Вы находитесь на странице: 1из 7

Input Shaping Control of Double- researchers have suggested adding feedback control to suppress

the double-pendulum dynamics 关8,9兴. Using feedback control is


Pendulum Bridge Crane Oscillations challenging due to the difficulty of measuring the payload motion.
Therefore, this paper does not advocate feedback control, but
rather command-shaping control that does not require sensors.
William Singhose For a crane controller to be practical for real-time oscillation
e-mail: singhose@gatech.edu suppression, it must filter out unwanted excitations from the
human-generated command signal. Such a modification can be
Dooroo Kim accomplished by convolving the human-generated command sig-
nal with a sequence of impulses 关10,11兴. The result of the convo-
Woodruff School of Mechanical Engineering, lution is then used to drive the crane motors. This input-shaping
Georgia Institute of Technology, process is demonstrated in Fig. 1 with a pulse command and an
input shaper containing three impulses. Note that velocity pulse
Atlanta, GA 30332 commands are very common in crane control because human op-
erators often press on-off buttons to drive cranes. The proper tim-
ing and scaling of the shaper impulses ensure that the system will
Michael Kenison arrive at the desired state without vibration. Input shaping has
Schlumberger Technology Corporation, proven effective for controlling oscillation of single-pendulum
Sugar Land, TX 77478 cranes 关5,12,13兴 and of single-pendulum cranes that undergo pay-
e-mail: kenison@sugar-land.oilfield.slb.com load hoisting 关14兴.
The earliest input-shaping techniques were developed in the
1950s 关11兴. The “posicast” control method developed by Smith
breaks commands into two smaller magnitude commands, one of
Large amplitude oscillation of crane payloads is detrimental to which is delayed by one-half period of the natural frequency, ␻n.
safe and efficient operation. Under certain conditions, the prob- The primary constraint equation used to calculate the command
lem is compounded when the payload creates a double-pendulum components ensures that there will be zero residual vibration
effect. Most crane control research to date has focused on single- when the system model is perfect. Therefore, posicast control is
pendulum dynamics. Several researchers have shown that single- now commonly referred to as zero vibration 共ZV兲 input shaping.
mode oscillations can be greatly reduced by properly shaping the Input shaping has been implemented on several large bridge
inputs to the crane motors. This paper builds on those previous
cranes at nuclear facilities 关12,14兴 and a 10-ton crane at Georgia
developments to create a method for suppressing double-
Tech 关5,15兴. The Georgia Tech crane has an overhead vision sys-
pendulum payload oscillations. The input shaping controller is
tem that can track the motions of the hook and payload. Figure 2
designed to have robustness to changes in the two operating fre-
shows the responses of the crane hook under standard operation
quencies. Experiments performed on a portable bridge crane are
and also when input shaping is enabled. The two responses are
used to verify the effectiveness of this method and the robustness
from the same human operator. Under normal operation the hook
of the input shaper. 关DOI: 10.1115/1.2907363兴
has large transient and residual sways. However, input shaping
virtually eliminates the detrimental oscillations.
Input shaping has many advantages over feedback controllers.
1 Introduction Input shaping does not require cameras or other sensors to deter-
mine the state of the crane and its payload. Therefore, input shap-
Manipulation of heavy objects at disaster sites, nuclear plants, ing is less expensive to implement. When using input shaping, the
warehouses, construction sites, and shipyards is often accom- system can have a faster settling time than with feedback control-
plished with cranes. Unfortunately, the natural sway of crane pay- lers alone 关16兴. Finally, input shaping cannot make a system go
loads causes safety hazards, time delays, and degradation of posi- unstable.
tioning accuracy. Much of the previous work on crane control has In this paper, the dynamics and control of bridge cranes that
concentrated on single-pendulum dynamics or on the single pen- exhibit double-pendulum dynamics are investigated. The impor-
dulum with changing length 共hoisting兲. tance of the double-pendulum effect is characterized as a function
If a computer controller is utilized and cable swing is consid- of the system parameters in Sec. 2. An input shaping scheme
ered in the control design, then time-optimal commands that result based on linear, time-invariant dynamics is then developed to
in zero residual vibration can be generated 关1兴. Hoisting of the mitigate the double-pendulum effects. As a first step in this devel-
payload during transverse motion increases the difficulty of con- opment, Sec. 3 demonstrates the effect of input shaping for only
trol because the oscillation frequency is time varying. Optimal the low mode of oscillation. While this significantly reduces the
controls based on time-varying and nonlinear models can be dif- oscillation, the second mode can still be problematic in a large
ficult to generate 关2,3兴. Even if optimal commands are generated, range of operating conditions. Therefore, Sec. 4 describes a
implementation may be impractical because the final setpoint method of input shaping to suppress both modes. Experimental
must be known at the outset of the motion. When feedback mea- results from a bridge crane are used to demonstrate the improved
surements are available, adaptive controllers and combination response provided by the proposed approach.
open- and closed-loop control is possible 关4,5兴.
If a crane behaves like a single pendulum, then experienced
crane operators can eliminate much of the payload sway by caus-
ing an oscillation during deceleration that cancels the oscillation
induced during acceleration. However, certain types of payloads
2 Double-Pendulum Dynamics
and riggings result in double-pendulum dynamics 关6,7兴. Under Figure 3 shows a schematic representation of a double-
these conditions, the manual method of eliminating oscillation pendulum crane. The crane is moved by applying a force, u共t兲, to
becomes very difficult, even for skilled operators. Therefore, some the trolley. A cable of length L1, hangs below the trolley and
supports a hook of mass, mh, to which the payload is attached
using rigging cables. Here, the rigging and payload are modeled
Contributed by the Dynamic Systems, Measurement, and Control Division of
ASME for publication in the JOURNAL OF DYNAMIC SYSTEMS, MEASUREMENT, AND CON-
as a second cable, of length, L2, and point mass, m p. Assuming
TROL. Manuscript received August 16, 2006; final manuscript received August 14, that the cable and rigging lengths do not change, the linearized
2007; published online May 1, 2008. Assoc. Editor: Nader Jalili. equations of motion, assuming zero initial conditions, are

