Sie sind auf Seite 1von 8

Aleksander Hac

Delphi Chassis,
Advanced Product/IVIaterials Technologies,
Dayton, OH 45401-1245

lljoong Youn
Hyundai Motor Connpany, Gyjnggi,
Republic of Korea

Hsien H. Chen
Vehicle Systems Research Department,
General Motors NAO R&D Center,
Warren, Ml 48090-9055

Control of Suspensions for


Vehicles With Flexible Bodies
Part II: Semi-Active Suspensions
Two methods of control of semi-active suspensions that specifically address the problem of structural vibrations of the vehicle body are considered. These control strategies are based on those developed for active suspension systems in Part I of this
study and rely on either modifications of suspension control forces that account for
body compliance or on the addition of a proof-mass actuator to reduce structural
vibrations. A half-car model that includes body compliance is used to evaluate the
effects of these control strategies on the performance of the suspensions with twostate and continuously modulated dampers., The performances of the systems are
evaluated in both the time and frequency domains. The effect of time delays in the
process of actuating the adjustable dampers is investigated. Significant reductions of
structural vibrations are observed when the nodes of body beaming modes are a
sufficient distance away from the suspension mounting points, and the time delays in
the control system are negligible. The results deteriorate markedly when two-state
dampers are used instead of continuously variable dampers or when a time delay in
excess of 5 ms is present in the control loop. When the preference in suspension
design shifts toward road holding it becomes increasingly difficult to improve the
vehicle structural response without sacrificing other aspects of performance.

Introduction

Semi-active suspensions have the potential to attain more


widespread use in mass-produced vehicles than fully active systems because of their lower cost and their negligible demand
for power. These suspensions use adjustable dampers as force
actuators, in which the rate of energy dissipation can be varied
either between two distinct levels ("on-off " dampers) or continuously within a given range of damping coefficients. In the
latter case, the force generated by such a damper can in principle
track the force that would be required by an active device as
long as it is dissipative, that is, does not require supply of
energy. Consequently, semi-active suspensions can in many respects achieve performance comparable to that of active suspensions.
An important aspect in which the performance of "semiactive" suspensions is not satisfactory, is high frequency harshness that has been observed in road tests and reported in analytical studies of semi-active suspensions (Miller and Nobles,
1990). This problem is particularly severe for systems involving
on-off dampers. Since two-state dampers can vary the damping
coefficient only between two levels, they have a tendency to
produce forces that have discontinuities at the instant when
switching between these two levels of damping occurs. Forces
with such characteristics contain substantial contributions in
the high frequency range of their spectrum. Such forces excite
structural vibrations of the body which are transmitted through
the floor, steering column, and body panels causing noise, rattles, and contributing to fatigue and deterioration of ride quality.
The results of analysis and simulations based on rigid-body
models fail to predict these vibrations.
Several studies (Allen and Karnopp, 1975; Ahmadian and
Marijoram, 1989; Khulief and Sun, 1989) have been published
in which the effects of semi-active suspensions on vehicle models including structural flexibility of the body were evaluated.
Contributed by the Dynamic Systems and Control Division for publication in
the JOURNAL OF DYNAMIC SYSTEMS, MEASUREMENT, AND CONTROL . Manuscript

received by the DSCD February 22, 1994; revised manuscript received March 7,
1995. Associate Technical Editor: G. Rizzoni.

They generally show some reduction in the levels of structural


vibration but these improvements were more modest than for
fully active suspensions. The control laws used by Allen and
Karnopp (1975) and Ahmadian and Marijoram (1989) were
based on "skyhook damping" strategy, which is optimal for a
one-degree-of-freedom vehicle model. Reductions in the levels
of structural vibration reported in these studies were the result
of better isolation of the vehicle body from the road inputs, so
that the beaming modes were less excited. No attempt has been
made to modify the control laws to make the controller more
aware of the flexible modes. Khulief and Sun (1989) applied
a full state feedback control law to semi-active suspension of
a vehicle with flexible body and demonstrated that modal damping of the elastic body modes can be significantly increased.
Since, however, the main focus of their investigation was on
modeling and numerical issues, they did not demonstrate the
effects of the proposed controller on other aspects of performance, such as the dynamics of the wheels.
The purpose of this study is to investigate whether the application of more sophisticated control strategies that include consideration of body compliance, can reduce structural vibrations
of the body without adversely affecting other aspects of performance. For this purpose a "model following" (Choek et al.,
1985) approach is used, in which the control laws developed
in Part I of this paper constitute the reference models. These
control laws, developed on the basis of well-established linear
control techniques, provide the desired suspension forces in the
form of feedback control laws. Damping coefficients are then
selected that minimize at any time instant the difference between the actual damping force and the desired control force.
The system in which a proof-mass actuator is used to control
structural vibrations, while the suspension controller is based
on a rigid-body model, is also investigated. As in Part I, decentralized control strategies with output feedback are considered
to provide a reasonable balance between complexity and performance.
To evaluate the degree of enhancement of the performance
brought about by proposed controllers, the behavior of the system is compared with that of the semi-active suspensions with

