Professional Documents
Culture Documents
1, JANUARY 2003
Abstract—This paper deals with modeling and control of a crane In our previous study [6], we applied a decentralized control
mounted on a tower-like flexible structure. A fast transfer of the system consisting of two individual controllers for the up-and-
load causes swaying of the load rope and vibration of the flexible down direction and the rotational direction. The reasons why a
structure. Our object is to control both the sway and the vibration
by the inherent capability of the tower crane. We have already suc- decentralized control system was applied are the following.
ceeded in realizing a decentralized control consisting of two indi- • We assumed there is little coupling between the up-and-
vidual controllers. This paper describes the design of a centralized down direction and the rotational direction of the tower
control system considering coupling of between the up-and-down crane.
(boom luff) direction and the rotational direction. This is because
there is a possibility that the control worsened due to the decentral- • It is easy to identify models for controller design and to
ized control system because it did not take into account coupling adjust the controller.
both directions. It is, therefore, necessary to compare it with a cen- • In an experiment, the quantity of computation is small.
tralized control system and to examine the advantage of a decen- Using a decentralized compensator designed for each
tralized control system precisely. From the results of the compar-
ison of experiments and analysis, it is shown that the decentral- reduced-order model, simulations and experiments were carried
ized control system has almost the same performance and stability out. It was verified that the simultaneous control of the boom
as the centralized one and that it is effective for the control of a angle, the rotational angle, the vibration of the tower and the
tower crane. sway of the load was possible by the inherent capability of the
Index Terms—Decentralized control, flexible structure, jib-type tower crane.
crane, analysis, position control, robust control, vibration con- However, there was a possibility that the control was made
trol. worse by the decentralized control system especially in the vibra-
tion control of the tower and the load because coupling between
I. INTRODUCTION each direction occurs. It is consequently necessary to examine
the advantage of the decentralized control system precisely.
(13) (23)
(24)
(14)
The torque put into the root of the boom is
(25)
(15) where the angle between the boom-support rope and the boom
where is the equivalent mass and spring coeffi- is
cient of the tower. Those are computed by a modal analysis of
the finite-element model of the tower part. (26)
2) Introduction of Equivalent Forces Relative to the Bending
of the Tower: In the actual tower crane, the tower is bent by the and the distance between the top of the gantry and the top of the
momental force around axis. However, from Assumption boom is
(1.6) the center of mass of the tower top is in the plane of the
and axis. Therefore, if the momental force arises because of (27)
the movement of the boom and the load, it does not affect the
Using Lagrange’s equation
top of the tower. Consequently, we consider equivalent forces
relative to the bending of the tower [7]. Considering the strain
energy in a cantilever beam, if the tower is subject to a couple
acting at the top, the deflection is (28)
where
(16)
(17)
(18) and using the small angle approximation for the sway of the load
(20)
(31)
TAKAGI AND NISHIMURA: CONTROL OF A JIB-TYPE CRANE MOUNTED ON A FLEXIBLE STRUCTURE 35
(32)
(33)
(a)
The state equation of the full-order model is obtained
(34)
where
(35)
(36)
(37)
(38)
where
(39)
Fig. 3. Simulation and experimental results using PID control : 45 ! 55(deg), : 0 ! 45(deg).
resolved by the simulation response of the full-order model. Relative coordinates and are calculated using as
The state equation is represented in the linear parameter-varying follows:
system by (29) and (31) is as follows:
(42)
(43)
(40)
up-and-down direction and Fig. 2(b) shows the model for the
rotational direction. The following are assumptions for this
modeling instead of (1.3).
(3.1) The tower part is a one-degree-of-freedom system
which can move in a controlled direction and in addition to
(1.1)–(1.6).
(3.2) The load can sway only in a controlled direction.
The governing equation of this model is calculated by La-
grange’s equation. The state equation and the output equation
of the model for the up-and-down direction represented in the
linear parameter-varying system by (29) and (31) is as follows: Fig. 4. Generalized plant.
(44)
(45)
The state equation and the output equation of the model for the
rotational direction are
(46)
(47)
TABLE II
SPECIFICATIONS OF EXPERIMENTAL SETUP
TABLE III
REFERENCES EXPERIMENTS
IV. EXPERIMENTS
Experimental Setup
Fig. 8 shows the schematic diagram of the experimental setup
and Table II shows its specification. The displacement of the top (a)
of the tower is assumed to be proportional to the strain, which is
detected by a strain gauge attached to the lower part of the tower.
The boom angle is detected by a potentiometer connected to the
shaft that is the center of rotation of the boom. The edge of the
boom is equipped with a potentiometer with a fork clipping the
load rope and the potentiometer detects the angle of the load
rope that is assumed to be straight [1]. The rotational angle is
detected by the rotary encoder in the motor of the rotation. The
continuous controller designed in Section III is digitized with a
sampling time of 4 ms.
Experimental Results
We perform the control experiments by giving objective
values for the boom angle and the rotational angle. The ramp (b)
reference input is applied because rapid motion by the step
reference input causes swaying of the load. Fig. 9 shows the
response using the objective values shown in Table III. The
boom angle is transferred to an angle counterclockwise from
the -axis
(50)
(a)
(a)
(b)
Fig. 11. Block diagrams of analysis. (a) Decentralized control. (b) Central-
ized control.
(51)
(b)
where is a nominal value. Furthermore, the governing equa-
Fig. 12. Results of analysis. (a) Variation of l . (b) Variation of _ .
tion of the crane (30) includes . The term is represented
below using the real number parameter
In the experiment in Section IV, the ramp reference of the rota-
(52) tional speed is 20 deg/s. Equation (54) considers the variation of
the rotational speed, which is almost two times faster than the
experiment. The LFT representation is derived in similar ways;
Using (51) and (52), the tower crane model can be transformed
the result of analysis is shown in Fig. 12(b).
into the descriptor form as follows:
The robust stability scarcely improves using the centralized
controller.
VI. CONCLUSION
(53)
This paper has presented a comparison between decentral-
After this equation is transformed into LFT representation [9],
ized control and the centralized control to investigate the effi-
the of the decentralized control system is computed using the
ciency of the decentralized control. The centralized control de-
feedback loop diagram of Fig. 11(a), where is the structure of
sign model and the controller were newly designed. We compare
the perturbations, means the controller for the up-and-down
the response of the control experiment and the result of anal-
direction and means the controller for the rotational direc-
ysis considering the parameter variation which has a consider-
tion. The of the centralized control system is computed using
able effect on the coupling between directions. The control ex-
the feedback loop diagram of Fig. 11(b). Fig. 12(a) shows the
periment showed that it was difficult to improve the control per-
results of the analysis in the case in which the parameter
formance using centralized control because control theory
varies. The solid line indicates the result of the centralized con-
had difficulty adjusting the control arbitrarily for the cross-term.
troller and the broken line indicates that of the decentralized one.
The analysis showed that the robust stability in the case in
In the case in which the variation of is considered, the range
which the parameter varied indicated no difference between the
of the variation is
decentralized control and the centralized one. Thus, the decen-
tralized control was sufficiently effective for the control of the
(54) tower crane.
42 IEEE TRANSACTIONS ON CONTROL SYSTEMS TECHNOLOGY, VOL. 11, NO. 1, JANUARY 2003