You are on page 1of 19

Nonlinear Dyn (2014) 78:609–627

DOI 10.1007/s11071-014-1464-4

ORIGINAL PAPER

A dynamic optimization theoretical method for heavy loaded


vibrating screens
Sergio Baragetti · Francesco Villa

Received: 19 November 2013 / Accepted: 20 May 2014 / Published online: 12 June 2014
© Springer Science+Business Media Dordrecht 2014

Abstract In the present work, an innovative design imental tests, to optimize the dynamic behavior of the
strategy for the optimization of the dynamic perfor- screen and to verify its fatigue resistance.
mances and the structural loads of heavy loaded vibrat-
ing screens is presented. A dynamic model of a vibrat- Keywords Vibrating screen · Numerical
ing screen for the selection of inert materials in an optimization · Dynamic optimization · FEM modeling ·
asphalt plant is proposed, and a numerical optimization Strain gage measurements
procedure is applied to selected design parameters and
geometrical features. The algorithm provides a tool to List of symbols
improve the dynamic behavior of vibrating screens of
different geometric and inertial properties. The results x (m) Horizontal coordinate of the center of
are analyzed, in order to find the parameters apt to min- gravity
imize the pitching angle of the examined screen during y (m) Vertical coordinate of the center of grav-
stationary working conditions, thus providing a bet- ity
ter material selection by reducing gravel throwback. θ (◦ ) Pitching angle of the vibrating screen
A numerical FEM model analysis and an experimental A (m) Horizontal distance between the center
strain-gage campaign have been conducted on a realiza- of gravity and the unload side spring sup-
tion of the vibrating screen, testing two optimized and port
un-optimized configurations, to verify the FEM model B (m) Horizontal distance between the center
results. The complete work gives the machine designer of gravity and the load side spring
a powerful tool, validated by means of full scale exper- A (m) Vertical distance between the center of
gravity and the unload side spring
support—positive in the downward
S. Baragetti (B) · F. Villa
Department of Engineering, University of Bergamo,
direction
Viale Marconi 5, 24044 Dalmine, Italy B  (m) Vertical distance between the center of
e-mail: sergio.baragetti@unibg.it gravity and the load side spring—
F. Villa positive in the upward direction
e-mail: francesco.villa@unibg.it α (◦ ) Inclination of the active force with res-
pect to the vertical
S. Baragetti
GITT, Centre on Innovation Management and Technology
e (m) Eccentricity of the direction of applica-
Transfer, University of Bergamo, Via Salvecchio 19, tion of the active force with respect to the
24129 Bergamo, Italy center of gravity

123
610 S. Baragetti, F. Villa

F0 (N) Active force design amplitude Ct (Ns/m) Total transverse damping coefficient of
n (rpm) Electric engines rotational speed the supports
ω (rad/s) Electric engines angular frequency Cc (Ns/m) Critical damping coefficient in the longi-
F0,x (N) Horizontal component of the active force tudinal direction
F0,y (N) Vertical component of the active force Cct (Ns/m) Critical damping coefficient in the trans-
F (N) FEM model active force amplitude verse direction
Fx (N) FEM model horizontal component of the
active force
Fy (N) FEM model vertical component of the 1 Introduction
active force
k (N/m) Single spring longitudinal stiffness The screening of solid particles by vibrating sieves is a
kt (N/m) Single spring transversal stiffness common industrial practice, which has been employed
E (MPa) Longitudinal elastic modulus of the in a wide range of applications, including food powder
spring processing, chemical processing and asphalt and aggre-
G (MPa) Transversal elastic modulus of the spring gates production systems. Vibrating screen devices are
d (m) Wire diameter of the spring used in order to sieve and select different particles to be
R (m) Radius of the spring helix used in industrial processes in which the final grain size
c = 2R/d Spring curvature ratio and composition of the particulate are vital. Although
m (kg) Global mass of the system the massive adoption of such devices in the production
J (kg m2 ) Global moment of inertia of the system sector, in-depth analyses of the dynamic behavior and
k1 (N/m) Total longitudinal stiffness of the springs the process effectiveness of vibrating screens are for-
related to the unload side supports eign to the research work connected with the design of
kt1 (N/m) Total transverse stiffness of the springs these machines. The dynamics of the particulate onto
related to the unload side supports the screen have been properly described in the work
k2 (N/m) Total longitudinal stiffness of the springs by Soldinger [1], which also elaborated an analytical
related to the load side supports model to estimate the transport velocity of rock mater-
kt2 (N/m) Total transverse stiffness of the springs ial over a vibrating screen of assigned circular motion
related to the load side supports [2]. Recent studies have been conducted, by adopting
K (N/m) Total longitudinal stiffness of the sup- Discrete Element Methods, in order to describe the par-
ports ticulate motions over vibrating sieves, thus identify-
K t (N/m) Total transverse stiffness of the supports ing some parameters for the optimization of the vibrat-
ωnx (rad/s) Natural angular frequency of the system ing screen efficiency, such as the screen-deck inclina-
in the x direction tion angle and the maximum vibration amplitude [3,4].
ωny (rad/s) Natural angular frequency of the system These results tend, however, to be influenced by the
in the y direction shape of the particles and of the vibrating deck, hence
c1 (Ns/m) Total longitudinal damping coefficient of are not universal. Dynamic analyses of the structural
the springs related to the unload side sup- response and shape optimization through Finite Ele-
ports ment Methods of vibrating screens for mining mater-
ct1 (Ns/m) Total transverse damping coefficient of ial have been conducted [5]. However, a study on the
the springs related to the unload side sup- optimization of the dynamics and the motion laws of
ports vibrating screens in order to obtain the desired operat-
c2 (Ns/m) Total longitudinal damping coefficient of ing behavior on such devices has not been systemati-
the springs related to the load side sup- cally developed yet.
ports In the present work, a study of the dynamics of a
ct2 (Ns/m) Total transverse damping coefficient of generic vibrating screen is proposed, by introducing
the springs related to the load side sup- an algorithm which can be used to obtain the desired
ports dynamic behavior by parametric optimization of the
C (Ns/m) Total longitudinal damping coefficient of design features, such as the spring support positioning
the supports and stiffness, and the variation of the amplitude, the

123
A dynamic optimization theoretical method 611

direction and the application point of the vibrating load.


