You are on page 1of 11

569

A method to determine the dynamic load distribution in


a chain drive

I Troedsson and L Vedmar*


Department of Machine Elements, Lund Institute of Technology, Lund, Sweden

Abstract: This paper presents a method to calculate the forces in a chain and, thus, the resulting load
distribution along the sprockets in a chain transmission working at a moderate or high speed. When
the chain drive is loaded, the rollers that contact the sprockets will move along the ¯ anks to diVerent
height positions. There are mainly two diVerent ways to determine the actual positions: to assume the
positions or to use force equilibrium and to calculate the positions. To ® nd the correct solution the
geometry and the force equilibrium are used which will give each roller’ s position, along the ¯ ank.
This method demands knowledge of all parts of the chain, even the slack part. Therefore it has been
necessary to model both the connecting tight and the slack spans in which power between the two
sprockets is transmitted. The gravitational force acting at the chain has been included in the complete
model so that the position of the rollers and the forces in the links at the slack span can be calculated.
The elastic deformation in the chain has also been included. The moment of inertia in the two
sprockets and in the outer geometry has been taken into account, but not the inertia forces in the
chain.

Keywords: chain drive, vibrations, dynamics, oscillations, transmission error, chain strength,
polygon action, roller chain

NOTATION Subscripts

d damping
a centre distance
g driven sprocket
C stiVness
i roller number
D damping constant
p driving sprocket
F force
sl slack span
g acceleration due to gravity
sp tight span
J moment of inertia
1 ¯ ywheel at driving side
m mass of the chain and rollers
2 ¯ ywheel at driven side
M torque
n number of rollers
Rd pitch radius
t undeformed pitch 1 INTRODUCTION
U gear ratio
x; y; z coordinates The chain drive is one of the oldest machine elements
z number of teeth but is still a very common way to transmit power. The
major advantages of using chain drives are the ¯ exibility
¬ pitch angle in selecting shaft centre, the ability to drive more than
² coordinate along the tooth ¯ ank one shaft, the high reliability and durability, the rela-
(non-dimensional) tively cheap components and the relatively high eY-
³ chain angle ciency. However, the chain drives introduce some
’ coordinate problems such as noise and vibrations. Even if the most
basic mechanics of the chain drive have been known for
a long time, research in the subject has been con-
The MS was received on 25 May 2000 and was accepted after revision centrated to the latter half of the twentieth century. In
for publication on 11 August 2000.
* Corresponding author: Department of Machine Elements, Lund most cases, this has been because of the increased
Institute of Technology, Box 118, S-221 00 Lund, Sweden. accessibility of powerful calculation tools which have
C04000 ß IMechE 2001 Downloaded from pic.sagepub.com at PENNSYLVANIA STATE UNIV on September 18, 2016 Proc Instn Mech Engrs Vol 215 Part C
570 I TROEDSSON AND L VEDMAR

