Download as pdf or txt
Download as pdf or txt
You are on page 1of 15

Proceedings of the XXVI Iberian Latin-American Congress on Computational Methods in Engineering CILAMCE 2005

Brazilian Assoc. for Comp. Mechanics & Latin American Assoc. of Comp. Methods in Engineering
Guarapari, Espı́rito Santo, Brazil, 19th –21st October 2005

Paper CIL0620

ANALYSIS OF TRANSIENT LOADS ON CABLE-REINFORCED CONVEYOR


BELTS WITH DAMPING CONSIDERATION

R. Pascual
V. Meruane
rpascual@ing.uchile.cl
vmeruane@ing.uchile.cl
Mechanical Engineering Department, Universidad de Chile, Beauchef 850, Santiago, Chile

G. Barrientos
gbarrien@udec.cl
Mechanical Engineering Department, Universidad de Concepción, Concepción, Chile

Abstract. This work presents a methodology to compute dynamic stress distributions on


large conveyor belts considering a viscous-damping model. Results could be very useful
at the conveyor design stage since at starts and stops of the system, large dynamic loads
may be encountered. Such loads may produce a catastrophic failure of the conveyor, with
important direct and downtime costs. We present a finite- element based methodology and
use the Newmark time-integration method to solve the transient responses. Our scheme
includes the effects of damping. A real-life industrial case is analyzed and three critical
situations are considered: normal stop, sudden stop, suddenly aborted start. Results are
shown in terms of dynamic loads at critical points of the system. The simulations show
that the damping effects are negligible in the analysis of real-life designs, thus allowing
the use of the analytical solution. For the analized case, maximum loads obtained with
the present methodology are smaller in 5% than those obtained with the conservative
approach.

Keywords: Structural dynamics, large conveyor belts, finite element, transient analysis.

1. INTRODUCTION
A large variety of industries use large conveyor belts to move raw material between
process units. Recent technology has been developed to meet the industry needs, leading
to systems with improved mechanical properties, increased strength and wear properties,
and other specific properties associated to each application. Conveyor belts are also of
great use in the mining industry. There, designs consider the use of steel-cable reinforced
belts in order to support the large static and dynamic loads that appear. At the design stage,
it is very important to estimate the dynamic loads that a conveyor belt may suffer in order
to specify system dimensions and material selection. There exists a tradeoff between the
investment in an oversized belt and the downtime cost associated to a catastrophic failure
of an undersized belt. Traditional design standards for belt conveyor with less than 1000
m. References (1)-(2) do not consider explicitly the elastic characteristics of the belt.
They are based on empirical data from conveyors of medium length, with no input from
current large distance (up to several kilometers) applications.
CILAMCE 2005 – ABMEC & AMC, Guarapari, Espı́rito Santo, Brazil, 19th – 21st October 2005

The situation may be modelled using the one-dimensional wave equation, whose
solution is well known when the initial velocity conditions are given. When, as in the
case of conveyor belts, initial conditions are given in terms of acceleration, obtaining an
analytical solution is much more complex. Previous works only deal with conservative
systems (3).
Harrison (4; 5; 6) establishes an analytical solution to the problem by modelling in
terms of stress waves that propagate in time trough the belt, and quantifies the dynamic
stress. His model is one-dimensional which does not permit the analysis of other loads
like shear loads and torsion loads.
The objective of this work is to evaluate the effect of damping on the dynamic re-
sponses of a large-scale conveyor belt. By doing so, we will be able to understand how
damping may be used to control and reduce the dynamic stress under transient load situ-
ations.
Wheeler (7) presents a method to calculate the flexure resistance at the design stage,
with the view of minimising these resistances to improve the efficiency of the belt. Plate
theory is used to approximate the deflection of the troughed conveyor belt between idler
sets. In reference (8), Wheeler analyzes the rotating resistance of belt conveyor idler rolls
occurs due to the friction of the rolling elements in the bearings, the viscous drag of the
lubricant and the friction of the contact lip seals. He presents models to estimate the
rotating resistance of idler rolls and discuss an apparatus designed to measure the rotating
resistance under simulated operating conditions.

