Fluid Power Control of a Hyper-Active Seat for Low-Frequency Vibration Suppression

ELSEVIER

Copyright © IFAC Advances in Automotive Control Salemo, Italy, 2004

IFAC PUBLICATIONS www.elsevier.comllocatelifac

FLUID POWER CONTROL OF A HYPER-ACTIVE SEAT FOR LOW-FREQUENCY VIBRATION SUPPRESSION Samuel Klooster· Kris Kozak· Joshua Vaughan· Peter Sanders· William Singhose·

• Georgia Institute of Technology Woodruff School of Mechanical Engineering Atlanta, GA , USA bill. singhose@me .gatech. edu

Abstract: Whole-body vibrations pose a problem to operators of off-road vehicles and heavy machinery. Current vehicle suspension systems including passive seats do not provide adequate vibration suppression for the operator. To deal with this, a three-degree-of-freedom Hyper-Active Seat is being developed at the Georgia Institute of Technology to suppress the vibrations felt by the operator. Two sets of experiments were conducted to evaluate the potential performance of the HyperActive Seat. First, the step response of the seat was determined for each of the three degrees of freedom. Second, the frequency response of the seat was generated to show the seat's capability of suppressing low-frequency vibrations. These two experiments characterized the roll-off frequency and damping ratio of the seat from which a system model was developed for each degree of freedom and compared to the experimental results. This model will aid in the development of a controller for vibration suppression. Copyright © 2004 IFAC Keywords: Active Seat, System Identification, Frequency Response, Modelling

1. INTRODUCTION

des of the human body. The human body usually has a resonant frequency between 4 and 5 Hz and a low tolerance for amplitudes between 0.5 and 1.5 g's (r.m.s.) (Gniady and Bauman, 1991). The combination of high-amplitude and low-frequency vibrations of the vehicle and the low tolerance of the human body causes serious health risks for operators who are subjected to prolonged exposure. Common health problems are developed in the lumbar spine area, shoulders, and upper back (Rehn et al. , 2002). Prolonged exposure to such vibrations can cause permanent damage to the spine, as well as increasing fatigue which may decrease the performance of the operator. Developing active seats to minimize these vibrations felt by the operator is a promising solution to this problem.

Terrain-induced low-frequency vibrations pose a problem to operators of off-road vehicles and heavy machinery. Whole-body vibration is experienced when the operator sits on a vibrating piece of machinery, such as a tractor, or bulldozer, for extended periods of time. This exposure time is often longer than the recommended standard set by ISO 2631 (2631-1 , 1997) . This standard recommends a maximum of 8 hours of exposure time for operators who are exposed to vibrations in the 4 to 8 Hz range with an amplitude greater than 0.064 g's (r.m.s.) . However, operators of mining, agricultural, and construction vehicles often work up to 12 hours per day and are exposed to vibration frequencies of 0 to 20 Hz at amplitudes of 0.2 to 1.2 g's (r.m.s.) (Gniady and Bauman, 1991).

Most active seat research concentrates on suppressing the vibration in the vertical direction and

These low-frequency vibrations are especially harmful because they excite the resonant frequen-

501

The format of this paper is first to describe the construction of the Hyper-Active Seat and show experimental results of the potential performance of the seat . Two experiments were conducted to evaluate the potential performance of the HyperActive Seat in the three cartesian directions . First , the open-loop step responses of the seat show the rise time of the seat. Second, the open-loop frequency responses of the seat shows the seat's potential for suppressing low-frequency vibrations. Then a model was developed from the step and frequency response data and compared to the experimental results. Analysis of the experimental data will evaluate the potential performance of the seat to suppress vibration. The development of a model will aid in the development of a controller to suppress these vibrations.

y

~~x Hydraulic Cylinders

Fig. 1. Sketch of Hyper-Active Seat

thus uses a one-degree-of-freedom model (Gniady and Bauman, 1991). Although the vertical direction contains the most intense vibrations felt by operators, they are subjected to vibration in other directions as well. An active seat with two or more degrees of freedom would be able to reject disturbances more effectively. Nevala, et al. have developed an active seat that controls the vertical and horizontal directions (Nevala et al., 1996) . Research has also been done on an active seat that controls the roll and pitch angles (Young and Suggs, 1973).

2. HYPER-ACTIVE SEAT CONSTRUCTION Construction of a 3 degree-of-freedom HyperActive Seat prototype, which will suppress vibrations in the vertical, Y , fore-aft , X, and pitch angle, rp, directions, has been completed. The design utilizes a parallel kinematical architecture in a variable geometry truss configuration. This architecture and configuration provides the desired three degrees of freedom while balancing the competing issues of actuator force transmission and workspace size. This configuration has the added benefit of containing no kinematic platform singularities within the workspace, which pose a safety hazard for an occupant.

