PERGAMON Computers and Structures 71 (1999) 303±332

Integral analysis of conveyor pulley using ®nite element

method
M. Ravikumar 1, Avijit Chattopadhyay *
Mechanical Handling Laboratory, Mechanical Engineering Department, Indian Institute of Technology, Madras 600 036, India
Received 22 January 1997; accepted 6 May 1998

Abstract

Pulleys are functionally critical items in belt conveyors. Earlier researchers have mostly analyzed the pulley in
parts rather than as an integral structure. Since the stresses and de¯ections of its various parts are inter-related, an
integral analysis is mandatory for reliable pulley design. A ®nite element software has been developed for the
analysis of pulleys utilizing a 3-noded thick conical shell element and a triangular ring element. This work considers
all the well-known assuctmptions on load variations along the pulley face-width and around its periphery. A semi-
analytical approach using Fourier analysis has been adopted. To validate the software, comparison has been carried
out with published experimental and theoretical investigations. A few parametric studies are also reported to
elucidate the mutual inter-dependence of stresses in the pulley components. # 1999 Elsevier Science Ltd. All rights
reserved.

Keywords: Pulleys; Belt conveyors; Finite element method

1. Introduction mation in treating the elastic coupling between them.

Speci®cally, their displacements are not coupled at
Pulleys are functionally very critical for belt con- their connections. This leads to signi®cant errors in
veyor operation. In view of its complex and varied the stress and strain ®elds about the connectors.
construction (Fig. 1), the standard practices (viz. Also, these methods cannot be extended to complex
CEMA MHEA [1, 2]) on pulley design are hardly ade- shapes as those of pulleys having tapered webs or
quate for sizing all its parts. As pointed out by many provided with taper-lock arrangements. On the con-
e.g. Refs [2, 3] the stresses and deformations in the rim, trary, the ®nite element method, though a little
web, hub, shaft, and their interconnections are tedious and time consuming, o€ers a practical sol-
mutually related. For an optimum design, all parts ution technique. But the available ®nite element
comprising the overall unitÐthe pulleyÐshould be studies [10±13] either do not provide adequate infor-
integrally analyzed. mation about the loadings, formulations and bound-
All the theoretical studies [4±9] using classical ana- ary conditions adopted or do not verify their
lytical approaches have considered the pulley in validity with respect to experimental data.
parts rather than as a single structure. In this Hence, with the requirement of larger and heavier
approach, computation is easy, as a close-form sol- conveyor systems, a properly veri®ed software is
ution exists and it takes less execution time. But the required for the design analysis of a complete pulley.
solutions near the connection region between the This paper discusses the details of a ®nite element
parts cannot be accurate because of the approxi- model developed [14] and con®rms its validity with
respect to published data. A few parametric studies on
varied con®guration of the parts are also reported to
* Corresponding author. illustrate the inter-relation of stresses in the pulley
1
Presently with Tata Consultancy Services, Madras components.

0045-7949/99/$ - see front matter # 1999 Elsevier Science Ltd. All rights reserved.
PII: S 0 0 4 5 - 7 9 4 9 ( 9 8 ) 0 0 1 4 5 - X
304 M. Ravikumar, A. Chattopadhyay / Computers and Structures 71 (1999) 303±332

Fig. 1. Typical sectional view of a conveyor pulley.

2. Loading details the belt wrap around its rim. Whereas the driven
pulley, having equal belt tension on either side
In a belt±pulley drive (with no-slip, and friction [Fig. 2(b)], can be considered to be subjected to only
coecient = m) the belt tensions (Fig. 2) are given by radial load.
the Euler's equation:
T1 ˆ T2 emyr : …1† 2.1. Loading around the circumference

The varying belt tension around a drive pulley No well established experimental or real-life data is
[Fig. 2(a)] results in radial and tangential loads over available on the actual load variation around the

Fig. 2. Belt tension around pulley.