2. ANALYTICAL MODEL
Figures 1 and 2 show a typical configuration of an industrial conveyor belt. It is
composed of a belt, a drive pulley that provides the necessary driving torque; a driven
pulley; rollers to support the belt; and a system to keep the tension which may be a tail
mass or a spring system. Under nominal conditions, the belt is loaded on the upper side
and unloaded on the lower side. During starting and stopping of a conveyor system, large
stress are on the belt caused by the motor pulley. On one side, we observe traction and
compression waves. Low frequency oscillations are also seen at the tail pulley and at the
tail mass(3). Experience shows that the vibration frequency induced on the tail pulley and
the tail mass is about 10 times smaller compared to the frequency of the elastic wave, so
we may consider negligible the first effect with respect to the second (3).
Harrison (3) argues that the propagation velocity of the elastic wave determines the
dynamic tension of the belt. For a cable-reinforced belt, such velocity is given by:
s s
Esc Esc Asc
vsc = = (1)
ρsc µsc
where µsc is the mass of the steel cable per unit length and Asc corresponds to its cross
section. If the cables are covered by an elastomeric layer, there exists a delaying effect
due to the mass of the rubber (experimentally verified by Harrison(4)). The rollers and
the raw material also delay the propagation of the wave. In order to consider these effects,
we define an equivalent mass per unit length:
µeq = µb + µil + µl (2)
CILAMCE 2005 – ABMEC & AMC, Guarapari, Espı́rito Santo, Brazil, 19th – 21st October 2005

Figure 1: Basic scheme of a conveying system

Figure 2: Belt cross section


CILAMCE 2005 – ABMEC & AMC, Guarapari, Espı́rito Santo, Brazil, 19th – 21st October 2005

where µb is the belt mass per unit length, µil is an equivalent rollers’ mass per unit length,
µl corresponds to the raw material mass per unit length. The equivalent density of the
loaded belt is then:
µeq
ρb = (3)
Asc
and the velocity of the wave is estimated as:
s s
Eb Esc ρsc ρsc µsc
r r
vcl = = = vsc = vsc (4)
ρb ρsc ρb ρb µeq

Equation (4) assumes that the material is uniformly distributed along the belt, with no
relative movement between the raw material and the belt. vcl represents to a lower bound
for the wave propagation velocity. If the conveyor is unloaded:
µsc
r
vcl = vsc (5)
µb + µil
We also must consider that the wave front has different velocities in each side of the
conveyor. For the return side, we have:
µsc
r
vr = vsc (6)
µb + µir
where µir is the mass per unit length on the return side.
In real-life applications, belts work partially loaded and load adhesion depends on
the nature of the raw material. For these reasons, Harrison (5) proposes a correction to
velocity vc given by:
q
vc = vcu − α (vcu − vcl ) (7)
q0
where α is the coupling factor between the raw material and the belt. He considers 0.8 for
granulated material and humid material, and 0.3 for rocks (diameter 0.2 m). If we consider
a homogeneous belt model, there will be a single wave propagation speed velocity, which
corresponds to the average value:
vr + vc
v0 = (8)
2
Once we have estimated the velocity at which the wave front advances, we may
evaluate the time interval needed for it to go trough the belt. If we consider constant
velocity propagation, it takes:
l 1
Tc =
2 vc
time units to go trough the load side of the belt when it is unloaded. l correspondes to the
belt length. When it is loaded it takes:
l 1
Tcl =
2 vcl
CILAMCE 2005 – ABMEC & AMC, Guarapari, Espı́rito Santo, Brazil, 19th – 21st October 2005

time units. We also know that it takes


l 1
Tr =
2 vr
time units to return to the driven pulley.
The cycle interval when the belt loaded and unloaded respectively:

