α)cosα + sinα − ]
2Rb

1 π(L − Bmc )E
= (34)
kh 4(1 − v2 )
To give explanation to above derivation, it is clear that the energy of defect gear based on cantilever beam of current meshing
position equals to the energy of the healthy beam minus the total energy of defects before current position. And also, because the
surface defect is located on meshing line, the length of contact line decreases, which equals to L − Bmc . So far, it obtains the second stage
model of surface defects with arbitrary spatial configuration.

(3) All the defects come out of meshing: the modification of bending stiffness, shearing stiffness, axial compressive stiffness and
Hertz contact stiffness are derived as below,
∫
1 2Ub α2
3{1 + [(α2 − α)sinα − cosα]cosα1 }2 (α2 − α)cosα
= 2 = dα
kb F − α1 2EL[(α2 − α)cosα + sinα]3
∫ Rb − Rr
3(Rb − Rb cosα1 cosα2 + x1 cosα1 )
+ √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ 3 dx1
0 2EL(Rb sinα2 + RR − RR2 − x21 )
(35)
∑m ∑ki Bi ∫ αis2 3{1 + [(α2 − α)sinα − cosα]cosα1 }2 (α2 − α)cosα
− j=1 L
( dα
i=1
− αis
1
2EL[(α2 − α)cosα + sinα]3
∫ αis
2 3{1 + [(α2 − α)sinα − cosα]cosα1 }2 (α2 − α)cosα
− tj 3 dα )
− αis
1
2EL[(α2 − α)cosα + sinα − 2Rb
]

6
X. Yang et al. Mechanical Systems and Signal Processing 218 (2024) 111564

Fig. 3. Stress analysis of gear cracking defects during meshing.

∫ α2
1 2Us 1.2(1 + v)cos2 α1 (α2 − α)cosα
= 2 = dα
ks F − α1 EL[(α2 − α)cosα + sinα]
∫ Rb − Rr
1.2(1 + v)cos2 α1
+ √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ dx1
0 EL(Rb sinα2 + RR − RR2 − x21 )
∑m ∑ki Bi ∫ − αis2 1.2(1 + v)cos2 α1 (α2 − α)cosα (36)
− j=1 L
( dα
i=1
− αis EL[(α2 − α)cosα + sinα]
1

∫ − αis
1 1.2(1 + v)cos2 α1 (α2 − α)cosα
− t j dα )
− αis
2 EL[(α2 − α)cosα + sinα − ]
2Rb
∫ α2
1 2Ua (α2 − α)cosαsin2 α1
= 2 = dα
ka F − α1 2EL[(α2 − α)cosα + sinα]
∫ Rb − Rr
sin2 α1
+ √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ dx1
0 2EL(Rb sinα2 + RR − RR2 − x21 )

∑m ∑ki Bi ∫ − αis2 (37)

(α2 − α)cosαsin2 α1
− ( dα
i=1 j=1 L i
− αs1 2EL[( α2 − α)cosα + sinα]
∫ − αis
2 (α2 − α)cosαsin2 α1
− tj dα
− αis1 2EL[(α2 − α)cosα + sinα − ]
2Rb

1 πLE
= (38)
kh 4(1 − v2 )
Similarly, the energy of defect gear based on cantilever beam of current meshing position equals to the energy of the healthy beam
minus the total energy of defects before current position. However, different with stage (2), there is no defect located on current
meshing line the length of contact line remains same with normal part.

2.2. TVMS model of cracking defects with arbitrary spatial configuration

For cracking defects, the failure parts of gear tooth are able to be regarded as surface defects remain on tooth. Specifically speaking,

7
X. Yang et al. Mechanical Systems and Signal Processing 218 (2024) 111564

although there is no defect part drops out from tooth itself, there are definite segments do not work and the energy of which also loses,
namely the dead zone. The boundary of failure part of cracking defect should be modified based on the crack shape and its limiting line,
which is shown as solid red line in Fig. 3. The limiting line utilized in proposed model is exponential limiting line. The model is based
on Saint Venant’s principle that the normal stress distributed along both sides of the crack decreases in form of exponent. And the
vibration differential equation of a cracked beam could be expressed as,

EI0 Q1 wiv + 2EI0 Qʹ1 wʹ + EI0 Qʹʹ1 wʹʹ + Mẅ = 0 (39)

ʹʹ

− 2λDf − 1
Q1 (x) = [1 + C1 exp( )] (40)
2hx
3
where C1 = (2h(2h− xh) )3 − 1. I0 is the cross sectional moment of inertia of healthy beam, M represents unit of mass, w the deflection, and
x c

Q1 (x) is the attenuation function, Df is the distance between current meshing position and the crack, hc represents the depth of crack,
λ = 0.667 is the coefficient affected by crack stress decrement, whose value is evaluated by experiment.
Based on Saint Venant’s principle, the attenuation function Q1 (x) also describes the relation between the cross sectional moment of
inertia of the cracked and healthy beam as Ic = Q1 (x)I0 . Thus, it could be obtained that,
2hx − t(x) 3
Q1 (x) = ( ) (41)
2hx

