Archive of Applied Mechanics 74 (2005) 489--508 © Springer-Verlag 2005

DOI 10.1007/s00419-004-0363-4

Model of contact between rollers and sprockets

in chain-drive systems
S. L. Pedersen

489
Summary A model of a roller-chain drive is developed and applied to the simulation and
analysis of roller-chain drives of large marine diesel engines. Two different ways of modelling
the contact between the rollers and sprockets are presented; one using a circular shaped tooth
profile and the other using the shape of a real tooth profile. The main components of the
roller-chain drive model include the sprockets with different sizes and the chain made of rollers
and links, which are represented by rigid bodies, mass particles and spring-damper assemblies,
respectively. The contact between the rollers and the sprockets are represented by a continuous
contact force. The models proposed effectively represent the polygonal effect always present in
this type of drive. The numerical results from the simulations are compared with analytical
results, for simplified models. The model with a real tooth profile proves superior to the model
with a circular tooth profile.

1
Introduction
The work presented is a further development of the methods used in [11]. The purpose is to
suggest a model that improves the simulation-based prediction of roller-chain forces. In
particular, the model should improve numerical stability, link engagement with sprockets and
the contact forces between rollers and sprockets.
The dynamics of the roller-chain drives are characterized by a complex behavior with impacts
between the chain links and sprockets, and by discontinuities in the system components’
velocities giving rise to transversal and longitudinal vibrations of the spans of the chain. These
events are the factors responsible for part of the noise presented by mechanical devices that use
roller chains and ultimately by the wear of the roller-chain drives. Though roller chains have
been used for a long time as a reliable mechanical component to transmit power and to handle
materials mechanically, most studies of their dynamical behavior were done in the last decades,
see e.g. [4, 5, 12, 14, 17, 18]. For earlier work, see e.g. [2, 9]. The main reason for this situation is
that the dynamic behavior is very complex, making it impossible to find general analytical
procedures able to describe the problem thoroughly. With the development of fast computers,
some recent efforts have been put forward in order to better understand different aspects of
these mechanical components, see e.g. [17]. In a review of the state-of-the-art in [18], many of
the investigations that have been carried out are summarized and it states that integrated models
describing the full dynamics of the system are necessary, in order to accurately describe the
interrelation between the different elements in the chain drive system.
The major difficulties in the study of the roller-chain drives are related to the roller-chain
wrapping around the sprockets, which forms a polygon. This effect, called polygon action,
together with the impact between rollers and sprockets are responsible for the noise and

Received 12 June 2004; accepted for publication 14 October 2004

S. L. Pedersen (✉)
Department of Mechanical Engineering, Solid Mechanics,
Technical University of Denmark,
Nils Koppels Allè, Building 404,
2800 Kgs.Lyngby, Denmark
e-mail: slp@mek.dtu.dk
The applications represent the ongoing work that resulted from many
fruitful discussions with Per Rønnedal, Mikkel Preem et al. of MAN
B&W Diesel A/S, Copenhagen, Denmark.
 vibration of the roller-chain drive. The influence on the system dynamics of the impulsive forces
that act on the rollers and sprockets at the moment of their engagement, which results in a
discontinuity of the roller velocity has been the subject of different investigations, see e.g. [3, 16].
The combined action of the polygonal effect and the roller impact leads to the development of
longitudinal and transverse vibrations of the chain strands in which the flexibility of the links
plays an important role.
In [17], a discrete model is proposed, which takes into account the polygonal action. The
driving span of the chain is modeled by lumped masses connected by linear springs and this
model considers the coupling between the longitudinal and transversal vibrations and the
moving boundary conditions. However, the effect of chain guides in the drive and the angular
speed fluctuation of the driving sprocket are not considered. In [5], a methodology is presented,
490 where the roller–sprocket and the guide-chain contacts are treated as unilateral constraints, see
also [13].
In the present study, two different models of contact between the rollers and sprockets are
proposed and compared with analytical results, for simplified models, see [1]. The roller-chain
drive model has one driving sprocket and one or more driven sprockets. The complete chain is
modeled by lumped masses connected by spring-damper uniaxial elements. The methodology
proposed for the roller-sprocket contacts is penalty forces, which use the continuous force model
proposed in [8]. Through the application of this, it is shown that the interrelated dynamics of the
elements in the chain drive system is captured and the contact problem is fully characterized.
The model with a real tooth profile proves superior to the model with a circular tooth profile.
The numerical results obtained from the simulations with the two different contact models
are compared with each other and with analytical results for simplified models, whereas the
necessary experimental validation is left for future studies.