 M. Ravikumar, A. Chattopadhyay / Computers and Structures 71 (1999) 303±332 305

(drive) pulley circumference. From the Eulerian con- Both ®rst order (m = 1) and third order (m = 3) har-
cept, the variation is considered to be exponential; for monics are considered in the present work. Besides
practical purposes it is often assumed to be linear. these sinusoidal variations, a uniform load distribution
Both the types of variations are studied here. For any along the face-width has also been considered as fol-
angle, y (y1EyE y2), the tension is given by: lows:
exponential T…y† ˆ T2 e m…y2 ÿy† ; …2†
T…z† ˆ T…y†=b: …7†
T1 ÿ T2
linear T…y† ˆ T2 ‡ …y2 ÿ y†; …3†
…y2 ÿ y1 †
for 0E y < y1 and y2 < yE2p, T(y) = 0.
The corresponding radial force per unit area on the
rim surface is obtained as 3. Finite element formulation

T…y† Pulley structure is a solid of revolution, symmetric

Pr …y† ˆ : …4†
R
b about its shaft axis and, its material properties can be
From the above, the tangential force per unit area can considered to be so. But the loading on it is asym-
be expressed thus: metric (Fig. 2). Hence, the stress variations and dis-
placements are three dimensional rather than axially
Pt …y† ˆ m
Pr …y†: …5† symmetric. Analyzing such problems three dimension-
ally is computationally expensive. Instead, a semi-ana-
lytical approach based on the Fourier series [15] is
2.2. Loading along the face-width used. Here, the load is expanded into a Fourier series
of n symmetric and anti-symmetric load components.
The load variation along the pulley face-width (The radial loading on a driven pulley is constant over
depends on the belt carcass, troughing angle of the the belt wrap and is symmetric about the y = 08 plane.
approaching belt, etc. Various types of idealization For such a situation, the anti-symmetric terms shall
have been adopted by earlier workers [4, 10]: not exist.) The series represents a periodic function of
the circumferential angle. Each component analysis
. uniform distribution; remains mathematically two dimensional. The sym-
. ®rst order sinusoidal distribution; and metric and anti-symmetric displacement components
. third order sinusoidal distribution. are evaluated separately considering the corresponding
load components, using the strain±displacement
Commonly, the load distribution (Fig. 3) is expressed
matrices discussed below. According to the principle of
by the sum of a few harmonics of sinusoidal function
superposition, the cumulative solutions for all the com-
of axial coordinate (z) as given by:
ponents represent the response of an axisymmetric
pT…y† mp…b=2 ÿ z† body to the asymmetric load, from which the stresses
T…z† ˆ sin : …6†
2b b are evaluated.

Fig. 3. Belt tension variation along the face-width.

306 M. Ravikumar, A. Chattopadhyay / Computers and Structures 71 (1999) 303±332

3.1. Element details the n harmonics) can be expressed as the sums of sym-
metric and anti-symmetric components:
Two types of axisymmetric elements have been
X
3 X
3
adopted for this work: u0 ˆ Ni u 0i cos ny ‡ Ni u0i sin ny;
iˆ1 iˆ1
1. conical shell element, since it is computationally e- X
3 X
3
cient; and u0 ˆ Ni u 0i cos ny ‡ Ni u 0i sin ny;
2. triangular ring element, as it can approximate com- iˆ1 iˆ1
plicated shapes. X
3 X
3
v0 ˆ Ni v0i sin ny ÿ Ni v0i cos ny;
3.1.1. Conical shell element iˆ1 iˆ1

This is a 3-noded isoparametric thick conical shell X

3 X
3
v0 ˆ Ni v0i sin ny ÿ Ni v 0i cos ny;
element [Fig. 4(a)] developed based on the Reisner±
iˆ1 iˆ1
Naghdi shell theory [16]. Here, the displacement
X
3 X
3
equations are w0 ˆ Ni w 0i cos ny ‡ Ni w0i sin ny;
U…s; y; z† ˆ u0 …s; y† ‡ zu0 …s; y†; iˆ1 iˆ1
X
3 X
3
V…s; y; z† ˆ v0 …s; y† ‡ zv0 …s; y†; w0 ˆ Ni w 0i cos ny ‡ Ni w 0i sin ny;
2
z 00 iˆ1 iˆ1
W…s; y; z† ˆ w0 …s; y† ‡ zw0 …s; y† ‡ w …s; y† …8† X
3 X
3
2 00
w00 ˆ Ni w 00i cos ny ‡ Ni wi sin ny …9†
0 0
where u and v represent rotations of the tangents to iˆ1 iˆ1
the reference surfaces of the shell about s and y coordi-
nates and w 0 and w0 are terms contributing to a non- where the single barred variables (u 0i, v0i, etc.) rep-
vanishing transverse normal strain [16]. The seven resent the symmetric amplitudes, while the double
dependent variables u0, v0, w0, u 0 , v 0 , w 0 and w0 are the barred variables …u0i ; v0i , etc) are the additional anti-
seven degrees of freedom. For a general case of asym- symmetric amplitudes. And Ni (i = 1, 2, 3) are shape
metric loading, the displacement relations (for each of functions given by the expressions:

Fig. 4. Details of elements with coordinate systems: (a) view from right end; (b) 3-noded thick conical shell; (c) triangular ring.
 M. Ravikumar, A. Chattopadhyay / Computers and Structures 71 (1999) 303±332 307

x2 ÿ x x2 ‡ x
N1 ˆ ; N2 ˆ 1 ÿ x2 ; N3 ˆ ; ‰x ˆ s=lŠ: …10†
2 2
Substituting the above symmetric and anti-symmetric displacement components in the general strain±displacement
relations, the following matrices are obtained for the corresponding cases.

8 9
> X 3  
> @Ni @Ni 0 >
>
>
> u0i ‡ z ui cos ny >
>
>
> @s @s >
>
>
> iˆ1 >
>
>
> >
>
>
> X3      >
>
>
> 1 n u w n u0
w0
z2
w 00 >
>
>
> N v ‡
0i
sin f ‡
0i
‡ z N v 0
‡ i
sin f ‡ i
‡ i
cos ny >
>
>
> 1 ‡ z r
i 0i
r l r
i i
r l 2 l >
>
>
> l iˆ1 >
>
8 9 >
> >
>
> e s > >
> X3   >
>
>
> > > >
>
>
> e y >>
>
>
>
> 0 00
wi ‡ zwi cos ny
>
>
>
< e > = >
< iˆ1
>
=
z
ˆ ; …11†
> g sy >
> > > >X 3      >
>
>
> >
> >
> @Ni @Ni 0 1 n v0i n 0 v0i >
>
>
> g > > v0i ‡ z v ÿ u0i ÿ sin f ‡ z ÿ ui ÿ sin f sin ny >
: sz > ; > >
> @s @s i 1 ‡ lz r r r r >
>
>
g yz >
> iˆ1 >
>
>
>   >
>
>
> X 3
@N @N z2
@N >
>
>
> i i 0 i 00 0 >
>
>
> w0i ‡ z wi ‡ wi ‡ ui cos ny >
>
>
> @s @s 2 @s >
>
>
> iˆ1 >
>
>
>   >
>
>
> X3 >
>
>
> 1 n n 0 2 n 00 v0i 0 >
>
>
: z
ÿ N w
i 0i ÿ z N w
i i ÿ z N w
i i ÿ ‡ v i sin ny >
;
1 ‡ l iˆ1 r r 2r l

8 9
> X 3  
>
>
> @Ni @Ni 0 >
>
>
> u0i ‡ z ui sin ny >
>
>
> @s @s >
>
>
> iˆ1 >
>
>
>      >
>
>
> 1 X3
n u w n u 0
w 0
z 2
w 00 >
>
>
> 0i 0i >
>
> ÿ Ni v0i ‡ sin f ‡ ‡ z ÿ Ni vi ‡ sin f ‡ 0 i i
‡ i
sin ny >
>
>
>
> 1 ‡ z r r l r r l 2 l >
>
8 9 > >
>
l iˆ1 >
>
>
> e > >
> X3   >
>
>
>
s >
> >
> >
>
>
> e y >
> >
> w 0
‡ zw 00
sin ny >
>
>
< = < > > i i >
=
ez iˆ1
ˆ      : …12†
>
> gsy >> >
> X 3
@N @N 1 n v n v 0 >
>
>
> >
> > >
> i i 0 0i 0 i >
>
>
> g > v0i ‡ z v u0i ÿ sin f ‡ z u ÿ sin f cos ny >
: sz > ; > >
> @s @s i 1 ‡ lz r r r i r >
>
>
gyz >
> iˆ1 >
>
>
>   >
>
>
> X3
@N @N z 2
@N >
>
>
> i i 0 i 00 0 >
>
>
> w 0i ‡ z w i ‡ w i ‡ u i sin ny >
>
>
> @s @s 2 @s >
>
>
> iˆ1 >
>
>
>   >
>
>
> X3 >
>
>
> 1 n n 0 2 n 00 v0i 0 >
>
>
: z
N i w 0i ‡ z N i wi ‡ z N i w i ÿ ‡ vi cos ny >
;
1 ‡ l iˆ1 r r 2r l

For rim, f = 0, s = z and z = r; and for web, f = 908, s = r and z = z.