made it possible to solve some of the complex problems was lumped at the centre of each roller. The rollers were
associated with chain drives. Much of the research has connected by springs. This model also allows transverse
been concentrated on either solving the internal geo- movement of the chain in the tight span but neglects the
metric problems, such as load distribution between the slack span.
chain and sprocket, or solving the eVects of a chain drive All the above-mentioned dynamic models have
on a transmission. assumed that the end rollers of the tight span land on
In the 1950s and 1960s, Binder [1] and Rachner [2] the pitch circle of the two sprockets and all have
presented the ® rst step to an analysis of forces and assumed that the eVect of the slack span can be
stresses in a chain drive, by assuming that the contact neglected. This means that none of these models ful® ls
angle relative to the symmetry line of each tooth was the equilibrium equations for the sprockets, which
constant. In the early 1980s, Naji [3] and Naji and demands knowledge of the forces in both the entrance
Marshek [4, 5] complemented Binder’ s and Rachner’ s and the exit parts of the chain.
research by theoretical and experimental investigations A static model developed by the present authors was
studying contact phenomena under quasi-static condi- presented in an earlier paper [12]. That model was
tions due to non-trivial geometry of the sprocket tooth. obtained from a chain drive containing two sprockets
In the 1990s, Kim and Johnson [6, 7] improved the and one chain. The two sprockets were connected by
model employed by Naji [3] and Naji and Marshek’ s two spans: one tension and one slack span. It was
[4, 5] model using the full ANSI standard geometry. necessary to include these in the analysis to ful® l the
However, none of these workers has combined elastic equilibrium equations for the chain and sprockets. The
deformation with non-trivial geometry and none of rollers that contact each sprocket can move along the
them has used a complete gear set with two sprockets tooth surface and their positions are given by the static
and two connecting chain spans. Furthermore they have equilibrium equations. In this report the European
only solved the problem for static conditions. geometry of the chain and sprocket is used, diVerent
The polygon eVect, which has been known almost from most of the above-mentioned reports.
since the ® rst chain drive was developed, introduces The purpose of this paper is to combine the earlier
problems due to the error in gear ratio during one pitch. static model with dynamics to ® nd out how the geo-
For a chain drive with two sprockets and one chain, the metry and slack span aVect the amplitude and the load
torque and speed on the driven sprocket will ¯ uctuate if distribution along the sprockets and the forces in the
the driving sprocket drives with a constant speed and chain.
torque. If the speed is high enough, this variation in gear
ratio introduces vibrations which must be treated in a
dynamic model. Binder [1] and Rachner [2] calculated 2 DYNAMIC MODEL
the gear ratio assuming the tight span was rigid, with the
pitch radius given by the chain pitch. They investigated A chain drive is used to transmit power between dif-
the variation in gear ratio for the chain drive for the two ferent parts in a complete transmission. Since the
cases where the length of the span is equal to an integer dynamic behaviour of the chain drive is aVected by its
number of chain pitch or an odd number of half a pitch. surrounding parts, it is necessary to include these in a
An early attempt to calculate the amplitude of the dynamic model. The chain drive itself contains a chain
vibration was made by Turnbull and Fawcett [8] in 1975. which connects the driving and driven sprockets. These
They carried out an approximate harmonic analysis of are connected to the driving and driven mechanisms
the acceleration of the two sprockets, by using a Fourier with some kind of shaft. These mechanisms contain
analysis of an equivalent four-bar mechanism. They some inertia as well as some variations in the driving
neglected the mass in the chain and assumed that the and braking torque. Since the surrounding systems are
deformation in the chain was small, the chain was not generally known, they are assumed to be made up by
positioned along the pitch circle on the sprockets, and one inertia and one stiVness, separated from each other,
the tight span was always parallel to the common tan- on both sides of the chain drive.
gent of the two sprockets pitch circles. Turnbull et al. [9] A chain can be used to transmit power between sev-
also conducted experimental investigations into the eral sprockets, but the chain drive treated in this report
dynamic behaviour. contains only three parts: one driving sprocket, one
A dynamic model which used a discrete dynamic driven sprocket, and the chain which transmits the
method was developed by Fawcett and Nicol [10] in power between them. The chain which connects the two
1979. They modelled a moving chain with a series of sprockets with each other has two spans: one slack and
links. The boundary condition for a link leaving and one tensioned. The tensioned span transmits most of the
entering the span was also introduced. power, while the slack span transmits very little power.
In 1992 a more complete model was presented by The rollers which are in contact with the sprockets can
Veikos and Freudenstein [11]. The chain was there move along the tooth surface and their position is given
modelled as a series of rollers where half the link mass by force equilibrium and by the fact that the distance

Proc Instn Mech Engrs Vol 215 Part C Downloaded from pic.sagepub.com at PENNSYLVANIA STATE UNIV on September 18, 2016 C04000 ß IMechE 2001
A METHOD TO DETERMINE THE DYNAMIC LOAD DISTRIBUTION IN A CHAIN DRIVE 571