Tl = Tcl + Tr
Tu = Tc + Tr

We may also define the propagation interval as:


2l
τ= (9)
v0
According to wave theory (14), the maximum dynamic load occurs at the origin (see
figure 3) and at the instant:
τ l
tf max = = (10)
2 v0
2.1 System inertias
The mass of the rotating parts (pulleys, gears, couplings) are considered as an equiv-
alent mass in traslation, which has the same amount of kinetic energy as the rotating
system:
1 2
+ Ig ng ωg2 + Ic nc ωc2

Tsr = Im nm ωm (11)
2
where nm is the number of motors, with a moment of inertia Im , ng is the number of gears,
with a moment of inertia Ig , nc is the number of couplings, with a moment of inertia Ic ,
and ωi is the angular velocity of the component i.
The equivalent traslation kinetic energy of the rotative parts is:
1
T = m1 vb2 (12)
2
Using (11) and (12) we obtain:
1  2
+ Ig ng ωg2 + Ic nc ωc2

m1 = 2
Im nm ωm (13)
vb

The mass of the driving pulley m2 and from the driven pulley m3 are computed using the
same scheme. In order to compute the energy at the rollers, we use their mass on both
sides of the belt (µil and µr ). The length of both sides is the same so we may define:
l
m4 = (µr + µil ) (14)
2
CILAMCE 2005 – ABMEC & AMC, Guarapari, Espı́rito Santo, Brazil, 19th – 21st October 2005

For the belt, we compute its equivalent mass from µb and define:
m 5 = µb l (15)
For the raw material we use:
l
m 6 = µl (16)
2
Now, we are in position to estimate the total mass of the system as:
5
X
meq = mi + α0 m6 (17)
i=1
where α0 correspond to the fraction of the nominal load capacity being transported.
Equation (17) corresponds to the mass in movement; nevertheless it does not repre-
sent the mass that effectively transmits the wave front. For that we consider:
mef = meq − m1 − m2 (18)
which allow us to estimate the effective average mass per unit length:
mef
µef = (19)
l
2.2 Belt loads
In stationary regime, the belt is subjected to static loads f1 and f2 , respectively. In
transient situations, we need to add the dynamic loads fd . On the lower side and upper
sides of the belt respectively:
fr = f2 + fd (20)
fc = f1 + fd (21)
Harrison (1) determines max(fd ) using the method of impedance:

∂u
max(fd ) = mef v0 (22)
∂t max
where u is the axial displacement.

3. NUMERICAL MODEL
The model for the belt is based on the work of Harrison (14). It assumes the belt as
a bar clamped on both extremes to a moving rigid structure, as it is shown in figure 3.
The bar has length l. The system has an initial traslation velocity (i.e., nominal conveyor
operation), which may also be cero (starting). Points A and B do not have relative motion
with respect to the reference structure.
Existing analytical solutions consider conservative systems where the mass is uni-
formly distributed along the belt; as a consequence, it is there assumed constant propa-
gation velocity on both load and discharge parts of the belt. On the other hand, the use
of a finite elements model allows us to model variable mass distribution and damping. In
such a way, we take into account that the upper part of the belt carries raw material, while
the lower part does not. These arguments justify the use of numerical solutions to the
problem.
CILAMCE 2005 – ABMEC & AMC, Guarapari, Espı́rito Santo, Brazil, 19th – 21st October 2005

Figure 3: Mathematical model of the belt

3.1 Finite elements model


Following the finite elements method (9), we discretize the belt into n bar elements,
each connected to the next trough a node. For this kind of element, the displacement field
is approximated with a first order polinomial. The element stiffness and mass matrices
are given by:
 
EA 1 −1
Ke = (23)
le −1 1
 
ρAle 1 −1
Me = (24)
2 −1 1
where E is the Young modulus, A is the cross section of the element, ρ is the mass
per unit volume. We also assume that movement is exclusively axial.
The elementary damping matrix Ce is built considering a single viscous damper with
constant c:
 
