See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/315530097

Residual Swing Suppression of Suspended Loads using Parametric

Acceleration Profiles

Conference Paper · January 2010

DOI: 10.2316/P.2010.675-014

CITATIONS READS

0 32

1 author:

Giovanni Incerti
Università degli Studi di Brescia
69 PUBLICATIONS   67 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Dynamic Simulation of Industrial Cam Devices with Velocity Control View project

All content following this page was uploaded by Giovanni Incerti on 06 June 2019.

The user has requested enhancement of the downloaded file.

Proceedings of the 29th IASTED International
Conference on

Modelling, Identification,
and Control
Editor: K. Hangos

February 15 - 17, 2010

Innsbruck, Austria

ISBN:
CD: 978-0-88986-819-9

675MIC Start >>

 IASTED Home ACTA Press Home Contact
Us

Modelling, Identification,
and Control
International Program Committee
(MIC 2010)
Additional Reviewers
February 15 - 17, 2010
Information on Publication
Innsbruck, Austria
Foreword

Table of Contents

Author Index

Main Page

SPONSOR
Conferences The International Association of Science and Technology for Development (IASTED)

Membership EDITOR
K. Hangos – Hungarian Academy of Sciences, Hungary
Publications
PLENARY SPEAKER
FAQs
R. Rajamani – University of Minnesota, USA
Search
INTERNATIONAL PROGRAM COMMITTEE
S. Alwash – Jordan University of Science & Technology, Jordan
J.C. Amaro Ferreira – ISEL, Portugal
A. Ashimov – Institute of Problems of Informatics and Control, Kazakhstan
F. Assous – Ariel University Center, Israel
H. Attia – National Research Council Canada, Canada
J. Boaventura – University of Tras-os-Montes and Alto Douro, Portugal
V. Bobal – Tomas Bata University in Zlin, Czech Republic
G. Bontempi – Université Libre de Bruxelles, Belgium
S.R. Das – University of Ottawa, Canada
B. Dobrucky – University of Zilina, Slovakia
E. Furutani – Kyoto University, Japan
T. Hagiwara – Kyoto University, Japan
R. Henriksen – Norwegian University of Science and Technology, Norway
H. Hu – University of Essex, UK
I. Jesus – Institute of Engineering of Porto, Portugal
B. Joseph – University of South Florida, USA
V. Jotsov – Bulgarian Academy of Sciences, Bulgaria
D. Jovcic – University of Aberdeen, UK
H. Kawabe – Hiroshima Institute of Technology, Japan
T. Kawabe – University Tsukuba, Japan
K. Khalili – University of Birjand, Iran
T. Kulworawanichpong – Suranaree University of Technology, Thailand
C. Kyrtsou – University of Macedonia, Greece
K. Li – Queen's University Belfast, UK
W.M. Lippe – University of Münster, Germany
K.L. Man – University College Cork, Ireland
M. Marcoková – University of Zilina, Slovakia
G. Merkuryeva – Riga Technical University, Latvia
E.D. Moreno – Federal University of Sergipe, Brazil
P. Nahodil – Czech Technical University in Prague, Czech Republic
G. Nikolakopoulos – University of Patras, Greece
C. Pinto – Higher Institute of Engineering of Porto, Portugal
M.M. Polycarpou – University of Cyprus, Cyprus
N. Prljača – University of Tuzla, Bosnia and Herzegovina
I.J. Ramírez-Rosado – University of Zaragoza, Spain
M. Rodrigues – Sheffield Hallam University, UK
N.J. Roussos – Hellenic Navy Research Center, Greece
B. Sohlberg – Dalarna University, Sweden
P. Sowa – Silesian University of Technology, Poland
A. Swierniak – Silesian University of Technology, Poland
J.A. Tenreiro Machado – Instituto Superior de Engenharia do Porto, Portugal
A. Tornambè – University of Rome Tor Vergata, Italy
H. Trinh – Deakin University, Australia
K. Tsakalis – Arizona State University, USA
R. Ulson – State University of São Paulo - UNESP, Brazil
G. Varga – University of Miskolc, Hungary
J. Vittek – University of Zilina, Slovakia
Q.G. Wang – National University of Singapore, Singapore
E.E. Yaz – Marquette University, USA
W. Yu – CINVESTAV-IPN, Mexico
V. Zanchi – University of Split, Croatia
L. Zhang – Harbin Institute of Technology, PRC