where t(x) is the defect depth defined by the exponential limiting line on current meshing position. Substitute Eqs. (40)–(41), t(x) is
calculated correspondingly. In proposed model, we define the boundary of crack area on every single slice along tooth width. As long as
the tooth crack is continuous and smooth, the limiting surface of the crack could be defined accordingly. By now, the dead zone of the
crack tooth is described completely by exponential limiting line model. More details about calculation the exponential limiting line can
be found in [24].
Similarly, for the minute integrate element, assuming that the initial and terminal position corresponding rotation angles of driving
gear are αiq1 and αiq2 , the defect depth is tj , which is calculated by exponential limiting model, and the defect width is Ci , where i = 1,
2, ..., m and j = 1, 2, ..., k, m and k are decided by integration step. The value of Ci could also be calculated by Eqs. (40) and (41). The
bending stiffness and shearing stiffness are modified due to cracking defects. The axial compressive stiffness, Hertz contact stiffness and
foundation deflection stiffness does not change.
The bending energy, shearing energy of the defect are derived as,
m ∑ ki ∫ dis 2 ∫ dis 2
∑ Ci 2 [Fb (d − x) − Fa h] 2 [Fb (d − x) − Fa h]
Ub d = ( dx − dx) (42)
i=1 j=1
L dis 2EIx d dis 2EIsx d
1 1

m ∑ ki ∫ dis 2 ∫ dis 2
∑ Ci 2 1.2(Fcosα1 ) 2 1.2(Fcosα1 )
Us d = ( dx − dx) (43)
i=1 j=1
L dis 2GAx d dis 2GAsx d
1 1

where Ix d , Ax d ,Isx d , Asx d , ds1 and ds2 are derived similar as Section 2.1.
According to meshing position, there are three stages during the whole engagement. The models are conducted as below address
three stages:

(1) No defect enters into meshing: the TVMS model perform the same as healthy model.
(2) Defect (within limiting line) enters into meshing and located on meshing position: the modification of bending stiffness,
shearing stiffness are derived as below,
∫
1 2Ub α2
3{1 + [(α2 − α)sinα − cosα]cosα1 }2 (α2 − α)cosα
= 2 = dα
kb F − α1 2EL[(α2 − α)cosα + sinα]3
∫ Rb − Rr
3(Rb − Rb cosα1 cosα2 + x1 cosα1 )
+ √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ 3 dx1
0 2EL(Rb sinα2 + RR − RR2 − x21 )
(44)
∑mc ∑ki Ci ∫ αiq2 3{1 + [(α2 − α)sinα − cosα]cosα1 }2 (α2 − α)cosα
− j=1 L
( dα
i=1
− αiq
1
2EL[(α2 − α)cosα + sinα]3
∫ αiq
2 3{1 + [(α2 − α)sinα − cosα]cosα1 }2 (α2 − α)cosα
− tj 3 dα)
− αiq
1
2EL[(α2 − α)cosα + sinα − 2Rb
]

8
X. Yang et al. Mechanical Systems and Signal Processing 218 (2024) 111564

Table 1
Compound faults TVMS function of healthy components: surface defect.
Stage 1 Stage 2
⃒ ⃒ i ⃒ ⃒ i
1 L 1 ⃒⃒α2 ∑mc ∑ki Bi 1 1 ⃒⃒− αs2 1 1 1 ⃒⃒α2 ∑m ∑ki Bi 1 1 ⃒⃒− αs2 1
= − i=1 j=1 (1 − ) + (49) = − i=1 j=1 (1 − ) + (53)
kbs1 L − Bmc kb1 ⃒− α1 L hx3j kb1 ⃒− αis kbr kbs2 kb1 ⃒− α1 L hx3j kb1 ⃒− αis kbr
1 1
⃒α2 ⃒ i ⃒ ⃒ i
1 L 1 ⃒⃒ ∑mc ∑ki Bi 1 1 ⃒⃒− αs2 1 1 1 ⃒⃒α2 ∑m ∑ki Bi 1 1 ⃒⃒− αs2 1
= − i=1 j=1 (1 − ) + (50) = − i=1 j=1 (1 − ) + (54)
kss1 L − Bmc ks1 ⃒− α1 L hxj ks1 ⃒ i − α s1 ksr kss2 ks1 ⃒− α1 L hxj ks1 ⃒− αis ksr
1
⃒ ⃒ αis2 ⃒ ⃒ i
1 L 1 ⃒⃒α2 ∑mc ∑ki Bi 1 1 ⃒⃒− 1 1 1 ⃒⃒α2 ∑m ∑ki Bi 1 1 ⃒⃒− αs2 1
= − i=1 j=1 (1 − ) + (51) = − i=1 j=1 (1 − ) + (55)
kas1 L − Bmc ka1 ⃒− α1 L hxj ka1 ⃒− αis1 kar kas2 ka1 ⃒− α1 L hxj ka1 ⃒− αis kar
1

1 L 1 1 1
= (52) = (56)
khs1 L − Bmc kh1 khs2 kh1

∫ α2
1 2Us 1.2(1 + v)cos2 α1 (α2 − α)cosα
= 2 = dα
ks F − α1 EL[(α2 − α)cosα + sinα]
∫ Rb − Rr
1.2(1 + v)cos2 α1
+ √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ dx1
0 EL(Rb sinα2 + RR − RR2 − x21 )
∑mc ∑ki Ci ∫ − αiq2 1.2(1 + v)cos2 α1 (α2 − α)cosα (45)
− j=1 L
( dα
i=1
− αiq EL[(α2 − α)cosα + sinα]
1

∫ − αiq
2 1.2(1 + v)cos2 α1 (α2 − α)cosα
− tj dα)
− αiq
1 EL[(α2 − α)cosα + sinα − ]
2Rb

where mc is subscript of the current width length of defect. To give explanation to above derivation, the energy of defect gear based on
cantilever beam of current meshing position equals to the energy of the healthy beam minus the total energy of defects before current
position. And the length of contact line remains same with normal part. So far, it obtains the second stage model of cracking defects
with arbitrary spatial configuration.