3.1.2. Triangular ring element

For the triangular ring element [Fig. 4(b)] [15] the displacements are, as before
X X
Uˆ u cos ny ‡ u sin ny;
X X
Vˆ v sin ny ÿ v cos ny;
X X
Wˆ w cos ny ‡ w sin ny: …13†

Assuming linear displacement models for shape function and using isoparametric form, the strain±displacement
matrices for the symmetric and anti-symmetric cases are:
308 M. Ravikumar, A. Chattopadhyay / Computers and Structures 71 (1999) 303±332
8 9
> @u >
>
> cos ny >
> has been undertaken to select the number of Fourier
>
> @z >
>
>
>   >
> harmonics n.
>
> w v >
>
>
> ‡ n cos ny >
>
>
> >
>
8 9 >
> r r >
>
> 
e >>
> >
> 4.1. Rim
>
>
z >
>>
> @w >
>
>
> 
e y >
>>
> cos ny >
>
>
< = >>
< >
=
e r @r
ˆ   ; …14† Fig. 5 describes the axial and circumferential stresses
>
> g zy >> >
> u @v >
> around the rim periphery as evaluated from the present
>
> > > ÿ n ‡ sin ny >
>
> g > > > > r @z >
> software. Here, all the above mentioned types of load
: zr > ; > >
>  
>
>
>
g yr >
> @w @u >
> distributions along the face-width have been considered.
>
> >
>
>
> ‡ cos ny >
> Lange's experimental values are also shown therein. It
>
> @z @r >
>
>
>   >
>
>
> >
>
>
> w @v v >
: ÿ n ‡ ÿ sin ny > ;
r @r r
8 9
> @u >
>
> sin ny >
>
>
> @z >
>
>
>   >
>
>
> w v >
>
>
> ÿ n sin ny >
>
>
> >
>
8 9 >
> r r >
>
> e z >
>
> >
>
>
> >
> >
> @w >
>
>
> eŠ >
> >
> sin ny >
>
>
< = y> >
< >
=
er @r
ˆ   : …15†
>
> gzy >
> >
> u @v >
>
> >
> >
> n ‡ cos ny >
>
>
>
> g > > r @z >
: zr >; > >
>  
>
>
>
gyr >
> @w @u >
>
>
> >
>
>
> ‡ sin ny >
>
>
> @z @r >
>
>
>   >
>
>
> >
>
>
> w @v v >
: n ‡ ÿ cos ny > ;
r @r r

3.2. Boundary conditions

Many [4, 7, 13] have analyzed the rim as a cylindrical

shell simply supported at the rim±web junction. For a
preliminary analysis to start with, a similar approach is
followed here. Both circumferential and radial degrees
of freedom at these support points are assumed to be
zero.
Subsequently, for integral analysis, the pulley shaft is
considered to be simply supported at the bearings. Both
circumferential and radial degrees of freedom are sup-
pressed at this location. Because of geometric symmetry
about the mid-section, only one-half of the pulley is
analyzed by suppressing the axial degree of freedom for
all nodes at this section.

4. Validatory studies

Detailed analysis has been carried out on a pulley

model studied by Lange [4]. His work being restricted Fig. 5. Rim stresses around the circumference at mid-section
to driven pulley (i.e. T1 = T2) only, analysis on a drive for di€erent loading considerations; (a) conical shell element;
pulley (i.e. T1 > T2) has been undertaken on another (b) simply supported rim. . . . Experiment, Uniform dis-
con®guration for which theoretical investigation have tribution, - - - - 3rd order sinusoidal distribution, ±±± 1st order
been reported by Das and Pal [7]. A convergence study sinusoidal distribution.
 M. Ravikumar, A. Chattopadhyay / Computers and Structures 71 (1999) 303±332 309