ADDITIONAL REVIEWERS
M. Bakosova – Slovakia
S. Ben Taieb – Belgium
C. Botan – Romania
I. Damousis – Greece
M. de la Sen – Spain
P. Dostal – Czech Republic
H. Elsidani Elariss – Lebanon
L.A. Fernandez-Jimenez – Spain
N. Ferreira – Portugal
G. Georgoulas – Greece
F. Gurleyen – Turkey
M. Hansen – Norway
D. Honc – Czech Republic
G. Hovland – Norway
G. Incerti – Italy
H.R. Karimi – Norway
E. Khmelnitsky – Israel
M. Kubalcik – Czech Republic
R. Lobato – Brazil
R. Matusu – Czech Republic
A. Pantelous – UK
D. Puangdownreong – Thailand
A. Shamekh – UK
D. Strunin – Australia
M. Sulitka – Czech Republic
S. Tokat – Turkey
V. Troynikov – Australia
J. Vojtesek – Czech Republic
F. Yang – USA

INFORMATION ON PUBLICATION
For each IASTED conference, the following review process is used to ensure the highest level of
academic content. Each full manuscript submission is double blind reviewed by a minimum of two
separate reviewers on the International Program Committee/Additional Reviewers list. The review
results are then compiled. If there are conflicting reviews, the paper is sent to a third reviewer.

Photo Credits:
Background: Techno/Lynne Lancaster.
Top: Building 3D Wireframe/Barun Patro; Middle: Future/Iva Villi; Bottom: Switch/Luciano
Tirabassi.

Copyright © 2010 ACTA Press

ACTA Press ACTA Press ACTA Press

P.O. Box 5124 B6, Suite #101, 2509 Dieppe Ave SW P.O. Box 354
Anaheim, CA 92814–5124 Calgary, Alberta T3E 7J9 CH–8053 Zurich
USA Canada Switzerland

Publication Code: 675

E. Briano, C. Caballini, P. Giribone, and R. Revetria (Italy)

675-073
Modeling the Impacts of Deregulation and Competitive Entry on Cable Service
K.H. Tiedemann (Canada)

675-074
Modeling the Impact of Voluntary Time of Use Rates for Business and Industry
K.H. Tiedemann (Canada)

675-081
Modeling the Impact of Residential Time of Use Rates
K.H. Tiedemann (Canada)

675-082
Intellectual Property (IP) Cores and a Logic Fault Test Simulation Environment
M.H. Assaf, L.-A. Moore (Trinidad and Tobago), S.R. Das (Canada, USA), S.N. Biswas, A.R. Applegate
(USA), and E.M. Petriu (Canada)

Process and Energy Systems

675-023
Modeling and Control of an Ill-Conditioned Distillation Column
V.T. Minh and I. Bin Ibrahim (Malaysia)

675-034
An Averaged Model of Dispersion of Pollutant in a Channel: Logarithmic Flow
D.V. Strunin (Australia)

675-035
Modelling of a Nonlinear Multivariable Boiler Plant using Hammerstein Model: A
Nonparametric Approach
S.Z. Rizvi and H.N. Al-Duwaish (Saudi Arabia)

675-037
Adaptive Predictive Control of Three - Tank - System
M. Kubalčík and V. Bobál (Czech Republic)

675-039
Energy Optimal Control of a Cooling System - Limits of Performance
F. Kitanoski and A. Hofer (Austria)

675-046
Multiple Model Predictive Control of the Fed-Batch Reactor
J. Novak, P. Chalupa, and V. Bobal (Czech Republic)

675-066
A Model based Approach to Design of Passive and Active Heave Compensation of Crown
Mounted Drilling Equipment
M. Ottestad, K. Olav Haland, and M.R. Hansen (Norway)