(3) All the defects come out of meshing: the modification of bending stiffness, shearing stiffness are derived as below,
∫
1 2Ub α2
3{1 + [(α2 − α)sinα − cosα]cosα1 }2 (α2 − α)cosα
= 2 = dα
kb F − α1 2EL[(α2 − α)cosα + sinα]3
∫ Rb − Rr
3(Rb − Rb cosα1 cosα2 + x1 cosα1 )
+ √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ 3 dx1
0 2EL(Rb sinα2 + RR − RR2 − x21 )
(46)
∑m ∑ki Ci ∫ αis2 3{1 + [(α2 − α)sinα − cosα]cosα1 }2 (α2 − α)cosα
− j=1 L
( dα
i=1
− αis
1
2EL[(α2 − α)cosα + sinα]3
∫ αis
2 3{1 + [(α2 − α)sinα − cosα]cosα1 }2 (α2 − α)cosα
− tj 3 dα )
− αis
1
2EL[(α2 − α)cosα + sinα − 2Rb
]

∫ α2
1 2Us 1.2(1 + v)cos2 α1 (α2 − α)cosα
= 2 = dα
ks F − α1 EL[(α2 − α)cosα + sinα]
∫ Rb − Rr
1.2(1 + v)cos2 α1
+ √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ dx1
0 EL(Rb sinα2 + RR − RR2 − x21 )
∑m ∑ki Ci ∫ − αis2 1.2(1 + v)cos2 α1 (α2 − α)cosα (47)
− j=1 L
( dα
i=1
− αis EL[(α2 − α)cosα + sinα]
1

∫ − αis
1 1.2(1 + v)cos2 α1 (α2 − α)cosα
− tj dα)
− αis
2 EL[(α2 − α)cosα + sinα − ]
2Rb
It should be noted that only the upper limit of i changed comparing with the second stage. So far, the TVMS against arbitrary defect
spatial configuration is formulated.

9
X. Yang et al. Mechanical Systems and Signal Processing 218 (2024) 111564

Table 2
Compound faults TVMS function of healthy components: cracking defect *.
Stage 1 Stage 2
⃒ ⃒ i ⃒ ⃒ i
1 1 ⃒⃒α2 ∑mc ∑ki Ci 1 1 ⃒⃒− αs2 1 1 1 ⃒⃒α2 ∑m ∑ki Ci 1 1 ⃒⃒− αs2 1
= − i=1 j=1 (1 − ) + (57) = − i=1 j=1 (1 − ) + (59)
kbq1 kb1 ⃒− α1 L hx3j kb1 ⃒− αis kbr kbq2 kb1 ⃒− α1 L hx3j kb1 ⃒− αis kbr
1 1
⃒ ⃒ i ⃒α2 ⃒− αis
1 1 ⃒⃒α2 ∑mc ∑ki Ci 1 1 ⃒⃒− αs2 1 1 1 ⃒⃒ ∑m ∑ki Ci 1 1 ⃒⃒ 2 1
= − i=1 j=1 (1 − ) + (58) = − i=1 j=1 (1 − ) + (60)
ksq1 ks1 ⃒− α1 L hxj ks1 ⃒− αis ksr ksq2 ks1 ⃒− α1 L hxj ks1 ⃒− αis ksr
1 1

Fig. 4. Framework of gear compound faults modeling.

3. Framework of gear compound fault modeling

In this section, a universal gear compound faults modeling framework is elaborated, as shown in Tables 1 and 2 and Appendix 2.
Comparing Eqs. (12)–(15) with Eqs. (31)–(38) and Eqs. (44)–(47), it can be found that there exists non-linear relationship between the
healthy and fault model. Now we try to illustrate the relationship by functional expressing fault model with healthy model
components.
Firstly, for bending stiffness, the Eq. (6) could be modified to a function of separate components in Eq. (1), i.e.,
⃒α mc ∑ki ⃒− αis
1 L 1 ⃒⃒ 2 ∑ Bi 1 1 ⃒⃒ 2 1
= − (1 − ) ⃒ + (48)
kbs1 L − Bmc kb1 − α1
⃒
i=1 j=1
L hxj kb1 −
3
αis kbr
1

10
X. Yang et al. Mechanical Systems and Signal Processing 218 (2024) 111564

Fig. 5. 3D model of compound faults gear tooth and their perspective views. (a) Pitting and spalling compound faults; (b) spatial cracks compound
faults; (c) pitting and spatial crack compound faults; (d) pitting and spalling coupled with different spatial cracks compound faults.

where kbr is the stiffness between basic circle and tooth root circle, kb1 is the stiffness calculated based on basic circle. Besides, we
2h− t
defined hxj as effective tooth depth coefficient, and hxj = 2h j . It is clear that hxj represents the proportion of effective tooth depth.
The framework of gear TVMS modeling aginst arbitrary spatial configuration of defect is shown in Fig. 4

11
X. Yang et al. Mechanical Systems and Signal Processing 218 (2024) 111564

Fig. 6. TVMS of compound faults gear: (a) pitting coupled with spalling compound faults; (b) spatial cracks compound faults; (c) pitting and spatial
crack compound faults; (d) pitting and spalling coupled with different spatial cracks compound faults.