Journal of Dynamic Systems, Measurement, and Control MAY 2008, Vol. 130 / 034504-1
Copyright © 2008 by ASME

Downloaded From: https://dynamicsystems.asmedigitalcollection.asme.org on 06/19/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use


Fig. 1 Input shaping a pulse input

冉 冊 冉 冊
␪¨ 1共t兲 = −
g
L1
␪1 +
gR
L1
␪2 −
u共t兲
L1
Fig. 4 Variation of low and high frequencies

共1兲
␪¨ 2共t兲 = 冉 冊 冉
g
L1
␪1 − +冊
g gR gR
+
L2 L2 L1
␪2 +
u共t兲
L1 In this section, we examine a subset of crane manipulation tasks
where the crane is used to move payloads that remain near the
where ␪1 and ␪2 describe the angles of the two pendulums, R is ground. That is, the sum of the two cable lengths stays constant.
the ratio of the payload mass to the hook mass, and g is the Figure 4 shows the two oscillation frequencies as a function of the
acceleration due to gravity. Note that double-pendulum crane dy- mass ratio and the rigging length, L2, when the total length 共sus-
namics can also arise when the payload is a long distributed mass pension cable plus rigging兲 is held constant at 6 m. The low fre-
connected to a massless hook. The equations of motion of such a quency changes very little and corresponds closely to the fre-
crane differ from those given in Eq. 共1兲. However, the input shap- quency of a single 6 m pendulum. On the other hand, the second
ing techniques presented in this paper will still be effective, pro- mode has a strong dependence on the rigging length.
vided that the two oscillation frequencies in such a system can be The low frequency is maximized when the two cable lengths
reasonably approximated. are equal 共3 m in this case兲. This maximum condition can be
The linearized frequencies of the double-pendulum dynamics proven by differentiating the expression for ␻1 given in Eq. 共2兲
modeled in Eq. 共1兲 are
with respect to L2 and then setting the result equal to zero. Note

␻1,2 = 冑冑 g
2
共1 + R兲 冉 1
+
1
L1 L2
⫿␤ 冊 共2兲
that over the wide range of parameter values shown in Fig. 4, the
low frequency only varies between 0.2040 Hz and 0.2521 Hz.
This is a ⫾11% variation about the median value of 0.2281 Hz. In
where contrast, the second mode deviates ⫾78% over the same param-

