You are on page 1of 13

A New Pulley Stress Analysis Method Based on Modified Transfer Matrix

07/07/2014

A New Pulley Stress Analysis Method Based on Modified Transfer Matrix

Print

Xiangjun Qiu and Vinit Sethi, USA


Summary
A new pulley stress analysis m ethod is presented. It shall be referred to as the Modified Transfer Matrix (MTM) m ethod. This m ethod is based on a
reform ulation of transfer m atrices for the pulleys cylindindrical shell, end-disk plate with non-uniform thick ness and shaft by using finite elem ent
concepts. It com bines the strength of both classical stress analysis m ethods and finite elem ent m ethods. It proves to be an efficient and effective
approach in determ ining the stresses in a pulley. A pulley stress analysis software program nam ed PSTRESS 3.0 has been developed based on this
new m ethod. At the end of the paper, a num erical ex am ple of the pulley stress analysis, using PSTRESS 3.0, is given. The result is satisfactorily
com pared with that obtained in a finite elem ent m odel (ANSYS) solution with a very fine m esh.
Nomenclature
x

Cartesian coordinate in horizontal direction

y
z
r

Cartesian coordinate in vertical direction


Cartesian coordinate in pulley ax ial direction
r cylindrical coordinate in pulley radial direction

P
M
q
Ws

cylindrical coordinate in pulley circum ferential direction


shear force acting on shaft cross section
bending m om ent acting on shaft cross section
distributed load acting on shaft ax is
transverse displacem ent of shaft neutral ax is

rotational angle of shaft cross section

EI
a
G

shaft bending stiffness


shaft cross section area
shear m odulus of shaft m aterial
shaft state variable vector
m atrix containing coefficients of governing O DEs for
shaft bending deform ation
vector representing the non-hom ogeneous term of the governing O DEs for shaft bending
deform ation
transfer m atrix for the governing O DEs for shaft
bending deform ation
identity m atrix

A
B
T
I
KBM
U BM

TMB beam elem ent stiffness m atrix


TMB beam elem ent displacem ent vector

F BMex t TMB beam elem ent ex ternal force vector


F BMint TMB beam elem ent internal force vector
Qr
Q
Mr
M
Mr
g
u
v
w
E
t
c
p
D
m
um

transverse shear force acting on the disk cross section


perpendicular to radial direction
transverse shear force acting on the disk cross section
perpendicular to circum ferential direction
bending m om ent acting on the disk cross section
perpendicular to disk radial direction
bending m om ent acting on the disk cross section
perpendicular to disk circum ferential direction
twisting m om ent acting on the disk cross sections perpendicular
to disk radial direction and disk circum ferential direction
ex ternal transverse force acting on the neutral surface of the disk
pulley disk and shell displacem ents in pulley ax ial direction
pulley disk and shell displacem ents in pulley circum ferential direction
pulley disk and shell displacem ents in pulley radial direction
YO UNG'S m odulus of disk or cylindrical shell of pulley
PO ISSO N'S ratio of disk or cylindrical shell of pulley
thick ness of disk or cylindrical shell of pulley
coefficient to describe the geom etry of disk variable thick ness
ex ponential num ber to describe the geom etry of disk variable
thick ness
cross section bending stiffness of disk or cylindrical shell
FO URIER com ponent num ber
FO URIER com ponent of u

vm

FO URIER com ponent of v

wm

FO URIER com ponent of w

gm

FO URIER com ponent of g

Q rm

FO URIER com ponent of Q r

Qm

FO URIER com ponent of Q

M rm

FO URIER com ponent of M r

Mm

FO URIER com ponent of M

M rm

FO URIER com ponent of M r

rotational angle of disk or cylindrical shell of pulley

disk bending state variable vector of FO URIER com ponent m

http://www.ckit.co.za/secure/conveyor/papers/troughed/pulley/pulley.htm

1 / 13

A New Pulley Stress Analysis Method Based on Modified Transfer Matrix


m rm

disk circum ferential harm onic resultant bending m om ent

Vrm

disk circum ferential harm onic resultant transverse shear force

Am

m atrix containing coefficients of governing O DEs for disk bending deform ation of FO URIER
com ponent m

Bm

vector representing the non-hom ogeneous term of the governing O DEs for disk bending
deform ation of FO URIER com ponent m

U BDm

07/07/2014

TMB disk bending elem ent displacem ent vector

KBDextm TMB disk bending elem ent stiffness m atrix


F BDextm TMB disk bending elem ent ex ternal force vector
F BDintm TMB disk bending elem ent internal force vector
Nr

norm al force acting on the disk cross section perpendicular to radial direction

norm al force acting on the disk cross section perpendicular to circum ferential direction

Nr

in-plane shear force acting on the disk cross sections perpendicular to disk radial direction
and disk circum ferential direction

Nrm

FO URIER com ponent of Nr

Nm

FO URIER com ponent of N

Nrm
m

FO URIER com ponent of Nr

Cm

m atrix containing coefficients of governing O DEs for disk In-plane deform ation of FO URIER
com ponent m

disk plane-stress state variable vector of FO URIER com ponent m

U PNm

TMB disk plane-stress elem ent displacem ent vector

KPNm

TMB disk plane-stress elem ent stiffness m atrix

F PNextm

TMB disk plane-stress elem ent ex ternal force vector

F PNintm TMB disk plane-stress elem ent internal force vector


U DKm

TMB disk elem ent displacem ent vector

KDKm

TMB disk elem ent stiffness m atrix

F DK

extm

TMB disk elem ent ex ternal force vector

F DKintm TMB disk elem ent internal force vector


R
N1

cylindrical shell radius

N2

norm al force acting on the cylindrical shell cross section perpendicular to pulley
circum ferential direction
m em brane shear force acting on the cylindrical shell cross sections perpendicular to pulley
ax ial direction and circum ferential direction
transverse shear force acting on the cylindrical shell cross section perpendicular to pulley
ax ial direction
transverse shear force acting on the cylindrical shell cross section perpendicular to pulley
circum ferential direction
bending m om ent acting on the cylindrical shell cross section perpendicular to pulley ax ial
direction
bending m om ent acting on the cylindrical shell cross section perpendicular to pulley
circum ferential direction
twisting m om ent acting on the cylindrical shell cross sections perpendicular to pulley ax ial
direction and circum ferential direction