1 −1
Ce = c (25)
−1 1

After discretization, we solve the matrix equation:

Mẍ + Cẋ + Kx = f (26)

where f is the vector that describes the external loads, M is the mass matrix, C is the
damping matrix, K is the stiffness matrix and x is the displacement vector. In order to
solve (26) we use the Newmark method (9).
We simulated for three transient situations, as it is shown in figure 4:

1. Normal stop
vb
ü0 =
Td
CILAMCE 2005 – ABMEC & AMC, Guarapari, Espı́rito Santo, Brazil, 19th – 21st October 2005

v (m/s)
vb
v* C

B
A

t0 t1 Td time

Figure 4: A.- Normal stopping. B.- Suddenly aborted start. C.- Smooth stopping

drive
1800 m pulley

770 m 9º

take-up
pulley
1º tensor
pulley

Figure 5: General scheme

2. Sudden stop
v∗
ü0 = (27)
T1
3. Smooth stop
 
dv(t) vb π πt
ü0 (t) = = sin (28)
dt 2 Td Td

4. ILLUSTRATIVE EXAMPLE
The case that we study corresponds to an actual belt conveyor of a copper mining at
the north of Chile. The system is used to transport large-diameter rocks and is shown in
figure 5. Model parameters were obtained from reference (10).
We have,
fef = f1 − f2 = 1.117 106 N

4.1 Acceleration profiles


Normal stopping Assuming that no special braking system exists, and that only the static
load apply, the speed will reduce linearly to 0. In order to compute Td , we use the identity:
Z Td
I= fef dt = meq · 0 − meq vb
0
CILAMCE 2005 – ABMEC & AMC, Guarapari, Espı́rito Santo, Brazil, 19th – 21st October 2005

Definition Symbol Value Unit


Nominal belt speed vb 4.75 m/s
belt half length l/2 2561 m
belt mass per unit length µb 118 kg/m
cable mass per unit length µsc 53 kg/m
raw material mass per unit length µl 287 kg/m
rollers mass per unit length (upper side) µil 67 kg/m
rollers mass per unit length (lower side) µir 20 kg/m
Belt nominal capacity q0 4.92 106 Kg/h
Driving pulley mass 27.3 103 kg
Other masses 45.5 103 kg
Motor inertia Im 150 kgm2
Gears inertia Ig 10 kgm2
Coupling inertia Ic 10 kgm2
Static load f1 1.417 106 N
Static load f2 0.300 106 N
Motors nm 4
Gears ng 4
Couplings nc 4
Table 1: Model parameters
Element Symbol Value (103 Kg)
Drive pulley m2 27.3
Driven pulley m3 45.0
Rollers m4 222.8
Belt m5 604.4
Raw material m6 735.0
Table 2: System Inertias

then,
m e vb
Td = (29)
fef
From equation (13) we have:
m1 = 359.7 103 Kg
The rest of mass parameters is detailed in table (2).
So the total mass is:
X
me = mi = 1994.7 103 Kg
i

and
m e vb 1994.7
Td = = 4.75 = 8.5 s
Te 1117.0
CILAMCE 2005 – ABMEC & AMC, Guarapari, Espı́rito Santo, Brazil, 19th – 21st October 2005

v (m/s)

⎛ v* ⎞ ⎛ v* ⎞
arctan⎜⎜ ⎟⎟ + arctan⎜⎜ ⎟⎟
v*

⎝ T0 ⎠ ⎝ T1 ⎠

T0 T1 time

t0 t1

Figure 6: Aborted start. Speed of drive pulley vs time.