␤= 冑 共1 + R兲2 冉 1
+
1
L1 L2
冊 冉 冊
2
−4
1+R
L 1L 2
共3兲
eter range.
These results seem to indicate that an oscillation control
scheme would need more robustness to variations in the second
The frequencies depend on the two cable lengths and the mass mode than in the first mode. However, if the amplitude of the
ratio. It is of interest to investigate how the frequencies change as second mode is very small compared to the first mode, then the
a function of the system parameters. Such information can be used controller does not need to address the second mode. The relative
to design an effective input-shaping controller. contribution of the two modes can be examined by breaking the
overall dynamic response into components arising from ␻1 and
␻ 2.
The responses of the two swing angles, ␪1 and ␪2, to an impulse
of magnitude A, introduced at time t0, are

A␻1共1 + ␻22L1␣兲
␪1共t兲 = sin共␻1共t − t0兲兲
k
A␻2共1 + ␻21L1␣兲
− sin共␻2共t − t0兲兲 共4兲
k

A␻1 A␻2
␪2共t兲 = sin共␻1共t − t0兲兲 − sin共␻2共t − t0兲兲 共5兲
Fig. 2 Typical hook responses k k

where

− g共1 + R兲
␣= and k = ␤L1g 共6兲
␻21␻22L1L2
If we assume small angles, then the impulse response of the
payload in the horizontal direction can be approximated as

x共t兲 = C1 sin共␻1t + ␺1兲 + C2 sin共␻2t + ␺2兲 共7兲

Fig. 3 Double-pendulum crane where

034504-2 / Vol. 130, MAY 2008 Transactions of the ASME

Downloaded From: https://dynamicsystems.asmedigitalcollection.asme.org on 06/19/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use


Fig. 6 Simulated hook response, R = 2

dulum was set to 2 and the length ratio was 1. Figure 6 also shows
the predicted response when a ZV shaper designed to suppress the
low mode is used to modify the commands. The simulated re-
sponse shows that the low mode has been eliminated, but some
Fig. 5 Ratio of high-mode amplitude to low-mode amplitude second-mode oscillations remain. The oscillation amplitude is 9%
of the unshaped case.
Figure 7 shows the experimentally measured responses using
␻1L1共1 + ␻22␣共L1 + L2兲兲 approximately the same parameter settings as in the simulations
C1 = shown in Fig. 6. The parameters are approximate because suspen-
k

冑冉兺 冊 冉兺 冊
sion length, rigging length, and masses cannot be measured with
n 2 n 2 total accuracy. Furthermore, actuator nonlinearities in the experi-
⫻ A j cos共␻1t j兲 + A j sin共␻1t j兲 共8兲 mental setup were not integrated into the simulations. These tests
j=1 j=1 were conducted with a computer generating the commands so that
the crane would accurately move 30.5 cm. The experiments con-
− ␻2L1共1 + ␻21␣共L1 + L2兲兲 firm the predicted large reduction in the low-mode amplitude
C2 = when ZV shaping is utilized. However, the residual oscillation is
k

冑冉兺 冊 冉兺 冊
only reduced down to 22% of the unshaped case due to small
n 2 n 2 nonzero initial conditions, modeling errors, and nonlinear actuator
⫻ A j cos共␻2t j兲 + A j sin共␻2t j兲 共9兲 dynamics. The experiments and simulations demonstrate the im-
j=1 j=1 portant result that single-mode ZV shaping will attenuate the first
The two coefficients, C1 and C2, indicate the contributions of each mode, but not necessarily the second mode.
mode to the overall payload response.
The goal of this work is to design an input shaper to move 4 Input Shaping for Double Pendulums
double-pendulum cranes with very little residual vibration. To do When the second mode causes the payload oscillation to exceed
this, we need to limit the maximum amplitude of the residual tolerable levels, then it must be taken into account when designing
vibration from a series of impulses. Because ␻1 ⫽ ␻2 and the an input shaper. There are a number of methods for designing
damping for each mode is approximately zero, the maximum am- multimode input shapers 关22–24兴. In this section, a technique is
plitude can be found by adding the maximum amplitudes from developed to directly target the two predicted frequencies of a
each mode as follows: double-pendulum crane. In order to determine the input shaper
Vamp = 兩C1兩 + 兩C2兩 共10兲 impulse amplitudes and time locations, a set of constraint equa-
tions is formulated and then satisfied.
Using this decomposition, the contribution of the second mode
becomes apparent and indicates when two-mode input shaping is 4.1 Residual Vibration Constraints. The maximum residual
required. Figure 5 shows the ratio of the high-mode contribution vibration amplitude from a series of impulses given in Eq. 共10兲
to the low-mode contribution for a large range of length and mass can be used as a constraint equation by requiring the vibration
ratios, again assuming an overall length of 6 m. The surface indi- amplitude, Vamp, to be less than some tolerable threshold, Vtol. It
cates that double-pendulum input shaping will be necessary for has been shown that robustness can be improved if the vibration is
systems with low payload-to-hook mass ratios. The second-mode limited to a small value, rather than forced to be exactly zero
contribution is particularly large when the suspension and rigging 关19,25兴.
lengths are approximately equal. In previous work on two-mode shapers, the vibration was not