S
Q1
Q2
M1
M2
M 12

norm al force acting on the cylindrical shell cross section perpendicular to pulley ax ial direction

fz

ex ternal load acting on the neutral surface of shell in pulley ax ial direction

f
fr

ex ternal load acting on the neutral surface of shell in pulley circum ferential direction
ex ternal load acting on the neutral surface of shell in pulley radial direction

f zm

FO URIER com ponent of f z

fm

FO URIER com ponent of f

f rm
N1m

FO URIER com ponent of f r


FO URIER com ponent of N1

N2m

FO URIER com ponent of N2

M 1m

FO URIER com ponent of S

M 2m

FO URIER com ponent of M 1

M 12m

FO URIER com ponent of M 2

V1m

FO URIER com ponent of M 12

xm

FO URIER com ponent of elem ent boundary equivalent transverse shear force

Jm

cylindrical shell state variable vector of FO URIER com ponent m

Im

m atrix containing coefficients of governing O DEs for cylindrical shell of FO URIER com ponent
m
vector representing the non-hom ogeneous term of the governing O DEs for cylindrical shell of
FO URIER com ponent m

M 12
U CSm

TMB cylindrical shell elem ent displacem ent vector

Kcs m

TMB cylindrical shell stiffness m atrix

F CSextm TMB cylindrical shell elem ent ex ternal force vector


F CSintm TMB cylindrical shell elem ent internal force vector
m

state variable vector of FO URIER com ponent m

S
Hm

ax ial or radial coordinate

Lm

vector representing the non-hom ogeneous term of the governing O DEs of a pulley
com ponent

Um

generalized displacem ents of state variable vector of a pulley com ponent

m atrix containing coefficients of governing O DEs of a pulley com ponent

http://www.ckit.co.za/secure/conveyor/papers/troughed/pulley/pulley.htm

2 / 13

A New Pulley Stress Analysis Method Based on Modified Transfer Matrix

07/07/2014

Fm

generalized forces of state variable vector of a pulley com ponent

Tm

transfer m atrix for the governing O DEs of a pulley com ponent

Dm

TMB elem ent displacem ent vector of a pulley com ponent

Km

TMB elem ent stiffness m atrix of a pulley com ponent

F int m

TMB elem ent internal force vector of a pulley com ponent

F ext m TMB elem ent ex ternal force vector of a pulley com ponent
1. Introduction
An engineered class belt conveyor pulley typically consists of a cylindrical shell, two end disk s with variable thick ness, a shaft, and two lock ing devices
connecting end disk s to the shaft as shown in Fig, 1. The pulley is usually subjected to severe bending due to very high belt tensions and lock ing
assem bly pressures, In the design of such a pulley, it is necessary to tak e into account the possibility of fatigue failure. Costly failures in large
conveyor pulleys have led designers to seek detailed stress fatigue or endurance analysis. To date, two types of approaches for pulley stress
analysis have been reported in the literature. O ne is the classical m echanics approach developed by LANGE [1] and SCHMO LTZI [2]. The other is the
finite elem ent m ethod (FEM) em ployed by VO DSTRICL [3], DANIEL [4] and SETHI et al. [5]. Both types of approaches have advantages and
disadvantages.