A numerical algorithm is developed, in order to analyze
the way in which the selected design parameters influ-
ence the vibrating screen dynamic output variables of
interest, such as the maximum vertical and horizontal
displacement amplitudes, and the maximum pitching
angle reached in the dynamic excitation of the screen.
Finally, a FEM model and a strain gage measurement
campaign have been conducted over an actual realiza-
tion of the vibrating screen. A complete procedure to
analyze, optimize and provide for structural verifica-
tion of a generic vibrating screen is hence presented
in this paper, giving the machine designer an effective
tool to forecast the dynamic and structural behavior of Fig. 1 3d model of the inclined vibrating screen, showing its
main components, i.e., the flanks, the bridge, the eccentric motors
such devices.
and the crossbeams
The deck to be optimized is a typical heavy loaded,
inclined vibrating screen, adopted in batch asphalt
plants. Such devices are typically used to select the loads induced by the eccentric motors, which are con-
appropriate grain sizes for the mixtures, which are tained in the symmetry plane of the device, the system
mixed to bitumen to form the final composition of the will be described by a three-equation model. The main
product. The inclination of the sieving nets provides geometrical features of the system are represented in
an increased screening capability with a reduced sur- Fig. 2.
face. The device, presented in Fig. 1, is a low carbon The active force, induced by the eccentric motors, is
steel carpentry structure, its main elements being the modeled using the data given by the motor constructor:
flank sides, the bridge supporting the eccentric elec-
F0 = 260,000 N (1)
tric engines, and the crossbeams, which link the flanks
of the device and provide support for the sieving nets. n = 1,000 rpm (2)
The screen is joined to the plant tower structure with 2π n
ω= = 104.7 rad/s (3)
cylindrical helical springs, which form the support of 60
the device. The actual design allows the selection of 6 The modulus of the force is subdivided into its compo-
different grain sizes. nents, considering that the force is inclined of an angle
α = 34.5◦ with respect to the coordinate system.
F0x = F0 sin α = 147,266 N (4)
2 Analytical and numerical methods
F0y = F0 cos α = 214,273 N (5)
2.1 Analytical description of the system The springs are at first modeled with a perfectly elastic
behavior [6–8], the longitudinal and transversal stiff-
In the present subsection, a law of motion for the device ness being obtained from:
center of gravity will be deduced, considering both a
Gd
purely elastic system as well as a description which k= (6)
includes hysteretic damping of the vibrating screen 8 c3 i
springs. The natural frequencies of the system will be E d4 E
kt = = k (7)
then obtained. 64 i R 3 G
To describe the dynamics of the screen, a rigid body The vibrating screen supports can hence be modeled as
model will be outlined, since the torsional and the flex- perfectly elastic, as in the scheme of Fig. 3.
ural stiffness of the vibrating frame are assumed to Notice that the weight of the system is not consid-
be certainly superior with respect to the supporting ered, since it influences only the static position of the
springs. Due to the symmetry of the problem, ensured center of gravity, but does not modify the dynamics of
by the geometrical symmetry of the screen and of the the body. The equations of the system are hence:

123
612 S. Baragetti, F. Villa

Fig. 2 Layout of the


vibrating screen and
geometrical description of
the supports and active force
location with respect to the
system center of gravity

Fig. 3 Loads and reactions


on the vibrating
screen—perfectly elastic
model

 
m ẍ + (kt1 + kt2 ) x + kt1 A − kt2 B  θ = F0x sin (ωt) m ÿ + K y + H θ = Fy (12)
(8) J θ̈ + Ht x + H y + (D + Dt ) θ = −eF (13)
m ÿ + (k1 + k2 ) y + (k1 A − k2 B) θ = F0y sin (ωt) with obvious meaning of the introduced terms.
(9) The optimum working condition for the sieving
 
J θ̈ + kt1 A − kt2 B  x + (k1 A − k2 B) y process has been selected as the zero-pitching config-
  uration, which minimizes gravel throwback, i.e.,
+ k1 A2 + k2 B 2 + kt1 A2 + kt2 B 2 θ
θ =0 (14)
= −eF0 sin (ωt) (10)
θ̇ = 0 (15)
Equations (8–10) can be rewritten in a compact
θ̈ = 0 (16)
form as:
In fact, the presence of particle throwback reduces the
m ẍ + K t x + Ht θ = Fx (11) capacity of a continuous vibrating screen, while a pos-

123
A dynamic optimization theoretical method 613

itive throw would be detrimental to the screening effi- being A x , A y denoted as:
ciency, according to [9] and [10].
Inserting such hypotheses in Eqs. (11–13) leads to F0x /K t
Ax =  2 (32)
the system: 1 − ωωnx
m ẍ + K t x = Fx (17) F0y /K
Ay =  2 (33)
m ÿ + K y = Fy (18)
1 − ωωny
J θ̈ + Ht x + H y = −eF (19)

which is composed of ordinary second order differen- The description of the dynamic system can be refined,
tial equations, and can be resolved in closed form. The by adding the contribution of the hysteretic damping
solution for Eqs. (17–18), as reported in [11], is: of the springs, which can be modeled by appropriate
coefficients and inserted in the model, as in Fig. 4.
x = A1 sin (ωnx t) + A2 cos (ωnx t) The coefficients c1 and c2 are related to the lon-
F0x /K t gitudinal damping of the system, due to the hysteretic
+  2 sin (ωt) (20)
effects of the springs, and are referred to the unload and
1 − ωωnx
load sides, respectively. Similarly, ct1 and ct2 are the
   
y = A3 sin ωny t + A4 cos ωny t transversal damping coefficients. The damped system
F0y /K is described by the following equations:
+  2 sin (ωt) (21)
1 − ωωny  
m ẍ + (ct1 + ct2 ) ẋ + ct1 A − ct2 B  θ̇ + (kt1 + kt2 ) x
√  
where ωnx = K t /m is the natural frequency of the + kt1 A − kt2 B  θ = F0x sin (ωt) (34)

system in the horizontal direction, and ωny = K /m m ÿ + (c1 + c2 ) ẏ + (c1 A−c2 B) θ̇ +(k1 +k2 ) y
is the natural angular frequency in the vertical direction. +(k1 A−k2 B) θ = F0y sin (ωt) (35)
  