2
Model
2.1
The model of the roller-chain drive
A typical roller-chain drive is composed of one chain that wraps around two or more sprockets,
see Fig. 1. The roller-chain is made of alternating inner and outer links assembled in pivots by
bearing pins and bushes. For the purposes of the study presented here, the mass of the chain is
assumed to be lumped at the roller locations, see Fig. 2. Springs and dampers with constant
stiffness and damping coefficients model the links. In this model, the clearances between pin and
bushing are neglected as well as the rotational inertia of the rollers about their center of gravity.
Figure 2a (from [15]) shows the real links, with alternating inner and outer links. The inner
and outer links are assembled in pivots by bearing pins and bushes. Figure 2b shows the

Fig. 1. A roller-chain drive system with four

sprockets
Fig. 2. The roller links in a the reality (figure from [15]) and b the simplified model

simplified model of the chain. The initial distance between the rollers, when there is no tension 491
in the spring, is referred to as the chain pitch P.
When the chain is lying on a sprocket, it forms polygons instead of circles, which causes a
periodic fluctuation of linear velocity together with periodic transverse displacements of both
ends of the chain. This is known as the polygonal action or polygonal effect, see e.g. [2]. Due to
the polygonal effect, the velocities of the driven sprockets will fluctuate even if the velocity of the
driving sprocket is constant. The polygonal effect is thus responsible for the transverse and
longitudinal vibrations that develop in the chain. The excitation resulting from the impact of the
roller when it seats on the sprocket and the polygonal effect are the responsible parts for the
noise and vibration of the roller-chain drive. In order to have a useful model of the vibration
behavior of the chain and the tension fluctuation, special attention must be paid to a correct
engagement and disengagement description.
The model developed here is in 2D and both the tight sides and the slack sides of the chain are
represented. No assumption is made that limits the number of sprockets in the drive, the
minimum number of teeth engaged at a certain time or the variance of the angular velocity of
the driving sprocket. Moreover, the location of the center of the sprockets can be fixed or be time
dependent. This allows the introduction of active tensioners in the roller-chain drive. Let the
roller chain be modeled as a system of particles connected by translational springs and dampers.
The rollers that are not seated in the sprockets constitute a particle system free in space. The
equations of motion of the rollers are

mr q̈r = ƒr , (1)

where mr is the mass of a link and q̈r is the accelerations of the rollers. The right-hand side ƒr is
the vector with the forces acting in the individual chain links due to the flexibility of the links.
For the rollers in contact with a sprocket the equations of motion are also given by Eq. (1), but
with the contact forces added to the right hand side.
Each sprocket has an individual number of teeth nt , a pitch circle radius Rs and a pitch angle

= 360◦ /nt , see Fig. 3. The local coordinate system (s , s ) is placed in the center of the sprocket,
with the s axis going through the lowest part of a tooth and it follows the sprocket when it

Fig. 3. The numerical model of the

sprocket and its parameters
 rotates. The rotation of the local coordinate system (s , s ), with respect to the global coordinate
system (x, y) is given by the angle s . For a sprocket having the external forces ƒs , made up from
the forces and moment with which the chain affects the sprocket and from optional directly
applied forces and/or moment, the equations of motions are

Ms q̈s = ƒs , (2)

where matrix Ms is the mass matrix including the mass moment of inertia and q̈s contains the
accelerations.

2.2
492 The continuous contact force method
A procedure to describe the capture of the roller by a sprocket is to use a non-linear force contact
model that can take into account the impact that occurs during engagement. Based on the
Hertzian contact theory, see e.g. [6], several models have been proposed, which take into account
the energy dissipation due to impact, friction, vibrations or localized deformations. In [8], an
impact–contact model that provides an accurate description of the physical phenomena and is
numerically stable, has been proposed. The contact model proposed here is based on the model
by Lankarani and Nikravesh and it takes into account the specifics of the roller-chain dynamics.
The contact force is given by


3(1 − e2 )v
ƒc = Kg  1+
n
n, (3)
4v (−)
where Kg is the generalized coefficient of stiffness,  is the indentation, the power n is 1.5 for
metallic surfaces, e is the coefficient of restitution, v is the relative velocity, v (−) is the relative
approach speed, (−) indicates that it is the value just as impact begins, and n is the normal unit
vector to the contact surface. For the rollers in contact with a sprocket, the contact force ƒc is
added to the right hand side of Eq. (1). The contact forces on the sprockets (−ƒc ) and the
moment caused by the contact forces are added to the right hand side of Eq. (2).

2.3
Circular tooth profile
Figure 4 shows the model of the circular tooth profile. The radius of the circular tooth profile is
Rt and the roller radius is Rr , yielding a clearance of
R = Rt − Rr . The center of the circular
tooth profile is placed Rs +
R from the center of the sprocket. This is done in order to prevent all
the rollers in contact with a sprocket from disengaging at the same time. The unit vectors ur and
ut are given by
   
cos(s ± j
) − sin(s ± j
)
ur = , ut = (4)
sin(s ± j
) cos(s ± j
)

where the ± depends on whether the sprocket rotates counterclockwise (+) or clockwise (−).
The vector from the sprocket center to the roller center is in the local coordinate system
(s , s ) given by sr , where the prime indicates local coordinate system as done in [10]. The

Fig. 4. The numerical model of the circular tooth

profile
position of the sprocket center is in the global coordinate system (x, y) given by the vector rs and
the center of the tooth profile is in the local coordinate system given by vector st . The center of
the tooth profile is, therefore, in the global coordinate system given by
rt = rs + As st , (5)
where the transformation matrix As is given by


cos(s ) − sin(s )
As = , (6)
sin(s ) cos(s )

and
  493
cos(±j
)
st = (Rs +
R) . (7)
sin(±j
)