To continue research in this area, a three-degreeof-freedom Hyper-Active Seat is being developed at the Georgia Institute of Technology shown in Figure 1. It is actuated using three linear fluid power actuators that form a parallel manipulator with three Revolute-Prismatic-Revolute links (3RPR). It has the ability to suppress vibrations in the vertical, Y , fore-aft , X , and pitch angle , rp directions (or vertical, side-to-side and roll if the top of the seat is rotated by 90 degrees).

The actuator hardware used for this prototype was provided by Deere and Company, and consists of three hydraulic cylinders (with a stroke of 6.35 cm and a 3.81 cm diameter bore shown in Figure 1), and three proportional flow control valves. This hardware is currently used in John Deere's one-degree-of-freedom Active Seat™ (Dufner and Schick, 2002). These actuators are powered by a 2.27 * 10- 3 m 3 / s, variable displacement pump running at a maximum of 3.45 * 103 kPa across the cylinders.

Vibration suppression in the fore-aft and pitch angle direction are generally neglected, however , agricultural equipment operated with heavy mounted implements, such as a plow or trailer, oscillates in these directions (Marsili et al. , 2002 ; Mehta et al. , 2000; Prasad et al., 1995). Although the primary application of the Hyper-Active Seat is for agriculture and construction vehicles, the seat may be adaptable to suppress vibrations in passenger vehicles.

2. 1 Control System

Movement control of the Hyper-Active Seat in the three cartesian directions (vertical, Y , fore-aft , X , and pitch angle, rp) required the knowledge of the leg lengths. To obtain these lengths, encoders are used to measure the angles between the actuators and the top and bottom of the seat. The angle measurements where then used to estimate the leg lengths of the seat.

One added advantage to the geometry of the Hyper-Active Seat is that it can operate as a safety seat , a seat that moves during a vehicle collision to decrease occupant injuries, by changing pitch angle (Klooster and Singhose, 2003 ). By actively (or passively) tilting the front of the seat upwards , the normal and frictional forces on the occupant can be increased preventing them from sliding beneath the steering wheel (Bohrnler, 1997; Mikami, 1994; Simon, 1969).

To move the Hyper-Active Seat in the three cartesian directions, the inverse kinematics are calculated using the following equations:

502

.'? 0.8 -·

y

'-' Q)

""0

.:: 0.6 --

x Fig. 3. Vertical Unloaded Step Response

Fig. 2. Diagram of 3RPR Mechanism II = [(X

+ dsin'P -

a1 COS'P - X01)2

+(Y - dcosc.p - a1 sin'P - yod 2]1 / 2 (1) l2

= [(X + dsin'P -

a1 cos'P - X02)2

+(Y - dcos'P - a1 sin'P - Y02)2]1 / 2 (2)

+ dsin 'P + a2 cos'P - X03)2 +(Y - dcos'P + a2 sin'P - Y03)2]1 / 2, (3)

l3 = [(X

where the coordinates and kinematic parameters are shown in Figure 2. These inverse kinematic equations transform the desired vertical, Y , foreaft, X, and pitch angle, 'P commands into the corresponding desired leg length commands. In order to know the exact position of the HyperActive Seat in the cartesian works pace, the forward kinematics are used. The forward kinematical equations, which give X, Y , and 'P in terms of the leg lengths, can be solved for using (1-3).

Fig. 4. Vertical Loaded Step Response and the effective loaded weight of the seat is 86 kg.

3.1 Step Response

To evaluate the potential of the Hyper-Active Seat to suppress disturbances encountered by the vehicle, and to create a model of the system, the step response and frequency response were explored. The open-loop step response of the seat characterizes the rise time of the seat , the damping ratio, as well as the residual vibration, and the open-loop frequency response of the seat indicates what bandwidth of frequencies the seat can suppress. The analysis of the data collected during the experiments led to the development of a simple model for each of degree of freedom.

The open-loop step response of the Hyper-Active Seat gives information, such as damped natural frequency and damping ratio , to develop a model which will be used to determine what range of vehicle vibrations can be cancelled. A step of 1 in. was input into the vertical and horizontal directions, while a step of 0.15 rad was input into the pitch angle direction. To obtain the step response, the maximum possible voltage was given to the system for 0.050 sec. Analysis of the step response data indicated the damped natural frequency and damping ratio for each of the three directions. These values are tabulated in Table l. Analysis of the step response data revealed the addition of mass to the seat reduced the damped natural frequency and damping ratio.