The initial conditions of the system are: J θ̈ + ct1 A − ct2 B ẋ + (c1 A−c2 B) ẏ
   
x =0 (22) + c1 A2 +c2 B 2 +ct1 A2 +ct2 B 2 θ̇ + kt1 A − kt2 B  x
 
ẋ = 0 (23) + (k1 A − k2 B) y k1 A2 + k2 B 2 + kt1 A2 + kt2 B 2 θ
y=0 (24) = −eF0 sin (ωt) (36)
ẏ = 0 (25)
which can be rewritten in a simplified form as:
By substituting them in Eqs. (20, 21) we obtain:
F0x /K t ω m ẍ + Ct ẋ + Ct θ̇ + K t x + Ht θ = Fx (37)
A1 = −  2 (26) 
m ÿ + C ẏ + C θ̇ + K y + H θ = Fy (38)
ωnx
1 − ωωnx
J θ̈ + Ct ẋ 
+ C ẏ + (E + E t ) θ̇
A2 = 0 (27)
+Ht x + H y + (D + Dt ) θ = −eF0 sin (ωt) (39)
F0y /K ω
A3 = −  2 (28)
ω
1 − ωωny ny The solution of these equations can be obtained by imp-
osing the optimum zero-pitch working conditions—
A4 = 0 (29)
equations (14–16)—according to [11]:
The law of motion of the center of gravity for a perfectly  
Ct
elastic, rigid body vibrating screen model can hence be x = exp − ωnx t (A1 sin (ωnx t)+ A2 cos (ωnx t))
written as: Cct
  F0x /K t
ω + sin (ωt −φx )
x = A x sin (ωt) − sin (ωnx t) (30)  2 2
ωnx 1− ωωnx + 2 CCctt ωωnx
2
 
ω  
y = A y sin (ωt) − sin ωny t (31)
ωny (40)

123
614 S. Baragetti, F. Villa

Fig. 4 Dynamic model of


the vibrating screen
including damping
hysteretic effects

 
C      external force with angular frequency ω = 104.7 rad/s,
y = exp − ωny t A3 sin ωny t + A4 cos ωny t
Cc the following result occurs:
F0y /K  
+ sin ωt −φ y C γK 1
 2 2 = ξy = √ = 0.008 (46)
2 Cc ω 2 mK
1− ωωny + 2 CCc ωωny
The law of motion of the center of gravity can hence
(41) be approximated, after the transient state is terminated,
Ct and C are the equivalent damping coefficients from with the following equations:
the hysteretic effect of the material, as defined in [12], x = A x sin (ωt + φx ) (47)
and they can be written as:  
y = A y sin ωt + φ y (48)
γ Kt
Ct = = ct1 + ct2 (42) where A x and A y are defined as in Eqs. (32) and (33).
ω
γK The first natural frequencies for the damped system,
C= = c1 + c2 (43) defined as in [11], are approximated as in Eqs. (49,
ω
where γ is the loss factor due to hysteresis, an intrinsic 50), due to the limited contribution of hysteretic damp-
property of the material, that for steels can be assumed ing, remembering that these expressions are obtained
between 0.05 and 0.1, and ω is the angular frequency under the zero-pitching conditions, as defined in
of the active force. (14–16).
 
Cct and Cc are the critical damping coefficients of  
Kt Ct 2 ∼ K t
the system in the two directions, defined as: ωnx = − = (49)
m 2m m
Cct = 2 m K t (44)  
√  
Cc = 2 m K K C 2∼ K
(45) ωny = − = (50)
m 2m m
By confronting the hysteretic damping of the material
with the critical, minimum transitory damping of the To obtain the generic natural frequencies of the dynamic
system, it can be demonstrated that, for the typical char- system, the eigenvalue problem linked to (11–13) must
acteristics of asphalt plant vibrating screens, the ratio be resolved [13]. Expressing system (11–13) in matrix
between these damping coefficients is negligible. For form leads to:
⎡ ⎤
example, if considering the longitudinal motion, with m
an overall spring stiffness K = 1,000 N/mm, a loss fac- M=⎣ m ⎦ (51)
tor coefficient γ of 0.1, a mass m = 3,400 kg and an J

123
A dynamic optimization theoretical method 615

⎡ ⎤
Kt 0 Ht being qx = x, qẋ = q̇x = ẋ, q y = y, q ẏ = q̇ y =
K =⎣ 0 K H ⎦ (52) ẏ, qθ = θ, qθ̇ = q̇θ = θ̇ .
Ht H (D + Dt ) By adopting this model, a numerical solution of the
thus leading to the eigenvalue problem: undamped system is obtained. The dynamics of the
 Kt  undamped system are useful to investigate the behavior
 −λ 0 Ht 
   m m 

of the system in the starting transient of the vibrating
 −1 
M K − λI  =  0 K
m − λ H
m
=0
 screen. In order to describe the motion of the device dur-
  ing its actual operating conditions, a numerical model
 Ht H (D+Dt )
− λ
J J J which includes the hysteretic damping of the springs
(53) has to be reproduced. Starting from Eqs. (37–39), the
 