Fig. 1: Cross-section of pulley assem bly


The classical m echanics approach developed by LANGE and SCHMO LTZI is an approx im ate analytical approach, providing a closed-form solution for
stresses in a pulley. The advantages of this m ethod are that it is easy to program and tak es a very short ex ecution tim e to obtain a solution, The
disadvantage is that the stress solution is not accurate at the locations near the connection region between the shell and end disk s because of its
poor approx im ation in treating the elastic coupling between these com ponents. Specifically, the displacem ent of the end disk and shell are not
coupled at their connection. This leads to significant errors in the stress and strain field about the connectors as will be shown.
The FEM has just the opposite advantages and disadvantages of the LANGE classical m ethod. The m ajor advantage of FEM is its ease of treating
com plex geom etry and boundary conditions. The m ajor disadvantage is its long ex ecution tim e coupled with its need for an ex perienced user to
generate a proper finite elem ent m esh.
In this paper, a new m ethod called the Modified Transfer Matrix (MTM) m ethod is presented, This m ethod circum vents the disadvantages of both the
LANGE classical m ethod and the conventional FEM. The MTM m ethod proves to be a very effective and efficient approach in providing an accurate
pulley assem bly stress solution for any loading condition pulley.
Historically, the transfer m atrix m ethod was developed several decades ago [6] and was very popular in solving one dim ensional static and dynam ic
problem s before the advent of the FEM. Even today, this m ethod is still useful in providing dosed form solutions to certain elasticity problem s with
sim ple boundary conditions [7]. Although there is a lim itation in handling com plicated boundary conditions, such as the boundary conditions for a
pulley, the solution obtained by using the transfer m atrix m ethod is ex act. In this paper, it is shown that the lim itation of the transfer m atrix m ethod
can be overcom e if the transfer m atrix is reform ulated by using finite elem ent concepts. The reform ulated transfer m atrix is essentially a special
finite elem ent, The new m ethod using these special finite elem ents, called transfer m atrix based (TMB) finite elem ents, is capable of solving a class
of structural elasticity problem s (including the elasticity problem of a pulley), whose governing differential equations can be reduced to a set of
ordinary differential equations (O DEs). Regardless of how few of these TMB finite elem ents are used in a m odel, the solution obtained by this MTM
m ethod is generally very accurate due to the nature of the transfer m atrix m ethod.
Based on the MTM m ethod, a com puter program for pulley stress analysis nam ed PSTRESS 3,0 has been developed, This program can provide stress
solutions and perform fatigue analysis for m ost pulleys, with the characteristic geom etry shown in Fig. 1, The pulley can be subjected to any type of
non-uniform surface pressure, sheer loading, and prescribed lock ing pressure.
In section 2, the general ideas for deriving TMB elem ents for beam (i.e. shaft), end-disk plate with variable thick ness, and cylindrical shell are
presented. In section 3, the assem bly of these elem ents to m odel a pulley in PSTRESS 3.0 is discussed. In section 4, an ex am ple of a belt conveyor
pulley is num erically solved by PSTRESS 3.0 and the results are com pared with those obtained using a finely m eshed FEM (ANSYS) solution.
2. TMB Finite Elements for Beam, Disk Plate and Cylindrical Shell
Stresses and displacem ents in a pulley can be ex pressed in term s of FO URIER series with respect to the circum ferential angle because of the pulley's
ax isym m etric geom etry. Each FO URIER com ponent of the solutions can be determ ined by solving a set of corresponding governing differential
equations, which are uncoupled with the governing equations for other Fourier com ponents. In Appendices A, B and C, it is shown that the governing
equations for Fourier com ponents for shaft, end-disk plate with non-uniform thick ness, arid cylindrical shell of a pulley can be reduced to a set of
O DEs of the first order respectively. The general solutions to these O DEs can be ex pressed in term s of the transfer m atrix . By following the procedure
described in Appendix D, the general solutions can be reorganized in a finite elem ent form as below
KmD m = F int m + F ext m (1)
where Km is the TMB elem ent stiffness m atrix , D m is the elem ent displacem ent vector, F int m is the elem ent internal force vector, F ext m is the elem ent
ex ternal force vector, and the subscript "m " denotes the FO URIER com ponent num ber, The detailed procedures of developing TMB elem ents for
shaft, end-disk and cylindrical shell are given in Appendices A, B and C, respectively.
Rem ark s;
As seen in the above discussion, the general solution to the governing differential equations for a pulley can be finally transform ed into a finiteelem ent form . This allows us to ex ploit m any finite elem ent analysis (FEA) capabilities to resolve pulley stresses using our MTM m ethod, The m ost
valuable FEA capability to be em ployed is the way of treating com plicated boundary conditions. Therefore, using the MINI m ethod, we can easily tak e
into account the elastic coupling between the rim and the end-disk by following FEA assem bly procedures, and im plem ent the lock ing assem bly
pressure by using the FEA approach of treating m echanical Interference between two bodies. Both of these problem s cannot be easily or precisely
handled by m ost classical m ethods.
3. A ssembly of TMB Elements for a Pulley Model in PSTRESS 3.0
In PSTRESS 3.0, the above derived TMB beam , disk plate and cylindrical shell elem ent stiffness m atrices are brought together to font a global
stiffness m atrix for a pulley in essentially the sam e way as that in conventional FEM, However, care m ust be tak en at two locations, where special
elem ent assem bly m ethods are required.
The first location is the connection region between the shell and the disk shown in Fig, 2a, where the finite dim ension of the joint has to be tak en
into account in the stiffness m atrix . The conventional way of treating the whole region as a single node would cause significant error in stress
solutions near this region. O ne of the reasonable ways of providing correct elastic stiffness to connect the rim and the disk is treating the whole
region as a special elem ent by applying a substructure m ethod. In PSTRESS 3.0, such a special elem ent shown in Fig. 2b is developed.

http://www.ckit.co.za/secure/conveyor/papers/troughed/pulley/pulley.htm

3 / 13

A New Pulley Stress Analysis Method Based on Modified Transfer Matrix

07/07/2014

Fig. 2: Connecting elem ent between rim and end-disk


The second location is the connection point between the lock ing device and the shaft, O ne thing that m ust be k ept in m ind when assem bling
elem ents in this location is that the final solution consists of m any FO URIER com ponents, of which the shaft elem ent only contributes to the FO URIER
com ponents of m = 0,1 and -1. In fact, in a pulley structure, the shaft deform ation is governed only by these three FO URIER com ponents due to its
slender geom etry. For the com ponent of m = 0, the corresponding shaft deform ation can be ex actly m odeled by a TMB cylindrical shell elem ent
presented in Appendix C. W hen a TMB beam elem ent described in Appendix A is used, representing the shaft deform ation of com ponents of m = -1
and 1 (i.e., bending deform ation), the following constraints on the deform ation of the connecting point m ust be im posed:
W m = -m Vm

(2)

U m = rshaft m

(3)

Ws = Wm
s = m

(4)
(5)

W s and s are the shaft deflection and rotational angle respectively at the connecting point, where the subscript s denotes shaft deform ation. The
general definitions of W s and s are given in Appendix A.
U m, Vm, W m and m are disk plate displacem ents and rotational angle at the connecting point, where the subscript m denotes the FO URIER
com ponent num ber: m = -1 or 1. The general definitions of U m, Vm, W m and m, are given in Appendix B. rshaft is the shaft radius at The connection
point.
Such constraints are easy to im plem ent in a finite elem ent m odel by using FEA static condensation or penalty m ethods [10]. In PSTRESS 3.0, the
static condensation m ethod is em ployed. Finally, it m ust be pointed out that the TMB beam elem ent stiffness m atrix and corresponding nodal forces
m ust be m ultiplied by a factor of 2 before they are assem bled into global equations, due to the difference between actual forces and harm onic
forces.
4. Numerical Example
Consider the belt conveyor pulley shown in Fig, 1, which is supported on two bearings and subjected to a lock ing pressure of 115.71 m Pa (16,778 psi)
at the interface between the lock ing device and disk hub. The shell circum ferential surface pressure and shear loading between circum ferential angles
of 83 and 254% are developed from unequal belt tensions T 1 = 1,017.8 k N (228,800 lb) and T 2 = 632.98 k N (142.300 lb) shown in Fig. 3. The
m aterial properties and geom etrical param eters of the pulley are given in Table 1, This pulley is analyzed by using PSTRESS 3.0, ANSYS 4.4 and
CDIs derivation of LANGES classical m ethod, respectively. Because of sym m etry, only one quarter of pulley cross section is m odeled.