675-032
Verification of Fluid Structure Interaction Model of Pulsating Flow in a Silicone Pipe
N. Veshkina, D. Obidowski, J. Świniarski, and I. Zbiciński (Poland)

675-113
Control of Heat Exchanger using Predictive Approach
P. Chalupa, J. Novák, and V. Bobál (Czech Republic)

Mechanical and Electrical Applications

675-014
Residual Swing Suppression of Suspended Loads using Parametric Acceleration Profiles
G. Incerti (Italy)

675-019
Developing a Pen for Tremor Patients
A. Kumagai, Y. Suh, T. Tracy, K. Naritomi, and K. Pierson (USA)
Proceedings of the 29th IASTED International Conference
Modelling, Identification and Control (MIC 2010)
February 15 - 17, 2010 Innsbruck, Austria

RESIDUAL SWING SUPPRESSION OF SUSPENDED LOADS

USING PARAMETRIC ACCELERATION PROFILES
Giovanni Incerti
Department of Mechanical and Industrial Engineering
University of Brescia
Via Branze, 38
Brescia, Italy
email: [email protected]

ABSTRACT
The paper presents a motion planning method for swing
suppression of suspended loads transported by bridge or
gantry cranes. Acting on the reference motion command
of the trolley and using a parametric acceleration profile, it
is possible to avoid the generation of residual oscillations,
without changing the motion time and the total displace-
ment which must be covered by the trolley. This technique
can be easily implemented on an actual crane, provided
that a servomotor with position and/or velocity feedback
is available for the motion control of the trolley.

KEY WORDS
Crane dynamics, Motion planning, Residual oscillations

1 Introduction Figure 1. An example of a bridge crane.

It is common knowledge that bridge and gantry cranes are

employed in many industrial plants or warehouses to lift
and transport goods or machine parts (Fig.1). The necessity such as the rope length, the traveling time and distance of
to speed up the materials handling requires higher veloci- the trolley, the damping factor of the system, etc.
ties and, at the same time, an efficient control of the load Since over twenty years extensive research has been
swinging. carried out for the reduction of load swing of cranes [1-10]:
These oscillations arise especially during the accel- the major contributions concern the developments of math-
eration and deceleration phases of the trolley motion; in- ematical models (even for the dynamic analysis of the load
deed, if the starting and braking intervals are too short, in the three-dimensional space), the theoretical and experi-
the high acceleration levels generate very intensive inertia mental study of open-loop and closed-loop control tech-
forces, which can determine large amplitude oscillations of niques, the robustness of the methods under parameters un-
the loads and consequently a dangerous situation for per- certainty, etc.
sons which operate in the working area of the crane. More- The present paper aims at giving a theoretical con-
over the damping effects, due to air resistance acting on a tribution to the above mentioned subject and proposes a
suspended load, are very low and therefore it is difficult to solution based on the synthesis of optimized motion pro-
quickly eliminate an oscillatory motion produced, for ex- files, which must be used as input functions for the motion
ample, by a rapid deceleration of the trolley. controller of the trolley. The procedure requires a simula-
In many practical cases the swing control of the load tion model in order to foresee the dynamic behaviour of the
is obtained by a skilled operator, which manually operates load, when different acceleration profiles are employed to
the crane in such a way as to avoid or to reduce at minimum drive the trolley.
the risk conditions. However, the use of servo-controlled The problem here presented is developed and solved
motors, even on cranes and hoisting machines, allows the as a classic constrained optimization problem, where it is
crane designers to reduce the load oscillations and to im- necessary to minimize a suitable target function, related to
prove the machine dynamics through an accurate selection the vibration amplitude of the suspended load.
of the motion profiles. Of course, this motion planning re- The following paragraphs given mathematical details
quires the knowledge of a series of numerical parameters, about the dynamic model of the mechanical system and

675-014 147
 trolley load; the corresponding velocity components can be subse-
quently calculated by differentiation with respect to time.
c Using the symbols and the sign conventions indicated in
Fig. 2, we have:
 
xP = x + L sin ϑ ẋP = ẋ + Lϑ̇ cos ϑ
=⇒ (2)
L yP = L cos ϑ ẏP = −Lϑ̇ sin ϑ
ϑ yP
Taking into account Eqs. (2), the kinetic energy assumes
x the following expression:
1 1
P T = m(ẋ2P + ẏP2 ) = m(ẋ2 + L2 ϑ̇2 + 2Lẋϑ̇ cos ϑ) (3)
2 2
mg where m is the load mass.
xP The potential energy, due to gravity field, is:

V = mg(L − yP ) = mgL(1 − cos ϑ) (4)

where g is the acceleration due to gravity; the null value of

Figure 2. Dynamic model of a bridge crane in the xy plane. V has been conventionally fixed for ϑ = 0.
The Rayleigh function is defined by the following
equation:
1
D = cϑ̇2 (5)
about the functions used as reference for the trolley motion; 2
in the final part of the paper the results obtained through Substituting Eqs. (3), (4) and (5) into Eq. (1) and dividing
several computer simulations and optimizations are pre- all terms by mL2 , we obtain the motion equation of the
sented and discussed. load:
c g ẍ(t)
ϑ̈ + ϑ̇ + sin ϑ = − cos ϑ (6)
mL2 L L
2 Dynamic model of a bridge crane Now, if the small oscillations hypothesis is assumed,
that is sin ϑ ' ϑ and cos ϑ ' 1, the differential equation
The simplified model of a bridge crane with a swinging (6) becomes linear and it can be solved without particular
load is schematically represented in Fig. 2. The dynamic difficulties:
analysis is carried out in the two-dimensional space; we as-
c g ẍ(t)
sume that the rope is perfectly rigid, with negligible mass ϑ̈ + ϑ̇ + ϑ = − (7)
and constant length L. Moreover we suppose that the trol- mL2 L L
ley is able to correctly execute the required motion profile At this point, Eq. (7) can be rewritten in the standard form:
x(t); this hypothesis is usually satisfied, provided that a
ẍ(t)
servomotor equipped with a position and/or a velocity feed- ϑ̈ + 2ξωn ϑ̇ + ωn2 ϑ = − (8)
back loop is used to drive the trolley. L
The energy dissipations, that take place during the where ωn is the natural frequency of the system and ξ its
load oscillations, are modeled by a viscous damper with damping ratio; these constants are defined by the following
damping constant c; the numerical value of this parameter expressions:
can be determined through a free oscillation experiment, by r
measuring the amplitude reduction at every cycle. g c
ωn = ξ= √
The motion equation of the load can be easily deduced L 2mL gL
through a Lagrangian approach. If ϑ indicates the angular If the trolley displacement ẍ(t) is known, the solution of
displacement of the rope from the vertical position, we can Eq. (8) can be calculated through the convolution integral;
write:   for null initial conditions (ϑ(0) = 0, ϑ̇(0) = 0) and consid-
d ∂T ∂T ∂D ∂V
− + + =0 (1) ering and under-damped system (ξ < 1) (this hypothesis is
dt ∂ ϑ̇ ∂ϑ ∂ ϑ̇ ∂ϑ always satisfied for a swinging load) we have:
where the symbols T and V indicate respectively the ki- Z t
1
netic and the potential energy of the load, whereas the sym- ϑ(t) = − F(t, τ ) dτ (9)
bol D indicates the Rayleigh function, which is introduced ωd L 0
in order to take into account the energy dissipations due to p
where ωd = ωn 1 − ξ 2 is the damped natural frequency
air resistance and frictional effects. of the system and F(t, τ ) is defined as:
To calculate the kinetic energy, it is at first neces-
sary to determine the cartesian coordinates (xP , yP ) of the F(t, τ ) = ẍ(τ )e−ξωn (t−τ ) sin[ωd (t − τ )] (10)

148
Through an opportune selection of the function x(t), it is where h is the total traveling distance that must be covered
possible to suppress or reduce the load oscillations without by the trolley. If the maximum positive and negative accel-
augmenting the motion times. The next paragraph gives erations are different, two distinct acceleration coefficients,
some mathematical details about the class of parametric respectively indicated by the symbols ca+ and ca− , are de-
functions used to achieve this result. fined. These characteristic parameters can be expressed as
function of the coefficients δi (the detailed analytical ex-
pression can be found in [12] for acceleration profiles with
3 Mathematical definition of the motion sinusoidal connections); therefore using compact mathe-
command matical expressions, we can write:

To define the motion of the trolley, it is convenient to adopt cv = p(δ) ca+ = q(δ) ca− = r(δ) (11)
a class of functions whose profile can be easily modified
by changing some numerical parameters, named shape co- where δ = {δ1 δ2 . . . δ7 }T is a vector containing the shape
efficients, without altering the continuity condition on the coefficients and p, q ed r are three functions of the seven
displacement, velocity and acceleration at the initial and fi- scalar variables δi (i = 1, 2, . . . , 7).
nal instants. Clearly it is always possible to reduce the number
of parameters that define the acceleration profile, provided
For example, if we use the so called trapezoidal mo-
that suitable mathematical relationships are defined among
tion laws, the acceleration profiles are defined by seven nu-
the coefficients δi . For example, Fig. 4 shows two simpli-
merical values δi = ti /tf (i = 1, . . . , 7), which repre-
fied acceleration profiles, only dependent by two parame-
sent fractions of the total motion time tf . These functions,
ters (indicated with the symbols λ and µ) and without the
commonly used to design cam mechanisms, [11], can be
central interval (δ4 = 0). For the profile in Fig. 4a the vec-
effectively employed to define the motion command of the
tor δ is defined as follows:
trolley as well. Fig. 3a,b shows two examples of trape-
zoidal acceleration profiles, having respectively linear and δ = {λ (0.5 − 2λ) λ 0 µ (0.5 − 2µ) µ}T
sinusoidal connections between the constant acceleration
&x&
intervals. δ 1 δ 2δ δ
3 δ 4 δ5 6 δ 7 where 0 ≤ λ, µ ≤ 0.25; for the profile in Fig. 4b we have:

&x& δ = {λ 0 (0.5 − λ) 0 (0.5 − µ) 0 µ}T

δ1 δ2 δ3 δ4 δ5 δ6 δ7

where 0 ≤ λ, µ ≤ 0.5. Note that for λ = µ = 0.25, both

functions generate the same acceleration profile.
t /tf &x&

&x&
0.5 − 2μ
t /tf μ μ

λ 0.5 − 2λ λ μ 0.5 − 2μ μ t /tf

a)
&x& λ 0.5 − 2λ λ t /tf
δ1 δ2 δ3 δ4 δ5 δ6 δ7

&x&
δ1 δ2 δ3 δ4 δ5 δ6 δ7
a)
&x&
t /tf
&x&
0.5 − μ μ
b)
t /tf λ 0.5 − λ 0.5 − μ μ t /tf
Figure 3. Trapezoidal acceleration profiles: a) acceleration
profile with linear connections; b) acceleration profile with λ 0.5 − λ t /tf
sinusoidal connections. b)

Figure 4. Acceleration profiles with sinusoidal connections

defined by two parameters λ e µ.
To compare different motion commands it is useful
to calculate two non-dimensional parameters, respectively
named velocity coefficient cv and acceleration coefficient
ca , and defined by the following relationships: 4 Swing suppression
ẋmax ẍmax This section presents some basic ideas that can be exploited
cv = ca =
h/tf h/t2f to suppress or reduce the oscillations of the suspended load.