may be observed that the peak stresses calculated pulley with exponential and linear increase in tension
assuming uniform distribution are quite close to the ex- are furnished in Fig. 9. Both the cases provide almost
perimental values as against those obtained considering the same results throughout.
®rst and third order sinusoidal variations. This is A comparison of the stresses obtained by considering
because the tensioning arrangement in the experimental the conical shell and triangular ring elements is reported
set-up provided uniform load along the pulley face. in Fig. 10. Both match quite closely. But, it is noted
Fig. 6 compares Lange's experimental plots of stresses that computation with a conical shell element is about 3
along the face-width at di€erent peripheral angles with times faster than with the triangular element and it
the prediction of this software. While the agreement is requires about 2 times less memory.
reasonable in most of the cases, for y = 958, the trend Fig. 11 provides a comparison of the stresses around
of experimental plot di€ers marginally. However, in the the rim periphery calculated by assuming the rim as
absence of data at all points, the experimental graph simply supported at the rim±web junction, and consider-
appears to be approximate. ing the rim as an integral part of the entire pulley.
The stress variations around the rim at the mid-sec- Lange's experimental values are also superimposed
tion due to various wrap angles (60, 120, 180, and 2408) thereon. It is interesting to observe that the results from
are furnished in Fig. 7. In line with the observations analysis with simply supported consideration is in very
reported earlier [7, 13], the stress magnitude tends to go good agreement with those from the integral analysis
up with reduced wrap angle due to localization of the and experiment.
applied load. The location of maximum stress also
depends signi®cantly on the wrap angle. 4.2. Web
Fig. 8 compares the rim stress variation along the
face-width as calculated for a drive pulley reported in The web stresses are closely dependent on the overall
Ref. [7]. The stresses around the periphery of the drive rigidity of the pulley structure [3]. Hence, unlike in the

Fig. 6. Maximum rim stresses along the face-width. . . . Experiments, FEM.

310 M. Ravikumar, A. Chattopadhyay / Computers and Structures 71 (1999) 303±332

Fig. 7. Rim stresses around the circumference at mid-section for various wrap angles. Yr = 608, - - - - - Yr = 1208,
± ± ± ± Yr = 1808,Ð Ð Yr = 2408.
 M. Ravikumar, A. Chattopadhyay / Computers and Structures 71 (1999) 303±332 311

rim, the stresses in the web need to be analyzed consid- not reported in Lange's work, and hence are not com-
ering the entire pulley as an integral structure. For a pared. Fig. 14 shows the de¯ections at those points as
complex shape such as an entire pulley, the shell element compared with measured values.
can not be used. Therefore, the triangular ring element
has been adopted for the integral analysis. 4.3. In¯uence of aspect ratio
Fig. 12 indicates that the radial and circumferential
stresses in the web compare well with Lange's exper- For a reliable ®nite element analysis, usually the
imental data. The radial stresses at various radii on the aspect ratios should be within 1±3. Since the pulley is
web (presented in Fig. 13) also agree closely with the ex- constituted of parts mostly having thin cross sections, it
periments. Circumferential stresses at these points are is very convenient to have meshes with higher aspect
ratios. This reduces the total number of elements, mini-
mizing the required computational memory and time. A
comparative study with 3 di€erent ratios, 1.25, 2 and
2.5 was made to ascertain a suitable value. Veri®cation
with experimental investigations on web stresses (Fig. 15)
shows that analysis with an aspect ratio of around 1.25
produces good results. Hence, aspect ratios of 1±1.5
have generally been adopted in the areas of interest.
Fig. 16 shows a typical mesh pattern used for estimating
the stresses at the taper-lock interfaces with shaft and
hub.

5. Design studies

A pulley having 800 and 100 mm rim and shaft diam-

eters, for a 1000 mm wide belt, has been taken as the
basic model for this parametric study. The thicknesses
of rim and web, etc., have been calculated from the
MHEA practice, which gives the maximum information
in this regard. A taper-lock hub connection of 100 mm
bore (DOBIKON 1012-100  145) [17] has been
adopted, whose full torque capacity has been assumed
to be working on the shaft. Based on this torque, the
belt tensions (T1 and T2) have been worked back con-
sidering a wrap angle of 1808 and a friction coecient
of 0.3 between the pulley and belt. The pulley has been
assumed to be driven, subjected to the total belt tension
(T1 + T2) of 149.704 kN.
Since the stress calculation procedure for the shaft is
well-known, stresses in the rim and webÐbeing the
other critical parts of a pulleyÐhave, in particular, been
reported here. The following aspects have been investi-
gated:

1. in¯uence of di€erent web locations referred to the

belt edge;
2. e€ects of providing sti€ening rings inside the rim;
3. changes in stresses due to di€erent shaft diameters;
4. stresses in tapered web pro®les.