Fig. 3: Load on pulley due to belt tension


In the PSTRESS m odel, the rim is m odeled with 2 TMB elem ents, the disk is m odeled with 6 TMB elem ents, the shaft is m odeled with 3 TMB
elem ents, and 71 FO URIER com ponents are used. The reason for using m ore TMB elem ents for the disk is the necessity of tak ing account of the
non-uniform thick ness of the disk . In the ANSYS FEA m odel, 5,000 ax isym m etric structural solid elem ents (with non ax is-sym m etric loading) are
em ployed in the 2-0 cross-section. The use of the ANSYS FEM pack age to analyze a pulley is discussed in [5].
Material Property
Young's m odulus, MPSI
Poisson's ratio
Rim Geometry
Rim length, inches
Rim outer diam eter, inches
Rim thick ness, inches
Belt width, inches
Disk Geometry
Lock ing device width, inches
Hub outer diam eter, inches
Hub inner diam eter, inches
Hub width, inches
Fillet radius at hub, inches
Filet radius at rim , inches
Disk thick ness between hub and rim , inches
Radius
Thickness
1.
15.610
1.740
2.
16.268
1.487
3.
19.321
1.409
4.
22.400
1.217
5.
22.742
1.199
Shaft Geometry
Diam eter, inches
Shaft length, inches
Distance between bearing centers, inches
Distance between disk centers, inches

30
0.3
82
54
1.5
72
3.3
27.6
20.27
6.7
3.67
1.20

16.535
121
104
74.13

Table 1: Material properties and geom etrical param eters.

http://www.ckit.co.za/secure/conveyor/papers/troughed/pulley/pulley.htm

4 / 13

A New Pulley Stress Analysis Method Based on Modified Transfer Matrix

07/07/2014

Figs. 4-7 show the PSTRESS num erical results com pared with the ANSYS results. From these figures, it is seen that at location A of the rim and
location D of the disk the agreem ent between the results O f the MTM m ethod and the results of the conventional FEM is good. At location B of the
rim and location C of the disk the agreem ent is still good, but som e inaccuracy is observed, The reason m ay be that locations B and C are within the
connection region between the rim and disk , where the 3-D stress state is m ore significant and cannot be fully tak en into account in 2-D shell and
plate theories. According to St. VENANTS principle and our ex perience, This 3-D stress state has only a very localized effect on pulley stress solution
when the thick nesses of rim and disk are relatively sm all com pared with the length and radius of the rim . It m ust be noted that the MTM m ethod is
m uch m ore efficient than the conventional FEM. PSTRESS 3.0 tak es approx im ately 30 seconds to obtain a solution on an IBM PC 486, including the
fatigue analysis. The FEM (ANSYS) solution tak es 12-24 hours on an IBM RISC 6000 work station.

Fig. 4: Stresses at Location A (inside of rim )


PSTRESS 3.0 Analysis

Fig. 5: Stresses at Location B (inside of rim )


PSTRESS 3.0 Analysis

Fig. 6: Stresses at Location C (inside of disk )


PSTRESS 3.0 Analysis

Fig. 7: Stresses at Location D (inside of disk )


PSTRESS 3.0 Analysis

Figs. 8-11 show the com parison of num erical results between LANGES solution and the FEM solution. Ex cept at location A, LANGE's solution does not
agree with the FEM solution. The poor agreem ent is due to the errors in treating the elastic coupling between the rim and the end-disk in LANGEs
m ethod.

Fig. 8: Stresses at Location A (Inside of rim )


LANGE Analysis

Fig. 9: Stresses at Location B (Inside of rim )


LANGE Analysis

Fig. 10: Stresses at Location C (Inside of disk )


LANGE Analysis

Fig. 11: Stresses at Location D (Inside of disk )


LANGE Analysis

Figs. 12 and 13 show both the PSTRESS and ANSYS results at two corners of the interface between the lock ing device and the shaft (locations E-E of
Fig. 1).At these two locations, 3-D stress state is m uch m ore significant than at locations B and C. In order to produce m ore accurate stress solutions
at the two corners, we introduce stress concentration factors in zero and first order FO URIER com ponent solutions by using our em pirical form ulae
built in PSTRESS 3.0. As seen in Figs. 12 and 13, these corrected solutions agree well with ANSYS solutions.

Fig. 12: Stresses at Location E (inside of pulley)


PSTRESS 3.0 Analysis

Fig. 13: Stresses at Location E (outside of pulley)


PSTRESS 3.0 Analysis

5. Conclusions
A new pulley stress analysis m ethod which is based on reform ulated transfer m atrix has been developed. An accurate solution can be obtained by
using this m ethod. Three transfer-m atrix -based elem ents for the shaft, disk plate and cylindrical shell have been developed. A num erical ex am ple
has been given, which dem onstrates the m erits of this new m ethod.
References

1.
2.
3.

LANGE, H.: Investigations on Stress in Belt Conveyor Pul leys; Doctoral thesis, Technical University Hannover, 1963.
SCHMO LTZI, W .: The Design of Conveyor Belt Pulleys with Continuous Shafts; Doctoral thesis, Technical University, Hannover 1974.
V0DSTRCIL, P.: Analysis of Belt Conveyor Pulley Using Finite Elem ent Method; Proc. 4th i nt. Conf. in Australia on Finite Elem ent Methods,
University of Melbourne, Aug. 18-20, 1982.

4.

DANIEL, W .J.T.: Developm ent of a Conveyor Pulley Stress Analysis Pack age; Proc. nt. Conf. on Bulk Material Storage, Handling and
Transportation, Newcastle, Aug. 22-24, 1983.

http://www.ckit.co.za/secure/conveyor/papers/troughed/pulley/pulley.htm

5 / 13

A New Pulley Stress Analysis Method Based on Modified Transfer Matrix

5.

07/07/2014

SETHI, V. and NO RDELL, L.K.: Modern Pulley Design Techniques and Failure Analysis Methods; Proceedings of SME Annual Meeting & Ex hibit,
Peno Nevada, USA, Feb. 15-18,1993.

6.
7.

PESTEL, S.C. and LECKIE, F.A.: Matrix Methods in Elastom echanics; New York McGraw-Hill Book Co., 1963.
YEH KAIYUAN: General Solution on Certain Problem s of Elasticity with Non-Hom ogeneity and Variable Thick ness, The Advances of Applied
Mathem atics and Mechanics, Vol. 1; China Academ ic Publishers, pp 240-273, 1987.

8.
9.
10.