3 Input Shaping for One Mode


There are numerous ways to design single-mode input shapers.
The interested reader should consult one of the many references
关10,11,17–20兴. In this section, ZV input shapers are used in both
simulation and in experiments on a portable bridge crane. The
crane has been used to teach advanced controls courses at Georgia
Tech, Georgia Tech Lorraine in France 关21兴, Kumoh National In-
stitute of Technology in Korea, and at the Tokyo Institute of Tech-
nology in Japan.
Figure 6 shows the simulated response of the hook when the
trolley is moved 30.5 cm 共1 ft兲 using standard on/off velocity
commands generated when a human operator pushes the control
buttons. For these simulations, the mass ratio of the double pen- Fig. 7 Experimental response, R = 2

Journal of Dynamic Systems, Measurement, and Control MAY 2008, Vol. 130 / 034504-3

Downloaded From: https://dynamicsystems.asmedigitalcollection.asme.org on 06/19/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use


Fig. 8 Suppressing two frequency ranges Fig. 9 Length-ratio sampling „nominal value= 2.4…

limited to a physical dimensional limit, such as 1 cm. Instead, the its insensitivity—the nondimensional frequency range over which
vibration of each mode was limited to a small percentage of the it suppresses vibration.
vibration occurring without the use of input shaping. While lim- In Fig. 8, modes at 1 Hz and 2.75 Hz are both suppressed with
iting percentage vibration is theoretically convenient, it makes it an insensitivity of 0.4. The low-range suppression is from
challenging to enforce a performance requirement expressed as a 0.8 Hz to 1.2 Hz. Dividing this spread by the nominal 1 Hz value
physical dimension such as, “limit the residual oscillation to less gives I1 = 0.4. The high mode is suppressed from
than 1 cm.” Fortunately, Eq. 共10兲 allows us to directly limit the 2.2 Hz to 3.3 Hz. When this 1.1 Hz range is divided by the nomi-
overall residual amplitude to a specific dimensional value by writ-
nal 2.75 Hz value, the second-mode insensitivity is once again
ing it in the following form:
calculated as I2 = 0.4.
Vtol 艌 兩C1兩 + 兩C2兩 共11兲 The analysis in Sec. 2 showed that the high mode is more
dependent on the physical parameters than the low mode. Given
In this case, Vtol can be expressed with a unit of length, such as
that there will always be errors in the parameter estimates, the
centimeter or inch.
second-mode suppression range may need to be wider. This is
4.2 Amplitude Constraints. The vibration caused by an in- easily done within this framework. Furthermore, the suppression
put shaper can be limited by Eq. 共11兲. However, if the input shaper ranges do not need to be symmetrical about the estimated frequen-
impulse amplitudes are not constrained, then their values can cies. For example, if the normal operation of the crane is more
range between positive and negative infinities. There are two pos- likely to cause the second mode to increase, rather than decrease,
sible solutions to this problem: limit the magnitude of the im- then the suppression range could be skewed to higher frequencies.
pulses to less than a specific value or require all the impulses to Finally, the tolerable vibration limit, Vtol, can be set to any value
have positive values. To streamline the presentation, the shapers and can even have different values for each mode.
discussed in this paper will contain only positive impulses as fol- Double-frequency-range suppression can also be accomplished
lows: in a manner that is directly related to the physical parameters of
the crane. Instead of creating suppression ranges from estimated
Ai ⬎ 0, i = 1, . . . ,n 共12兲 frequencies, we can vary the suspension-to-rigging length ratio
where n is the number of impulses in the shaper. Note that the over the range experienced during normal crane operation. The
methods presented here are applicable to shapers containing nega- constraint equations can then be used to limit the oscillation
tive impulses. If negative impulses are allowed, then the rise time across this operating range. This approach is demonstrated in Fig.
will improve, but potential drawbacks such as excitation of un- 9 for the case when the nominal length ratio is 2.4. The length
modeled high modes and actuator saturation must be addressed. ratio insensitivity is defined as the nondimensional range of length
Techniques for managing the challenges of negative input shapers ratios over which the vibration remains below the tolerable limit.
have been well documented 关20兴. A second amplitude constraint For example, the length-ratio insensitivity for the case shown in
must be enforced so that the shaped command reaches the desired Fig. 9 is 1.0 because the oscillation is limited from 0.5 to 1.5
setpoint; the impulse amplitudes must sum to 1 as follows: times the nominal length ratio of 2.4. Note that additional con-
n straints can easily be added to accommodate changes in mass or