These experiments were conducted with the HyperActive Seat unloaded and loaded, due to the uncertainty of an operator's weight. Knowledge of the seat's performance at two weights allows for the extrapolation of the data to other weights. The effective unloaded weight of the seat is 40 kg,

Figure 3 and Figure 4 show the unloaded and loaded system step response in the vertical direction. The non-linearities of the system are evident in the residual vibration, because the responses do not show a constant damped natural frequency, as evident after the first peak overshoot.

3. EXPERlMENTS

503

Table 1. Experimental Data and Model Predictions Damped :-latural Frequency (Hz )

Damping Ratio

Roll-Off frequency (Hz)

Model Roll-Off Frequency (Hz)

Vertical

Cnloaded Loaded

12.0 10.0

0.24 0.17

12.0 10.3

12.0 8.3

Model Damping Ratio 0.26 0.17

Horizontal

Unloaded Loaded

5.0 3.6

0.22 0.20

6.0 3.5

5 .2 3.6

0.25 0.22

Pitch Angle

Unloaded Loaded

7.1 5 .0

0.17 0.16

12.0 5.0

12.5 8.6

0.20 0.18

0.2 :---;---,--_,_--,-.,--,.,....-r--,-.,.....,-r-r---,-,-,

0.8

S

,--.,

,g.

c

~0.6 ...

oJ" d:':~~~~~~~~~~=~ ...... ··:··'···~--···~;~r~~~;v~ ~e~t" -

o 'L':: ~ 0.1 .··.·

"0

.3

i. 0.4 E -<

c..

· k

-
0.2

- - :YIodel

.

.

. ., ... -

. I

Fig. 5. Horizontal Unloaded Step Response

Fig. 7. Pitch Angle Unloaded Step Response 0. 16,.-.--,-~~.,....,....,....,..,.....,.~,.......,...,-,-.,....,....,...,..........,~~---r-''''''

:.:: :r' ::

.:.L.iil L: . :!.L ..... L.i.!..ULL.l ..LU.. ... L . t

·

. ,

... .

_ _ ~:~:~r~:. {··:':·

1.2 r-----,-,..---r-~....,..,---r-,-,_,_-,.._.,__._-_r_r__,__,.,...,

Io

: : :

.

:? ......., 0.8

: ..... .

0.1-. -" :::

] 0.08 k ·~·

o

~0.06 H·:·

"0

E 0.6 : ::::::::::: ~-__M_o_de_I_ _ _..... ! OA~..
-< 0.04-;. 0.02-~

0.2 ",; .. ;..;..•..:.. :.! .. i.. L.! .. ~ .;.. !.. L.;.1. ).)...:...•...:. ..•..:...;..:...:...~ ..;.. : . . . : : .

O~-~~~-~~~~~w-~~--~~

o

..

0.5

1.5

0.1

0.2

0.3

OA

0.5

0.6

0.7

Time (sec)

Time (sec) Fig. 8. Pitch Angle Loaded Step Response Fig. 6. Horizontal Loaded Step Response

3.2 Frequency Response

The unloaded and loaded system step response in the horizontal direction is shown in Figures 5 and 6. The rise time and settling time are greater in the horizontal direction since the actuators cannot generate as much force horizontally due to the geometry.

Frequency response data was collected along the three cartesian degrees of freedom: vertical, horizontal, and pitch angle. The frequency response gives a clear understanding of the roll-off frequency of the seat in each of the three directions as shown in Table 1. Knowledge of the roll-off frequency indicates the bandwidth the seat can attenuate vibrations.

Figure 7 and Figure 8 show the unloaded and loaded system step response in the pitch angle direction. The step responses is unlike that of the vertical and horizontal directions because it is rotation and not translation. An FFT of the unloaded pitch angle step response revealed two damped natural frequencies at 7.1 Hz and 12.5. The interaction of these frequencies reduce the amount of residual vibration.

Figure 9 and Figure 10 show the frequency response of the unloaded and loaded Hyper-Active Seat in the vertical direction, while Figures 11 and 12 show the frequency response of the (unloaded and loaded) Hyper-Active Seat in the horizontal direction. The Hyper-Active Seat has a much lower roll-off frequency in the horizontal direction compared to the vertical direction, due to the

504

j

j

40 r-----.-T>-n~r_------~~~--~ ... ". . . . . ..

. :!

.. ·..... L.;...i .~ . .:.L

30

Q,)

~c

0

1"

~

20

S

1O r"

Q,)

~

0

So-1O

~ -20

'" -20 ~

~

... .:.... ..... ,

0.1

.