D + Dt system can be rewritten as a first order differential prob-
(K t /m − λ) (K /m − λ) −λ
J lem of the form (55), by adopting the following matrix
H2 H2 expression:
− (K t /m − λ) − (K /m − λ) t = 0 (54)
mJ mJ ⎧ ⎫ ⎡ 0 1 0 0 0 0



q̇x ⎪

Equation (54) is a cubic equation in λ, and will be ⎪ ⎪ ⎢ − K t − Ct 0 − mt ⎥


⎪ q̇ẋ ⎪⎪ ⎢ m 0 − Hmt
C

solved in the numerical model. ⎪
⎪ ⎪
⎪ ⎢ m ⎥
⎨ q̇ ⎬ ⎢ 0 0 0 1 0 0 ⎥
y ⎢
=⎢ ⎥
⎪ q̇ ẏ ⎪ C ⎥

⎪ ⎪ ⎪
⎪ ⎢⎢ 0 0 − K
m − C
m − H
m − m ⎥
2.2 Numerical rigid body model ⎪
⎪ ⎪ ⎥

⎪ q̇ ⎪ ⎣ 0 0 0 0 0 1 ⎦
⎩ ⎪
θ

q̇θ̇ Ht Ct C D+Dt E+E t
Since the analytical solutions of Eqs. (11–13) and − J −J −J −J − J − J
H
⎧ ⎫ ⎧ ⎫
(37–39), related to the undamped and the damped sys- ⎪ qx ⎪ ⎪ 0 ⎪

⎪ ⎪
⎪ ⎪ ⎪ ⎪

tem, respectively, can be found with a certain effort ⎪
⎪ qẋ ⎪⎪ ⎪
⎪ Fx /m ⎪ ⎪

⎪ ⎪
⎪ ⎪
⎪ ⎪

only by assuming the zero-pitching conditions (14–16), ⎨q ⎬ ⎨0 ⎬
y
a numerical solution of the system is proposed hereby. × + (57)

⎪ q ẏ ⎪
⎪ ⎪
⎪ Fy /m ⎪ ⎪
The solution has been implemented in a numerical algo- ⎪
⎪ ⎪ ⎪
⎪ ⎪ ⎪ ⎪


⎪ qθ ⎪ ⎪ ⎪0 ⎪

rithm, generating a software which easily obtains the ⎪
⎩ ⎪ ⎭ ⎪ ⎩ ⎪

dynamics of the system. The equation solutions were qθ̇ −Fe/J
numerically determined using an explicit Runge–Kutta where the appropriate coefficients are the same of
solver for non-stiff problems [14]. In order to use this the system (37–39). The damping coefficients can be
tool, the system must be put in the form: obtained using the relation defined in [12]:

ẏ = f (y, t) γ k1
(55) c1 = (58)
y (0) = y0 ω
γ k2
The undamped system (8–10) can be hence expressed c2 = (59)
ω
as:
⎧ ⎫ ⎡ ⎤⎧ ⎫ The transversal damping can be easily obtained by rela-
⎪ q̇x ⎪ 0 1 0 0 0 0 ⎪ qx ⎪

⎪ ⎪
⎪ ⎢ Kt ⎥⎪⎪ ⎪ ⎪ tion (7) as:

⎪ q̇ẋ ⎪
⎪ ⎢ − m 0 0 0 − Hmt 0⎥⎪ ⎪ qẋ ⎪⎪

⎪ ⎪ ⎥⎪ ⎪
⎨ q̇ ⎪ ⎬ ⎢⎢ 0 0 0 1 0 0 ⎥

⎨ q

⎬ c1t =
E
c1 (60)
=⎢ ⎥
y y
⎢ G
⎪ q̇ ẏ ⎪
⎪ ⎪ ⎢ 0 0 −m 0 −m
K H
0⎥⎥⎪⎪ q ẏ ⎪


⎪ ⎪
⎪ ⎢ ⎥⎪⎪ ⎪ ⎪
E
c2t = c2

⎪ θ⎪
q̇ ⎪
⎪ ⎣ 0 0 0 0 0 1⎦⎪ ⎪ qθ ⎪ ⎪ (61)

⎩ ⎭ ⎪
⎩ ⎪
⎭ G
q̇θ̇ − HJt 0 − HJ 0 − D+DJ
t
0 q θ̇ The numerical model is then verified by creating
⎧ ⎫
⎪ 0 ⎪ an equivalent rigid body model, and comparing the

⎪ ⎪


⎪ Fx /m ⎪



obtained results. A CAD model of a vibrating screen
⎨ ⎬ is loaded into the rigid body solver, and the longitu-
0
+ (56)

⎪ Fy /m ⎪
⎪ dinal and tangential stiffness behaviors of the springs

⎪ ⎪


⎪ 0 ⎪
⎪ are reproduced, by introducing vertical and horizontal
⎩ ⎭
−Fe/J linear spring models in correspondence of the support

