Professional Documents
Culture Documents
Modeling and Energy-Based Sway Reduction Control For Tower Crane Systems With Double-Pendulum and Spherical-Pendulum Effects
Modeling and Energy-Based Sway Reduction Control For Tower Crane Systems With Double-Pendulum and Spherical-Pendulum Effects
Modeling and Energy-Based Sway Reduction Control For Tower Crane Systems With Double-Pendulum and Spherical-Pendulum Effects
spherical-pendulum effects
Abstract
As typical underactuated systems, tower crane systems present complicated nonlinear dynamics. For simplicity, the pay-
load swing is traditionally modeled as a single-pendulum in existing works. Actually, when the hook mass is close to the
payload mass, or the size of the payload is large, a tower crane may exhibit double-pendulum effects. In addition, existing
control methods assume that the hook and the payload only swing in a plane. To tackle the aforementioned practical
problems, we establish the dynamical model of the tower cranes with double-pendulum and spherical-pendulum effects.
Then, on this basis, an energy-based controller is designed and analyzed using the established dynamic model. To further
obtain rapid hook and payload swing suppression and elimination, the swing part is introduced to the energy-based con-
troller. Lyapunov techniques and LaSalle’s invariance theorem are provided to demonstrate the asymptotic stability of
the closed-loop system and the convergence of the system states. Simulation results are illustrated to verify the correct-
ness and effectiveness of the designed controller.
Keywords
Modeling, spherical-pendulum effect, double-pendulum effect, energy-based control, convergence
Creative Commons CC BY: This article is distributed under the terms of the Creative Commons Attribution 4.0 License
(http://www.creativecommons.org/licenses/by/4.0/) which permits any use, reproduction and distribution of the work without
further permission provided the original work is attributed as specified on the SAGE and Open Access pages (https://us.sagepub.com/en-us/nam/
open-access-at-sage).
142 Measurement and Control 53(1-2)
dpm
vm = + v3pm
dt
2 3 2 3
x_ + l1 u_ 1 C1 C2 l1 u_ 2 S1 S2 + l2 u_ 3 C3 C4 l2 u_ 4 S3 S4 x + l1 S1 C2 + l2 S3 C4
6 7 6 7
=6 _ _ 7 + 0 0 f_ T 36 l1 S2 + l2 S4 7
4 l1 u2 C2 + l2 u4 C4 5 4 5
_ _ _ _ l1 C1 C2 + l2 C3 C4 ð7Þ
l1 u1 S1 C2 l1 u2 C1 S2 l2 u3 S3 C4 l2 u4 C3 S4
2 3
x_ + l1 u_ 1 C1 C2 l1 u_ 2 S1 S2 + l2 u_ 3 C3 C4 l2 u_ 4 S3 S4 f_ ðl1 S2 + l2 S4 Þ
6 7
=6 _ _ _
4 l1 u2 C2 + l2 u4 C4 + fðx + l1 S1 C2 + l2 S3 C4 Þ
7
5
_ _ _ _
l1 u1 S1 C2 l1 u2 C1 S2 l2 u3 S3 C4 l2 u4 C3 S4
∂E
_ 1 C2
= x2 f_ + l21 f_ S22 + S21 C22 + l22 f_ S24 + S23 C24 + 2l1 l2 f_ ðS2 S4 + S1 C2 S3 C4 Þ + 2l1 xfS
∂f_
_ 3 C4 + xl1 u_ 2 C2 + 2l2 xf_ u_ 4 C4 l1 xS ð12Þ
+ 2l2 xfS _ 2 l2 xS _ 4 l21 u_ 1 C1 S2 C2 l1 l2 u_ 1 C1 C2 S4 + l21 u_ 2 S1
+ l1 l2 u_ 2 ðS1 S2 S4 + C2 S3 C4 Þ l1 l2 u_ 3 S2 C3 C4 l2 u_ 3 C3 S4 C4 + l1 l2 u_ 4 ðS2 S3 S4 + S1 C2 C4 Þ + l2 u_ 4 S3
2 2
144 Measurement and Control 53(1-2)
2
d ∂E x + l21 S22 + l21 S21 C22 + l22 S24 + l22 S23 C24 + 2l1 l2 S2 S4 €
= m2 f
dt ∂f_ + 2l1 l2 S1 C2 S3 C4 + 2l1 xS1 C2 + 2l2 xS3 C4 + 2l2 xu_ 4 C4
+ 2m2 l1 C1 C2 u_ 1 l1 S1 S2 u_ 2 + l2 C3 C4 u_ 3 l2 S3 S4 u_ 4 xf_ m2 ðl1 S2 + l2 S4 Þ€
x
!