4. Model validation

The model proposed in present work address irregular shapes of defects. In order to depict complex shape of defects, three-
dimensional modeling software is combined to model. We firstly use Solidworks 3D modeling software to model complex charac­
teristics of defects, and then take advantages of the Application Program Interface (API) macro-process technique to secondary develop
SW, which enables us to acquire the 3D coordinates of defect shape and position. In this section, four types of compound faults
modeling are established, including compound surface defects, compound cracking defects, compound surface and cracking defects
without defect overlap with each other and compound surface and cracking defects with defects overlap with others.
A pair of involute spur gears is used for modeling, and the geometrical parameters are listed in Appendix 1. Fig. 5 shows the 3D
model of gear faults. Pitting coupled with spalling compound faults is illustrated in Fig. 5(a)–(d) demonstrates the spatial cracks
compound faults, pitting and spatial crack compound faults, pitting and spalling coupled with different spatial cracks compound faults,
respectively.
As we derived in Section 3, the relative position between the surface defects and cracks is quite important for the computation of
TVMS. Before the compute steps, it is vital to clarify whether the current defects are located within certain crack limit line. There exist
various appearances of compound faults. The designed specific relative positions that presented in Fig. 5 may reflect the certain
scenario of real world insufficiently. Even so, the shape characteristics of pitting, spalling and cracks are designed veritably according
to real world scenarios [33–37]. The pitting and spalling faults are generated by approximating the actual shape and position. In
existing studies, it is found that the crack propagation paths tend to be smooth, continuous, and rather straight with only a slight
curvature [38]. Hence, the crack curve is spline-driven continuous smooth curve and he crack surface is smooth and continuously
differentiable.

12
X. Yang et al. Mechanical Systems and Signal Processing 218 (2024) 111564

Fig. 7. Comparison of different types of compound faults: (a) overall view; (b) zoomed view 1; (c) zoomed view 2; (d) zoomed view 3.

13
X. Yang et al. Mechanical Systems and Signal Processing 218 (2024) 111564

Fig. 8. Finite element model of compound faults gear pair:(a) overall view; (b) pitting coupled with spalling compound faults; (c) spatial cracks
compound faults; (d) pitting and spatial crack compound faults; (e) pitting and spalling coupled with different spatial cracks compound faults.

Using the traversal technology of Solidworks API, the 3D surface points coordinate data could be exported and stored in.txt format
files, which is convenient for simulation numerical solving.

4.1. TVMS comparison

The comparison between proposed models and normal results are demonstrated in Fig. 6. An entire process of engagement of a
certain gear tooth is depicted. The normal stiffness is displayed as blue line and the fault stiffness are presented as colorful line. It
should be noticed that the calculation method consider the elastic deformation of gear, and there exists a transition process between
single and double meshing zone instead of abrupt transformation. The detailed method can be found in [31]. From Fig. 6(a), it is
observed that the stiffness declines suddenly whether pitting or spalling occurs, which is due to the abrupt change of the contact
meshing line length. Nevertheless, Fig. 6(b) shows that the stiffness decreases smoothly when crack occurs. This is because that the
generation of crack start from the surface of gear, furthermore, the crack curved surfaces are also smooth and continuously differ­
entiable. From Fig. 6(c) it could be found that the reduction of stiffness independent with each other when there is no overlap part

14
X. Yang et al. Mechanical Systems and Signal Processing 218 (2024) 111564

Fig. 9. Comparison of TVMS by proposed method and FEM method: (a) pitting coupled with spalling compound faults; (b) spatial cracks compound
faults; (c) pitting and spatial crack compound faults; (d) pitting and spalling coupled with different spatial cracks compound faults.

among defects. While overlaps exist, the stiffness in defect area coupled with each other and is dominated by the largest size of defect.
Particularly, the stiffness further reduces markedly on the meshing position of surface defects based on the domination stiffness. Fig. 7
illustrates the comparison and detail views of all types of compound faults results and further clarifies the phenomenon stated above.

4.2. FEM validation

A FEM model is established to validate the mesh stiffness equations derived by the proposed analytical model. As shown in Fig. 8(a),
in order to improve the computational efficiency, only several teeth are established in the finite element model. Due to irregular shapes
of defects, it is unable to meshing the model by structure hexahedron elements. Tetrahedron elements are used in FEM models. To
tackle above limitation as much as possible, we increase the element seed density on defect areas to improve the calculation precision
as shown in Fig. 8(b)–(e).
Fig. 9 shows the comparison results of TVMS obtained by FEM and proposed method of the axis-fixed gear tooth pair in the first
stage. From Fig. 9, it shows that the stiffness results obtained from the proposed method and the FEM match well, but errors exist. The
errors are probably related to the decrease of computational accuracy utilizing tetrahedron elements, hardware limitation, etc. In
general, the error of the two methods is small, which verifies the effectiveness and accuracy of proposed model.

5. Stiffness sensitivity analysis

In this section, the stiffness sensitivity analysis of the proposed model will be further studied. By observing the formalization of the

15
X. Yang et al. Mechanical Systems and Signal Processing 218 (2024) 111564

Fig. 10. Sensitivity of TVMS to defects extension along tooth width and tooth thickness direction.

16
X. Yang et al. Mechanical Systems and Signal Processing 218 (2024) 111564

Fig. 11. Comparison of mean stiffness sensitivity ηm s d in different defect geometrical features of: (a) defect 1 increases along tooth thickness and
tooth width, respectively; (b) defect 2 increases along tooth thickness and tooth width, respectively.

stiffness model in section 3, it can be found that after the different types are unified, the stiffness decline of a certain meshing position
can be quantitatively analyzed. In detail, the defect parameters in a certain direction of the tooth width, tooth thickness and meshing
position are fixed, and the stiffness sensitivities to defects extension along the tooth width and tooth thickness are analyzed. Two
defects are set to help analysis. The first one occupies 1/6–1/3 of the whole meshing duration and the second one occupies 1/2–3/4 of
the whole meshing duration.