The vector st becomes

st =As st = (Rs +
R)ur , (8)
and vector rt becomes
rt = rs + (Rs +
R)ur . (9)
The roller position is in global coordinate system (x, y) given by the vector rr . The vector etr from
the center of the tooth profile to the roller center is given by
etr = rr − rt . (10)
The pseudopenetration in the tooth is given by
 = etr  −
R , (11)
where etr  is the two norm of vector etr . The normal unit vector n to the contact surface is then
given by
etr
n=− . (12)
etr 
The point of application of the contact force on the roller is the center of the roller. The vector
from the sprocket center to the point of application of the contact force on the sprocket, is in
global coordinates given by
sps = rr − rs , (13)
and in the local sprocket coordinates Eq. (13) is given by

sps = AsT sps . (14)
The velocity of the contact point on the roller is
ṙrp = ṙr , (15)
and the velocity of the contact point on the sprocket is
p
d(rs + As ss ) 
ṙsp = = ṙs + s Bs sps , (16)
dt
where s = ˙s and the matrix Bs is given by


− sin(s ) − cos(s )
Bs = . (17)
cos(s ) − sin(s )

The relative velocity v at the contact point is given by

v = (ṙrp − ṙsp )n . (18)
 The contact force of the roller can now be found using (3), the normal vector n in equations
given in (12), the indentation  in (11) and the relative velocity v in (18). The moment on the
center of the sprocket caused by the contact force is given by
ns = −ŝps · ƒc , (19)
p
where the hat (ˆ) indicates that the vector is perpendicular to ss , rotated in the counterclockwise
direction.

2.4
Real tooth profile
494 The standard tooth form (Type II per ASA B29.1-1950) is shown in Fig. 5, taken from [1].
According to American standards, the dimensions shown in Fig. 5 are defined below (in inches).
P is the chain pitch, nt is the number of teeth, Dr = 2Rr is the roller diameter, Dt = 2Rt is the
seating curve diameter. The angles A and B in Fig. 5 are defined by
60◦
A = 35◦ +
nt
35◦ (20)
B = 18◦ −
nt

The different lengths shown in Fig. 5 are defined by

Dt = 1.005Dr + 0.003, M = 0.8Dr cos A, T = 0.8Dr sin A, RE = 1.3025Dr + 0.0015 ,

180◦ 180◦
W = 1.24Dr cos , V = 1.24Dr sin ,
nt nt
(21)
RF = Dr [0.8 cos B+ 1.24 cos C − 1.3025] − 0.0015 ,
  2 180◦ 180◦
H = RF F 2 − 1.24Dr − P2 , S = P2 cos + H sin ,
nt nt
where the angle C is given by
64◦
C = 17◦ − . (22)
nt
Separating the tooth profile into areas, we obtain seven areas where the roller–sprocket contact
can occur, see Fig. 6.
The contact areas can be defined by local vectors given in the local coordinate system (t , t ),
see Figs. 7 and 9 for the angles involved.

Fig. 5. The standard sprocket tooth

form (Type II per ASA B29.1–1950)
 495
Fig. 6. The real tooth profile separated in seven contact areas

Fig. 7. The ten reference points used

to describe the shape of the tooth
profile


Fig. 8. The local vector sƒ∗ used to
describe the shape of the tooth profile

All the points denoted with ∗ in superscript are related to the left side of the tooth and those
without it are related to the right side (relative to the t axis). The vectors in Fig. 7 are given by

sb = {W, V}T , sb ∗ = {−W, V}T , sc = {−M, T}T , sc ∗ = {M, T}T , (23)

with W, V, M, and T defined in (21). The local coordinate system placed at the center of the jth
tooth is rotated with an angle t = s − /2 ± j
with respect to the global coordinate system,
where again the ± depends on the rotational direction of the sprocket. The transformation
 Fig. 9. Angles used to describe the
shape of the tooth profile
496

matrix At from the local tooth coordinates system to the global coordinate system, is given by


cos(t ) − sin(t )
At = . (24)
sin(t ) cos(t )
The global vectors to the ten reference points shown in Fig. 7 are all found in the same way, e.g.,
the global vector rc∗ to point c ∗ is given by
rc∗ = rt + At sc∗ . (25)
The local vector sƒ ∗ shown in Fig. 8 is given by
 P    
 − 2 cos
2 − H sin
2 
sƒ∗ =     . (26)
 P
− 2 sin
2 + H cos
2
The local vector from point b∗ to point ƒ ∗ is given by
  
s ∗x
sbƒ ∗ = sƒ ∗ − sb∗ = sbƒ
 . (27)
bƒ ∗ y

The global vector to the center of the tooth is given by

rt = rs + Rs ur . (28)
Equation (28) is similar to Eq. (9), but with a difference of
R, that was added for numerical
reasons in the circular tooth profile.
The angles shown in Fig. 9 are given by
1a = + A, 2a= 2 − A, 1c = 2 − A, 2c= 2 − A+ B ,
1c∗ = + A − B, 2c∗ =+ A, 2b =  − A+ B, 1b∗ = A − B ,
with A and B defined in (20). The angles at point b and b∗ are found by
sbƒ ∗ y
tan 2b∗ = ,
sbƒ ∗ x
(29)
1b = 2 − 2b∗ .
When the roller gets in contact with the sprocket the contact area has to be found. First of all it is
checked if the roller is in contact with area 4 or in contact with the left or the right side of the
tooth center line, see Fig. 10.
The condition for the roller being either on the left side or the right side of the tooth center
line is