between them is one chain pitch. The chain pitch is 3 THE CHAIN DRIVE
aVected by the elastic deformation and thus dependent
on the stretching force in the link. To ful® l the com- The chain drive can be divided into four diVerent parts:
patibility equations for the ® rst and the last roller in the tight span, the slack span, and the chain in contact
contact with each sprocket, the position of the ® rst and with the driving and driven sprocket respectively. The
the last roller in each span and the force in the con- part of the chain in contact with one of the two
necting links have to be known. Therefore both the slack sprockets transmits power between the chain and the
span and the tight span have to be modelled. The eVects sprocket through normal forces between the teeth and
from the driven side of the transmission as well as the the rollers, while the spans transmit the power between
driving side are modelled as a ¯ ywheel that is connected them. The load distribution between the chain and the
with the sprocket by a shaft. sprocket depends on which point of the tooth ¯ ank each
To describe mathematically the chain drive shown in of the rollers contact. Since the chain load in the tight
Fig. 1, a few parameters have to be introduced. The span is much greater than in the slack span, most of the
¯ ywheel at the driving side is indicated by the subscript 1 power is transmitted between the sprocket through the
and the ¯ ywheel at the driven side by the subscript 2. tight span. The stretching force in the tight span is
The driving sprocket is indicated by the subscript p, and higher and thus dominates over the gravitational force
the driven sprocket by the subscript g. The moment of in this part. In the slack span most of the stretching
inertia is J, the driving torque is Mdrive and the driven
force comes from the gravitational force. The four parts
torque is Mbrake. The centre distance between the two
of the chain drive can, for convenience, be treated
sprockets is a, the number of rollers in the chain n, the
separately but, since they aVect each other, the com-
gravitational force acting at one roller mg, and the
patibility has to be considered. A method to solve this
number of teeth at the sprockets zp and zg .
problem for a static case, has been presented by the
A shaft, the connected ¯ ywheel and sprocket all have
present authors in reference [12]. The geometry used in
the same rotational axis. The two shafts, their connected
this report is the European geometry de® ned by DIN
¯ ywheels and sprockets are assumed to be parallel and
8187.
to be located in the plane perpendicular to the gravita-
The gravitational force acting on the chain is taken
tional direction. The two sprockets and the chain are
into account but not the inertia forces. The chain is
assumed to be in the same plane and all variations, on
modelled as massless springs which connect the rollers.
the sprockets and the chain, in the direction along the
rotation axis of the sprockets are neglected. All the links Half of the mass at the two connecting links together
are assumed to have equal stiVnesses and equal unde- with the mass of one roller are lumped at the centre of
formed pitches. The diVerence in pitch for the sprocket each of the rollers where the corresponding gravitational
is also neglected. All links are assumed to have equal forces act. It is assumed that all the rollers have the same
masses; the same is true for the rollers. The eVect of the masses and the same undeformed pitches. All links are
mass of the link on the equation of motion is neglected modelled with equal stiVnesses.
and the sprockets are assumed to be rigid. The friction is At the part of the chain which is in contact with the
neglected when calculating the position of the rollers. It sprocket, each roller can move along the tooth surface
is furthermore assumed that there are bearings at both and the position will be given by the geometry of the
sides of both the ¯ ywheels and the sprockets. The surfaces, the distance to the surrounding rollers and the
bending deformation in the shafts is small and can force equilibrium. The distance between the contacting
therefore be neglected. All bearings are assumed to be rollers is dependent on the stretching force in the links
without any deformation. and the rollers will therefore contact at diVerent points
on the tooth ¯ ank. This means that the position of each
roller is dependent on the torque acting at each sprocket
as well as the length of the spans and the stretching force
in them.
To make it easy to describe the ¯ ank in the x 0 and y 0
coordinates, a non-dimensional coordinate 04²46 has
been introduced (Fig. 2). It varies linearly between
every two succeeding integers and describes the centre
position of a roller, with the radius Rr , which moves
along the tooth ¯ ank. The opposite half of the tooth is
described by re¯ ecting the x 0 coordinates in the y 0 axis.
The position of an arbitrary roller can ® nally be
described by rotating the coordinates a number of pitch
angles. The complete sprocket can thus be rotated to
Fig. 1 Chain drive and surrounding transmission every arbitrary position.

C04000 ß IMechE 2001 Downloaded from pic.sagepub.com at PENNSYLVANIA STATE UNIV on September 18, 2016 Proc Instn Mech Engrs Vol 215 Part C
572 I TROEDSSON AND L VEDMAR

thereafter calculated, which gives a new roller position.


This calculation is repeated until the position given by
two following iterations are equal.
In Fig. 3 the forces on a roller in a general sprocket
position are shown. The angle ’ is used to describe the
position of the sprocket. The rotation coordinates to the
® rst contacting tooth’ s centre-line on the sprocket are ’p
and ’g . To describe the position of a speci® c roller the
angle ’i is used. The angles between the y axis and the
® rst roller of each sprocket are ’1;p and ’1g . With the
forces in Fig. 3, the moment on the sprocket from the
chain becomes

