07/07/2014
y
z
r
P
M
q
Ws
EI
a
G
A
B
T
I
KBM
U BM
vm
wm
gm
Q rm
Qm
M rm
Mm
M rm
http://www.ckit.co.za/secure/conveyor/papers/troughed/pulley/pulley.htm
1 / 13
Vrm
Am
m atrix containing coefficients of governing O DEs for disk bending deform ation of FO URIER
com ponent m
Bm
vector representing the nonhom ogeneous term of the governing O DEs for disk bending
deform ation of FO URIER com ponent m
U BDm
07/07/2014
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
inplane shear force acting on the disk cross sections perpendicular to disk radial direction
and disk circum ferential direction
Nrm
Nm
Nrm
m
Cm
m atrix containing coefficients of governing O DEs for disk Inplane deform ation of FO URIER
com ponent m
U PNm
KPNm
F PNextm
KDKm
F DK
extm
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
fm
f rm
N1m
N2m
M 1m
M 2m
M 12m
V1m
xm
FO URIER com ponent of elem ent boundary equivalent transverse shear force
Jm
Im
m atrix containing coefficients of governing O DEs for cylindrical shell of FO URIER com ponent
m
vector representing the nonhom ogeneous term of the governing O DEs for cylindrical shell of
FO URIER com ponent m
M 12
U CSm
Kcs m
S
Hm
Lm
vector representing the nonhom ogeneous term of the governing O DEs of a pulley
com ponent
Um
http://www.ckit.co.za/secure/conveyor/papers/troughed/pulley/pulley.htm
2 / 13
07/07/2014
Fm
Tm
Dm
Km
F int m
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.
http://www.ckit.co.za/secure/conveyor/papers/troughed/pulley/pulley.htm
3 / 13
07/07/2014
(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.
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
http://www.ckit.co.za/secure/conveyor/papers/troughed/pulley/pulley.htm
4 / 13
07/07/2014
Figs. 47 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 3D stress state is m ore significant and cannot be fully tak en into account in 2D shell and
plate theories. According to St. VENANTS principle and our ex perience, This 3D 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 1224 hours on an IBM RISC 6000 work station.
Figs. 811 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 enddisk in LANGEs
m ethod.
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 EE of
Fig. 1).At these two locations, 3D 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.
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 transferm 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. 1820, 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. 2224, 1983.
http://www.ckit.co.za/secure/conveyor/papers/troughed/pulley/pulley.htm
5 / 13
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. 1518,1993.
6.
7.
PESTEL, S.C. and LECKIE, F.A.: Matrix Methods in Elastom echanics; New York McGrawHill Book Co., 1963.
YEH KAIYUAN: General Solution on Certain Problem s of Elasticity with NonHom ogeneity and Variable Thick ness, The Advances of Applied
Mathem atics and Mechanics, Vol. 1; China Academ ic Publishers, pp 240273, 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 SKYKRIEGER. S.: Theory of Plates and Shells; McGrawHill 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.
(A1)
(A2)
(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 crosssection 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)
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 closedform transfer m atrix
1 zz 0
T(z) =
0 1
0 0
0 0
zz 0
(zz 0) (zz 0)
+
k aG
2EI
2EI
(zz 0)
zz 0
2EI
1
zz 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
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 ialtype 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 rM
+
+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 forcedeform 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
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 closedform 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 inplane loading. Loads Nr, N, and Nr are shown in their positive sense.
Fig. B2: Inplane 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)
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
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 +
wmsinm
m =1
m =0
Nr =
Nmcosm +
N=
(B36)
m =1
Nrmcosm +
vmcosm
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)
(B46)
(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 inplane 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 planestress 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 inplane 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)
(B54)
(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
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.
a.
b.
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.
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 forcedisplacem 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=
+
(12)
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)
vmcosm
(C16)
wmsinm
(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
07/07/2014
N1 =
N1mcosm +
m =0
N2 =
N2mcosm +
m =0
S=
Smsinm +
M 1mcosm +
M 2mcosm +
(C19)
Smcosm
(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 cylindricalshell 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, enddisk 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
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
(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 closedform 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)
(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)
T
UF
[ FintL L ] = [ T UU
][ Fintint L0 ] + P m
FU T FF
(D13)
UL
0 T UF
F int L
[ 0 T FU ][ U 0 ] = [ I T FF ][ Fint 0 ] + P m
(D14)
(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 finiteelem 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
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