5.1. Stiffness sensitivity to the extension of defects width and thickness

Firstly, we fix the width of defects as 20 %, 40 %, 60 % of the whole tooth width, and the stiffness reduction sensitivity to defects
extension along tooth thickness direction obtained as orange line shown in Fig. 10. While the depth of defects as 20 %, 40 %, 60 % of
the whole tooth thickness, and the stiffness reduction sensitivity to defects extension along tooth width direction obtained as blue line
shown in Fig. 10. Fig. 10(a), (c), (e) show the results of the end position of defect 1 and Fig. 10(b), (d), (f) shows the results of the end
position of defect 2.
The results of two types of cases are compared, as shown in the Fig. 10. It can be seen that defect thickness extension decreases the
gear stiffness more than defect width extension in the slight degree of the defect occurrence. As the defect develops to a certain degree,
the influence of defect width extension on the stiffness decline gradually increases. At this stage, defect width extension decreases the

17
X. Yang et al. Mechanical Systems and Signal Processing 218 (2024) 111564

Fig. 12. Sensitivity of TVMS to defects extension path on various meshing positions.

gear stiffness more than defect thickness extension. Specifically, as shown in Fig. 10(b), (d), (f), when defect 2 extends to the severe
degree of the whole extension process, defect thickness extension decreases the gear stiffness more than defect width extension again.
Besides, it is necessary to express the sensitivity of different types of gear fault to stiffness in comprehensive ways. In this section,
the effects of gear defects to system stiffness are expressed both by specific decreased TVMS value, and the stiffness sensitivity co­
efficients ηm ins
s d and ηs d . The method used for stiffness sensitivity analysis proposed by [40] is referred here for illustration.

m(Khw ) − m(Kcw d ) m(Kcw d )

ηms d = = 1− (61)
m(Kw )
h m(Khw )

ins(Khw ) − ins(Kcw d ) ins(Kcw d )

ηins
s d = = 1− (62)
ins(Kw )
h ins(Khw )

where Khw is the comprehensive TVMS of healthy gear obtained by improved potential energy method, and Kcw d is the comprehensive
TVMS of defect gear. m(x) is defined to calculate the mean value of one whole meshing period of TVMS, while ins(x) is the instan­
taneous value of TVMS in certain moment during meshing process. Thus, the mean stiffness sensitivity coefficient ηm s d and the
instantaneous stiffness sensitivity coefficient ηins
s d are defined, which represent the change ratio of mean stiffness and instantaneous
stiffness to health gear affected by gear defects.
The results of ηms d is shown in Fig. 11. See Fig. 11, the black, blue and red line represent the width and thickness of tooth that are
fixed as 20 %, 40 %, and 60 % of the whole width and thickness on meshing position, corresponding to the results shown in Fig. 10.
Stiffness sensitivity to different defect extension path is illustrated below. As shown in Fig. 11, while defect extending along the

18
X. Yang et al. Mechanical Systems and Signal Processing 218 (2024) 111564

Fig. 13. Comparison of instantaneous stiffness sensitivity ηms d on various meshing positions of: (a) defect 1 increases along tooth width; (b) defect 1
increases along tooth thickness, respectively; (c) defect 2 increases along tooth width; (b) defect 2 increases along tooth thickness.

tooth width direction gradually, the stiffness decreases gently, and the slope of stiffness presents a decrease first and increase then
performance. While as defect extending along the tooth thickness direction gradually, the slope of stiffness also presents a decrease first
and increase then performance. However, comparing with each other, it can be found that defect extending along the tooth thickness
exists a distinct inflection point. When defects thickness extension does not come into the inflection point, the slope of stiffness in­
creases rather slowly. After defects thickness extension goes through the inflection point, the slope of stiffness suddenly increases and
the reduction of which growing even fast.
Hence, the performance of stiffness sensitivities to different defects extension types shows that the extension of defects along tooth
width direction is generally a gradual changing process, while the extension of defects along tooth thickness direction presents distinct
two stages. In the first stage, the defects develop slowly along the tooth thickness direction and the extension along tooth width di­
rection dominate the development of defects. While when the process of defects thickness growth gets through the inflection point, the
defects develop along the tooth thickness direction accelerate abruptly and dominate the development of defects in this stage.

5.2. Influence of meshing position

It can be observed that the appearance of the inflection point performs differently on different meshing position with comparison to
Fig. 10(a)–(f). It inspires us to investigate the influence of meshing position to the stiffness sensitivity. Therefore, the stiffness sen­
sitivities under different meshing positions are further depicted in this section. The results are shown in Fig. 12.
As shown in Fig. 12, we demonstrate the results of 20 %, 40 %, 60 %, 80 % and 100 % position of the defects meshing process. The
results of ηins
s d is shown in Fig. 13, which is the instantaneous stiffness sensitivity corresponding to the results shown in Fig. 12. It can be

19
X. Yang et al. Mechanical Systems and Signal Processing 218 (2024) 111564

found in every figure that the slope of stiffness performs flatter with gear mesh progressing, i.e., the curvature of curve reduces, which
can be both observed in two direction analysis. Therefore, the development of defects near the tooth root grows more rapidly compared
with the tooth tip especially in the end phase of extension. Furthermore, it can be obtained from Fig. 13(b) and (d) that the inflection
point appears earlier in entire defects extension process with the increasing of meshing progress. That is to say, the gear tooth meshing
progress more, the earlier the defect thickness extension dominate the development of faults. The variation of magnitude and trend of
stiffness sensitivity remains is in accordance with the analysis conducted in Section 5.1, which further validates the given explanations.