123
616 S. Baragetti, F. Villa

points. Stiffness k is assigned to the vertical spring ele- changes. The undamped system is optimized by iter-
ments, while the horizontal ones have a stiffness equal ating on the eight chosen parameters, and simulat-
to kt . A damped system is also reproduced and com- ing the undamped dynamics for every spot. The opti-
pared with the damped results from the present analyt- mized parameters are taken as the combination of val-
ical model. ues which minimizes the maximum regime amplitude
of θ in the dynamic process. Optimization by the use of
the damped model is feasible at the price of higher com-
2.3 Sensitivity analysis and design optimization putational costs, due to the need to simulate an extended
time interval for every spot of the optimization process.
In order to design an optimized vibrating screen, the However, a sensitivity analysis on the influence of the
verified damped numerical model will be used as a basis various parameters over the displacement outputs of the
for the generation of an optimization algorithm, which system has been carried out using the damped system,
will be adopted to improve the screen dynamics, and and proved to agree satisfactorily with the undamped
in particular to reduce the pitching angle θ , to mini- results, thus justifying the optimization of the system
mize gravel throwback during sieving. The parameters with the undamped model, as will be showed in Sect. 4.
set which will be adopted for the optimization is pre-
sented here below, coherently with the dimensions set
in Figs. 2 and 3:
2.4 FEM model
• A—horizontal distance between the center of gravity
and the unload side spring support A finite element model of the vibrating screen was real-
• B—horizontal distance between the center of gravity ized in order to provide a verification of the stress state
and the load side spring support of the vibrating screen at the maximum amplitude of
• A —vertical distance between the center of gravity the external force, and to determine the natural fre-
and the unload side spring support—positive in the quencies and deformation modes of the screen. The
downward direction geometry was discretized with three-dimensional, lin-
• B  —vertical distance between the center of grav- ear tetrahedral elements, as it can be seen in Fig. 5.
ity and the load side spring support—positive in the Suitable mesh refinements were adopted near disconti-
upward direction nuity regions, as in Fig. 6. The material was described
• α—inclination of the active force with respect to the using an isotropic and linear elastic model. The con-
vertical straints were set in order to reproduce the supports and
• F—external force amplitude the symmetry along the vertical axis, as reported in
• k1 , k2 —total longitudinal stiffness of the unload and Fig. 5.
load side springs, respectively. The external force modulus was taken as the maxi-
• kt1 , kt2 —total transversal stiffness of the unload mum force generated by the vibrating motors, i.e., the
and load side springs, respectively. 73 % of the total design force. The load was subdivided
The optimization parameters are all contained in the and applied at the bolt holes in the motor supporting
K matrix (52) and in the known term of the undamped deck beam, as in Fig. 7, considering the symmetry of
system (11–13). In particular, matrix K can be rewritten the problem. The load was equally distributed on the 6
in order to emphasize some of these parameters as: holes, and the load in the central hole was subdivided
⎡ ⎤ by an half for symmetry reasons. The angle was chosen,
kt1 +kt2 0 kt1 A −kt2 B 
⎢ ⎥ from the optimized model results, as α = 30◦ .
K =⎣ 0 k1 +k2 k1 A−k2 B ⎦
Keeping these consideration in mind, the applied
kt1 A −kt2 B  k1 A−k2 B k1 A2 +k2 B 2 +kt1 A2 +kt2 B 2
forces came from the subsequent relations:
(62)
F = 190,216 N ∼
= 0.73F0 (63)
remembering that kt = GE k. ◦
α = 30 (64)
When performing the sensitivity analysis and the 
numerical optimization, matrix (62) must be recal- Fx = 16 F cos α = 27,455 N
on holes 1−2 (65)
culated every time one of the contained parameters Fy = 16 F sin α = 15,851 N

123
A dynamic optimization theoretical method 617

Fig. 5 FEM Model and constraints: a Tetrahedral mesh, b Horizontal and vertical constraints on the supports, c Symmetry constraints
on the vertical axis of the vibrating screen

Fig. 6 Mesh refinement


around structural
discontinuities and details

FEM analysis results. For every significant spot, two


strain gages were mounted, one orthogonally with the
other, and oriented along the principal stress directions,
verified via FEM analysis, forming an half Wheat-
stone bridge which was balanced in acquisition. Half-
bridge shear strain gages were mounted on the springs
to reconstruct the force exchanged with the supports
and the longitudinal displacement, to obtain the pitch
angle θ . The test were conducted in an empty config-
uration of the vibrating screen, without the gravel and
the sieve nets. The screen developed for the testing was
equipped with flanges which permitted to change the
Fig. 7 Load definition on the deck beam: a Bolt holes number- anchoring on the spring supports on the flank of the
ing, b Load definition machine, thus allowing to test both the optimized and
the anti-optimized configuration with the same device

Fx = 1
12 F cos α = 13,728 N and test bed. The spring supports on the load side were
on holes 3−4 (66) equipped with one extra spring slot, in order to change
Fy = 1
12 F sin α = 7,926 N
the stiffness of the load side. The experimental set up
is displayed in Fig. 8.
3 Experimental methods The stress reconstruction from the measured linear
deformations was obtained using the following rela-
A realization of the optimized vibrating screen was tions for plane stress [15–18]:
mounted on a test bed and equipped with strain gage
sensors. Linear strain gages were mounted on the E
σ1 = (ε1 + υε2 ) (67)
regions of the screen which resulted critical from the 1 − υ2

123
618 S. Baragetti, F. Villa

Fig. 8 Experimental test set up: a Global layout of the testing machine, b Linear strain gages mounted orthogonally on the upper flange
of the side flank, c Shear strain gages mounted on a supporting spring

Table 1 Experimental test plan


Test No. Strain gages zone IDs Geometric configuration K load side (N/m) K unload side (N/m)

1 F1L–F1R–F2L–F2R Optimized 660,000 440,000


2 F3L–F3R–F4L–F4R Optimized 660,000 440,000
3 S1L–S1R–S2L–S2R–B1L–B1R–B2L–B2R Optimized 660,000 440,000
4 S1L–S1R–S2L–S2R–B1L–B1R–B2L–B2R Optimized 440,000 440,000
5 S1L–S1R–S2L–S2R B3L–B3R Unoptimized 440,000 440,000
6 S1L–S1R–S2L–S2R B3L–B3R Unoptimized 660,000 440,000

E where D = 129 mm is the spring diameter. Once the


σ2 = (ε2 + υε1) (68)
1 − υ2 forces acting on the springs are known, the pitching
where 1 and 2 indicate the two orthogonal directions. angle θ can be easily calculated. If δ1 is the downward
The von Mises stress in the spot were then calculated movement of the vibrating screen coinciding with the
using the relation: global compression force F1 on the unload side, and
 δ2 is the analogous downward vertical displacement
σvm = σ12 + σ22 − σ1 σ2 (69) which generates the compression force F2 , the overall
pitching angle θ can be reconstructed from:
As for the shear strain gages, the shear stress acting on
θ = θ2 − θ1 = Bδ2 − Aδ1 (73)
the spring was reconstructed by the constitutive equa-
tion: Since the acquisition channels were limited, 4 test run
were performed on the optimized vibrating screen and
τ = Gγ (70)
2 in a non-optimal configuration. Apart from the strain
being γ the measured shear strain. From the shear stress gages positions, the stiffness of the load side springs
value, the torsional moment acting on the spring can be was changed by inserting or removing the central spring
reconstructed [7]: in the load side support (see Fig. 8a). An overall scheme
of the tests is presented in Table 1, while the locations of
π d 3τ the strain gages are presented in Table 2. The definition
Mt = (71)
16 of the exact positions of the strain gages on the flanks
with d = 21 mm as the wire diameter of the spring. and the deck beam is given in Fig. 9.
The longitudinal force which is exchanged with the In Table 3, a comparison between the actual geo-
supports can be obtained hence from: metric realization of the screen and the calculated opti-
mized and de-optimized geometric and stiffness val-
D ues, obtained by the model presented in Sect. 2.3, is
Mt = F (72)
2 exposed.