2l 2 2
C S2 C2
_ 2 + 2l2 S1 C1 C2 u_ 1
u
+ 2m2 x + l1 S1 C2 + l2 S3 C4 + l2 C4 u_ 4 x_ f_ + m2 1 1 1 2
f_
+ 2l22 C23 S4 C4 u_ 4 + 2l22 S3 C3 C24 u_ 3
l1 S1 + l2 S1 S2 S4
m2 l1 C1 C2 ðl2 S4 + l1 S2 Þ€ u1 m2 l2 C3 C4 ðl1 S2 + l2 S4 Þ€u3 + m2 l1 €u2 ð13Þ
+ l2 C2 S3 C4 + xC2
+ m2 l2 l2 S3 + l1 S2 S3 S4 + l1 S1 C2 C4 + 2xfC _ 4 €u4 + m2 l1 S1 C2 ðl1 S2 + l2 S4 Þu_ 2
1
+ m2 l2 S3 C4 ðl1 S2 + l2 S4 Þu_ 23 + m2 l2 l1 S2 S3 C4 l1 S1 C2 S4 2xfS _ 4 u_ 2
4
+ 2m2 l1 C1 l1 C22 + l2 S2 S4 u_ 1 u_ 2 m2 l2 x_ u_ 4 C4 + 2m2 l2 C3 S4 ðl1 S2 l2 S4 Þu_ 3 u_ 4
+ 2m2 l1 l2 f_ u_ 1 C1 C2 S3 C4 + 2m2 l1 l2 f_ u_ 2 ðC2 S4 S1 S2 S3 C4 Þ + 2m2 l1 l2 f_ u_ 3 S1 C2 C3 C4
+ 2m2 l1 l2 f_ u_ 4 ðS2 C4 S1 C2 S3 S4 Þ
where M denotes the slew control torque and Mf represents the friction torque.
In a similar way, the following results are obtained as follows
€ + ðMt + m1 + m2 Þ€
ðm1 l1 S2 + m2 l1 S2 + m2 l2 S4 Þf x + ðm1 + m2 Þl1 €u1 C1 C2 ðm1 + m2 Þl1 €u2 S1 S2
+ m2 l2 €
u3 C3 C4 m2 l2 €
u4 S3 S4 ðm1 + m2 Þl1 u_ 2 S1 C2 ðm1 + m2 Þl1 u_ 2 S1 C2 m2 l2 u_ 2 S3 C4
1 2 3
2 ð16Þ
m2 l2 u_ 24 S3 C4 2ðm1 + m2 Þl1 u_ 1 u_ 2 C1 S2 ððMt + m1 + m2 Þx + ðm1 + m2 Þl1 S1 C2 + m2 l2 S3 C4 Þf_
2m2 l2 u_ 3 u_ 4 C3 S4 2ðm1 + m2 Þl1 f_ u_ 2 C2 2m2 l2 f_ u_ 4 C4 = F Ff
l1 ððm1 + m2 ÞðxC2 + l1 S1 Þ + m2 l2 ðS1 S2 S4 + C2 S3 C4 ÞÞf € ðm1 + m2 Þl1 x€S1 S2 + ðm1 + m2 Þl2 €u2
1
€3 S13 S2 C4 + m2 l1 l2 €
m2 l1 l2 u u4 ðC13 S2 S4 + C2 C4 Þ + 2ðm1 + m2 Þl1 x_ fC _ 2
2
+ l1 f_ ðm1 + m2 Þ xS1 S2 l1 C21 S2 C2 m2 l2 ðC2 S4 S1 S2 S3 C4 Þ + ðm1 + m2 Þl21 u_ 21 S2 C2
ð18Þ
+ m2 l1 l2 u_ 2 C13 S2 C4 + m2 l1 l2 u_ 2 ðC13 S2 C4 C2 S4 Þ m2 l1 l2 u_ 1 u_ 3 C13 S2 C4 + 2m2 l1 l2 u_ 3 u_ 4 S13 S2 S4
3 4
+ 2ðm1 + m2 Þl21 f_ u_ 1 C1 C22 + 2m2 l1 l2 f_ u_ 3 C2 C3 C4 + 2m2 l1 l2 f_ u_ 4 ðS1 S2 C4 C2 S3 S4 Þ + ðm1 + m2 Þgl1 C1 S2
= df1 l2 u_ 2
1
€ ðxC4 + l1 S2 S3 S4 + l1 S1 C2 C4 + l2 S3 Þ m2 l2 x€S3 S4 + m2 l1 l2 €
m2 l2 f u1 S13 C2 S4 + m2 l1 l2 € u2 ðC13 S2 S4 + C2 C4 Þ
2
+ m2 l22 €u4 + m2 l2 f_ xS3 S4 l1 S2 C4 + l1 S1 C2 S3 S4 l2 C23 S4 C4 + m2 l1 l2 u_ 21 C13 C2 S4
ð20Þ
_ 4 2m2 l1 l2 u_ 1 u_ 2 S13 S2 S4
+ m2 l1 l2 u_ 2 ðC13 C2 S4 S2 C4 Þ + m2 l2 u_ 2 S4 C4 + 2m2 l2 x_ fC
2 2 3