兺A =1
the total suspension length.
i 共13兲
i=1 4.4 Minimization of Shaper Duration. Due to the transcen-
dental nature of the oscillation constraint equations, there are an
4.3 Robustness Constraints. The residual vibration con- infinite number of solutions. To select among these solutions and
straint of Eq. 共11兲 can be used to limit the vibration at a single set ensure that the rise time is as fast as possible, the shaper duration
of frequencies 共␻1 and ␻2兲. If the actual crane frequencies coin- must be made as short as possible. Therefore, the final necessary
cide with those used in Eq. 共11兲 to design the shaper, then the design constraint minimizes the time of the final input shaper
oscillation will be eliminated. However, to ensure robustness to impulse as follows:
modeling errors and parameter variations, the oscillation must re-
main small over a neighborhood of frequencies that surround the min共tn兲 共14兲
modeling frequencies. Robustness can be assured by using mul- To summarize the design process, two-mode SI input shapers
tiple versions of Eq. 共11兲 to limit the vibration at several points are designed by satisfying Eqs. 共12兲–共14兲 while Eq. 共11兲 is en-
near the modeling frequencies. This process is demonstrated in forced over two frequency ranges that contain the expected fre-
Fig. 8 for a two-mode system. In this case, the vibration has been quencies at several points throughout the expected range of length
limited at three frequencies near the low mode and five frequen- ratios. An alternate approach, which was illustrated in Fig. 9, is to
cies near the high mode. Because this approach allows the de- satisfy Eqs. 共12兲–共14兲 while Eq. 共11兲 is enforced over the expected
signer to specify the frequency range over which the vibration is range of length or mass ratios. These are straightforward numeri-
suppressed, the resulting shapers are called specified insensitivity cal optimizations. The input shapers designed for this paper were
共SI兲 shapers 关26兴. The robustness of such a shaper is measured by obtained using the MATLAB optimization toolbox.

034504-4 / Vol. 130, MAY 2008 Transactions of the ASME

Downloaded From: https://dynamicsystems.asmedigitalcollection.asme.org on 06/19/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use


Fig. 10 Effect of second-mode insensitivity on shaper Fig. 12 Effect of vibration limit on shaper duration
duration

from zero to some small value results in substantial time savings.