149
Particular attention will be paid to the so called overshoot- Using the the general formula for differentiation under the
ing effect, that is the residual oscillation that take place integral sign1 and remembering that F(t, τ )|τ =t = 0 (see
when the trolley motion ends and the load is ready to be put Eq. (10)), we have:
down. As it has been already outlined, the swing suppres- Z t
sion will be achieved through a suitable modification of the 1
ϑ̇(t) = − G(t, τ ) dτ
motion command, without reducing the average velocity of ωd L 0
the trolley, that is without changing the trolley stroke h and
the motion time tf . Clearly it is important to verify if the where the function G(t, τ ) is :
new motion command, besides suppressing the overshoot-
∂F(t, τ )
ing effect, is able to guarantee low amplitude oscillations G(t, τ ) = = ωn ẍ(τ )e−ξωn (t−τ ) cos[ωd (t−τ )+ε]
∂t
of the load during the motion phase of the trolley as well.
The residual oscillation is possible because some me- with ε being a phase angle given by:
chanical energy (kinetic and potential) is still present inside p
the system at the final time instant t = tf ; therefore, if this tan ε = ξ/ 1 − ξ 2
energy is totally eliminated or minimized through an accu-
rate planning of the trolley motion, the residual oscillation Hence, at the final time instant, tf we obtain:
of the load will disappears or it will have a small amplitude.
G
Starting from this simple consideration, we can af- ϑ̇f = − (16)
firm that the mechanical energy corresponding to the final ωd L
instant plays the role of a target function, whose value must where the constant G is defined as:
be minimized by means of suitable optimization techniques Z tf
(see [13] for details).
G= G(tf , τ ) dτ
For the system in Fig. 2 the total mechanical energy 0
is given by the following expression, which can be derived
from Eqs. (3) and (4): Now, taking into account the expressions (15) and (16),
and executing the necessary simplifications, Eq. (14) can
1
Etot = T + V = m(ẋ2 + L2 ϑ̇2 + 2Lẋϑ̇ cos ϑ) be rewritten in the form:
2
+mgL(1 − cos ϑ)
"  2 #
1 m G
(12) Etot,f = F2 + = γΨ
Since the trolley velocity ẋ is null for t = tf , we obtain 2 (1 − ξ 2 ) ωn
from Eq. (12):
where
1
Etot,f = mL2 ϑ̇2f + mgL(1 − cos ϑf ) (13) 1 m

G
2
2 γ= Ψ = F2 +
where the subscript f indicates the values of the variables 2 (1 − ξ 2 ) ωn
at the final instant t = tf . If we introduce the approxima-
We observe that γ is a constant depending on the load
tion cos ϑf ' 1 − ϑ2f /2 (valid for small amplitudes), and
mass m and on the assumed damping factor ξ of the sys-
express the acceleration g as function of the natural fre-
tem, whereas Ψ depends on the definite integrals F and
quency (g = ωn2 L), Eq. (13) can be rewritten in the form:
G, whose values change when the acceleration law of the
1 trolley ẍ(t) is modified. If we use the trapezoidal func-
Etot,f = mL2 (ϑ̇2f + ωn2 ϑ2f ) (14)
2 tion in Fig. 3, Ψ is a function of the seven variables δi
The value of ϑf on the right-hand side of Eq. (14) can be (i = 1, . . . , 7), used to define the acceleration profile;
calculated from Eq. (9), by imposing t = tf ; now, if we therefore we can write:
define the integral: Etot,f
= Ψ(δ1 , δ2 , . . . , δ7 ) = Ψ(δ)
Z tf γ
F = F(tf , τ ) dτ
1 For a two variable function f (t, τ ), the general formula for differen-
0
tiation under the integral sign is:
the final value ϑf is expressed by: Z b(t)
d
F f (t, τ )dτ =
ϑf = − (15) dt a(t)
ωd L Z b(t)
∂
= f (t, τ )dτ + f (t, b(t))b0 (t) − f (t, a(t))a0 (t)
a(t) ∂t
The final value of the angular velocity ϑ̇f , that appears
on the right-hand side of Eq. (14), can be obtained from If a(t) = 0 e b(t) = t the above-mentioned formula can be simplified as
follows:
Eq. (9), by differentiation with respect to time: Z t Z t
d ∂
 Z t  f (t, τ )dτ = f (t, τ )dτ + f (t, τ )|τ =t
1 d dt 0 0 ∂t
ϑ̇(t) = − F(t, τ ) dτ
ωd L dt 0