6. Conclusion

A gear compound faults TVMS model against arbitrary defect spatial configuration is proposed in this work. Based on this, the
universal framework of compound faults gear modeling is conducted. 3D models of different types of compound gear faults are built to
gain the simulation results, which is further compared with the results of the FEM model for validation Then, the relation of stiffness
sensitivity to defect extension path is conducted. The sensitivity of TVMS to the defect extension path along gear tooth width, thickness
and meshing path are analyzed finally.
The results of stiffness sensitivity analysis show that: 1) the extension of defects along tooth thickness exists an inflection point
distinct from tooth width extension. 2) During earlier stage of extension, the defects develop slowly along the tooth thickness direction
and the extension along tooth width direction dominate the development of defects. After the inflection point, the defects develop
along the tooth thickness direction accelerate abruptly and dominate the development of defects. 3) The development of defects near
the tooth root grows more rapidly compared with the tooth tip especially in the end phase of extension. 4) The more the gear tooth
meshing progresses, the earlier the defect thickness extension dominate the development of faults. The above analysis would be helpful
to the conduction of gear degeneration model and is able to further provide theory foundation to degradation mechanism of machinery
system.
In the future, the proposed method is to be extended to the modeling of internal gear like the ring gear in planetary gear system.
And the compound faults model will be applied to complex multistage gear systems to address the coupled dynamic behaviors due to
compound faults. Various compound faults will be applied to different stages and different gear positions. In that case, the coupling
effect of faults between different stages is expected to be found and discussed, providing guidance for condition monitoring and fault
diagnosis of compound multistage transmission systems.

CRediT authorship contribution statement

Xiao Yang: Writing – review & editing, Writing – original draft, Methodology, Conceptualization. Xiang Li: Writing – review &
editing, Supervision. Yaguo Lei: Writing – review & editing, Supervision, Conceptualization. Bin Yang: Writing – review & editing,
Supervision. Huan Liu: Writing – review & editing, Methodology, Conceptualization. Xuanyu Gao: Writing – review & editing.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to
influence the work reported in this paper.

Data availability

The data that has been used is confidential.

Acknowledgements

This research was supported by the National Science Fund for Distinguished Young Scholars (52025056), SanQin Scholar Inno­
vation Team, Shaanxi Science and Technology Innovation Team (2023-CX-TD-15) and the Fundamental Research Funds for the
Central Universities.

Appendix 1. Key parameters of gear pair in case study

Parameter Value Parameter Value Parameter Value Parameter Value

m 1.5e-3 Z1 20 L 1e-2 c* 1
α π/9 Z2 40 h*a 1 E 2.1e11

Appendix 2. Variants of compound fault modeling framework

(1) Variant of equations (57)–(58)

20
X. Yang et al. Mechanical Systems and Signal Processing 218 (2024) 111564

⃒α mc ∑ki ⃒− αis
1 L 1 ⃒⃒ 2 ∑ Ci 1 1 ⃒⃒ 2 1
= − (1 − ) ⃒ +
kbq1 L − Bmc kb1 − α1
⃒
i=1 j=1
L hxj kb1 −
3
αis kbr
1

⃒α mc ∑ki ⃒− αis
1 L 1 ⃒⃒ 2 ∑ Ci 1 1 ⃒⃒ 2 1
= − (1 − ) ⃒ +
ksq1 L − Bmc ks1 − α1
⃒
i=1 j=1
L hxj ks1 −
3
αis ksr
1

(2) Variant 1 of equations (49)–(52)

⃒α mc ∑ki ⃒− αis mc ∑ki ⃒− αis
1 L 1 ⃒⃒ 2 ∑ Bi 1 1 ⃒⃒ 2 ∑ Ci 1 1 ⃒⃒ 2 1
= − (1 − ) − (1 − ) +
kbs1 L − Bmc kb1 ⃒− α1 i=1 j=1
L hx3j kb1 ⃒− αis i=1 j=1
L hx3j kb1 ⃒− αis kbr
1 1

⃒α mc ∑ki ⃒− αis ki
mc ∑ ⃒− αis
1 L 1 ⃒⃒ 2 ∑ Bi 1 1 ⃒⃒ 2 ∑ Ci 1 1 ⃒⃒ 2 1
= − (1 − ) − (1 − ) ⃒ +
kss1 L − Bmc ks1 ⃒− α1 i=1 j=1
L hxj ks1 ⃒− αis i=1
L hxj ks1
j=1 − αis ksr
1 1

⃒α mc ∑ki ⃒− αis
1 L 1 ⃒⃒ 2 ∑ Bi 1 1 ⃒⃒ 2 1
= − (1 − ) ⃒ +
kas1 L − Bmc ka1 − α1
⃒
i=1 j=1
L hxj ka1 − αis kar
1

1 L 1
=
khs1 L − Bmc kh1

(3) Variant 2 of equations (49)-(52)

⃒α mc ∑ki ⃒− αis
1 L 1 ⃒⃒ 2 ∑ Bimax 1 1⃒ 2 1
= − (1 − ) ⃒⃒ +
kbs1 L − Bmc kb1 − α1
⃒
i=1 j=1
L hxjmax kb1 −
3
αis kbr
1

⃒α mc ∑ki ⃒− αis
1 L 1 ⃒⃒ 2 ∑ Bimax 1 1⃒ 2 1
= − (1 − ) ⃒⃒ +
kss1 L − Bmc ks1 − α1
⃒
i=1 j=1
L hxjmax ks1 − αis ksr
1

⃒α mc ∑ki ⃒− αis
1 L 1 ⃒⃒ 2 ∑ Bi 1 1 ⃒⃒ 2 1
= − (1 − ) +
kas1 L − Bmc ka1 ⃒− α1 i=1 j=1
L hxj ka1 ⃒− αis kar
1

1 L 1
=
khs1 L − Bmc kh1

(4) Variant 1 of equations (53)–(56)