4.5 Trade-Offs in Shaper Design. The duration of the two- Further increases in Vtol continue to yield improved rise time, but
mode SI shaper increases with the amount of desired insensitivity. at a reduced rate.
A plot of shaper duration versus second-mode frequency insensi- Using a two-mode SI shaper instead of a one-mode SI shaper
tivity is shown in Fig. 10 for different values of Vtol in both the obviously incurs a time penalty in shaper duration. It is important
low- and high-mode ranges. Fig. 10 was generated using typical to know if this time penalty is a significant drawback. Figure 13
frequencies of an industrial crane at Georgia Tech 共0.25 Hz and shows the percent increase in shaper duration for a typical two-
0.35 Hz兲 关5,15兴. Furthermore, the low-mode insensitivity, I1, was mode SI shaper when compared to a single-mode SI shaper de-
set to 0.10 共⫾5 % 兲. It is apparent that there are certain ranges in signed for the low mode. To generate the curve, shapers were
which a substantial increase in insensitivity causes only a small designed with a frequency insensitivity of 0.4 for the high mode
increase in shaper duration. This knowledge can be used to make and the insensitivity of the low mode was varied. The time penalty
an intelligent choice for the desired insensitivity level. For ex- in shaper duration incurred by adding the high-mode suppression
ample, a second-mode insensitivity of I2 = 0.4 共⫾20% 兲 with a Vtol is very small for low-mode insensitivity higher than 0.2. As a
of 5% for both low and high modes may be marginally sufficient result, suppressing the vibration of the high mode can be accom-
for a particular application. However, given the results shown by plished with very little cost.
the solid line in Fig. 10, the engineer should increase the required
insensitivity to 0.7, as only a negligible time penalty is incurred. 5 Experimental Verification
A three-dimensional plot of the shaper duration as a function of The procedure described above was used to design an input
frequency insensitivity for both the low and high modes is shown shaper for the portable crane with a payload-to-hook mass ratio of
in Fig. 11. The data were generated using Vtol = 0.05 共5%兲 for both 2 and a suspension-to-rigging length ratio of 1. The shaper was
the low and high frequency ranges. When both insensitivities are designed to accommodate a ⫾5% variation in the low mode
low 共⬍0.1兲, the shaper duration is a little over 3 s in duration. 共0.55 Hz兲 and a ⫾10% variation in the high mode 共1.7 Hz兲. The
When both I1 and I2 are greater than 0.1, I1 has the dominant

冋 册
resulting input shaper contains five impulses:

冋册
effect on shaper duration. That is, I2 can be very large without
causing the shaper duration to increase substantially, as was Ai 0.1661 0.2321 0.2035 0.2321 0.1661
shown previously in Fig. 10. On the other hand, when I2 is held = 共15兲
ti 0 0.3475 0.7569 1.1664 1.5138
constant and I1 is increased, the shaper duration increases more
rapidly. The sensitivity curve for this shaper is the solid line in Fig. 14.
Another important input shaper design parameter is the level of The curve shows the normalized vibration amplitude for the
tolerable residual vibration. Every system has some amount of double-pendulum system and indicates that both modes of the
vibration that is acceptable. If the shaper is designed to meet this double-pendulum crane will be suppressed. Figure 15 shows the
performance specification, rather than greatly exceed it, then the experimental response when the double-pendulum shaper is uti-
shaper duration can be significantly decreased. This effect was lized. As predicted by the sensitivity curve, both the first and
shown somewhat in Fig. 10, but it is clearly demonstrated in Fig. second modes are effectively suppressed by the input-shaping pro-
12 where the shaper duration is plotted as a function of the toler- cess.
able vibration amplitude, Vtol. Five separate values of the desired Recall that for low mass ratios, the second mode will be more
frequency insensitivity 共I1,2 = 0.2– 0.6兲 for both the low and high important and the need for two-mode shaping will increase. When
modes are shown. In each case, increasing the tolerable vibration the mass ratio is lowered to 0.5, the two frequencies shift to ap-

Fig. 11 Effect of first- and second-mode insensitivities on Fig. 13 Effect of first-mode insensitivity on shaper duration
shaper duration „I2 = 40% …

Journal of Dynamic Systems, Measurement, and Control MAY 2008, Vol. 130 / 034504-5

Downloaded From: https://dynamicsystems.asmedigitalcollection.asme.org on 06/19/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use


Fig. 14 Sensitivity curves for the two-mode SI shapers
Fig. 17 Effect of length ratio on oscillation amplitude, R = 2

proximately 0.58 Hz and 1.13 Hz. Note that the low mode is vir- tude of unshaped and SI shaped vibration over a large range of
tually unchanged, but the second mode has shifted considerably. length ratios for the case when R = 2. The experimental results

冋 册
For these values, the two-mode SI shaper is clearly demonstrate the desired robustness.