+ 2m2 l1 l2 f_ u_ 1 C1 C2 C4 + 2m2 l1 l2 f_ u_ 2 ðC2 S3 S4 S1 S2 C4 Þ + 2m2 l22 f_ u_ 3 C3 S24 + m2 gl2 C3 S4 = df2 l22 u_ 4
with frof , krf , frox , and krx are friction-related where q denotes the state vector and MðqÞ stands for
parameters. the inertia matrix, which are defined as follows
In practical applications, the payload is always
beneath the jib. Therefore, the following assumption is q = ½f x u1 u2 u3 u4 T
reasonable. 2 3
M11 M12 M13 M14 M15 M16
6 M12 Mt + m1 + m2 M23 M24 M25 M26 7
6 7
Assumption 1: The payload swing angles satisfy22–24 6 M13 M23 M33 0 M35 M36 7
Mð qÞ = 6
6 M14
7
6 M24 0 M44 M45 M46 7 7
4 M15 M25 M35 M45 m2 l22 C24 0 5
p p M16 M26 M36 M46 0 m2 l22
\ u1 , u2 , u3 , u4 \ ð23Þ
2 2
with
0 1
ðMt + m1 + m2 Þx2 + m2 l21 + m1 l21 S22 + S21 C22 + J
B C
M11 = @ + m2 l22 S24 + S23 C24 + 2m2 l1 l2 ðS2 S4 + S1 C2 S3 C4 Þ A, M12 = ðm1 l1 S2 + m2 l1 S2 + m2 l2 S4 Þ,
+ 2ðm1 + m2 Þl1 xS1 C2 + 2m2 l2 xS3 C4 + 2m2 l2 xu_ 4 C4
M13 = l1 C1 C2 ðm2 l2 S4 + ðm1 + m2 Þl1 S2 Þ, M14 = l1 ððm1 + m2 Þðl1 S1 + xC2 Þ + m2 l2 ðS1 S2 S4 + C2 S3 C4 ÞÞ,
_ 4 ,
M15 = m2 l2 C3 C4 ðl1 S2 + l2 S4 Þ, M16 = m2 l2 l2 S3 + l1 ðS2 S3 S4 + S1 C2 C4 Þ + 2xfC
M23 = ðm1 + m2 Þl1 C1 C2 , M24 = ðm1 + m2 Þl1 S1 S2 , M25 = m2 l2 C3 C4 , M26 = m2 l2 S3 S4 ,
M33 = ðm1 + m2 Þl21 C22 , M35 = m2 l1 l2 C13 C2 C4 , M36 = m2 l1 l2 S13 C2 S4 , M44 = ðm1 + m2 Þl21 ,
M45 = m2 l1 l2 S13 S2 C4 , M46 = m2 l1 l2 ðC13 S2 S4 + C2 C4 Þ
146 Measurement and Control 53(1-2)
Taking the time derivative of equation (24) and sub- M = kpf ef kdf f_ khf u_ 21 + u_ 22 + u_ 23 + u_ 24 f_ + Mf
stituting equations (15)–(20) into the resulting equation ð30Þ
yields
F = kpx ex kdx x_ khx u_ 21 + u_ 22 + u_ 23 + u_ 24 x_ + Ff
1 _ ðqÞq_ + ðm1 + m2 Þgl1 S1 C2 u_ 1 ð31Þ
V_ = q_ T MðqÞ€
q + q_ T M
2
+ ðm1 + m2 Þgl1 C1 S2 u_ 2
+ m2 gl2 S3 C4 u_ 3 + m2 gl2 C3 S4 u_ 4 Stability analysis
T
q + Cðq, q_ Þq_ Þ + ðm1 + m2 Þgl1 S1 C2 u_ 1
= q_ ðMðqÞ€ In this section, the stability of the closed-loop system
+ ðm1 + m2 Þgl1 C1 S2 u_ 2 and the convergence of the states are provided. The fol-
lowing theorem summarizes the main results of this
+ m2 gl2 S3 C4 u_ 3 + m2 gl2 C3 S4 u_ 4 paper.
= q_ T ðU GðqÞÞ + ðm1 + m2 Þgl1 S1 C2 u_ 1