123
A dynamic optimization theoretical method 619

Table 2 Strain gage


Strain Gage ID Description Configuration
localization and zone ID
F1L(R) Left (right) upper flange of the flank—unload side LT
F2L(R) Left (right) internal side of the flank—unload side LT
F3L(R) Left (right) upper flange of the flank—load side LT
F4L(R) Left (right) internal side of the flank—load side LT
S1L(R) Left (right) spring—unload side S
S2L(R) Left (right) spring—load side S
B1L(R) Left (right) deck beam—upper central region L
L longitudinal, LT B2L(R) Left (right) deck beam—upper side region L
longitudinal and transversal, B3L(R) Left (right) deck beam—proximity of the motor anchoring plate LT
S Shear

Fig. 9 Strain gage zone IDs identification

Table 3 Discrepancies
Parameter Optimized screen Optimized screen De-optimized screen De-optimized screen
between the optimized and
algorithm values actual realization algorithm values actual realization
de-optimized models and
their realizations A 800 mm 805 mm 1,000 mm 1,140 mm
B 1,400 mm 1,395 mm 1,000 mm 1,060 mm
A 600 mm 622 mm 1,000 mm 1,050 mm
B −100 mm −122 mm −500 mm −550 mm
K1 380,000 N/m 440,000 N/m 520,000 N/m 440,000 N/m
K2 520,000 N/m 660,000 N/m 520,000 N/m 440,000 N/m
α 30◦ 30◦ 40◦ 40◦
F 190,216 N 133,150 N 157,860 N 133,150 N

The number of springs which were mounted on 4 Results and discussion


every side can be easily determined, if we consider that
the measured longitudinal stiffness of a single spring 4.1 Introduction
was 110,000 N/m. The external force module was cho-
sen as the 70 % of the maximum motor force module In the present section, a comparison between the
(63): numerical dynamic model and the rigid body model are
presented, thus verifying the numerical model which
Ftest = 0.70 F = 133,150 N (74) has been used in the optimization phase. Sensitivity

123
620 S. Baragetti, F. Villa

Fig. 10 Comparison between undamped models of the vibrating screen: a Rigid body model, b Numerical model; the time histories
of the x and y coordinates of the center of gravity and the pitching angle θ are reported for each model

Table 4 System output


D.O.F Natural Natural Maximum Maximum
comparison for the
frequencies frequencies displacements displacements
unoptimized undamped
rigid body numerical rigid body numerical
screen—transitory
model model model model
maximum values
x f x = 2.864 Hz f x = 2.907 Hz 0.025 m 0.029 m
y f y = 2.426 Hz f y = 2.405 Hz 0.034 m 0.036 m
θ f θ = 4.549 Hz f θ = 4.581 Hz 0.69◦ 1.13◦

123
A dynamic optimization theoretical method 621

Fig. 11 FFT of the


Numerical model output for
the three displacement
histories: a y vertical
displacement, b x vertical
displacement, c θ angular
coordinate. Natural angular
frequencies from the
eigenvalue analysis are
reported in grey, the
continuous line indicating
the related coordinate

analyses of the model which led to the optimization cies, is showing a reduced amplitude for the angular
results are displayed, adopting both the undamped and coordinate, with respect to the other displacements.
damped simulations. The results of the FEM model are A comparison between the numerical and rigid body
presented, highlighting the regions of interest which damped models was also performed, in order to evalu-
were instrumented with strain gages in the experimen- ate the behavior of the developed model with the inclu-
tation phase. Finally, a comparison between the FEM sion of the hysteretic damping. The results are dis-
values of the Von Mises stress σvm and the recon- played in Fig. 12, and the maximum values at t = 50 s
structed σvm values from the strain gage measurements in Table 5, where it can be seen that the rigid body model
in different operating conditions is presented. presents a more intense transitory, although the regime
values present a good match. The numerical undamped
model will be hence considered verified, and will be
used as an optimization tool for the vibrating screen.
4.2 Comparison and validation between numerical
and rigid body models
4.3 Sensitivity analysis and optimization
The simulation time histories of the output variables,
i.e., the center of gravity displacements x and y and the To prove the adequacy of the undamped model for the
pitching angle θ obtained from the rigid body model optimization procedure, a sensitivity analysis of the
are displayed in Fig. 10, and compared to the results of system was conducted over the whole range of the cho-
the numerical model. The main parameters values are sen parameters both for the damped and the undamped
reported in Table 4. systems, comparing the x, y and θ outputs between the
As it can be seen from Fig. 10 and Table 4, the com- two formulations. In order to visualize the results of
parison between the two models shows a good resem- the analysis, only one or two parameters at a time were
blance. Some differences can be, however, recognized: changed during the simulation. The results proved the
a certain time shift between the models, inferior to effectiveness of using the undamped optimized model
0.2 s, and a fading of the high frequencies linked to for the analysis, which produced the overall best para-
the external forcing in the θ output of the numerical meters, which have already been presented in Table 3.
model. In the other outputs, the fluctuations of the exter- The results of the sensitivity analysis over the θ para-
nal applied force, which has a frequency of 16.67 Hz, meter have been presented in Fig. 13 for comparison.
are clearly displayed. The phenomenon is reported in As it can be seen from the comparison between the
Fig. 11, where the Fast Fourier Transform of the model two models in Fig. 13, the regions of optimization for
output is compared with the frequencies obtained by the θ show the same overall behavior if considering both
eigenvalue analysis. The effect of the external forcing the damped and the undamped model. Obviously, the
at 16.67 Hz, which can be clearly seen in Fig. 11 as the maximum angles reached by the undamped model are
separated peak with respect to the eigenvalue frequen- higher due to the presence of spikes in the optimization