BO YCE, W .E. and DIPRIMA, R.C.: Elem entary Differential Equations; Fifth Edition, John W iley & Sons Inc., New York , 1992.
TIMO SHENKO , S. and W O INO W SKY-KRIEGER. S.: Theory of Plates and Shells; McGraw-Hill Book Co., Second Edition, 1959.
CO O K, R.D., MALKUS, D.S. and PLE5HA, M.E.: Concepts and Applications of Finite Elem ent Analysis; Third Edition, John W iley & Sons Inc.,
1969.

A ppendix A
TMB Finite Element for Beam
Fig. A1 shows forces that act on a differential beam . Loads P, M, and q are shown in their positive sense. z is the ax ial coordinate.

Fig. A1: Forces that act on a differential elem ent of beam


The equilibrium equations are
dP/dz = -q
dM/dz = P

(A1)
(A2)

By using TIM0SHENKO s beam theory, we have


dW s/dz = - s + P/k aG
d s/dz = M/EI

(A3)
(A4)

where the subscript s denotes the shaft deform ation, W s is the shaft neutral ax is displacem ent, - s is the slope due to bending, dW s/dz is the slope
of the center line of the beam , k is a shape factor equal to 0.75 for circular cross section, EI is the bending stiffness, a is the cross-section area, and
G is the shear m odulus.
Eqs. (A1)-(A4) can be written in a m atrix form
d/dz = A + B

(A5)

= (W s, s, P, M) T
B = (0, 0, -q, 0) T

(A6)
(A7)

where

and

A=

0
0
0
0

-1
0
0
0

1/k aG
0
0
1

0
1/EI
0
0

(A8)

where is called state variable vector.


According to the O DE theory [8], the general solution to Eq. (A5) can be ex pressed as
z

(z) = T(z)(z 0) + T(z)

T(s)-1B(s)ds

(A9)

z0

where z 0 z z L, z 0 and z L are the coordinates corresponding to two ends of the beam , and T(z) is the transfer m atrix satisfying
dT/dz = AT and T(z 0) = l

(A10)

where l is the identity m atrix . Following the general procedure described in [8]. we can obtain the following closed-form transfer m atrix
1 z-z 0
T(z) =

0 1
0 0
0 0

z-z 0

(z-z 0) (z-z 0)
+
k aG
2EI
2EI
(z-z 0)
z-z 0
2EI
1
z-z 0

EI
0
1

(A11)

where z 0 z z L. Following the procedure described in Appendix D, we can obtain the following finite elem ent equation, which is equivalent to Eq.
(A9)
KBMU BM = F BMint + F BMext

(A12)

where
U BM = (W s (ZL), s(ZL), W s(Z0), s(Z0)) T

(A13)

F BMext is the beam elem ent ex ternal force vector, F BMint is the beam elem ent internal force vector, and K BM is the beam elem ent stiffness m atrix ,
which can be ex pressed:

http://www.ckit.co.za/secure/conveyor/papers/troughed/pulley/pulley.htm

6 / 13

A New Pulley Stress Analysis Method Based on Modified Transfer Matrix

KBM =

07/07/2014

12EI
L(1+)
6EI
L(1+)
-12EI
L(1+)
6EI
L(1+)

Sym m etric
(4+)EI
L(1+)
-6EI
L(1+)
(2-)EI
L(1+)

12EI
L(1+)
-6EI
L(1+)

(A14)

(a+)EI
L(1+)

W here
= 12EI/Lk aG and L = ZL - Z0

(A15)

Rem ark s:
It is not surprising to note that the TMB finite elem ent stiffness m atrix for a beam derived as Eq. (A14) is identical to the beam stiffness m atrix
derived by using conventional finite elem ent m ethod because the polynom ial-type trial function used in conventional FEM ex actly represents the
actual beam displacem ent. However, for other types of structures such as plate and shell, the TMB elem ent stiffness m atrices m ay not be the sam e
as the conventional finite elem ent stiffness m atrices.
A ppendix B
TMB Finite Element for Disk Plate with Variable Thickness
B.1 TMB Bending Element
Fig. B1 shows a differential elem ent of plate subjected to bending loads. x , y and z are Cartesian coordinates with z coinciding with pulley ax ial
direction, x horizontal direction, and y vertical direction. r and are disk cylindrical coordinates. Loads Q r, Q , M r, M and M r are shown in their positive
sense.

Fig. B1: Bending forces that act on a different elem ent of plate
The equilibrium equations are:
Qr =
Q =
Q r

M r
r
M r
r
Qr

M r-M

+
+2
-

r
Mr
r
1
r

1 M r
r
1 M
r
Q
= g(r, )

(B1)
(B2)
(B3)

where g is the ex ternal transverse force acting on the neutral surface of the plate, and equations representing the force-deform ation relationship are:

[ ur + ( 1r ur + 1r u )]
u
1 u
1 u
M = -D [
+
+
r
r r
r ]
1 u
1 u
M = D(1-)
[ r r r ]

M r = -D

(B4)
(B5)
(B6)

where U is the plate neutral surface transverse displacem ent (see Fig. B1), is the PO ISSO N's ratio,
D = Et/12(1 - )

(B7)

E is the YO UNG's m odulus, t is the thick ness of the plate, which is a function of r, W ithin the elem ent, we assum e
t = crp

(B8)

where a and p are constants. Also, we assum e the following FO URIER series for the com ponents of displacem ent and forces

u =

u mcosm +

m =0

g =

g mcosm +

Q rmcosm +

Q msinm +

M rmcosm +

m =1

http://www.ckit.co.za/secure/conveyor/papers/troughed/pulley/pulley.htm

(B10)

Q r,-msinm

(B11)

Q ,-mcosm

(B12)

M r,-msinm

(B13)

M ,-msinm

(B14)

m =1

M mcosm +

m =0

Mr =

g -msinm

m =0

m =0

M=

m =1

m =1

Mr =

(B9)

m =1

m =0

Q =

u -msinm

m =1

m =0

Qr =

m =1

M rsinm +

M r,-mcosm (B15)

m =0

7 / 13

A New Pulley Stress Analysis Method Based on Modified Transfer Matrix

07/07/2014

where u m, P m, Q rm, Q m, M rm, M m, M rm (m =O .1,2,...) are functions of r only. Substituting Eqs. (89)-(B15) into Eqs. (B1)-(B6), we have the following
ordinary differential equations
Q rm = M' rm +
Q m = M' rm +
Q ' rm
M rm = -D