'"

' ' ':'''';''':' '':- '.'-:'';''

.........:... .

,

. . . .... . · .... · · .. -:. .. ~·-7·~·7·~~· .. ·.. · .... ~ ..

4%~.1--~~~~·~·~1~~--~~~~~

1

Frequency (Hz)

Frequency (Hz)

Fig. 9. Vertical Unloaded Frequency Response

Fig. 12. Horizontal Loaded Frequency Response

, :

40 ~~'Tn-~=====.=.~ . .~ .. ==~

20r---.~-'~,".---r_~~~~--r-r... ... ~:

30

10

~ 20

".....

a:l Q,)

Q,)

] -20

0

..

.~ -10 .. ....... ~ .....~.... ;...~.. ~..;.;. ~ .; ......... L .... i.... ~ ..;.. ;..;.; .~.~.. ... .. ~"

'"

~ -20 -30

o

S-1O

S 10 ]

Hyper-Aclive Seal

.... :.... ..; .. ..;, .. ~ ...~. ~,.~ .;.~ ..

-30

-40

-

.. ' - - Model

. ~ -30

"-1

. . . . .;. ..;..;..:. ,:.:;. ... ...;. ..;. . ;.;.:. :;.;:. . {"l .........:..... L.l..L~.:U ................;....;.;.. ;.-,,-,~.. ... .. .'\~

'"

~-40

-50

. i

-WO~.I--~~~·~·~ · ·~I--~~·~ · ~ · ~~IO~~~

-40.~~~~~· ~··~--~~~·~·~ · ~ ·· ~~~

0.1

1

........i......:....~ ...j .. ~. ~.~. h......... ~......:....~ .. ~ .. ~ ... ,~.: -:.......... ~. ........ ~ ..... ~.... ~ .. ~ .. 1- ~+1-f" '"'' .. ~ .....~ ...~ ..~ .. ~. U. i·;··, .... ·.~.

10

Frequency (Hz)

Frequency (Hz)

Fig. 13. Pitch Angle Unloaded Frequency Response

Fig. 10. Vertical Loaded Frequency Response 40r-----.--.~Tn----r_~~~Tr.,--~

30 ..-.

,

'.

. . ":

_H)~r~Ac~l~'~ Seat .~,

":"': '

: : : : : ::

I

.. "':" ... .. ...~. --Model

10

~ ~: L: ..:·:.··r::. }J.tH:!:!T···:..··I·:. ·l..::r::[:.I:"l:j:l.l...:....:J

S -10

······,····,··,···;···~rT·····)···,···:··,· ';:.;.; ..... '1

~ -20

. . .....

~

c ~

~

0

: ::

"..... a:l

I

0

Q,)

So -30

-10

-20 ......... L .....~ ...l .. i. ..~ . ~ . L~l .....

'"

~-40

::::: \

-50

-3%~.I--~~~~·~·~·~ : ~~~~~~:~:~ ·:+.io~~

.

:

...

,

...

.

. . . :.. . .

....... ~ ..... ~.... ~. '~"l ' ~"~'~'~ ..... .... ~... .. ~.... ~ .. ~ .. ~ '~ ' ~'1 ' :'"

.

... ~ .. ...:..

..

.... ~..... ~....~ .. ~ .. j ' ~"~' ~'~ ......... ~... .. ~.. "f' 'i--j · t h· ~"" ..... ~ ~ :.~.~).~.... ... . ~... "-:-. ": .. ~ .. ~ ' ~':' :.. ......... :....

.t ..... ..

. . . :: ..

'

-WO~.l--~~~~~I~~~~~~IO~~~

Frequency (Hz)

Frequency (Hz) Fig. 11 . Horizontal Unloaded Frequency Response Fig. 14. Pitch Angle Loaded Frequency Response geometry ofthe seat. The stiffness of the system in the horizontal direction is lower due to increased rotation of the revolute joints shown in Figure 2.

4. MODELLING A simple model of the seat was developed from the data for each of the three degrees of freedom and not the whole seat due to coupling between the actuators. Each direction was modelled as third order system, of the form:

Figures 13 and 14 show the frequency response of the unloaded and loaded Hyper-Active Seat pitch angle. The roll-off frequency of the pitch angle was lower than the vertical direction, due to the geometry and motion of the seat. The stiffness of the seat is lower because of the increased rotation of the revolute joints. Also, the pitch angle is a rotation and not a translation, so the inertia properties are different.

cl£.