Fig. 8. Rim stresses at the outer and inner surfaces at y = 1208;

rim diameter = 626 mm, face-width = 712 mm, and thick- 5.1. Web location
ness = 12 mm; T1 = 13.223 kN, T2 = 6.029 kN. Axial
inside, - - - - Circumferential inside, Y Y Y Ref. 7 (Das and The overhang of the rim beyond the web is usually
Pal). decided empirically, ensuring that the hub does not
312 M. Ravikumar, A. Chattopadhyay / Computers and Structures 71 (1999) 303±332

Fig. 9. Rim stresses around the circumference at mid-section for linear and exponential variation in belt tensions. Linear,
Y Y Y Exponential.
 M. Ravikumar, A. Chattopadhyay / Computers and Structures 71 (1999) 303±332 313

Fig. 10. Rim stresses around the circumference at mid-section Fig. 11. Rim stresses around the circumference at mid-section
with di€erent elements; simply supported rim. . . . for di€erent supporting conditions. . . . Experiments,
Experiments, - - - - Triangular ring element, Conical Y Y Y Rim simply supported at webs, Integral pulley.
shell element.
A comparison of the stresses in the rim mid-section
project beyond the rim. MHEA shows that when the around the periphery due to the three web locations is
distance between the webs equals one-half the bearing shown in Fig. 18. Similarly, Fig. 19 depicts the maxi-
centers, the web stress is highest. Since the location of mum stresses in the web. The maximum radial stresses
the web in¯uences the stresses in other parts of the in the web obtained from this study were found to com-
pulley as well, a rational approach is needed to ®nd out pare well with the magnitude calculated from the
the web location, minimizing the stresses in all related MHEA formula for the assumed loading on this model
parts. Hence, the study has been undertaken with the pulley. It is clear that the ®rst location causes the mini-
following three locations of the web, shown in Fig. 17. mum stresses. It is important to observe from these
plots that, though the web stresses at the outer radii are
1. Web placed along the ideal belt edge; very low, they increase considerably near the hub±web
2. web shifted outside, near to the pulley edge; and junction. As reported elsewhere [3, 4, 14], most pulley
3. web shifted inside, away from the pulley edge. failures are initiated from this region only.
314 M. Ravikumar, A. Chattopadhyay / Computers and Structures 71 (1999) 303±332

5.2. Sti€ening rings

For wide faced pulleys, a common industrial practice

is to provide one or two sti€ening ring(s) inside the rim
circumference. No theoretical or experimental investi-
gation has been reported on the in¯uence of such stif-
fener(s). A study has been undertaken with the same
pulley model with its web along the ideal belt edge, and
a 20 mm thick sti€ening ring at its center. For all practi-
cal purposes, besides the number of rings, its radial
width (decided by the inner diameter) contributes the
most to sti€en the rim. So three cases of rings with 590,
405, and 225 mm inner diameters (approximately 90, 60,
and 30% of the annular gap between rim and shaft)
have been studied. The web stresses are almost indepen-
dent of the inner diameter of the rings. For 405 and
225 mm diameter rings, no signi®cant di€erence in stress
levels is noted. Hence, results on 225 mm diameter rings
are not discussed. Quite obviously, as seen from Fig. 24,
the peak stresses at the mid-section of the rim are
reduced due to the rings. Another interesting obser-
vation is that the occurrence of maximum stresses takes
place at 0 and 1808 circumferential angles, as opposed
to approximately 70 and 1108 for the case with a plain
rim without sti€ener (Fig. 5). It is important to note
from Fig. 25 that, on providing a sti€ener, the web
stresses also increase.

5.3. Shaft

The relation between shaft diameter and stresses in

other pulley parts have not been quantitatively pub-
lished so far. Hence, studies have been carried out with
the same pulley con®guration but with shafts having
100, 90 and 135 mm diameters. While the ®rst matches
the taper-lock adopted, the second is barely adequate
for the assumed loading from stress considerations, and
the last is obtained from CEMA recommendations tak-
Fig. 12. Maximum web stresses along the radius. . . .
ing de¯ection into account. The stresses in the rim with
Experiments, FEM
di€erent shaft sizes are almost the same. Hence they are
not discussed here. However, the web stresses (Fig. 26)
reduce with increasing shaft diameter, especially near
The stresses at the rim±web junction are also com- the hub±web junction.
pared in Fig. 20 for the three web locations. The corre-
sponding stresses at the shaft±hub interface with the 5.4. Tapered webs
plain hub is shown in Fig. 21. The three con®gurations
have further been studied providing a taper-lock con- Some of the popular pro®les of tapered webs are
nection between the hub and shaft. Figs. 22 and 23 pre- shown in Fig. 27. Unlike the pro®les of both type B and
sent the stresses at the hub/taper-lock and shaft/taper- C webs, which have a weak zone `a' near the hub [4],
lock interfaces, respectively. The stress pro®les appear type A appears to be safer. Published investigation on
to qualitatively match well with the manufacturer's this third type (A) being few, a demonstrative study has
data [17]. From these overall comparisons, it may be been carried out with two web pro®les of this shape
concluded that locating the web along the ideal belt with a small variation in the thicknesses as shown in
edge will result in minimum stresses in all critical Fig. 28. It is observed from the stress plots (Fig. 29)
points. that, by varying the thickness gradually, the web can be
M. Ravikumar, A. Chattopadhyay / Computers and Structures 71 (1999) 303±332 315