Theorem 1: The proposed method in equations (30)
+ ðm1 + m2 Þgl1 C1 S2 u_ 2
and (31) can drive the jib to desired angular position
+ m2 gl2 S3 C4 u_ 3 + m2 gl2 C3 S4 u_ 4 and the trolley to the desired position, respectively,
= M Mf f_ + F Ff x_ df1 l2 C2 u_ 2 1 2 1
and at the same time, it can suppress and eliminate
the hook and the payload swing, in the following
df1 l21 u_ 22 df2 l22 C24 u_ 23 df2 l22 u_ 24
sense
ð25Þ T
lim f x f_ x_ = ½fd xd 0 0T ,
where Cðq, q_ Þ denotes the centripetal-Coriolis matrix, t!‘
U stands for the control vector, and GðqÞ represents the lim ½u1 u2 u3 u4 T = ½0 0 0 0T , ð32Þ
_ (q) t!‘
gravity vector, and the property jT ½(1=2)M T
_ = 0, 8j 2 R6 25,26 is used. The detail expres-
C(q, q)j lim u_ 1 u_ 2 u_ 3 u_ 4 = ½0 0 0 0T
t!‘
sions of U and GðqÞ are given as follows
2 3 2 3 Proof: The following Lyapunov candidate function
M 0 is introduced
6F 7 60 7
6 7 6 7
60 7 6 ðm1 + m2 Þgl1 S1 C2 7 1 1
6 7
U = 6 7 , G ð qÞ = 66 7 Vall ðtÞ = V + kpf e2f + kpx e2x ð33Þ
7 2 2
60 7 6 ðm1 + m2 Þgl1 C1 S2 7
40 5 4 m2 gl2 S3 C4 5
0 m2 gl2 C3 S4 Differentiating equation (33) with respect to time,
It follows from equation (25) that the tower crane and substituting the results of equations (25)–(27) into
system with M and F as the inputs, f_ and x_ as the out- the resulting equation, one has
puts, is passive and dissipative. 2
V_ all ðtÞ = kdf + khf u_ 21 + u_ 22 + u_ 23 + u_ 24 f_
According to the structure of equation (25), an
energy-based controller is proposed as kdx + khx u_ 21 + u_ 22 + u_ 23 + u_ 24 x_ 2
df1 l2 C2 u_ 2 df1 l2 u_ 2 df2 l2 C2 u_ 2 df2 l2 u_ 2
M = kpf ef kdf f_ + Mf ð26Þ 1 2 1 1 2 2 4 3 2 4
40 ð34Þ
F = kpx ex kdx x_ + Ff ð27Þ
It follows from equations (33) and (34) that the
where ef and ex stand for the error signals, the first closed-loop system is Lyapunov stable at the desired
terms are used to guarantee accurate positioning, which equilibrium point,27–29 and it is obtained that
will be discussed in the subsequent section. They are
defined as _ x,
Vall ðtÞ 2 L‘ ) f, _ u_ 1 , u_ 2 , u_ 3 , u_ 4 , ef , ex , f, x, M, F 2 L‘
ð35Þ
ef = f fd ð28Þ
27–29
Then, using LaSalle’s invariance theorem,
e x = x xd ð29Þ
Theorem 1 can be proven.