[ u"

Mm = - D

[ u"m

1 (M - M ) - m M
rm
m
rm
r
r
2 M + m M
rm
m
r
r
1
m
+ Q rm Q m = gm
r
r
1
m
+
u' m u
r
r m
+ 1 u' m - m u m
r
r
m u' + m u
m
m
r
r

)]

M rm = D(1-)

(B16)
(B17)
(B18)
(B19)
(B20)
(B21)

where m = 0, 1, 2, ..., and the prim e represents derivative with respect to r. Introducing the following state variables
m = (u m, m, rm, M rm) T

(B22)

W here
m = u' m

(B23)

m rm = -2r M rm
Vrm = 2r (Q m - m /r M rm)

(B24)
(B25)

Elim inating five variables, M rm, M m, M rm, Q rm and Q m am ong Eqs. (B16)-(B21) and (B23)-(B25), we can obtain the following m atrix -form equation
d m/dr = Am m + Bm (m = 0, 1, 2, ...)

(B26)

W here Am =

0
1
m
_
r
r
2Dm (2 - 2 + m - m ) -2Dm (3 - 2 - )
r
r
2Dm (-3 + 2 + )
2D(1 - + 2m - 2m )
r
r
Bm = (0, 0, 2rg m, 0) T

0
1
2Dr
_1 m
r
r
(1 - )
-1
r
0

(B27)

(B28)

Therefore, following the procedure described in Appendix D, we can obtain the following finite elem ent equation for a circular plate ring (r0 r rL)
with variable thick ness and subjected to harm onic bending load
KBD mU BDm = F BDint m + F BDext m

(B29)

W here
U BDm = (U Lm, Lm, U om, 0m) T

(B30)

KBD m is the plate elem ent bending stiffness m atrix , F BDext m is the elem ent ex ternal force vector, F BDint m is the elem ent internal force vector, the
subscript L and 0 denote locations at r = rL and r0, respectively, and m denotes the FO URIER com ponent num ber. Due to the com plex ity in deriving
the transfer m atrix for Eq. (B26), it is m uch m ore difficult to obtain a closed-form ex pression for KBD m of Eq. (B29) than for KBM of Eq. (A12).
Instead, we can very accurately calculate KBD m by using our com puter program of PSTRESS 3.0 m entioned in Rem ark ii of Appendix D.
B.2 TBM Plane Stress Element
Fig. B2 shows a differential elem ent of plate subjected to in-plane loading. Loads Nr, N, and Nr are shown in their positive sense.

Fig. B2: In-plane forces that act on a differential elem ent of plate
The definitions of x , y, Z, r and are the sam e as in section B.1. The equilibrium equations are
Nr
r
Nr

+
+2

Nr - N
r
Nr

1 Nr
=0
r
1 N
+
=0
r
+

(B31)
(B32)

The equations for force-deform ation relationship are


Et
(1 - )
Et
N=
(1 - )
Et
Nr =
2(1 + )
Nr =

http://www.ckit.co.za/secure/conveyor/papers/troughed/pulley/pulley.htm

1 v

( r + r + r ) (B33)
w 1 v w
( r + r + r ) (B34)
v 1 w v
( r + r - r ) (B35)
8 / 13

A New Pulley Stress Analysis Method Based on Modified Transfer Matrix

07/07/2014

where the definitions of , E and t are the sam e as in section B.1, arid v and w are the displacem ents of the neutral surface in circum ferential and
radial directions, respectively.
W e assum e the following FO URIER series for the com ponents of displacem ents and forces

vmsinm +

v=

m =1

m =0

Nr =

m =0

(B37)

Nr,-msinm

(B38)

N,-msinm

(B39)

Nrmcosm

(B40)

m =1

Nrmsinm +

w-msinm

m =1

m =0

Nr =

Nmcosm +

N=

(B36)

m =1

Nrmcosm +

v-mcosm

m =0

wmcosm +

w=

m =1

m =0

where vmwm, Nrm, Nm and Nrm(m = 0,1,2, ...) are functions of r only. Substituting Eqs. (B36)-(B40) into Eqs. (B31)-(B35), we have the following
ordinary differential equations
1
r

N' rm +

2
r

N' rm +
Et
Nrm =
(1 - )
Et
Nm =
(1 - )
Et
Nrm =
2(1 + )

[
[
[

m N =0
(B41)
rm
r

(Nrm - Nrm)+
m
r

Nrm -

Nm = 0 (B42)

w' m + m vm + 1 wm
r
r
w' m + m vm + 1 wm
r
r
v' m - m wm - 1 vm
r
r

)] (B43)
] (B44)
] (B45)

where m = 0, 1, 2, 3,.... Introducing the following state variables


m = (wm, vm, 2rNrm , 2rNrm) T

(B46)

and elim inating Nm am ong Eqs. (B41)-(B45). we obtain


d m/dr = C m m (m = 0, 1, 2, ...)

(B47)

where

Cm =

_
_m
10
r
r 2Etr
m
1
1+
0
r
r
2Etr
2Et 2Etm _ 1 - _ m
r
r
r
r
2Etm 2Etm m
_2
r
r
r
r

(B48)

Thus, following the sim ilar procedure in deriving Eq. (B29) , we can finally derive the following TMB finite elem ent equation for a circular ring (r0 r
rL) with variable thick ness, subjected to in-plane harm onic loading
KPN m U PNm = F PNint m + F PNext m

(B49)

where
U PNm = (W Lm, VLm, W 0m, V0m) T

(B50)

KPN m is the elem ent plane-stress stiffness m atrix , F PNext m is the elem ent ex ternal force vector, F PNint m is the elem ent internal force vector, and the
subscript L and C denote locations at r = rL and r0 respectively, and m denotes the FO URIER com ponent num ber.
Com bining Eq. (B29) and Eq. (B49), we can form a TMB shell elem ent for disk plate, which is subjected to both bending and in-plane loading
KDK m U DKm = F Dkint m + F DKext m

(B51)

where KDK m is the disk elem ent stiffness m atrix , U DKm is the elem ent displacem ent vector, F DKext m is the elem ent ex tem al force vector, and F Dkint m
is the elem ent internal force vector. They can be calculated by the following form ulae
U DKm = (U Lm, VLm, W Lm, Lm, U 0m, V0m, W 0m, 0m) T (B52)
KDK m = ST1KBD mS1 + ST2KPN mS2

(B53)

F DKint m = ST1F BDint m + ST2F PNint m

(B54)

F DKext m = ST1 F BDext m + ST2 F PNext m

(B55)

S1 =

1
0
0
0

0
0
0
0

0
0
0
0

0
1
0
0

0
0
1
0

0
0
0
0

0
0
0
0

0
0
0
1

(B56)

Remarks:

iv.