Fig. 13. Web stresses at di€erent radii. . . . Experiments, FEM.

316 M. Ravikumar, A. Chattopadhyay / Computers and Structures 71 (1999) 303±332

Fig. 14. Web de¯ection at di€erent radii. . . . Experiment, FEM.

 M. Ravikumar, A. Chattopadhyay / Computers and Structures 71 (1999) 303±332 317

Fig. 15. In¯uence of aspect ratio on web stresses. . . . Experiments, - - - - Aspect ratio = 1.25, Aspect ratio = 2, ±±±Aspect
ratio = 2.5.
318 M. Ravikumar, A. Chattopadhyay / Computers and Structures 71 (1999) 303±332

Fig. 16. Typical mesh considered for analyzing stresses at taper-lock interfaces.
M. Ravikumar, A. Chattopadhyay / Computers and Structures 71 (1999) 303±332 319

Fig. 17. Di€erent web locations.

320 M. Ravikumar, A. Chattopadhyay / Computers and Structures 71 (1999) 303±332

Fig. 18. Variation in rim stresses at mid-section around the circumference for di€erent web locations. Web locationÐ1,
- - - Web locationÐ2, ±±±Web locationÐ3.
 M. Ravikumar, A. Chattopadhyay / Computers and Structures 71 (1999) 303±332 321

Fig. 19. Variation in maximum web stresses for di€erent web locations. - - - Web locationÐ1, Web locationÐ2, ±±±Web
locationÐ3.
322 M. Ravikumar, A. Chattopadhyay / Computers and Structures 71 (1999) 303±332

Fig. 20. Variation in stresses at rim±web junction for di€erent web locations. Web locationÐ1, - - - Web locationÐ2,
±±±Web locationÐ3.
 M. Ravikumar, A. Chattopadhyay / Computers and Structures 71 (1999) 303±332 323

Fig. 21. Variation in stresses at plain-hub/shaft interface for di€erent web locations, at y = 1808. - - - Web locationÐ1, ±±±Web
locationÐ2, Web location 3.
324 M. Ravikumar, A. Chattopadhyay / Computers and Structures 71 (1999) 303±332

Fig. 22. Variation in stresses at hub/taper-lock interface for di€erent web locations, at y = 1808. - - - Web locationÐ1, Web
locationÐ2, ±±±Web locationÐ3.
 M. Ravikumar, A. Chattopadhyay / Computers and Structures 71 (1999) 303±332 325

Fig. 23. Variation in stresses at shaft/taper-lock interface for di€erent web locations, at y = 1808. - - - Web locationÐ1, Web
locationÐ2, ±±±Web locationÐ3.
326 M. Ravikumar, A. Chattopadhyay / Computers and Structures 71 (1999) 303±332

Fig. 24. Variation in rim stresses around the circumference at mid-section due to di€erent sti€eners. Sti€ener caseÐ1,
- - - Sti€ener caseÐ2, ±±±Without sti€ener.
 M. Ravikumar, A. Chattopadhyay / Computers and Structures 71 (1999) 303±332 327

Fig. 25. Variation in web stresses due to di€erent sti€eners. With sti€ener, - - - Without sti€ener.
328 M. Ravikumar, A. Chattopadhyay / Computers and Structures 71 (1999) 303±332

Fig. 26. Variation in web stresses due to di€erent shaft diameters. 90 mm F, - - - 100 mm F, ±±±135 mm F.
M. Ravikumar, A. Chattopadhyay / Computers and Structures 71 (1999) 303±332 329

Fig. 27. Typical tapered webs; (a) type A; (b) type B; (c) type C.
330 M. Ravikumar, A. Chattopadhyay / Computers and Structures 71 (1999) 303±332

Fig. 28. Di€erent pro®les of tapered webs.