< 0 : right
d T ut = (30)
> 0 : left ,
where the unit vector ut is given by Eq. (4) and vector d is given by
d = r r − rt . (31)
 Fig. 10. The position of the roller with respect to
the tooth center line
497

Fig. 11. The seating curve, contact area 4

2.4.1
Contact with the seating curve
Area 4 is referred to as the seating curve and contact with this area is analysed independent of
whether the roller is on the right or the left side of the tooth center line, see Fig. 11.
The vector n4 , see Fig. 11 is given by
n4 = −ur , (32)
and d is given in (31). The pseudo penetration is calculated by
4 = d − (Rt − Rr ) . (33)
The conditions that have to be fulfilled in order for contact to occur with area 4 are

 >0 ,
Contact conditions for area 4: 4 (34)
0 ≤ |40 | ≤ |4 | .
2a − 1a 
where 4 = = − A and 40 is given by
2 2

d T n4
cos(40 ) = , (35)
d4 

2.4.2
Contact with the topping curve
Areas 1 and 7 are referred to as the topping curves. The contact with area 1 is similar to the
contact with area 7, therefore, only contact with area 1 is described here, see Fig. 12. The vectors
shown in Fig. 12 are given by
n1 = At n1 , (36)
d1 = rr − rb∗ ,
 Fig. 12. The topping curves, contact
areas 1 and 7

where the local vector n1 = {cos(1b∗ + 1 ), sin(1b∗ + 1 )}T , with 1 = (2b∗ − 1b∗ )/2. The
498 pseudopenetration is given by
1 = (RF + Rr ) − d1  , (37)
where the radius RF is given in (21). The conditions that have to be fulfilled in order for contact
to occur with area 1 are

 >0
Contact conditions for area 1: 1 (38)
0 ≤ |10 | ≤ |1 | ,

where 10 is given by

d1T n1
cos(10 ) = . (39)
d1 

2.4.3
Contact with the working curve
Areas 3 and 5 are referred to as the working curves. The contact with area 3 is similar to the
contact with area 5, therefore, only contact with area 3 is described here, see Fig. 13. The vectors
shown in Fig. 13 are given by
n3 = At n3 ,
d3 = rr − rc∗ , (40)
where the local vector n3 = {cos(1c∗ + 3 ), sin(1c∗ + 3 )}T , with 3 = (B/2). The
pseudopenetration is given by
3 = d3 + Rr − RE , (41)
where the radius RE is given in (21). The conditions that have to be fulfilled in order for contact
to occur with area 3 are

 >0
Contact conditions for area 3: 3 (42)
0 ≤ |30 | ≤ |3 | ,

where and 30 is given by

d3T n3
cos(30 ) = . (43)
d3 

Fig. 13. The working curves, contact

areas 3 and 5
 Fig. 14. The straight portions, contact
areas 2 and 6

2.4.4
Contact with the straight portion
Areas 2 and 6 are referred to as the straight portions. The contact with area 2 is similar to the
contact with area 6, therefore, only contact with area 2 is described here, see Fig. 14. The vectors
shown in Fig. 14 are given by
499
1
t2 = (re∗ − rd∗ ) ,
re∗ − rd∗ 
n2 = −tˆ2 , (44)
d2 = rr − r , d∗

where the vectors re∗ and rd∗ are the global vectors to the points e∗ and d ∗ . The
pseudopenetration is calculated by

2 = d2T n2 − Rr . (45)

The conditions that have to be fulfilled in order for contact to occur with area 2 are

 2 > 0
Contact conditions for area 2: d2T t2 > 0 .
 T
d2 t2 < re∗ − rd∗ 

The contact force of the rollers is given by Eq. (3), with the normal vector n for the different
contact areas being

Area 1 (or 7) : n = d1 /d1  ,

Area 2 (or 6) : n = n2 , (46)
Area 3 (or 5) : n = −d3 /d3  ,
Area 4 : n = −d/d .

The indentation  in Eq. (3) is given by i , where index i = 1, · · · , 7 refers to the area. The relative
velocity is calculated by Eq. (18), with the normal vector n given by Eq. (47). The moment applied
to the sprocket is calculated in a similar way as in Eq. (19).