Mp ˆ F 0;p R1;p cos…’1;p ¡ ’in;p †


X
n
‡ Fn;p Rn;p cos…’n;p ‡ ³p;out † ¡ mg Ri;p sin ’i;p
iˆ1

Mg ˆ ¡ F0;g R1;g cos…’1;g ¡ ’in;g †


X
n
¡ Fn;g Rn;g cos…’n;g ‡ ³g;out † ‡ mg Ri;g sin ’i;g
iˆ1

Fig. 2 De® nition of the non-dimensional coordinates ² …1†

where ³in and ³out are the chain angles, relative to the x
By using the ² coordinate the position in the next- axis, of the ® rst and last links respectively in contact
neighbouring roller can be described if the position of a with the sprocket.
speci® c roller, the force acting at the previous link, and The tight span is modelled as a series of links posi-
the distance between the rollers are known. The distance tioned along a straight line. The end points are the ® rst
between two rollers is calculated from the stretching connecting rollers at each sprocket. Since this span is
force in the link between them and the elastic defor- assumed to be stretched, the eVect from the gravitation
mation thereby given. Since the stretching force between force on the position of the rollers is neglected, but its
the two rollers depends on their position, the coordi- eVect on the stretching forces in the chain is taken into
nates of the second of the two rollers are ® rst calculated account.
using the undeformed link length. The force in the link is The position of the rollers at the slack span is calcu-
lated from the force equilibrium and the fact that the
distance between them is one chain pitch. The major
part of the stretching forces in the links comes from the
gravitational force. The stretching force is therefore
normally small and the elastic deformation is conse-
quently neglected, making the pitch equal to the unde-
formed pitch in this part of the chain. This pitch is
assumed to be equal for all links. However, if the dis-
tance between the end points of the slack span is greater
than the number of links in this span multiplied by the
chain pitch, a tight span is assumed and the elastic
deformations is then included in the analysis. The
equations for calculating the tight span, the slack span
and the position of the rollers in contact with one of the
rollers have been given in detail in reference [12].

4 FRICTIONAL FORCES

Most of the friction in a chain drive comes from the


Fig. 3 Sprocket in an arbitrary load position movement of the bushings, pins and rollers. This friction

Proc Instn Mech Engrs Vol 215 Part C Downloaded from pic.sagepub.com at PENNSYLVANIA STATE UNIV on September 18, 2016 C04000 ß IMechE 2001
A METHOD TO DETERMINE THE DYNAMIC LOAD DISTRIBUTION IN A CHAIN DRIVE 573

is a combination of Coulomb friction and viscous The damping in the two shafts comes from the internal
damping dependent on the lubrication. This is not damping in the material due to hysteresis. This damping
considered in this model. However, damping must be is smaller than the damping coming from the relative
introduced to produce a stable solution. Since most of movement between two surfaces. This kind of damping
the power is transmitted in the tight span, it is assumed could with good approximation be modelled as
that most of the damping also occurs at this side. The equivalent viscous damping. The damping moment
damping force is usually much smaller than the could then de described as
stretching force in the chain but is in some way related
to this. The damping force in the slack span is neglected, M1;damp ˆ Dshaft;1 …’_ p ¡ ’_ 1 †
since the stretching force at the slack span is smaller
than at the tight span. For convenience, a viscous M2;damp ˆ Dshaft;2 …’_ g ¡ ’_ 2 †
damping force is assumed, proportional to the relative
velocity between the two end points at the tight side. The …4†
damping force can then be described as
When solving the equation of motion the coordinates in
Fig. 4 have been used. The inertia in the chain is
Fd ˆ Dchain ‰R1;p ’_ p cos…’1;p ¡³in;p †¡R 1;g ’_ g cos…’1;g ¡³in;g †Š
neglected but not the gravitational force since otherwise
…2† it would be impossible to calculate the position of the
roller in the slack span. The outer systems on each side
of the chain drive are separated into one inertia and one
The damping in a chain has been studied by Rachner [2] elasticity. These represent the inertia and the elasticity of
and by Schlottmann et al. [13]. They studied the loga- the complete system in a simpli® ed form. The equation
rithmic decrement ¯ for a vertical hanging chain with a of motion for the model then becomes
mass at the end, and with the movement along the chain.
The logarithmic decrement ¯, and the damping constant
J1 ’1 ˆ Mdrive ‡ M1 ‡ M1;damp
D given thereby, are diVerent depending on what kind of
lubricant is used. For example, the logarithmic decre- Jp ’p ˆ ¡M1 ¡ M1;damp ‡ Mp ¡ Fd R1;p cos…’1;p ¡ ³in;sp †
ment for a chain lubricated with grease is twice that for a
chain lubricated with oil. Jg ’g ˆ Mg ¡ M2 ¡ M 2;damp ‡ Fd R1;g cos…’1;g ¡ ³in;sp †
In order to calculate the equivalent damping constant
in the tight span for the modelled chain drive, the J2 ’2 ˆ M2 ‡ Mbrake ‡ M2;damp
equation of motion was set up for a simpli® ed model
…5†
with two free sprockets and one tensioned chain span
tangentially connecting the two pitch circles. The
equation for this model could then be solved analytically where
and the damping constant Dchain could be calculated
from the logarithmic decrement as M1 ˆ Cshaft;1 …’p ¡ ’1 †

M2 ˆ Cshaft;2 …’g ¡ ’2 †
¯!0
Dchain ˆ …3†
p‰R 2d;p cos2 …’1;p ¡ ³sp †=Jp and the spring constants for the shafts are called Cshaft;1
‡R 2d;g cos2 …’1;g ‡ ³sp †=Jg Š and Cshaft;2 .