Gp( s) =

(2 + m ;;S + m k)' S S

(4)

where c, b, and k were determined by fitting the unloaded step and frequency responses using the

505

Institute of Technology for providing partial funding for this research.

optimization toolbox in MATLAB T M . A thirdorder model was selected due to the Hyper-Active Seats mechanical dynamics, namely inertia, being modelled as second-order, and the hydraulics of the seat being modelled as a simple integrator. An integrator was chosen for the hydraulics due to the evidence of a zero at the origin as seen in Figures 9-14. This is due to the distance the seat moves being the integral of the distance the hydraulic valve is open.

REFERENCES 2631-1 , ISO (1997). Mechanical variation and shock - evaluation of human exposure to whole-body vibration. part i: General requirements. Technical report. International Organization for Standardization. Bohmler, Klaus (1997). Restraining system for vehicle occupants. number 5,908,219. Dufner, D. and T . Schick (2002). John deere active seat: A new level of seat performance. In: AgEng2002: lntl. Conf. on Agricultural Engineering. Budapest, Hungry. p. 7. Gniady, John and John Bauman (1991). Active seat isolation for construction and mining vehicles. In: SAE Tech. Paper Series. pp. 1-6. Klooster, Samuel J. and William E. Singhose (2003) . A study of passenger seat parameters as a basis for active safety seat control. In: 2003 Mediterranean Conference on Control and Automation. Rhodes, Greece. p. 7. Marsili, A. , L. Regni, G. Santoro, P. Servadio and G. Vassalini (2002) . Innovative systems to reduce vibrations on agricultural tractors: Comparative analysis of acceleration transmitted through the driving seat. Biosystems Engineering 81(1) , 35-47. Mehta, C. R. , M. Shyam, Pratap Singh and R. N. Verma (2000) . Ride vibration on tractor-implement system. Applied Ergonomics 31 , 323-328. Mikarni, Tatuya (1994) . Seat with user protecting means. number 5,556,160. Nevala, Kalervo, Matti Kangaspuoskari and Tatu Leinonen (1996). Development of an active suspension mechanism for the seat vibration damping. In: Proceedings of the 4th lASTED lntl. Conf.: Robotics and Manufacturing. Honolulu, Hawaii, USA. pp. 337-339. Prasad, Niranjan, V. K. Tewari and Rajvir Yadav (1995) . Tractor ride vibration - a review. Journal of Terramechanics 32(4), 205-219. Rehn, B., 1. A. Bergdahl, C. Ahlgren, C. From, B. Jarvholm, R. Lundstrom, T. Nilsson and G. Sundelin (2002). Musculoskeletal symptoms among drivers of all-terrain vehicles. Journal of Sound and Vibration 253(1 ), 2129. Simon, Lewis B. (1969). Automatic vehicle occupant restraint. number 3,591 ,232. Young, Roy E . and C. W. Suggs (1973) . An active seat suspension system for isolation of roll and pitch in off-road vehicles. In: 1973 Annual Meeting of the American Society of Agricultural Engineers. Lexington, Kentucky, USA. p. 18.

The model of the seat was only fit for low frequencies and not valid for frequencies above the roll-off frequency. This was done since the low frequencies are more harmful to the operator. Once c, b, and k were determined for the unloaded seat, they were then calculated for the loaded case. The damping constant was increased by 30% due to the loading of the seat. This is justified because loading the seat increases the physical damping effects. Table 1 compares the roll-off frequencies and damping ratios predicted by the model versus the experimental data. The models predicted a decrease in roll-off frequency and damping ratio which correlates to the experimental data. A comparison of the models and the experimental results can be seen in Figures 3-14. Note that the pitch angle response of Figures 8 and 14 fit well for higher roll-off frequency, but does not take into account the lower frequency present. The model breaks down do to the presence of two frequencies. To increase the accuracy of the model a fifthorder model was tried, however this model did not capture the rise time of the step response. 5. CONCLUSION The architecture and components of a HyperActive Seat were discussed. The purpose of the Hyper-Active Seat is to suppress terrain-induced vibrations better than a one or two DOF seats. The seat also has the potential to use the pitch angle control for a safety seat implementation. The current performance ofthe seat was evaluated using step and frequency response in multiple degrees of freedom. From these results a third-order model was developed for each of the three degrees of freedom. This model predicts the response of the vertical and horizontal directions, and predicted the correct trends for the pitch angle. The model will be used in the development of a controller to suppress vibration. From the analysis of the data, it can be concluded the Hyper-Active Seat has the potential to suppress vibrations that are harmful to the operator. ACKNOWLEDGEMENTS We would like to thank the companies of the Fluid Power and Motion Control Center at the Georgia

506