where fd and xd stand for the jib target angular posi-
tion and the trolley target position, respectively.
Simulation results and analysis
To obtain rapid hook/payload sway suppression
and elimination, the hook and the payload swing- In this section, some simulation results are provided to
related information is added to the designed controller, validate the correctness and effectiveness of the pro-
and by doing so, equations (26) and (27) are improved posed energy-based controller. To this end, two groups
as follows of numerical simulation tests are given. In the first
Zhang et al. 147
(a) (b)
Figure 2. Group 1: simulation results of the PD controller (a): the jib slew angle, the trolley displacement, the slew control torque,
and the translation control force; (b): the hook swing angles, and the payload swig angles.
(a) (b)
Figure 3. Group 1: simulation results of the proposed controller (a): the jib slew angle, the trolley displacement, the slew control
torque, and the translation control force; (b): the hook swing angles, and the payload swig angles.
group, the proposed controller is compared with the Table1. Control gains.
proportional-derivative (PD) controller. Then, the
robustness of the proposed controller with respect to Controllers Control gains
sudden changes of payload mass, cable length, and ini-
PD controller kpf = 10:5, kdf = 18, kpx = 12, kdx = 15
tial payload swings, and external disturbances is veri- Proposed kpf = 13, kdf = 25, khf = 10,
fied in the second group. controller kpx = 6:5, kdx = 12, khx = 10
The parameters of the tower crane system with
PD: proportional derivative.
double-pendulum and spherical-pendulum effects are
set as follows
The initial jib slew angle, trolley translation position,
Mt = 3:5 kg, m1 = 0:5 kg, l2 = 0:5 m, hook swing angles, and payload swing angles are set as
J = 6:8 kg m2 , g = 9:8 m=s2 , frof = 5:2, fð0Þ = 0 deg, xð0Þ = 0 m, respectively.
krf = 1, ef = 0:01, frox = 5:4,
Simulation group 1: In this group, to better show the
krx = 1:5, ex = 0:01, df1 = 0:3, df1 = 0:3
control performance of the proposed energy-based
control method with hook and payload sway reduc-
Unless otherwise mentioned, m2 , l1 , u1 ð0Þ, u2 ð0Þ,
tion, we compare it with the PD controller.
u3 ð0Þ, and u4 ð0Þ are set as m1 = 1 kg,l2 =1 m, u1 ð0Þ =
0 deg,u2 ð0Þ=0 deg,u3 ð0Þ= 0 deg,u4 ð0Þ =0 deg, respec-
tively. The target jib slew angle and the desired trolley By trial and error method, the control gains of the
translation position are set as fd = 45 deg,fd =1 m, proposed control method and the PD control method
respectively. are chosen as shown in Table 1.
148 Measurement and Control 53(1-2)
(a) (b)
Figure 4. Group 2: simulation results of the proposed controller of Case 1. (a): the jib slew angle, the trolley displacement, the
slew control torque, and the translation control force; (b): the hook swing angles, and the payload swig angles.
(a) (b)
Figure 5. Group 2: simulation results of the proposed controller of Case 2. (a): the jib slew angle, the trolley displacement, the
slew control torque, and the translation control force; (b): the hook swing angles, and the payload swig angles.