Fig. 4 Transmission

C04000 ß IMechE 2001 Downloaded from pic.sagepub.com at PENNSYLVANIA STATE UNIV on September 18, 2016 Proc Instn Mech Engrs Vol 215 Part C
574 I TROEDSSON AND L VEDMAR

5 SOLVING PROCEDURE end positions is longer than the length of the remaining
links, a tight span is assumed. Otherwise the end posi-
The equations of motion are solved using the Runge± tions in the slack span are used to calculate the forces,
Kutta± Fehlman method. The torques Mp and Mg , and the directions of the forces in the ends of the span.
which have to be calculated for every time step, depend To verify that the number of rollers in contact with the
on the coordinates of the angles ’p and ’g and can be sprockets is correct, ® rst ful® lment of the equilibrium is
calculated if these are known. The equations concerning checked for the ® rst and last contacting rollers at each
the chain in reference [12] are also valid for this report, sprocket. If the contacting force at one of these rollers is
but the solving procedure is diVerent in some ways, since negative, it is necessary to decrease the number of con-
the angle coordinate ’g in that case was not given. In tacting rollers on that sprocket and to increase the
earlier sections the chain drive has been divided into number of rollers on the enclosing span. It is thereafter
four diVerent parts. The diVerent parts are driving also checked whether the ® rst and the last rollers at each
sprocket, driven sprocket, slack span and tight span. At span contact the sprockets. If this is the case, the num-
the part with the chain in contact with the sprocket it is ber of rollers that contact the sprocket has to be
possible to calculate the position of the next roller if the increased and the number of rollers in the enclosing span
position of the ® rst of the pair, and the stretching force is decreased. When all roller coordinates are known, the
in the link between them are known. However, the forces in the link after the last contacting rollers, F n;p
contacting force and the stretching force in the link are and Fn;g , are calculated from the condition in each
dependent on the position of the roller and the angle sprocket. These forces can then be compared with the
between the roller centre and the contact point between stretching forces in the same links, Fsl;1 and Fsl;n0 cal-
the roller and the contacting ¯ ank at the sprocket. This culated from the conditions valid for the slack span,
means that it is possible to locate the coordinates for all giving the diVerences in forces as Fn;p ¡ Fsl;1 and
rollers in contact with the sprocket as long as the tension F n;g ¡ Fsl;n . If these diVerences are too large, new values
force and the contacting point are both known. The for ²p and ²g are chosen. It is possible to solve the
stretching force in the ® rst link at each sprocket is problem by using the Newton± Raphson method and the
known from the compatibility with the tight and slack derivatives can then be calculated using the diVerence
spans; the other sprocket forces are not known. Since equation. However, it is sometimes necessary to intro-
the coordinate of the angles ’p and ’g will vary for every duce limitations in the derivatives in order to ensure that
time step the stretching forces will vary as well. When the system converges. In the program used to solve the
calculating the stretching forces in the tight span, the problem, a check is made that the stretching forces in
length has to be known. However, the positions of the the tight span do not reach values below zero, since then
® rst contacting rollers are unknown. Therefore values of the program cannot ® nd a solution. A numerical pro-
the non-dimensional coordinates ²p;1 and ²g;1 have to be blem in the solving procedure is that the stretching force
assumed. The coordinates for the ® rst contacting roller can reach a value below zero in one of the iterations,
can then be calculated. All the tension forces in the links caused by a large step in the value of ²p or ²g . An
at the tight span, with nsp;link links, can thereafter be investigation then has to be made into whether this is the
calculated, especially the ® rst and the last as shown in correct solution or a numerically false solution. This is
the following equations: done by using a constant-step method using small steps.
"q The calculation starts from the sprocket position just