⃒α mc ∑ki ⃒− αis mc ∑ki ⃒− αis
1 1 ⃒⃒ 2 ∑ Bi 1 1 ⃒⃒ 2 ∑ Ci 1 1 ⃒⃒ 2 1
= − (1 − ) − (1 − ) +
kbs1 kb1 ⃒− α1 i=1 j=1
L hx3j kb1 ⃒− αis i=1 j=1
L hx3j kb1 ⃒− αis kbr
1 1

⃒α mc ∑ki ⃒− αis mc ∑ki ⃒− αis

1 1⃒2 ∑ Bi 1 1 ⃒⃒ 2 ∑ Ci 1 1 ⃒⃒ 2 1
= ⃒⃒ − (1 − ) ⃒ − (1 − ) ⃒ +
kss1 ks1 − α1 i=1 j=1
L hxj ks1 − αis i=1 j=1
L hxj ks1 − αis ksr
1 1

⃒α mc ∑ki ⃒− αis
1 1 ⃒⃒ 2 ∑ Bi 1 1 ⃒⃒ 2 1
= − (1 − ) ⃒ +
kas1 ka1 − α1
⃒
i=1 j=1
L hxj ka1 − αis kar
1

1 1
=
khs1 kh1

(5) Variant 2 of equations (53)-(56)

⃒α mc ∑ki ⃒− αis
1 1 ⃒⃒ 2 ∑ Bimax 1 1⃒ 2 1
= − (1 − ) ⃒⃒ +
kbs1 kb1 − α1
⃒
i=1 j=1
L hxjmax kb1 −
3
αis kbr
1

21
X. Yang et al. Mechanical Systems and Signal Processing 218 (2024) 111564

⃒α mc ∑ki ⃒− αis
1 1⃒2 ∑ Bimax 1 1⃒ 2 1
= ⃒⃒ − (1 − ) ⃒⃒ +
kss1 ks1 − α1 i=1 j=1
L hxjmax ks1 − αis ksr
1

⃒α mc ∑ki ⃒− αis
1 1 ⃒⃒ 2 ∑ Bi 1 1 ⃒⃒ 2 1
= − (1 − ) ⃒ +
kas1 ka1 − α1
⃒
i=1 j=1
L hxj ka1 − αis kar
1