冋册Ai
ti
=
0.1889 0.3111 0.3111 0.1889

0 0.4804 1.0044 1.4849


共16兲 6 Conclusions
For certain payloads and rigging configurations, bridge cranes
The sensitivity curve for this shaper is shown by the dashed line in can exhibit significant double-pendulum dynamics. The second
Fig. 14. The oscillation suppression characteristics near the low mode becomes important when the mass of the hook is significant
mode are nearly identical to the SI shaper for R = 2; however, the when compared to the mass of the payload. Furthermore, the
second-mode suppression range has shifted to a lower frequency second-mode contribution is maximized when the suspension
range to match the decreased second-mode frequency of the R length and rigging length are equal. When the second mode is
= 0.5 case. The experimental responses shown in Fig. 16 demon- important, an input shaper can be designed to suppress the multi-
strate that the two-mode SI shaper works very well on the R mode vibration. Furthermore, the input shaper can be made robust
= 0.5 configuration. to modeling errors and parameter variations by suppressing a
For any control technique to be practical, it must be robust to range of possible frequencies or anticipated parameter variations.
modeling errors. The degree of input-shaping robustness can eas- Experiments on a portable crane demonstrated the effectiveness,
ily be controlled by adjusting the insensitivity used to design the practicality, and robustness of the proposed approach.
shaper. As stated previously, the shapers in Eq. 共15兲 and 共16兲 were
designed to suppress vibration for low-mode variations of ⫾5% Acknowledgment
and high-mode variations of ⫾10% from the expected values. To The authors would like to thank Siemens Energy and Automa-
test if robustness was achieved, the lengths of the suspension and tion for providing the equipment and funding for this research. In
rigging cable were varied and the tests were repeated 共without addition, the authors thank the President’s Undergraduate Re-
changing the input shaper兲 to obtain the residual vibration ampli- search Awards Program at Georgia Tech for their sponsorship of
tude in the presence of modeling errors. Fig. 17 shows the ampli- this project.

References
关1兴 Auernig, J. W., and Troger, H., 1987, “Time Optimal Control of Overhead
Cranes With Hoisting of The Load,” Automatica, 23, pp. 437–446.
关2兴 Moustafa, K. A. F., and Ebeid, A. M., 1988, “Nonlinear Modeling and Control
of Overhead Crane Load Sway,” J. Dyn. Syst., Meas., Control, 110, pp. 266–
271.
关3兴 Sakawa, Y., and Shindo, Y., 1982, “Optimal Control of Container Cranes,”
Automatica, 18共3兲, pp. 257–266.
关4兴 Butler, H., Honderd, G., and Van Amerongen, J., 1991, “Model Reference
Adaptive Control of a Gantry Crane Scale Model,” IEEE Control Syst., 11,
pp. 57–62.
关5兴 Sorensen, K., Singhose, W., and Dickerson, S., 2007, “A Controller Enabling
Precise Positioning and Sway Reduction in Bridge and Gantry Cranes,” Con-
trol Eng. Pract., 15共7兲, pp. 825–837.
关6兴 Kenison, M., and Singhose, W., 1999, “Input Shaper Design for Double-
Fig. 15 Experimental hook response with double-pendulum SI Pendulum Planar Gantry Cranes,” in IEEE Conference on Control Applica-
shaper, R = 2 tions, Hawaii, pp. 539–544.
关7兴 Liu, D., Guo, W., and Yi, J., 2005, “GA-Based Composite Sliding Mode Fuzzy
Control for Double-Pendulum-Type Overhead Crane,” in Second International
Conference on Fuzzy Systems and Knowledge Discovery, Vol. 3617, pp. 792–
801.
关8兴 Lahres, S., Aschemann, H., Sawodny, O., and Hofer, E. P., 2000, “Crane
Automation by Decoupling Control of a Double Pendulum Using Two Trans-
lational Actuators,” in Proceedings of the 2000 American Control Conference,
Vol. 2, pp. 1052–1056.
关9兴 Tanaka, S., and Kouno, S., 1998, “Automatic Measurement and Control of the
Attitude of Crane Lifters: Lifter-Attitude Measurement and Control,” Control
Eng. Pract., 6共9兲, pp. 1099–1107.
关10兴 Singer, N., and Seering, W., 1990, “Preshaping Command Inputs to Reduce
System Vibration,” J. Dyn. Syst., Meas., Control, 112, pp. 76–82.
关11兴 Smith, O. J. M., 1958, Feedback Control Systems, McGraw-Hill, New York.
关12兴 Singer, N., Singhose, W., and Kriikku, E., 1997, “An Input Shaping Controller
Enabling Cranes to Move Without Sway,” in ANS Seventh Topical Meeting on
Fig. 16 Experimental hook response with double-pendulum SI Robotics and Remote Systems, Augusta, GA, Vol. 1, pp. 225–231.
shaper, R = 0.5 关13兴 Starr, G. P., 1985, “Swing-Free Transport of Suspended Objects With a Path-