Any other state variables than those defined in Eq. (B22) and Eq. (B46) cannot be em ployed because they m ay lead to incorrect elem ent

http://www.ckit.co.za/secure/conveyor/papers/troughed/pulley/pulley.htm

9 / 13

A New Pulley Stress Analysis Method Based on Modified Transfer Matrix

07/07/2014

stiffness m atrices and force vectors due to the violation of MAXW ELL'S reciprocal theorem .

v.

Stiffness m atrices obtained from transfer m atrices m ust be sym m etric. Any m istak es in choosing state variables, in deriving equations, and/or
in num erical program m ing m ay lead to asym m etric stiffness m atrices.

vi.

If the stress calculation is desired, we can follow the following steps

a.
b.

calculate values of state variables at elem ent nodes,


use transfer m atrix and Eq. (D3) to calculate values of state variables at any interior location of the elem ent, where stress values are
desired,

c.
d.
e.

calculate first derivatives of state variables at desired locations by using Eq. (B26) and Eq. (B47),
calculate M rm, M m, M rm, Nrm, Nm and Nrm by using Eqs. (B19)-(B21) and Eqs. (B43)-(B45), and
calculate FO URIER com ponents of stresses by using calculated forces in step d.

A ppendix C
TMB Finite Element for Cylindrical Shell
Fig. C1 shows forces acting on a differential elem ent of a cylindrical shell with constant thick ness t and radius R, z and are cylindrical coordinates.
Loads N1, N2, S, Q 1, Q 2, M 1, M 2 and M 12 are shown in their positive sense.

Fig. C1: Forces that act on a differential elem ent of cylindrical


The equilibrium equations, according to [9] are
N1

z
N2
_ N2

R
Q2 =
Q1 =

R
Q 1

z
M 12
z
M 1

+
+

S
R
S
z
Q 2
R
M 2

+ fz = 0

(C1)

+f=0

(C2)

+ fr = 0

(C3)
(C4)

R
M 12

(C5)

where f z, f and f r are ex ternal loads acting on the neutral surface of the shell in ax ial, circum ferential and radial directions respectively, and the
equations for force-displacem ent relationship, according to [9]. are
Et
(1-)
Et
N2 =
(1-)
N1 =

u
v
w
+
+
z
R
R
u
v
w
+
+
z
R
R
Et
u
v
S=
+
(1-2)
R
z
w
u
M 1 = -D
+
z
R
w
u
M 2 = -D
+
z
R
M 12 = -(1 - )D w
Rz

(
(

) (C6)
) (C7)
) (C8)
) (C9)
) (C10)
(C11)

where u, v and w are shell neutral surface displacem ents in ax ial, circum ferential and radial directions, respectively,
D = Et/12(1 - )
the definitions of E and are the sam e as in section B, and t is the constant shell thick ness. We assum e the following FO URIER com ponents of
displacem ents and forces

fz =

f zmcosm +

m =0

f=

f msinm +

f rmcosm +

u mcosm +

http://www.ckit.co.za/secure/conveyor/papers/troughed/pulley/pulley.htm

f ,-mcosm

(C13)

f r,-msinm

(C14)

u -msinm

(C15)

v-mcosm

(C16)

w-msinm

(C17)

m =1

vmsinm +

m =1

w=

m =1

m =0

v=

(C12)

m =0

m =0

u =

f z,-msinm

m =1

m =1

fr =

m =0

wmcosm +

m =0

m =1

10 / 13

A New Pulley Stress Analysis Method Based on Modified Transfer Matrix

07/07/2014
N1 =

N1mcosm +

m =0

N2 =

N2mcosm +

m =0

S=

Smsinm +

M 1mcosm +

M 2mcosm +

(C19)

S-mcosm

(C20)

M 1,-msinm

(C21)

M 2,-msinm

(C22)

m =1

M 12msinm +

M 12,-mcosm (C23)

m =0

Q 1mcosm +

m =0

Q2 =

N2,-msinm

m =1

m =1

Q1 =

m =0

M 12 =

(C18)

m =0

m =0

M2 =

N1,-msinm

m =1

m =1

M1 =

m =1

Q 1,-msinm

(C24)

Q 2,-mcosm

(C25)

m =1

Q 2msinm +

m =1

m =0

where all FO URIER coefficients are functions of z only. Substituting Eqs. (C12)-(C25) into Eqs. (C1)-(C11), introducing the following state variables
m = (U m, Vm, W m, m, 2RN1m, 2RSm, 2RV1m, 2RM 1m) T

(C26)

W here
m = -W ' m

(C27)

and
V1m = Q 1m + m /R M 12m

(C28)

and elim inating five variables, N2m, M 2m, M 12m, Q 1m and Q 2m we can finally obtain the following m atrix -form equation after a long tedious procedure
of m athem atical derivation
dm/dz = J mm + I m

(m = 0, 1, 2, ...)