2.5
Analytical contact forces
The analytical results for the contact force between the roller and sprocket are described in [1].
These analytical results, however, only include two sprockets and thereby two chain segments of
which only one of them is under tension and the other under no tension. In the case of more
sprockets connected and different tension occurring in the segments, the analytical results
presented here are extended, so both of the segments in contact with a sprocket may be under
tension.
The link forces ƒl in the nb links of rollers in contact with a sprocket are
 i  sin( ) nb −i+1
sin( )
(ƒl )i = qa + qb + qc , i = 1, · · · , nb (47)
sin( +
) sin( +
)

where the angle , referred to as the pressure angle, is defined by

120◦
= 35◦ − (48)
nt
 and the contact forces ƒc on the nb bedded rollers in contact with a sprocket are

 i−1  nb −i
sin( ) sin(
) sin( )
(ƒc )i = qa + qb
sin( +
) sin( +
) sin( +
)
sin(
) 

× + 2qc sin , i = 1, · · · , nb (49)
sin( +
) 2

where qa is the tension in one segment, qb is the tension in the other segment, is the pressure
angle given by Eq. (20),
is the pitch angle, nt is number of teeth on the sprocket and qc is the
centrifugal force given by
500
mr V 2
qc = , V = Rs , (50)
Pg

where mr is the roller mass, P is the pitch,  is the angular velocity of the sprocket, Rs is the
radius of the sprocket and g is the gravitational acceleration. In order to compare the analytical
result with the numerical result, we plot the nb analytical force results versus a time scale where
the time values are given by

t e − ts
(t)i = ts + (i − 1) · , i = 1, · · · , nb , (51)
nb − 1

where ts is the start time when the contact begins and te is the end time when the contact ends.
The start and end time, ts and te , are estimated from the numerical results. The tension forces in
the segments qa and qb are also estimated from the numerical results, by an average value of the
link force in a segment.

3
Example
Two different models have been suggested to describe the contact surface between rollers and
sprockets and these models are implemented in a simulation program. For both methods, the
equations of motion of the roller-chain drive system are given by Eqs. (1) and (2). Both methods
use the same contact force model to describe the contact between the rollers and sprockets,
given by Eq. (3). However, the calculation of the indentation, relative velocity and normal vector
used in Eq. (3) are calculated differently. For both models, the link force between two adjacent
rollers is the spring/damper force between the two lumped masses.
The results from the numerical simulation is to be compared with results from a simplified
analytical model. The contact force between the rollers and sprockets are in the analytical
simplified model given by Eq. (49) and the link forces are given by Eq. (47).

3.1
Application to chain-drive system
The methodologies are applied to the simulation of a two-stroke diesel marine engine
roller-chain drive. The drive is standard in these marine engines that have between four and 12
cylinders for a power range of 1,760–78,000 kW. The roller-chain drive is placed at the fore end
of the engine L60MC and is composed by four sprockets and a chain made of 122 links, see
Fig. 15, Table 1 and Table 2.
The sprocket number (3) in the top of the chain drive is part of the pre-tensioning system and
it is located 0.3366 m to the right of the crankshaft and 3.0944 m above it during normal
operating conditions, see Fig. 15. Each link of the chain, with a pitch of 0.0889 m and a mass of
3.01 kg, is modeled as a flexible element with a stiffness of 815 MN/m, according to experimental
data obtained by the manufacturer. The driver sprocket, which is the sprocket on the crankshaft,
rotates in the numerical simulations with a constant angular velocity of 120 rpm, which gives a
driving frequency of 120/60 = 2 Hz and a tooth frequency of 2 · 60 = 120 Hz. The time it takes for
one link to move one whole revolution in the chain drive is tc = ( 2Hz1
)(122/60) = 1.017 s.
 501

Fig. 15. a Blueprint of the chain drive (from MAN B&W Diesel A/S) and b the numbering of the sprockets
and segments

The purpose of the four different sprockets in the chain drive are

Sprocket 1 = Crankshaft sprocket (driver)

Sprocket 2 = Counterweight sprocket
Sprocket 3 = Tightener sprocket
Sprocket 4 = Counterweight sprocket

The data for the system used in the numerical simulation is given in the following four tables.
The critical damping is calculated by assuming the chain as being a long open string √of a
series of masses connected with springs; for a chain with nl number of links Dcr = 2nl mr K,
yielding for the test case Dcr ≈ 1.2 · 106 kg/s.
√ For a single mass connected to ground with a spring,
∗ ∗
the critical damping coefficient is Dcr = 2 mr K, yielding for the test case Dcr ≈ 1.0 · 105 kg/s. The
∗
used damping coefficient in the simulation corresponds to 0.002% of Dcr and 0.25% of Dcr .
From the data given in Table 3, an average generalized coefficient of stiffness is calculated and
the value used in Eq. (3) is Kg = 1 GN/m. The coefficient of restitution is a constant 0 ≤ e ≤ 1,
where e = 0 relates to a fully plastic contact and e = 1 relates to a fully elastic contact. In order to