123
622 S. Baragetti, F. Villa

Fig. 12 Comparison between damped models of the vibrating screen: a Rigid body model, b Numerical model; the time histories of x
and y coordinate of the center of gravity, and the pitching angle θ are reported for each model

Table 5 System output comparison for the optimized vibrating the motors support beam and the flank regions near the
screen damped model, regime maximum values (t = 50 s) vertical support of the beam, as showed in Fig. 14.
D.O.F. Regime maximum Regime maximum The maximum values of σvm are given in Fig. 14
displacements displacements as well, neglecting the numerical singularities which
rigid body model numerical model derive from modeling imperfection, such as the stress
x 0.0047 m 0.0047 m concentration around the bolt holes. The zones are
y 0.00272 m 0.00270 m defined coherently with the strain gages positioning ID
θ 0.017◦ 0.059◦ described in the experimental testing section.
The FEM model was used also to extract the first
structural natural frequencies, given in Fig. 15 along
space. The undamped model can be hence adopted to
with the relative deformation modes. The reported
be used in the optimization model.
modes highlight the vibrational behavior of the thinner
carpentry structures, which are easily excited at low
4.4 FEM results frequencies. However, the first mode of the carpentry,
found at 50.34 Hz, is well above the frequency of the
The results in terms of σvm stresses highlighted that external load, which is of 16.67 Hz, from Eq. (3), thus
the most critical regions of the vibrating screen were preventing resonant behavior.

123
A dynamic optimization theoretical method 623

Fig. 13 Sensitivity analysis results: a, b θmax optimization for e θmax optimization for eccentricity e—damped and undamped
A and B—a Undamped model, b Damped model, c, d θmax opti- models, f θmax optimization for force amplitude F and inclination
mization for A and B  —c Undamped model, d Damped model, α—damped and undamped models

4.5 Experimental test results As can be noticed from the comparison of optimized
screen test results (Test No. 1–3) from Table 6 with the
The regime maximum von Mises stresses, obtained FEM results reported in Fig. 14, a good agreement is
from Eqs. (67–69) using the linear strain gage mea- reached between the experimental measurements and
surements which were conducted on a realization of the FEM model for the higher stresses, measured in the
the vibrating screen, both in the optimized and unopti- upper side of the deck beam, obtaining a maximum dis-
mized configurations, are reported in Table 6. The test crepancy of 25 %. Good agreement is reached also in
characteristics for every run are reported in Table 1, the upper flange and the side of the unload flank zone,
and the strain gage location IDs in Fig. 9 and Table 2. especially in the left side where a 5 % error is found.

123
624 S. Baragetti, F. Villa

Fig. 14 Visualization of FEM results: global σvm stresses distribution on the model and closeup of the σvm contour near the strain gage
positions, with numerical values in MPa

Fig. 15 Deformation modes associated with the first 5 frequencies of the system: a First mode: 50.34 Hz, b Second mode: 62.91 Hz, c
Third mode: 63.67 Hz, d Fourth mode: 64.89 Hz, e Fifth mode: 65.99 Hz

The right region presents a gap of 25 % with respect to between results has been found in the load side region,
the FEM model, suggesting a slight asymmetry in the although it must be considered that the measured val-
actual realization of the screen. A certain difference ues were inferior to 3 MPa, thus increasing the influ-

123
A dynamic optimization theoretical method 625

Table 6 Regime
Test No. Zone ID σvm (MPa) Test No. Zone ID σvm (MPa) Test No. Zone ID σvm (MPa)
experimental results from
linear strain gage 1 F1L 1.9 3 B1R 10 5 B3L 12.5
measurements
F1R 1.5 B1L 9 B3R 12.5
F2L 6 B2R 0.4
F2R 4.7 B2L 0.4
2 F3L 1.1 4 B1R 10 6 B3L 12.5
F3R 1.4 B1L 9 B3R 12.5
F4L 2.8 B2R 0.4
F4R 2.8 B2L 0.4

Table 7 Experimental results from shear strain gage measure- conditions, a brushless stepper motor control could be
ments on the springs implemented in order to reduce the overstressing of the
Test No. τreg (MPa) Freg (N) τstop (MPa) Fstop (N) supports while the device is halting.
A sample of the directly measured shear strain
3 9.4 265 180 5,100
γ , related to test 3, is reported in Fig. 16a, show-
4 10.6 300 140 4,000
ing the spring overload during the machine stop. The
5 24.5 690 122 3,400 same behavior was detected on the linear strain gages
6 20 560 100 2,800 mounted on the sides of the motor deck beam, as indi-
cated in Fig. 16b, although the maximum strain value
ence of the measurement errors. A limited variation is was limited if compared to the stress measurements in
found in the stress comparison of the lateral region of the other significant regions of the screen, which did
the deck beam, i.e., in the B2L(R) measuring point, not show this amplification during the stop phase, as
with a 0.4 MPa gap between FEM and experiments. shown in Fig. 16c.
FEM simulations were always overpredicting the mea- Finally, the pitching angle θ was reconstructed from
sured result, thus providing a conservative model for the measured shear of the springs, using relation (73),
the vibrating screen verification. A certain asymmetry and compared with the predicted value for tests 3 and 4,
was found on the actual realization of the screen, espe- resulting in the data presented in Table 8. The predicted
cially regarding test 1. values were obtained by simulating the actual realized
The results obtained from the shear strain gages screen layout.
mounted on the springs are reported in Table 7. The The results clearly indicate the accuracy of the pre-
regime values of the maximum tangential stress τ and diction from the dynamic model. Obviously, configu-
of the relative vertical force F exchanged with the sup- ration 3 has a reduced pitching angle with respect to
port are indicated for the most loaded spring. The val- configuration 4, since its parameters are closer to the
ues of τ and F during the stop transitory are indicated ideal optimized model—see Tables 1 and 3.
as well, since experimental testing showed a dramatic
increase of the spring loads of up to 20 times the regime
values, due to the heavy screen vibrations when the 5 Conclusions
eccentric motors were halted.
From the results obtained in Table 7, it can be In the present work, a dynamic optimization proce-
observed that the optimized configurations 3 and 4 dure for vibrating screens of arbitrary geometry has
exchange roughly half of the maximum loads with the been outlined. The dynamic model has been exhaus-
supporting structure, if compared with the unoptimized tively verified both numerically and experimentally.
configurations 5 and 6. During the stop phase, however, Strain gage measurements have been carried out over
significant oscillations led to a noticeable load incre- an actual realization of the vibrating screen, both in an
ment of up to 20 times for the optimized configuration. optimized and anti-optimized configuration, measuring
Although the vibrating screen is a continuously work- the stresses values in the critical locations which were
ing machine, thus justifying a design based on regime highlighted by FEM analysis. A reconstruction of the