Clink after a new roller has contacted the driving sprocket and
F0;p ˆ …x 1;g ¡ x1;p †2 ‡ …y1;g ¡ y1;p †2 ¡ nsp tlink continues to the sprocket position, where the contacting
nsp;link
force on the ® rst roller at the driven sprocket becomes
#
nsp;link …nsp;link ¡ 1†mg less than zero. The number of rollers in the tight span is
¡ sin ³in;p then increased by one and the coordinates ’g and ’2 are
2Clink
decreased by one pitch angle. Then the calculation
continues until a new roller contacts the driving
F0;g ˆ F 0;p ‡ …nsp;link ¡ 1†mg sin ³in;p
sprocket. The number of rollers at the tight span is then
…6† decreased by one and the coordinate ’p and ’1 are
decreased by one pitch angle. It is veri® ed for every time
When the coordinates of the ® rst contacting roller at step that the free rollers in the spans do not interfere
each sprocket are known together with the force in the with any of the sprocket teeth.
link which connects them to the tight span, it is possible The results and the time that it takes to solve the
to calculate the position of the second contacting roller problem depend on the starting values. When starting
at each sprocket. By repetition, all contacting rollers can from rest with only the torques Mdrive and Mbrake acting
then be positioned at the sprockets. The position of the on the model, assuming that ’p ˆ ’1 and ’g ˆ ’2 , very
last contacting roller at each sprocket de® nes the end high oscillations will occur, which make it almost
position of the slack span. If the distance between these impossible to solve the problem. A better way to start

Proc Instn Mech Engrs Vol 215 Part C Downloaded from pic.sagepub.com at PENNSYLVANIA STATE UNIV on September 18, 2016 C04000 ß IMechE 2001
A METHOD TO DETERMINE THE DYNAMIC LOAD DISTRIBUTION IN A CHAIN DRIVE 575

the calculation is to assume that the same torque acts on with stiVness equal to the stiVness of the tight span and
both sides of the ¯ ywheels. This means that free pulleys:
’1 ¡ ’p ˆ Mdrive =Cshaft;1 and ’2 ¡ ’g ˆ Mbrake =Cshaft;2 s 


















when starting. These starting values result in a high- 1 Clink …R 2d;g Jp ‡ R2d;p Jg †
torsion oscillation in the shafts, high from the beginning, !3 ˆ …9†
zp navgJp Jg
but the amplitude of these oscillations decreases with
time. The time that it takes to achieve a steady state The studied chain drive has n ˆ 40 links. The numbers
condition is, however, quite long and the last method of teeth are zp ˆ 10 and zg ˆ 20, the pitches at the
results in a very long solution time. In order to obtain sprockets are tp ˆ tg ˆ 15:875 mm and the chain has the
better starting values a simpli® ed model was developed. pitch tlink ˆ 15:95 mm. The moments of inertia for the
This model contains the two ¯ ywheels, their connecting ¯ ywheels used are J1 ˆ J2 ˆ 0:2 kg m2 , the moment of
shafts and two sprockets. The chain is then modelled as inertia for the two sprockets is calculated by approx-
a tight span with no mass. The stiVness in this tight span imating them with a solid disc with the outer diameter
is set to the stiVness in the tight span in the presented equal to the pitch radius of each sprocket, the spring
model. The end points of the tight span are assumed to constants for the shaft, C1 ˆ C2 ˆ 155:6 N m/rad and
contact the sprockets at the pitch circles given by the the chain stiVness Clink =navg ˆ 17:7 MN/m. The approx-
pitch of the elastic deformed chain. This simpli® ed imate natural resonance speeds of revolution for the
model is given the same damping as the presented shaft become !1 ˆ 2:66 rad/s and !2 ˆ 88:9 rad/s, and
model. The simpli® ed model is then executed until a the approximate natural resonance speed of revolution
stable solution is obtained. Then the values of the for the chain is !3 ˆ 1770:6 rad/s.
speeds, the angles, the torque in the shafts and the force The calculation of the forces in the chain is performed
in the tension span are translated into the presented using the solving procedure described earlier. When the
model as starting values. calculation starts, the oscillations are quite high. Most
of these oscillations are caused by the initial conditions
and are a result of the solution of the homogeneous
equations of motion. However, these oscillations
decrease with time and, after a while, only the particular
6 RESULTS
solution will remain, resulting in an oscillation with
constant amplitude from pitch to pitch. Since the start-
If the polygon action is neglected by approximating the up phenomena are dependent on the start condition,
chain as a stiV belt, which acts at pulley radii equal to which is unknown, only the steady state condition is
the sprocket pitch circle radii, the natural resonance studied in this report.
speeds of revolution for the shafts can be calculated Using a constant rotational speed the calculations
approximately by solving the following equation: were started and carried out until a stable solution was
³
´ ³ ´ received. The constant torque M drive ˆ 10 N m and
!1;2 4 !1;2 2 Mbrake ˆ 20 N m were used for all cases. The amplitude
J1 J2 Jtot ‡…C1 Jtot J2 ‡ Ctot J1 J2 ‡ C2 J1 Jtot
zp zp of the oscillations in the chain is dependent on the
Á damping in the chain and shafts. The damping in the
zp zg chain is diVerent when the chain is lubricated by oil from
‡ J1 C 22 ‡ C 21 J2 ¡ J1 Ctot C2
z g zp that when the chain is lubricated by grease. Therefore
two diVerent values of the damping constant in the
!
chain were studied, representing typical values for these
¡ C1 C2 Jtot ¡ C1 Ctot J2 ˆ0 …7† two cases. The damping in the two shafts is set to
Dshaft ˆ 0:024 N m/s for both shafts in all cases. The
diVerence of the force extrema, Fmax ¡ Fmin , in the link
where which connects the driving sprocket and the ® rst roller
in the tight span, is shown in Fig. 5 for diVerent speeds.
J g zp J p z g For this case it is impossible to obtain a solution from
Jtot ˆ ‡
zg zp the model for speeds higher than !p ˆ 630 rad/s, since
the oscillation results in negative stretching forces in the
C 2 zp C 1 zg tight span and the model is not designed for this. In Fig.
Ctot ˆ ‡
zg zp 5 it could be seen that the system responses to
approximately !3 =3; !3 =4; !3 =5; . . .. The system there-
…8† fore contains an in® nite number of natural responses.
The system responses will also exist for !3 and !3 =2 even
The resonance speeds of revolution for the chain drive if they cannot be calculated for this case. The mean
can be approximately calculated using an elastic belt, stretching force Fmean in this case is approximately 400 N