034504-6 / Vol. 130, MAY 2008 Transactions of the ASME

Downloaded From: https://dynamicsystems.asmedigitalcollection.asme.org on 06/19/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use


Controlled Robot Manipulator,” J. Dyn. Syst., Meas., Control, 107, pp. 97– 关20兴 Singhose, W., Singer, N., and Seering, W., 1997, “Time-Optimal Negative
100. Input Shapers,” ASME J. Dyn. Syst., Meas., Control, 119, pp. 198–205.
关14兴 Singhose, W., Porter, L., Kenison, M., and Kriikku, E., 2000, “Effects of 关21兴 Lawrence, J., and Singhose, W., 2005, “Design of Minicrane for Education and
Hoisting on the Input Shaping Control of Gantry Cranes,” Control Eng. Pract., Research,” in Sixth International Conference on Research and Education in
8共10兲, pp. 1159–1165. Mechatronics, Annecy, France.
关15兴 Khalid, A., Huey, J., Singhose, W., and Lawrence, J., 2006, “Human Operator 关22兴 Hyde, J., and Seering, W., 1991, “Using Input Command Pre-Shaping to Sup-
Performance Testing Using an Input-Shaped Bridge Crane,” ASME J. Dyn. press Multiple Mode Vibration,” in IEEE International Conference on Robot-
Syst., Meas., Control, 128共4兲, pp. 835–841. ics and Automation, Sacramento, CA, pp. 2604–2609.
关16兴 Kenison, M., and Singhose, W., 2002, “Concurrent Design of Input Shaping 关23兴 Singh, T., and Heppler, G. R., 1993, “Shaped Input Control of a System With
and Proportional Plus Derivative Feedback Control,” ASME J. Dyn. Syst.,
Multiple Modes,” J. Dyn. Syst., Meas., Control, 115, pp. 341–347.
Meas., Control, 124共3兲, pp. 398–405.
关24兴 Singhose, W., Crain, E., and Seering, W., 1997, “Convolved and Simultaneous
关17兴 Pao, L., Chang, T., and Hou, E., 1997, “Input Shaper Designs for Minimizing
Two-Mode Input Shapers,” IEE Proc.: Control Theory Appl., 144, pp. 515–
the Expected Level of Residual Vibration in Flexible Structures,” in Proceed-
ings of the American Control Conference 1997, Albuquerque, NM, Jun. pp. 520.
3542–3546. 关25兴 Singhose, W., Seering, W., and Singer, N., 1990, “Shaping Inputs to Reduce
关18兴 Park, U. H., Lee, J. W., Lim, B. D., and Sung, Y. G., 2001, “Design and Vibration: A Vector Diagram Approach,” in IEEE International Conference on
Sensitivity Analysis of an Input Shaping Filter in the z-Plane,” J. Sound Vib., Robotics and Automation, Cincinnati, OH, IEEE, New York, Vol. 2, pp. 922–
243共1兲, pp. 157–171. 927.
关19兴 Singhose, W., Seering, W., and Singer, N., 1994, “Residual Vibration Reduc- 关26兴 Singhose, W., Seering, W., and Singer, N., 1996, “Input Shaping for Vibration
tion Using Vector Diagrams to Generate Shaped Inputs,” ASME J. Mech. Des., Reduction With Specified Insensitivity to Modeling Errors,” in Japan-USA
116, pp. 654–659. Symposium on Flexible Automation, Boston, MA, Vol. 1, pp. 307–313.

Journal of Dynamic Systems, Measurement, and Control MAY 2008, Vol. 130 / 034504-7

Downloaded From: https://dynamicsystems.asmedigitalcollection.asme.org on 06/19/2019 Terms of Use: http://www.asme.org/about-asme/terms-of-use

Вам также может понравиться