 M. Ravikumar, A. Chattopadhyay / Computers and Structures 71 (1999) 303±332 331

comprehensive knowledge of stresses and deformations

in all pulley parts, an integral analysis is essential.
The integral analyses with various con®guration of
pulley parts has been helpful in arriving at some broad
design guidelines. For an optimum design, the location
of the web should be in line with the belt edge. Since
the belt wanders on the pulley surface during operation,
a situation like case 3 [Fig. 17(c)] may arise, which
causes higher stresses in all parts. Hence, for compu-
tation purposes, the web should be assumed to be
located inside by twice the edge clearance. Since taper-
lock induces considerable stress in the shaft and hub,
this aspect should be considered in the shaft/hub design.
On providing sti€ening rings, the rim stresses reduce,
but the web may be subjected to critical stresses.
Increase in shaft diameter decreases the web stresses, as
with increased rigidity, the shaft bends less, and the ¯ex-
ural stress in the web is reduced. By using a suitable
tapered web pro®le it is possible to arrive at an almost
uniform stress distribution in the web. An optimization
study can be undertaken to ensure near-permissible
stresses throughout the web thereby reducing the pulley
weight and, at the same time, increasing the fatigue life
of the web. All these aspects should be taken into
account while deciding the pulley con®guration and
designing the parts.

References

[1] Conveyor Equipment Manufacturers Association U.S.A..

Speci®cation for welded steel conveyor pulleys with com-
pression type hubs. ANSI/CEMA B105. 1±1992.
Maryland, 1987.
[2] Mechanical Handling Engineers' Association U.K.
Recommended practice for troughed belt conveyors.
London, 1986.
[3] King TJ. Belt conveyor pulley designÐwhy the failures?.
Fig. 29. Stresses in di€erent tapered webs. - - - Web pro®leÐ Bulk Solids Handling 1986;6:339±47.
a, Web pro®leÐb. [4] Lange H. Untersuchungen zur Beanspruchung von
Foerderbandtrommeln. Dr -Ing dissertation. TH
Hannover, Germany, 1963.
[5] Scholich S. Spannungen und Verformungen an
stressed with near uniformity and the stress peak can be Gurtbandtrommeln. Bergbautechnik 1965;15:517±22.
brought down signi®cantly, eliminating any risky zone. [6] Schmoltzi W. Zur Auslegung von Trommeln mit
Durchgefuehrter Welle fuer Gurtfoerderanlagen. Dr -Ing
dissertation. TH Hannover, Germany, 1974.
[7] Das SP, Pal MC. Stresses and deformations of a conveyor
6. Conclusions power pulley under exponential belt tensions. Comp Struct
1987;27:787±95.
[8] Hong-Sheng Lu. Shell strength of conveyor belt pulleys:
The results obtained from the present work are found
theory and design. Int J Mech Sc 1988;30:333±45.
to match quite well with the available data. Depending [9] Longman D.J. Design of belt conveyor pulley drums. In:
upon the service (i.e. application and location) of pull- International Mechanical Engineering Congress. Perth,
eys, suitable loading considerations can be undertaken Australia, 1994:121±128.
by the software. It is also noted that, while simply sup- [10] Linder C. Berechnung einer Bandtrommel mit dem Finite-
ported consideration is adequate for rim analysis, for a Elemente-Program TPS 10. Braunkohle 1975;3:81±97.
332 M. Ravikumar, A. Chattopadhyay / Computers and Structures 71 (1999) 303±332

[11] Vodstrcil R. Analysis of a belt conveyor pulley using ®nite [14] Ravikumar M. Studies on conveyor pulley design using
element methods. In: 4th International Conference in ®nite element method. M.S. thesis. IIT Madras, India,
Australia on Finite Element Methods. Melbourne, 1995.
Australia, 1982:143±148. [15] Zienkiewicz O.C. The ®nite element method in engineering
[12] Qiu Xiangjun, Sethi V. A new pulley stress analysis science. London: McGraw Hill, 1971.
method based on modi®ed transfer matrix. Bulk Solids [16] Sivadas KR, Ganesan N. Vibration analysis of thick com-
Handling 1993;13:713±24. posite clamped conical shells of varying thicknesses. J
[13] Siva Prasad N, Sarma Radha. A ®nite element analysis for Sound Vibr 1992;152:27±37.
the design of a conveyor pulley shell. Comp Struct [17] BIKON-Technik GmbH Germany. Welle-Nabe-
1990;35:267±77. Verbindungen.