Table 1. Data for the sprockets

Sprocket
1 2 3 4
Pitch circle
radius, RS (m) 0.849325 0.425245 0.397000 0.425245
Position, xs (m) 0.0000 0.5560 0.3366 −0.5560
Position, ys (m) 0.0000 2.2020 3.0944 2.2020
Number of teeth, nt
60 30 28 30
Mass, ms (kg) 850 759 183 759
Mass moment of
inertia, Js (kg m2 ) 435 82 15 82
 Table 2. Data for the roller-chain
Value
Mass per. roller, mr (kg) 3.01
Chain pitch, L0 (m) 0.0889
Roller length (m) 0.02699
Roller radius (m) 0.027
Stiffness, K (MN/m) 815
Damping, D (kg/s) 250
Pre-tension (kN) 21
Number of links (nl ) 122

Table 3. Material data

502
Value
Poisson’s ratio,
0.3
Young’s modulus, E (N/m2 ) 2.06 · 1011
exponent, m 3
material property, h (m2 /N) 1.406 · 10−12

Table 4. Segment length and contact angles

Segment
1 2 3 4
Length (m) 1.8797 0.4104 1.2619 2.2312
Angle on sprocket i (rad) 0.3485 2.2748 2.3337 3.2010
Angle on sprocket j (rad) 3.4901 5.4164 2.3337 3.2010

include maximal energy dissipation in the model, e = 0 is used. Different values of e has been
applied resulting in similar conclusions as presented in this work, see Sect. 3.2.
In Table 4, the data for the four segments in the chain drive is given, where segment one is the
chain strand between sprocket one and sprocket two, segment two is the chain strand between
sprocket two and sprocket three, and so forth.

3.2
Numerical and analytical results
In the following, the link force between two adjacent rollers, the average link force (average over
all links in a segment) between the neighboring rollers in a segment and the transversal
oscillation of the middle of a segment between two sprockets will be tested. The test is done by
comparing the two different contact surface models, that is the circular tooth profile and the real
tooth profile. Further, the contact force between rollers and sprockets will be tested by comparing
a simplified analytical model with the two different models used for the numerical simulations.

3.2.1
The link forces
The variation of the link forces and analysis of the transversal oscillations of the chain strands
exemplifies the type of results necessary for the design of the roller-chain drives in terms of
fatigue and wear.
Figure 16 shows a comparison between the two methods, of the link force between two
adjacent rollers during one second of the simulation. The dashed line is the result obtained using
the real tooth profile and the solid line is the result obtained using the circular tooth profile. The
link force obtained using the real tooth profile has a noticeable variation when the link is in
contact with sprocket number one in the time period [9.3; 9.6 s] and with sprocket number three
in the time period [9.8; 9.9 s], whereas for the circular tooth profile, the large variations are in the
time periods before and after. The amplitude for the real tooth profile is also noticeably smaller
than for the circular tooth profile. We observe that in this case, the use of the real tooth profile
has a decreasing effect on the link force.
 503

Fig. 16. The link force between two adjacent rollers during 1 s. Dashed line for the real tooth profile and solid
line for the circular tooth profile

Fig. 17. An average link force between the neighboring rollers in segment one during 1 s. Dashed line for the
real tooth profile and solid line for the circular tooth profile

Figure 17 shows a comparison between the two methods, of the average link force (average
over all links in the segment) between the neighboring rollers in segment one during one second
of the simulation, that is, the free links situated in the chain strand between sprocket one and
two. The dashed line is the result obtained using the real tooth profile and the solid line is the
result obtained using the circular tooth profile. The mean value of the average link force for the
circular tooth profile is approx. 25 kN, whereas for the real tooth profile it is approx. 19 kN. The
amplitude of the average link force when using the circular tooth profile is up to approx. 13 kN,
whereas for the real tooth profile, it is up to approx. 6 kN. We observe that in this case the use of
the real tooth profile has a decreasing effect on the average link force as expected due to the
decreasing effect on the link force shown as described in the previous paragraph.
From Figs. 16 and 17 it can be seen that there is numerical noise on the results, which is partly
due to the low value of the damping coefficient D compared with the critical damping coefficient.
Higher values of D has been tested and they have a smoothing effect on the link forces, but
increases the simulation time. For higher values of D, however, the same conclusion with respect
to comparison of the two models are drawn.
504

Fig. 18. The oscillation of the middle of segment one about the initial position. Dashed line for the real tooth
profile and solid line for the circular tooth profile

3.2.2
The oscillations of chain strands
For the analysis of the transversal vibrations of the chain, the perpendicular distance between
the link closest to the middle of the initial chain strand line and the middle point of the initial
chain strand line is calculated.
Figure 18 shows a comparison between the two methods for the oscillation of segment one,
that is, the perpendicular distance between the middle of the segment and the initial chain
strand line. The dashed line is the result obtained using the real tooth profile and the solid line is
the result obtained using the circular tooth profile. The figure shows that the result from the two
methods are similar, the vibrations are with approximately the same major frequency. Again, as
for the link force, we observe that the real tooth profile has a smoothing effect on the result
compared to the circular tooth profile.
Figure 19 shows a comparison between the two methods for the oscillation of segment four,
that is the perpendicular distance between the middle of the segment and the initial chain strand
line. The dashed line is the result obtained using the real tooth profile and the solid line is the
result obtained using the circular tooth profile. The figure shows that the result from the two
methods are somewhat different in the size of the amplitudes of the vibrations. Again, as for the
link force, we observe that the real tooth profile has a smoothing effect on the result compared to
the circular tooth profile.