From which we may compute the constant negative acceleration to be used during
the integration:
−4.75
q̈ = = −0.5588 m/s2
8.5
Suddenly aborted start An aborted start corresponds to a sudden stop of the system
when it starts to accelerate to find its nominal speed (i.e., due to an overload trip). Accord-
ing to Harrison(10), the sudden change in the acceleration pattern induces an impulsive
effect on the belt, which takes it to a speed:
  ∗  ∗ 
v v
v = T1 tan arctan
∗∗
+ arctan (30)
T0 T1
which generates a dynamic load (see figure 6). We assume that there exists an accel-
eration between t0 y t0 + t1 :
v ∗∗
q̈a = −
T1
From the data:

v ∗∗ = 4.44 m/s

4.44
q̈a = − m/s2
6.8
Smooth stopping We consider a system where acceleration follows a sine function in
time (figure 4) and its given by equation (28). In such a way, no discontinuities appear in
the acceleration. In our case:
 
4.75 π πt
ü0 (t) = sin
2 8.5 8.5
CILAMCE 2005 – ABMEC & AMC, Guarapari, Espı́rito Santo, Brazil, 19th – 21st October 2005

Figure 7: Damping identification setup

The equivalent stiffness is estimated from the effective mass per unit length and the
propagation velocity:
k = µef v02 (31)
We also assume a constant propagation velocity v0 :
v0 = 2355 m/s
The stiffness of the belt is:
k = 1.74 109 N
For the numerical integration we use the Newmark method with integration parame-
ters β = 41 and γ = 21 . These values assure a robust convergence. In what follows, we
show the results for the different cases of study. We first perform an analysis with no
damping, an then we compare results using feasible damping values.

4.2 Damping identification


Figure 7 shows the experimental setup used to estimate the damping factor for the
first mode shape (elongation mode). The belt is in a free-free border condition and an
initial displacement condition. Assuming that the transient response was dominated by
the first mode shape, we obtained an estimation of ξ1 = 0.06, which is the expected range
for this kind of material (15).

4.3 Results
Conservative system We show results for the aborted start case. The belt conveyor starts
to increase to its nominal speed and when it reaches 80% the belt stops suddenly. Accord-
ing to Harrison, this can be modelled using a virtual speed of 4.44 m/s, so we have the
initial conditions (according to figure 3):
u̇(x, t) = 4.44 m/s
ü(0, t) = −0.6529 m/s2
ü(l, t) = −0.6529 m/s2
CILAMCE 2005 – ABMEC & AMC, Guarapari, Espı́rito Santo, Brazil, 19th – 21st October 2005

The maximum tension attains a value of 1.018 106 N at x = 0, instant 2.22 s. This
value is 16.8% higher than the one obtained with a normal stop. Figure 8(a) shows the
level of the dynamic loads for several instants.
Figure 8(b) shows the strain related displacements at several instants. The maximum
value attains 0.86 m at t = 2.22 s. We observe asymmetrical responses due to the correc-
tion of the mass distribution. Response amplitudes are increased with respect to the case
where no mass asymmetry is considered.

Damped system The stiffness of the belt was estimated from the average effective mass
of the system as well as with the average wave propagation velocity:
k = µef v02 = 313.8 · 23552 = 1.74 109 N/m (32)
Using this equivalent
√ stiffness and mass parameters we estimate the critical damp-
ing from cc = 2 km. From the experimental damping factor ξ = c/cc we obtain an
estimation for the damping parameter c. We have:
1.74 109
k= = 339711 N/m
5122
The belt mass is 6.044 105 Kg and

cc = 2 339711 · 604400 Kg/s
ξ = 0.4, and:
c = 3.62 105 Kg/s
We take this value as reference and we used it for the aborted start case. Figure 9 shows
the effect of several damping levels on the maximum load.
We observe that for damping values above c = 45 104 Kg/s the system shows no
vibrations and the load profile is similar to the one at x = 0.