1 1
=
khs1 kh1

References

[1] Z. Chen, W. Zhai, K. Wang, Vibration feature evolution of locomotive with tooth root crack propagation of gear transmission system, Mech. Syst. Sig. Process.
115 (2019) 29–44.
[2] X. Liang, M.J. Zuo, Z. Feng, Dynamic modeling of gearbox faults: a review, Mech. Syst. Sig. Process. 98 (2018) 852–876.
[3] Z. Chen, J. Ning, K. Wang, et al., An improved dynamic model of spur gear transmission considering coupling effect between gear neighboring teeth, Nonlinear
Dyn. 106 (2021) 339–357.
[4] Y. Lei, J. Lin, M.J. Zuo, et al., Condition monitoring and fault diagnosis of planetary gearboxes: a review, Measurement 48 (2014) 292–305.
[5] H. Ma, J. Zeng, R. Feng, et al., Review on dynamics of cracked gear systems, Eng. Fail. Anal. 55 (2015) 224–245.
[6] F. Chaari, T. Fakhfakh, M. Haddar, Dynamic analysis of a planetary gear failure caused by tooth pitting and cracking, J. Fail. Anal. Prev. 6 (2006) 73–78.
[7] X. Liang, H. Zhang, L. Liu, et al., The influence of tooth pitting on the mesh stiffness of a pair of external spur gears, Mech. Mach. Theory 106 (2016) 1–15.
[8] Y. Lei, Z. Liu, D. Wang, et al., A probability distribution model of tooth pits for evaluating time-varying mesh stiffness of pitting gears, Mech. Syst. Signal Process.
106(jun.) (2018) 355–366. DOI:10.1016/j.ymssp.2018.01.005.
[9] A.F. Rincon, F. Viadero, M. Iglesias, et al., Effect of cracks and pitting defects on gear meshing, Proc. Inst. Mech. Eng. C J. Mech. Eng. Sci. 226 (11) (2012)
2805–2815.
[10] F.S. Meng, G.A novel evolution model of pitting failure and effect on time-varying meshing stiffness of spur gears, Eng. Fail. Anal. 120 (1) (2021).
[11] F. Chaari, W. Baccar, M.S. Abbes, et al., Effect of spalling or tooth breakage on gearmesh stiffness and dynamic response of a one-stage spur gear transmission,
Eur. J. Mech.-A/solids 27 (4) (2008) 691–705.
[12] R. Ma, Y. Chen, Q. Cao, Research on dynamics and fault mechanism of spur gear pair with spalling defect, J. Sound Vib. 331 (9) (2012) 2097–2109.
[13] S.A. Abouel-seoud, E.S. Dyab, M.S. Elmorsy, Influence of tooth pitting and cracking on gear meshing stiffness and dynamic response of wind turbine gearbox,
Int. J. Sci. Adv. Technol 2 (3) (2012) 151–165.
[14] C. Zhe, H. Niaoqing, G. Fengshou, et al., Pitting damage levels estimation for planetary gear sets based on model simulation and grey relational analysis, Trans.
Can. Soc. Mech. Eng. 35 (3) (2011) 403–417.
[15] K. Chen, H. Ma, L. Che, et al., Comparison of meshing characteristics of helical gears with spalling fault using analytical and finite-element methods, Mech. Syst.
Sig. Process. 121 (2019) 121, https://doi.org/10.1016/j.ymssp.2018.11.023.
[16] L. Xiang, C. An, Y. Zhang, et al., Failure dynamic modelling and analysis of planetary gearbox considering gear tooth spalling, Eng. Fail. Anal. (2021), https://
doi.org/10.1016/j.engfailanal.2021.105444.
[17] Y. Luo, N. Baddour, M. Liang, A shape-independent approach to modelling gear tooth spalls for time varying mesh stiffness evaluation of a spur gear pair, Mech.
Syst. Signal Process. 120(APR.1) (2019) 836–852. DOI:10.1016/j.ymssp.2018.11.008.
[18] Z. Chen, Y. Shao, Dynamic simulation of spur gear with tooth root crack propagating along tooth width and crack depth, Eng. Fail. Anal. 18 (8) (2011)
2149–2164.
[19] Z. Chen, Y. Shao, Dynamic simulation of planetary gear with tooth root crack in ring gear, Eng. Fail. Anal. 31 (2013) 8–18.
[20] O.D. Mohammed, M. Rantatalo, J.-O. Aidanpää, Improving mesh stiffness calculation of cracked gears for the purpose of vibration-based fault analysis, Eng.
Fail. Anal. 34 (2013) 235–251.
[21] C. Hou, Y. Lu, Identification of cracks in thick beams with a cracked beam element model, J. Sound Vib. 385 (2016) 104–124.
[22] S. Christides, A. Barr, One-dimensional theory of cracked Bernoulli-Euler beams, Int. J. Mech. Sci. 26 (11–12) (1984) 639–648.
[23] S.H. Carneiro, D.J. Inman, Continuous model for the transverse vibration of cracked Timoshenko beams, J. Vib. Acoust. 124 (2) (2002) 310–320.
[24] L. Zongyao, Mechanism Research on Typical Faults and Phenomenological Models of Vibration of Planetary Gear Trains, Xi’an Jiaotong University, Xi’an, 2020.
[25] W. Yu, Y. Shao, C.K. Mechefske, The effects of spur gear tooth spatial crack propagation on gear mesh stiffness, Eng. Fail. Anal. 54 (2015) 103–119.
[26] Z. Chen, W. Zhai, Y. Shao, et al., Analytical model for mesh stiffness calculation of spur gear pair with non-uniformly distributed tooth root crack, Eng. Fail.
Anal. 66 (2016) 502–514.
[27] L. Wang, Y. Shao, Fault mode analysis and detection for gear tooth crack during its propagating process based on dynamic simulation method, Eng. Fail. Anal.
71 (2017) 166–178.
[28] T. Duan, J. Wei, Q. Yan, et al., Detecting the 3D spatial varying crack evolution-induced vibration of gearbox through a system level rigid-flexible coupling
model, in: Mechanism and Machine Theory: Dynamics of Machine Systems Gears and Power Trandmissions Robots and Manipulator Systems Computer-Aided
Design Methods, 2022, p. 174.
[29] T. Duan, J. Wei, Q. Yan, et al., Investigations on crack propagation and meshing characteristics of planetary gear train considering crack closure effect, Eng.
Failure Anal. 134 (2022) 106064. DOI:10.1016/j.engfailanal.2022.106064.
[30] X. Liang, M.J. Zuo, T.H. Patel, Evaluating the time-varying mesh stiffness of a planetary gear set using the potential energy method, Proc. Inst. Mech. Eng. C J.
Mech. Eng. Sci. 228 (3) (2014) 535–547.
[31] X. Yang, Y. Lei, H. Liu, et al., Rigid-flexible coupled modeling of compound multistage gear system considering flexibility of shaft and gear elastic deformation,
Mech. Syst. Sig. Process. 200 (2023) 110632.
[32] Y. Sun, H. Ma, Y. Huangfu, et al., A revised time-varying mesh stiffness model of spur gear pairs with tooth modifications, Mech. Mach. Theory 129 (2018)
261–278.
[33] Y. Yang, J. Tang, N. Hu, et al., Research on the time-varying mesh stiffness method and dynamic analysis of cracked spur gear system considering the crack
position, J. Sound Vib. 548 (2023) 117505.
[34] Z. Li, H. Ma, M. Feng, et al., Meshing characteristics of spur gear pair under different crack types, Eng. Fail. Anal. 80 (2017) 123–140.
[35] X. Liu, N. Li, Y. Lei, et al., Optimal weight impulse extraction: new impulse extraction methodology for incipient gearbox condition monitoring. Available at
SSRN 4695331.
[36] F. Huang, L. Cheng, L. Tang, H. Yang, Analysis on surface crack of 20CrMnMo carburized heavy-duty gear steel, Heat Treat. Met. 35 (2010) 110–113 (in
Chinese).

22
X. Yang et al. Mechanical Systems and Signal Processing 218 (2024) 111564

[37] Failure analysis gears-shafts-bearings-seals, Failure Analysis • Installation & Maintenance, Rexnord Industries, LLC, Gear Group, 1978, pp. 1–20.
[38] D.G. Lewicki, Gear crack propagation path studies-guidelines for ultra-safe design, J. Am. Helicopter Soc. 47 (1) (2002) 64–72.
[39] C. Xie, L. Hua, X. Han, J. Lan, X. Wan, X. Xiong, Analytical formulas for gear body-induced tooth deflections of spur gears considering structure coupling effect,
Int. J. Mech. Sci. 148 (2018) 174–190.
[40] H. Yang, W. Shi, L. Guo, et al., Study on mesh stiffness and sensitivity analysis of planetary gear system considering the deformation effect of carrier and bearing,
Eng. Fail. Anal. 135 (2022) 106146.

23