Fig. 19. The oscillation of the middle of segment four about the initial position. Dashed line for the real tooth
profile and solid line for the circular tooth profile
 505

Fig. 20. The frequency of the oscillation of the middle of segment one - For the circular tooth profile (left)
and for the real profile (right)

Figure 20 shows the frequency spectra from a FFT (fast fourier transformation) of the results
shown in Fig. 18, the left figure for the results obtained using the circular tooth profile and the
right figure for the results obtained using the real tooth profile. Both the frequency spectra shows
a major peak around 5.5 Hz, that for the circular tooth profile the peak is at approx 6.1 Hz and for
the real tooth profile the peak is at approx. 5.1 Hz. The other smaller peaks are not comparable.
The main excitation frequencies are the driver frequency 2 Hz and the tooth frequency 120 Hz,
but either of these frequencies are seen in Fig. 20.
The eigenfrequencies ƒn for a free undamped pretensioned string is given by

F
ƒn = n , (52)
4lt mt

where n is the eigenfrequency number, F is the pretensioning force, lt is the length of the string
and mt is the mass of the string, see e.g. [7]. For segment number one, the length is lt ≈ 1.8797 m,
corresponding to approx. 21 links which yields a string mass of mt ≈ 63.21 kg. For the circular
tooth profile, we found the mean average link force to be F ≈ 25 kN and for the real tooth profile
it is F ≈ 19 kN. Using Eq. (52) we obtain the first eigenfrequency for segment number one,
assuming it is comparable to a pretensioned string, that is for the circular tooth profile
ƒ1(c) ≈ 7.3 Hz and for the real tooth profile ƒ1(r) ≈ 6.3 Hz. For the numerical simulation, damping is
included in the longitudinal direction of the links, which lowers the eigenfrequency. It is
therefore likely that the 6.1 and 5.1 Hz corresponds to the first eigenfrequency of the chain
strand in the transversal direction.
Figure 21 shows the frequency spectra from a FFT of the results shown in Fig. 19, the left figure
for the results obtained using the circular tooth profile and the right figure for the results
obtained using the real tooth profile. Both the frequency spectra show a major peak about 5 Hz,
that is for the circular tooth profile the peak is at approx. 5.2 Hz and for the real tooth profile the
peak is at approx. 4.5 Hz. Both methods also have a peak at the tooth frequency 120 Hz. For the
circular tooth profile the other major peaks are at approx. 15.5, 25.5, 83 and 157 Hz and for the
real tooth profile the peaks around the same values are approx. 13, 33, 87 and 153 Hz.
For segment number four the length is lt ≈ 2.2312 m, corresponding to approx. 25 links which
yields a string mass of mt ≈ 75.25 kg. For the circular tooth profile we found the mean average
link force to be F ≈ 26 kN and for the real tooth profile it is F ≈ 20 kN. Using Eq. (52), we obtain
the first eigenfrequency for segment number four, assuming it is comparable to a pretensioned
string, that is for the circular tooth profile ƒ1(c) ≈ 6.2 Hz and for the real tooth profile ƒ1(r) ≈ 5.5 Hz.
It is therefore likely, due to the damping, that the 5.2 and 4.5 Hz corresponds to the first
eigenfrequency of the chain strand in the transversal direction.

3.2.3
The contact force
The variation of the contact forces between the rollers and sprockets exemplify another type of
results that can be derived from the numerical simulation program.
506

Fig. 21. The frequency of the oscillation of the middle of segment four – For the circular tooth profile (left)
and for the real profile (right)

Fig. 22. Contact force on a roller during one revolution. Dashed line for the real tooth profile and solid line
for the circular tooth profile

Figure 22 shows the contact force on a roller during one second of the simulation in the time
period [9;10 s]. The dashed line is the result obtained using the real tooth profile and the solid
line is the result obtained using the circular tooth profile. The following figures are from the
same simulation time period, but taken in the shorter time periods, where contact has occurred.
Figure 23 shows the contact force on a roller during contact with sprocket number one (left)
and sprocket number two (right). The dashed line is the analytical result calculated by Eq. (49),
where the average tension force is found by the real tooth profile method, the dashed-dotted line
is the analytical result where the average tension force is found by the circular tooth profile
method; the solid line is the numerical using the circular tooth profile and the solid-dotted line
is the numerical using the real tooth profile. The start and end time of the contact period are for
the analytical results estimated by the numerical results and same values are used for both
analytical results.
Figure 24 shows the contact force on a roller during contact with sprocket number three (left)
and sprocket number four (right). The dashed line is the analytical result where the average
tension force is found by the real tooth profile method; the dashed–dotted line is the analytical
result where the average tension force is found by the circular tooth profile method; the solid line
is the numerical result using the circular tooth profile and the solid-dotted line is the numerical
result using the real tooth profile. The start and end time of the contact period are for the
analytical results estimated by the numerical results and same values are used for both analytical
results.
The contact forces between the rollers and the sprockets are in better agreement with the
analytical expected results, when the real tooth profile is used. However, for the two
counterweight sprockets (two and four) where the contact period is short, the analytical result
does not compare well with either of the two methods. In the analytical results, it is assumed that
the angular velocity is constant, which is true in the numerical simulation only for the driver
sprocket one. It is also assumed that the angle where there is contact (the pressure angle) is
constant during the whole contact period, whereas in the numerical simulations, this varies.
These are some of the reasons for the difference between the results.
 507
Fig. 23. Contact force from sprocket one (left) and sprocket two (right). The solid line (numerical) and
dashed-dotted line (analytical) for the circular tooth profile. The solid-dotted line (numerical) and the dashed
line (analytical) for the real tooth profile