123
626 S. Baragetti, F. Villa

Fig. 16 Measurement
outputs recorded during test
3—ending transitory
beginning at approximately
230 s: a Shear strain, related
to the S1L strain gage, b
Linear strain, related to the
strain gage mounted on the
side of the beam deck
(B2L), c Linear strain,
related to the strain gage
mounted on the center of the
beam deck (B1L)

pitching angle allowed to verify the actual improve- • A numerical algorithm, which describes this model,
ment of the optimized configuration. By resuming the has been successfully created and validated against
results of the present work, we can state that: commercial software.
• An optimization technique based on such algorithm
• A dynamic model of vibrating screens, suitable to has been developed, identifying an optimized work-
different geometries and layouts, has been set up. ing condition which minimizes gravel throwback

123
A dynamic optimization theoretical method 627

Table 8 Pitching angle θ: comparison between values from the 4. Zhao, L., Zhao, Y., Liu, C., Li, J., Dong, H.: Simulation of
dynamic model and reconstructed values from experimental data the screening process on a circularly vibrating screen using
3D-DEM. Min. Sci. Technol. (Xuzhou, China) 21(5), 677–
Test No. Expected θ (◦ ) Measured θ (◦ ) 680 (2011)
5. Zhao, Y., Liu, C., He, X., Zang, C., Wang, Y., Ren, Z.:
3 0.0109◦ 0.0115◦
Dynamic design theory and application of large vibrating
4 0.0201◦ 0.0218◦ screen. Procedia Earth Planet. Sci. 1(1), 776–784 (2009)
6. Bazzaro, E., Gorla, C., Miccoli, S.: Lezioni di Tecnica delle
Costruzioni Meccaniche - Seconda Edizione. Schonenfeld
& Ziegler, Milano (1997)
by reducing the pitch angle θ , and decreases the 7. Davoli, P., Vergani, L., Beretta, S., Guagliano, M., Baragetti,
support loads in the regime conditions. S.: Costruzione di Macchine 1–2a Ed. Milano, McGraw-Hill
• After appropriate structural verifications via FEM Italia (2007).
8. Childs, P.R.N.: Mechanical Design, 2nd edn. Elsevier B-H,
modeling, a realization of the vibrating screen has Oxford (2004)
been tested in optimized and anti-optimized config- 9. Smith, M.R., Collins, L., Fookes, P.G.: Aggregates: Sand.
urations, showing the effectiveness of the adopted The Geological Society, Gravel and Crushed Rock Aggre-
procedure. gates for Construction Purposes. Bath (2001)
10. Bringiotti, M.: Frantoi e Vagli. Parma, Pei (2002)
• The experimental testing has highlighted huge 11. Hartog, J.P.D.: Mechanical Vibrations. Dover Publications,
vibrations when the screen was arrested, leading Mineola (NY) (1984)
to the highest recorded values of exchanged forces 12. Dimarogonas, A.: Vibration for Engineers. Prentice Hall,
with the supports. A dynamic control in the ending Upper Saddle River (NJ) (1996)
13. de Silva, C.W.: Vibration: Fundamentals and Practice. Tay-
transitory of the device is recommended in order lor & Francis, Boca Raton (2006)
to minimize the outstanding measured loads in the 14. Quarteroni, A., Sacco, R., Saleri, F.: Numerical Mathemat-
stop transient phase. ics. Springer, Berlin/Heidelberg (2007)
15. Young, W.C., Budynas, R.G., Sadegh, A.M.: Roark’s For-
mulas for Stress and Strain, 8th edn. McGraw-Hill, New
Acknowledgments The Authors wish to thank Bernardi York (2012)
Impianti for the support of the research project. 16. Comi, C., Corradi Dell’Acqua, L.: Introduzione alla Mecca-
nica Strutturale, 2nd edn. McGraw-Hill Italia, Milano (2007)
17. Pilkey, W.D.: Formulas for Stress, Strain, and Structural
References Matrices. Wiley, Hoboken, NJ (2005)
18. Terranova, A., Baragetti, S.: Progetto e calcolo di sistemi
meccanici. Milano, Hoepli (2008)
1. Soldinger, M.: Interrelation of stratification and passage in 19. D’Argenio, A., Ravasio, P.: Studio di vagli oscillanti per
the screening process. Miner. Eng. 12, 497–516 (1999) inerti: modelli analitici, numerici e riscontri sperimentali.
2. Soldinger, M.: Transport velocity of a crushed rock material M.Sc. Thesis—Università degli Studi di Bergamo, A.A.
bed on a screen. Miner. Eng. 15, 7–17 (2002) 2003–2004
3. Wang, G., Tong, X.: Screening efficiency and screen length
of a linear vibrating screen using DEM 3D simulation. Min.
Sci. Technol. (Xuzhou, China) 21, 451–455 (2011)

123

You might also like