C04000 ß IMechE 2001 Downloaded from pic.sagepub.com at PENNSYLVANIA STATE UNIV on September 18, 2016 Proc Instn Mech Engrs Vol 215 Part C
576 I TROEDSSON AND L VEDMAR

Fig. 5 DiVerence in the force extrema

and the doubled amplitude which is shown in Fig. 5 several response speeds. For lower speeds the response
given by static calculation according to the polygon speed for the outer system is dominant since that speed
eVect is around 20 N. It is also interesting to note that is close to the natural response speed. For higher speeds
the amplitude can be higher than the mean force with the response speed for the chain becomes more and
still positive stretching forces in all the links at the tight more important. The frequency from the chain is visible.
span. This is caused by the fact that the force diVerence As stated earlier, it is noticeable that the maximum force
Fmax ¡ Fmean is much greater than the force diVerence for higher speeds diVers more from the mean value than
Fmean ¡ Fmin . The highest of the two natural response the minimum force. The point where a roller enters the
speeds in the outer system is visible for the curves near slack span is noticeable in Fig. 7 at ’p =…2¬p † approxi-
!p ˆ 100 rad/s. Furthermore it can be seen that the mately equal to 0.2, and the point where a roller enters
amplitude of the force generally increases with higher the tight span is noticeable at ’p =…2¬p † approximately
speeds. equal to 0.7.
In Fig. 6 the variation of the gear ratio U ˆ !p =!g is Figure 8 shows the stretching force in a link when it
shown. The same damping is used as in Fig. 5. A static travels around the complete chain transmission which
calculation would give the variation in gear ratio, has n ˆ 40 links. The stretching force is shown versus a
Umax ¡ Umin , equal to 0.1 for comparison. As can be seen non-dimensional coordinate along the chain de® ned as
in Fig. 6, the variation in gear ratio goes asymptotically ’p =…2¬p †, from one to the number of links. This coor-
to 0.1 for speeds lower than 200 rad/s. The natural dinate has its origin at the point where a roller has just
response speed could also be seen in the variation in the entered the driving sprocket. The coordinate is equal to
gear ratio. The mean variation in the gear ratio seems to 1.0 when the ® rst roller comes into contact with the
increase with higher speeds. Comparing Figs 5 and 6 it driving sprocket. Since the ® gure shows the results for a
can be seen that the variation in the mean gear ratio does steady state condition the force for 40 < ’p =…2¬p † < 41
not increase as rapidly as the stretching force. is equal to the forces for 0 < ’p =…2¬p † < 1. It is remark-
In Fig. 7 the force in the link, which connects the ® rst able that the variation in the force in the tight span is
contacting roller at the driving sprocket and the tight very much aVected by the oscillations coming from the
span, is shown during one pitch, when Dchain ˆ 30 N s/m. relative movement between the sprockets, when the
It could be seen that the variation in the force increases forces in the slack span hardly varies at all. The ¯ uc-
with higher speeds. The variation in the force is a mix of tuation in the stretching force along the sprockets

Proc Instn Mech Engrs Vol 215 Part C Downloaded from pic.sagepub.com at PENNSYLVANIA STATE UNIV on September 18, 2016 C04000 ß IMechE 2001
A METHOD TO DETERMINE THE DYNAMIC LOAD DISTRIBUTION IN A CHAIN DRIVE 577