Fig. 24. Contact force from sprocket three (left) and four (right). The solid line (numerical) and dashed-
dotted line (analytical) for the circular tooth profile. The solid-dotted line (numerical) and the dashed line
(analytical) for the real tooth profile

4
Discussion
The modelling with a real tooth profile, was introduced in order to improve an earlier model. The
simulation time is more or less the same for both the circular tooth profile method and the real
tooth profile method, when the time used in order to find a static equilibrium before starting the
simulation, which is necessary for the circular tooth profile method, is not taken into account.
However, for a large example with more links and sprockets, the circular tooth profile method
fails to solve the problem. The modelling with the real tooth profile improved the numerical
simulations both related to numerical stability and the results of the contact force are in better
agreement with the analytical expected values.
Using the real tooth profile, we observe that much of the numerical noise that occurs when the
circular tooth profile is used, disappear. The frequencies of the oscillations of the middle of the
segments are somewhat similar, but the amplitudes are different in some of the segments. The
real tooth profile method is preferable to model the contact, because the model is numerically
more stable, the numerical noise on the results disappear and the results obtained with this
method are in better agreement with analytical results.

5
Conclusion
With the continuous contact force method, it is possible to perform a dynamic simulation of the
roller-chain/sprocket system, including the polygonal effect, which appears when the individual
chain links engage and disengage the sprockets, the roller impact at engagement, the flexibility
of the links and the coupling between axial and transverse vibrations. Moreover, the formulation
allows for the introduction of off-center sprockets in the roller-chain drives, which are sources of
 extra excitations on the drive. It also allows for introducing other effects, such as contact with
guide bars and tightener systems.

References
1. Binder, R.C.: Mechanics of the roller chain drive. Prentice Hall, Englewood Cliffs, 1956
2. Bouillon, G.; Tordion, G.V.: On polygonal action in roller-chain drives. J Eng Industry, ASME Trans 87
(1965) 243–250
3. Chew, M.: Inertia effects of a roller-chain on impact intensity. J Mech Transmissions, Automation Des
107 (1985) 123–130
4. Choi, W.; Johnson, G.: Vibration of the roller chain drives at low, medium and high operating speeds.
Vibration Mech Sys His Mech Des ASME, 63 (1993) 29–40
508 5. Fritz, P. Pfeiffer, F.: Dynamics of high-speed roller-chain drives Des Eng Tech Conf 3(-Part A) (1995)
573–584
6. Goldsmith, W.: Impact, the theory and physical behavior of colliding solids. Edward Arnold Ltd.,
London, 1960
7. Graff, K.F.: Wave motion in elastic solids. Dover New York, 1991
8. Lankarani, H.M.; Nikravesh, P.E.: Continuous contact force models for impact analysis in multibody
systems. Nonlin Dyn 5(1994) 193–207
9. Mahalingam, S.; Polygonal actions on chain drives. J Franklin Inst 265(1) (1958) 23–28
10. Nikravesh, P.E.: Computer-aided analysis of mechanical systems. Prentice Hall, Englewood Cliffs 1988
11. Pedersen, S.L.; Hansen, J.M.; Ambrósio, J. A. C.: A roller chain drive model including contact with
guide-bars In Jorge, A. Ambrósio, C. (ed.) Multibody Dynamics 2003, International Conference on
Advances In Computational Multibody Dynamics. ECCOMAS 16 pages on CD, July 2003
12. Pfeiffer, F.: Chain dynamics. Des Eng Tech Conf 3-(part A) (1995) 1187–1194
13. Pfeiffer, F.; Glocker, C.: Multibody dynamics with unilateral contacts. Wiley, New York, 1996
14. Troedson, I.; and Vedmar, L.: A dynamic analysis of the oscillations in a chain drive. J Mech Des 123
(2001) 395–401
15. Tsubaki: Tsubaki marine engine roller chain, maintenance manual. Catalog, vol. 1211. Tsubaki Chain Co.,
Japan, 1991
16. Turnbull, S.R.; Fawcett, J.N.: Dynamic behavior of roller chain drives. Inst. Mech. Engrs., Part C (1972)
(70/72) 29–35
17. Veikos, N.; Freudenstein, F.: On the dynamic analysis of roller-chain drivers Part i – theory. Mech Des
Syn ASME 46 (1992) 431–439
18. Wang, K.W.; Liu, S.P.: On the noise and vibration of chain drive systems. Shock Vibration Dig 23(4)
(1991) 8–13