150
Now the optimization problem can be formulated as fol- Fig. 4a; using the profile in Fig. 4b a minimum has been
lows: minimize the multi-variable function Ψ, considering found for λ = 0.373 and µ = 0.276.
the following constraints: Fig. 6 shows the effects of the optimization proce-
dure; by comparing the diagrams (d) and (f) with the dia-
δi ≥ 0 i = 1, . . . , 7 gram (b), we can observe that the residual vibration dis-
appears when the optimized commands are used: more-
in order to avoid that the minimization algorithm can assign over the same plots put in evidence that the difference
negative values, clearly not acceptable, to the shape coeffi- ∆ϑ = ϑmax − ϑmin during the motion time (0 ≤ t ≤ tf )
cients. Moreover, if we consider that these coefficients are does not suffer appreciable variations when different mo-
defined as ratio between the duration of the partial intervals tion command are employed.
ti and the total motion interval tf (see Fig. 3) it is needed
to assume the following condition as further constraint:
7
6 Conclusion
X
δi = 1
i=1
The computer simulations here presented show that the
proposed technique seems able to reduce the oscillations
Optionally we can add other inequalities, so that the max- of suspended loads, without reducing the motion time of
imum allowed values for the coefficients cv , ca+ and ca− , the trolley and its total stroke.
defined in section 3, are not exceeded. In this case it is The proposed approach can be implemented without
necessary to consider the functions p(δ), q(δ) ed r(δ) in- difficulties on an actual crane, since it requires only the use
dicated in Eq. (11) and impose the conditions: of a controlled servomotor to drive the trolley. The motion
planning is carried out off-line, through a simple mathemat-
p(δ) ≤ c∗v q(δ) ≤ c∗a+ r(δ) ≤ c∗a− ical model of the swinging load and a suitable optimization
algorithm, which modifies the trolley motion the until the
where the starred symbols indicate the limit values of the best performance (that is the suppression of the residual os-
corresponding coefficients. Obviously the optimization cillation) is obtained.
procedure can be also applied to motion commands defined The optimized motion profile can be successively
by a number of parameters less than seven, as in the case transferred into the controller memory, for example as data
of the profiles in Fig.4. file containing the trolley positions at regular time instants.
The modification of the reference motion command is
5 Numerical results obtained through a set of coefficients, whose values deter-
mine univocally the shape of the displacement, velocity and
This section shows some numerical results obtained acceleration functions. The continuity conditions of these
through the optimization technique described in the pre- functions at the boundaries of the motion interval are al-
vious paragraphs. The calculations have been carried out ways guaranteed, independently from the numerical values
for a load suspended from a massless rope having a length assigned to the shape coefficients.
L = 6 m; the trolley total displacement and the motion time It is also possible to express the maximum values of
are respectively h = 12 m and tf = 12 s. To take into ac- velocity and acceleration as function of these parameters
count the energy dissipations a damping ratio ξ = 2% has and to verify immediately if the calculated motion com-
been assumed. Using the above indicated
p data, the natural mand is compatible with the velocity and torque limits of
period of the oscillations Tn = 2π L/g is equal to 4.915 the motor installed on the trolley.
seconds, whereas the ratio between the natural period and In the future, an implementation of this technique on
the motion time is Tn /tf = 0.41. a small scale prototype of crane is scheduled; the experi-
The optimization procedure has been implemented mental results deriving from these test will allow us to val-
for both acceleration functions depicted in Fig. 4a and idate the theoretical results here presented.
Fig. 4b. In accordance with the criteria exposed in the pre-
vious section, the Ψ function (that is the final mechanical
energy of the swinging load divided by the coefficient γ) References
has been selected as objective to be minimized and it has
been calculated and plotted as function of the two variables [1] Singhose, W., Porter, L., Kenison, M., Kriikku, E., Ef-
λ and µ, which define the acceleration commands. (see fects of hoisting on the input shaping control of gantry
Fig. 5). cranes, Control Engineering Practice, vol. 8, 1159-
By means of a dedicated routine that employs the co- 1165 (2000).
niugate gradient algorithm [13], the values of the variables
corresponding to the minimum have been calculated; start- [2] Cartmell, M.P., On the Need for Control of Nonlinear
ing from λ = 0.25 and µ = 0.25 we have found a minimum Oscillations in Machine Systems, Meccanica, vol. 38,
for λ = 0.114 and µ = 0.134 with the command profile in 185-212 (2003).