Fig. 6 Variation in the gear ratio

Fig. 7 Force in the tight span at diVerent speeds

C04000 ß IMechE 2001 Downloaded from pic.sagepub.com at PENNSYLVANIA STATE UNIV on September 18, 2016 Proc Instn Mech Engrs Vol 215 Part C
578 I TROEDSSON AND L VEDMAR

Fig. 8 The stretching force in the chain

implies that the position where each roller contacts a REFERENCES


¯ ank diVers greatly.
1 Binder, R. C. Mechanics of Roller Chain Drive, 1956
(Prentice-Hall, Englewood CliVs, New Jersey).
2 Rachner, H.-G. Staalgelenkketten und Kettengetriebe, 1962
(Springer-Verlag, Berlin).
3 Naji, M. R. On timing belt and roller chain load
7 SUMMARY distribution. PhD dissertation, University of Houston,
Houston, Texas, 1981.
The presented model of a chain transmission combines a 4 Naji, M. R. and Marshek, K. M. Experimental determina-
static calculation of the forces at the sprocket and in the tion of the roller chain load distribution. Trans. ASME, J.
Mechanisms, Transmission Autom. Des., 1983, 105, 331±
chain, using the complete geometry, with dynamic
338.
oscillation of the sprockets. Since it is necessary to
5 Naji, M. R. and Marshek, K. M. Analysis of roller chain
model some part of the surrounding geometry, the sprocket pressure angles. Mechanisms Mach. Theory, 1984,
results also include the eVects from this. The calculation 19(2), 197± 203.
shows that older static models do not give suYcient 6 Kim, M. S. and Johnson, G. E. Mechanics of roller chain±
results for a chain drive acting at higher speeds. The sprocket contact: general modelling strategy. In ASME
developed model is used to calculate the forces acting in International Power Transmission and Gearing Confer-
the chain, at the roller and on the sprockets. The results ence, Vol. 2, DE-Vol. 43-2, 1992, pp. 689± 695 (American
show that the roller positions vary according to the Society of Mechanical Engineers, New York).
dynamic behaviour. It is obvious that the tension forces 7 Kim, M. S. and Johnson, G. E. Mechanics of roller chain±
in the tight span are more aVected by the variation, sprocket contact: observations about the contact phenom-
ena and load distribution. In ASME International Power
coming from dynamic eVects, than the slack span.
Transmission and Gearing Conference, Vol. 2, DE-Vol.
Although the force in the slack span is small, it is
43± 2, 1992, pp. 697± 702 (American Society of Mechanical
absolutely necessary to include this part in the analysis: Engineers, New York).
otherwise it will be impossible to determine the height 8 Turnbull, S. R. and Fawcett, J. N. An Approximate
position of the rollers on the sprocket. These positions Kinematic Analysis of the Roller Chain Drive, 1975, pp.
have a signi® cant in¯ uence on the load distribution and 907± 911 (Mechanical Engineering Publications Limited,
the dynamic load in the chain. London).

Proc Instn Mech Engrs Vol 215 Part C Downloaded from pic.sagepub.com at PENNSYLVANIA STATE UNIV on September 18, 2016 C04000 ß IMechE 2001
A METHOD TO DETERMINE THE DYNAMIC LOAD DISTRIBUTION IN A CHAIN DRIVE 579

9 Turnbull, S. R., Nicol, S. W. and Fawcett, J. N. An ical Design and Synthesis, DE-Vol. 46, 1992, pp. 431± 439
experimental investigation of the dynamic behaviour of a (American Society of Mechanical Engineers, New York).
roller chain drive. ASME paper 77-det-168, 1977. 12 Troedsson, I. and Vedmar, L. A method to determine
10 Fawcett, J. N. and Nicol, S. W. A theoretical investigation the static load distribution in a chain drive. Trans.
of the vibration of roller chain drives. In Proceedings of ASME, J. Mech. Des., September 1999, 121, 402±
the Fifth World Congress on the Theory of Machines and 408.
Mechanisms, Montreal, Canada, 1979, pp. 1482± 1485. 13 Schlottmann, G., Welke, L. and Klith, O. Experimentelle
11 Veikos, N. M. and Freudenstein, F. On the dynamic Parameterermittlung fur Zugmittelgetriebe. Maschinenbau
analysis of roller chain drives. Part 1: theory. In Mechan- Technik, September 1990, 39(9), 397± 400.

C04000 ß IMechE 2001 Downloaded from pic.sagepub.com at PENNSYLVANIA STATE UNIV on September 18, 2016 Proc Instn Mech Engrs Vol 215 Part C

You might also like