5. CONCLUSIONS
This article builds on previous research from Riffo (11), Fuentes (12) and Amigo
(13). The work analyzes the effect of non-uniformity in the distribution of mass along the
belt (charge and discharge sides) and also damping effects on transient response due to
suddenly aborted starts and stops. Comparing to previous research, we find a reduction of
only 4% in maximum reactions, showing that damping effects are negligible. Of course,
this result is valid for the configuration in study. To the authors knowledge no previous
results appear in the available literature. The implementation of the method is quite simple
and the strategy shows a lot of flexibility to handle different model parameters and border
and initial conditions.

Acknowledgements
The authors wish to acknowledge the partial financial support of this study by the
FOndo Nacional de DEsarrollo Científico Y Tecnológico (FONDECYT) of the chilean
government (projects 1020810 and 1030943).
CILAMCE 2005 – ABMEC & AMC, Guarapari, Espı́rito Santo, Brazil, 19th – 21st October 2005

0.8
.25
.75
1.5
0.6 2.22

0.4
Normalized force

0.2

-0.2

-0.4

-0.6

-0.8
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
x/l

(a) Forces

0.9
.25
.75
0.8 1.5
2.22

0.7

0.6
Displacements [m]

0.5

0.4

0.3

0.2

0.1

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
x/l

(b) Displacements

Figure 8: Aborted start. Strain related forces and displacements at several instants (sec-
onds).
CILAMCE 2005 – ABMEC & AMC, Guarapari, Espı́rito Santo, Brazil, 19th – 21st October 2005

0.03

0.025

0.02
Normalized force

0.015

0.01

0.005

0
4
0 3 10
105
4.5 105
-0.005 106

-0.01
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Normalized time

Figure 9: Dynamic load at x = l/2 for several damping levels

REFERENCES
[1] German Industrial Standard, 1982. Belt Conveyers for Bulk Materials, DIN 22101.

[2] CEMA, 1997. Handbook of Belt conveyors for Bulk Materials.

[3] Harrison, A., 1986. Stress front velocity in elastomer belts with bonded steel cable
reinforcing, Bulk Solids Handling, 6(1), 27-31.

[4] Harrison, A., 1996. Simulation of conveyor dynamics, Bulk Solids Handling, vol.
16, n. 1, pp. 33-36.

[5] Harrison A.,1996. 15 Years of conveyor belt nondestructive evaluation, Bulk Solids
Handling, vol. 16 , n. 1, pp. 13-19.

[6] Harrison A., 1998. On the appropriate use of dynamic stress models for conveyor
design, Bulk Solids Handling, vol. 8, n. 6, pp. 677-680.

[7] Wheeler, C.A., Roberts, A.W., Jones, M.G., 2004. Calculating the Flexure Resis-
tance of Bulk Solids Transported on Belt Conveyors, Particle & Particle Systems
Characterization, vol. 21, n. 4, pp. 340-347.

[8] Wheeler, C.A., 2003. Analysis of the Main Resistances of Belt Conveyors, Ph.D
Thesis, Chapter 5, The University of Newcastle, Australia.

[9] Bathe, K.J., 2003. Finite Element Procedures, Prentice-Hall of India.


CILAMCE 2005 – ABMEC & AMC, Guarapari, Espı́rito Santo, Brazil, 19th – 21st October 2005

[10] Harrison, A., 1988. Informe técnico a la empresa Chuquicamata, Chile.

[11] Riffo, F., 2000. Análisis numérico del comportamiento dinámico de una correa
transportadora. Thesis, Universidad de Concepción, Concepción.

[12] Fuentes J.P., 2002. Modelación numérica de las fuerzas dinámicas producidas en
una correa transportadora de gran longitud, Thesis, Universidad de Concepción,
2002.

[13] Amigo O., 1990. Analisis dinámico de correas transportadoras, Tesis de Magı́ster,
U. de Concepción.

[14] Harrison A., 1983. Criteria for minimizing transient stress in conveyor belts, Me-
chanics eng. Trans. I.E.Aust., vol. 8, n. 3, pp. 129-134.

[15] Harris, C.M., 1996. Shock and Vibration Handbook, 4th ed., Mc-Graw-Hill.

You might also like