(C29)

W here
0
m
R
0
0
Jm =

_m
R

_
R

0
_m
R

-1

0
0

Etm
2Etm
R
R
2Etm 2Et
tm 4
0
(1+
)
R
R
12R
0

12EtR

m
R

0
EtR
3(1+)R

R
0

1+
0
EtR
0 0

0
6(1 - )
0
EtR

_m
0
R
0

m
R

(C30)

and
I m = (0, 0, 0, 0, - 2Rf zm, -2Rf m, 2Rf rm, 0) T

(C31)

Therefore, according to the conclusions of Appendix D, the transfer m atrix for Eq. (C29) can be obtained by the com puter program developed in
PSTRESS 3.0, Based on the transfer m atrix , the TMB finite elem ent for cylindrical shell with z 0 z z L can be derived by following the procedure
described in Appendix D, and the result can be ex pressed by
KCS m U CSm = F CSint m + F CSext m

(C32)

where
U CSm = (U Lm, VLm, W Lm, Lm, U 0m, V0m, W 0m, 0m) T

(C33)

KCS m is the cylindrical-shell stiffness m atrix , F CSext m is the elem ent ex ternal force vector, F CSint m is the elem ent internal force vector, and the
subscripts L and 0 denote locations at z - z L, and z = z 0, respectively, and m denotes the FO URIER com ponent num ber.
If stress calculation for cylindrical shell is desired, we can follow the sim ilar procedure described in Rem ark vi of section B.
A ppendix D
Procedure of Deriving TMB Finite Element
In Appendices A, B and C, it is shown that the governing equations for FO URIER com ponents for shaft, end-disk plate with nonuniform thick ness, and
cylindrical shell of a pulley can be reduced to a set of O DEs of first order respectively, and these O DEs can be finally unified as the following form
d m/ds = H m(s) m + L m

(m = 0, 1, 2, ...)

(D1)

where the subscript m denotes the ordinary num ber of FO URIER com ponents, H m is a 2n x 2n m atrix . L m is a 2n x 1 vector containing ex ternal loads,
m is the state variable vector defined as

http://www.ckit.co.za/secure/conveyor/papers/troughed/pulley/pulley.htm

11 / 13

A New Pulley Stress Analysis Method Based on Modified Transfer Matrix

07/07/2014
m = (U Tm, F Tm) T

(D2)

U m is an n x 1 vector containing generalized displacem ents, F m an n x 1 vector containing corresponding generalized internal forces, and s is an ax ial
or radial coordinate of the pulley, which is assum ed within the range of s 0 s s L, where s 0 and s L are corresponding to two ends of a pulley
com ponent section (i,e, elem ent).
Because Eq. (D1) is linear, the general solution is obtainable. By using the theory of O DE [8], the general solution to Eq. (D1) can be written as
s

m(S) = T m(S) m(s 0) + T m(S)

T m(X) -1L m (x )dx

(D3)

s0
where T m(s) is a 2n x 2n m atrix called transfer m atrix satisfying
dT m/ds = H mT m and T m(s 0) = I

(D4)

where I is the identity m atrix , For any set of linear ordinary differential equations, which can be ex pressed in the form of Eq. (D1), there ex ists a
unique 2n x 2n transfer m atrix T m(s) satisfying Eq. (D4). If Eq. (D1) is a set of constant O DEs (i.e., H m is independent of s) or it can be transform ed
into a set of constant O DEs, a closed-form transfer m atrix T m(s) is obtainable. The general procedure of deriving T m(s) is discussed in m any
tex tbook s, such as the one by BO YCE and DIPRIMA [8]. In what follows in this section, it is shown that a special finite elem ent (i.e., TMB elem ent)
can be developed from Eq. (D3).
Let us consider a section of pulley, which is located in s 0 s s L, and let
T m(SL)=

T UU T UF

[T

m(SL)=

m(S0)=

FU

T FF

UL
F int L
Uo
-F int 0

(D5)

(D6)

(D7)

sL
P m = T m(s L)

T m(X) -1 B(x )dx

(D8)

s0
W here
U L = U m(SL)
U o = U m(S0)

(D10)

F int L = F m(SL)

(D11)

F int 0 = -F m(S0)

(D12)

(D9)

Then Eq. (D3) can be rewritten as


U

-T

UF
[ FintL L ] = [ T UU
][ Fintint L0 ] + P m
FU -T FF

(D13)

Reorganizing Eq. (D13), we have


I -T UU

UL

0 -T UF

F int L

[ 0 -T FU ][ U 0 ] = [ -I -T FF ][ Fint 0 ] + P m

(D14)

from which we can obtain


KmD m = F int m + F ext m

(D15)

W here

Km =

0 -T UF

[ -I

-T FF

]-1[

-T UU

-T FU

D m = (U TL, U T0) T
F int m =

F ext m =

F int L
F int 0
0

-T UF

-1

-T FF

(D16)

(D17)

]
]-1

(D18)

Pm

(D19)

It is seen that Eq. (D15) is form ulated in a finite-elem ent form , where Km is elem ent stiffness m atrix , D m is the elem ent displacem ent vector, F int m is
the elem ent internal force vector, and F ext m is the elem ent ex ternal force vector, The detailed procedures of developing TMB elem ents for shaft, enddisk and cylindrical shell are given in Appendices A, B and C, respectively.
Remarks:

i.

It is noted that in developing TMB finite elem ents the trial function for representing elem ent displacem ent field is not required. This is the
feature that distinguishes the MTM m ethod from the conventional finite elem ent m ethod.

ii.

In general, for any type of structures, whose governing differential equations can be reduced to the form of Eq. (D1), which can be further
transform ed into a set of constant O DEs, the corresponding transfer m atrix T m(s) of Eq. (D3) can be derived following som e standard
procedure described in [8]. In m ost cases, the procedure of deriving T m(S) can be fulfilled by using a com puter program , and P m defined in
Eq. (D8) can be evaluated analytically or num erically; and Therefore, a com puter program can be m ade, where the input is the m atrix H m and
the ex ternal loads, and the output is the TMB elem ent stiffness m atrix and elem ent ex ternal force vector. In this program , there is no

http://www.ckit.co.za/secure/conveyor/papers/troughed/pulley/pulley.htm

12 / 13

A New Pulley Stress Analysis Method Based on Modified Transfer Matrix

07/07/2014

discretization error in com puting TMB elem ent stiffness m atrix . As a FO RTRAN subroutine, such a com puter program has been developed in
PSTRESS 3.0.

http://www.ckit.co.za/secure/conveyor/papers/troughed/pulley/pulley.htm

13 / 13