151
 [3] Yi, J., Yubazaki, N., Hirota, K., Anti-swing and posi- [8] Chang, C., The switching algorithm for the control of
tioning control of overhead traveling crane, Informa- overhead crane, Neural Computing & Applications,
tion Sciences, vol. 155, 19-42 (2003). vol. 15, 350-358 (2006).

[4] Lee, H.H., A New Motion.Planning Scheme for Over- [9] Lee, H.H., Liang, Y., Segura, D., A New Sliding-Mode
head Cranes with High-Speed Hoisting, J. of Dy- Antiswing Trajectory Control for Overhead Cranes
namic Systems, Measurement and Control, vol. 126, with High-Speed Load Hoisting, J. of Dynamic Sys-
359-364 (2004). tems, Measurement and Control, vol. 128, 842-845
(2006).
[5] Gallina, P., Trevisani, A., Synthesis and Experimental [10] Kapucu, S., Yıldırım, N., Yavuz, H., Bayseç, S., Sup-
Validation of a Delayed Reference Controller for Ac- pression of residual vibration of a translating swing-
tive Vibration Suppression in Mechanical Systems, ing load by a flexible manipulator, Mechatronics,
J. of Applied Mechanics, vol. 72, 623-627 (2005). vol. 18, 121-128 (2008).

[6] Liu, D., Yi, J., Zhao, D., Wang, W., Adaptive sliding [11] Norton R.L., Cam design and Manufacturing Hand-
mode fuzzy control for a two-dimensional overhead book, Industrial Press, New York, 2002.
crane, Mechatronics, vol. 15, 505-522 (2005). [12] Magnani, P.L., Ruggieri, G., Meccanismi per mac-
chine automatiche (in Italian), UTET, Torino (1986).
[7] Chang, C., Hsu, S., Chiang, K., A practical fuzzy con-
trollers scheme of overhead crane, J. of Control The- [13] Rao, S.S., Engineering Optimization - Theory and
ory and Applications, vol. 3, 266-270 (2005). Practice, John Wiley & Sons, Inc., New York (1996).

a) b)

c) d)

Figure 5. Graphical representations of the function Ψ in the domain of the variables λ and µ: a, b) surface and contour plot
for acceleration command in Fig. 4a (0 ≤ λ, µ ≤ 0.25); c, d) surface and contour plot for acceleration command in Fig. 4b
(0 ≤ λ, µ ≤ 0.5).

152
 0.6 6
T T
0.5 5

0.4 4

0.3 3

0.2 2
trolley accel. [m/s2]

load oscill. [deg]

0.1 1

0 2 4 6 8 10 12 14 16 18 20 22 24 0 2 4 6 8 10 12 14 16 18 20 22 24
0.1 1

0.2 2

0.3 3

0.4 4

0.5 5

0.6 6

time [s] a) time [s] b)

0.6 6
T T
0.5 5

0.4 4

0.3 3

0.2 2
trolley accel. [m/s2]

load oscill. [deg]

0.1 1

0 2 4 6 8 10 12 14 16 18 20 22 24 0 2 4 6 8 10 12 14 16 18 20 22 24
0.1 1

0.2 2

0.3 3

0.4 4

0.5 5

0.6 6

time [s] c) time [s] d)

0.6 6
T T
0.5 5

0.4 4

0.3 3

0.2 2
trolley accel. [m/s2]

load oscill. [deg]

0.1 1

0 2 4 6 8 10 12 14 16 18 20 22 24 0 2 4 6 8 10 12 14 16 18 20 22 24
0.1 1

0.2 2

0.3 3

0.4 4

0.5 5

0.6 6

time [s] e) time [s] f)

Figure 6. Dynamic simulations with non optimal and optimal motion commands:
• Non-optimized trolley acceleration (a) and consequent load oscillation (b) calculated for λ = 0.25 and µ = 0.25.
• Optimized trolley acceleration (c) and consequent load oscillation (d) calculated with the profile in Fig. 4a for
λ = 0.114 and µ = 0.134.
• Optimized trolley acceleration (e) and consequent load oscillation (f) calculated with the profile in Fig. 4b for
λ = 0.373 and µ = 0.276.

153

View publication stats