The simulation results are shown in Figures 2 and 3. Case 2: Sudden change of the cable length l1 . The
It is seen that with similar transportation time for posi- cable length is changed from 1 to 1.5 m at t = 2 s,
tioning (both within 5 s), the proposed energy-based while its nominal value is still kept the same as it in
control method with hook and payload sway reduction simulation group 1.
shows better swing suppression and elimination effects. Case 3: Initial hook/payload swings. The initial pay-
More precisely, the hook/payload swing angles are sup- load swing angles are set as u1 (0) = 5 deg,
pressed and eliminated within a smaller range by the u2 (0) = 4 deg, u3 (0) = 6 deg, and u4 (0) = 8 deg,
designed controller (u1 max = 3:1 deg, u2 max = 1:9 deg, respectively.
u3 max = 4 deg, u4 max = 1:9 deg, u1res = 0:01 deg, Case 4: External disturbances. To emulate external
u2res = 0:01 deg, u3res = 0:3 deg, u4res = 0:1 deg) than disturbances, such as winds, various external distur-
by the PD control method (u1 max = 9:5 deg, u2 max = bances are added to the hook/payload swing angles.
4:9 deg, u3 max = 10 deg, u4 max = 5 deg, u1res = 2:2 deg, More precisely, impulsive disturbances with an ampli-
u2res = 3:8 deg, u3res = 2:5 deg, u4res = 4:1 deg). ui max , tude of 5 deg are added to u1 between 2 and 3 s,
uires , i = 1, 2, 3, 4, represent the maximum swing angles impulsive disturbances with an amplitude of 5 deg are
and the residual swing angles. added to u2 between 7 and 8 s, impulsive disturbances
with an amplitude of 10 deg are added to u3 between
Simulation group 2: In this group, to further verify 9 and 10 s, and impulsive disturbances with an ampli-
the robustness of the designed controller, the follow- tude of 6 deg are added to u4 between 3 and 4 s.
ing four cases are taken into consideration.
Case 1: Sudden change of the payload mass m2 . The The simulation results for the above four cases are
payload mass m2 is changed from 0.5 to 1 kg at shown in Figures 4–7 in order. By comparing Figures 4
t = 3 s, while its nominal value is still kept the same and 5 with Figure 3, it is seen that the overall control
as it in simulation group 1. performance of the proposed control method, including
Zhang et al. 149
(a) (b)
Figure 6. Group 2: simulation results of the proposed controller of Case 3. (a): the jib slew angle, the trolley displacement, the
slew control torque, and the translation control force; (b): the hook swing angles, and the payload swig angles.
(a) (b)
Figure 7. Group 2: simulation results of the proposed controller of Case 4. (a): the jib slew angle, the trolley displacement, the
slew control torque, and the translation control force; (b): the hook swing angles, and the payload swig angles.
positioning and swing suppression and elimination, is experimental results will be provided to prove the prac-
not affected much by sudden change of the payload tical control performance of the designed controller.
mass and the cable length, implying that it is robust
with respect to system parameters. It is noted from Declaration of conflicting interests
Figures 6 and 7 that the initial payload swing angles
and the unexpected external disturbances are attenu- The author(s) declared no potential conflict of interests with
respect to the research, authorship and/or publication of this
ated and damped out by the suggested control method.
article.
It can be concluded from the above analysis that the
proposed controller shows strong robustness.
Funding
The author(s) disclosed receipt of the following financial sup-
Conclusion port for the research, authorship, and/or publication of this
In this paper, using Lagrange’s method, we model the article: This work was supported by the National Key R&D
tower cranes with double-pendulum and spherical- Program of China under Grant No. 2018YFB1305400, the
pendulum effects. On this basis, the energy-based con- Key Research and Development (Special Public-Funded
Projects) of Shandong Province under Grant No.
troller with hook and payload reduction is proposed.
2019GGX104058, the National Natural Science Foundation
We use Lyapunov techniques and LaSalle’s invariance for Young Scientists of China under Grant No. 61903155,
theorem to validate the stability. Numerical results are and the Natural Science Foundation of Shandong Province
provided to validate the superior control performance under Grant No. ZR2019QEE019.
of the designed energy-based sway reduction controller.
To the best of our knowledge, it is the first closed-loop
method for tower cranes with double-pendulum and ORCID iD
spherical-pendulum effects. In our future work, the Menghua Zhang https://orcid.org/0000-0001-8588-3612
150 Measurement and Control 53(1-2)