518 / Vol. 118, SEPTEMBER 1996

Transactions of the ASME

Copyright 1996 by ASME


Downloaded From: http://asmedigitalcollection.asme.org/ on 12/16/2015 Terms of Use: http://www.asme.org/about-asme/terms-of-use

were present, their damping coefficients could be included in


the values of minimal damping Vk^m The forces produced by
the adjustable dampers are the products of the damping coefficients and the velocities across the dampers, which in turn can
be expressed as linear functions of the state variables. Hence
the portion of the control vector that corresponds to the suspension forces can be written as
Ui(t) = (diag v)Ex

(2)

where Ui(0 = [Uf(t), u,.it)]^, diag \{t) = diag (ui, V2), is a


2 x 2 diagonal matrix with elements Vi and V2 on the main
diagonal, and E is a constant 2 X matrix such that the product
of /tth row of E and the state vector x is the velocity across the
kth damper. For the system involving the proof-mass actuator,
the vector of control forces can be expressed as:
(3)

u(f) = [ u [ ( 0 , . ( f ) ] '
Fig. 1 Vehicle model with semi-active suspension

rigid-body-based controller and with a passive suspension. The


selection of an appropriate passive system as a basis for comparison is not obvious. For a passive system, the value of the
damping coefficient is the result of a compromise between the
requirements of ride comfort and road holding. Electronically
controlled suspensions, on the other hand, are capable of deciding between the requirements of ride comfort and road holding
on the basis of momentary information and can change their
characteristics accordingly. As a result, when no need for significant lateral force exists, the suspension can be programmed
to a soft mode, in which the damping of the wheels may be
significantly smaller than for a typical passive system. During
cornering, braking, acceleration, when significant cross-winds
are detected etc., the suspension can be switched into a firm
mode. In the comparison between passive and semi-active suspensions, two sets of damping coefficients were considered for
the passive suspension. They yield the same variances of tire
deflections as those produced by semi-active systems in the soft
and firm modes, respectively.
2

The Vehicle Model


A two-dimensional model of a vehicle with a semi-active
suspension is shown in Fig, 1. The fundamental assumptions
regarding the model as well as the vehicle parameters are the
same as for the active system considered in Part I. The only
difference is that the suspension control forces Uf and u, are
here generated by adjustable dampers that can produce only
dissipative forces. It is assumed that the time-varying damping
coefficients vi(t) and U2(0 can be arbitrarily regulated, which
in practice can be achieved only approximately. In the figure,
a proof-mass actuator (active vibration absorber) is attached to
the vehicle body to control the structural vibrations, although
for the greater portion of this study the system without an absorber is considered. We assume that the absorber remains a
fully active device since the power demand for this absorber is
much smaller than that of the active suspension. When continuously variable dampers are used, the damping coefficients can
be continuously varied within a given range of values

where M(r) refers to the force generated by the proof-mass


actuator. Consequently, the state equation describing the dynamics of the system can be written as
X = Ax -I- B

(diag v)Ex

+ Dvf.

(4)

Ua

In the above equation matrices A, B, and D are the same as


for the active system with the passive damping coefficients C/
and c, replaced by uim, and t)2,,, respectively. The state vector,
X, and the disturbance, w, are defined as in Part I. For the
system without an absorber, the state equation has the form:
X = Ax + B(diag v)Ex -I- Dw.

(5)

In Eq. (5) the state vector x and matrices A, B, D, and E are


properly modified to account for the lower dimension of the
system. Equations (4) and (5) are bilinear state equations due
to the presence of terms involving the product of the state vector
and the control variables Vi(t) and 1)2(0

3 Semi-Active Control
Before introducing specific control laws for semi-active suspensions, we summarize the control objectives and interpret
them in terms of requirements on the frequency domain characteristics. This will provide a better understanding of how these
objectives may be achieved through variation of damping.

3.1 Discussion of Control Objectives. As discussed


briefly in the Part I of this study, the design of a vehicle suspension involves a compromise among several, generally conflicting, objectives. For ride comfort the acceleration levels of the
passenger compartment should be kept low. Good road holding
ability requires minimization of the variation of the tire normal
force (which is proportional to tire deflection) around static
equilibrium, since adequate lateral forces can be assured only
when firm, uninterrupted contact between the tires and the road
is maintained (Hrovat, 1993). A further objective considered
in this study is to reduce structural vibrations of the body as
expressed by modal accelerations of the first two beaming
modes. At the same time, packaging requirements for the front
and rear suspensions and the proof-mass actuator must be satisfied, imposing constraints on the suspension deflections and the
relative displacement of the absorber with respect to the body.
k= 1,2
Vkit) =S Vs.,
Vk
(1) All these goals must be achieved within the limits on the dampwhere vtmax > ^nin s 0 for any k. In the case of on-off dampers, ing coefficients of the adjustable dampers and the force develi)t(f) can be switched only between two states v.,o/, and Vjirm- The oped by the proof mass actuator. A valuable insight into the
minimal damping coefficients can be usually much smaller than trade-offs among various control objectives can be gained by
for a typical passive system, but damping cannot be totally considering the frequency domain characteristics. Since a semieliminated. Without loss of generality, the passive suspension active suspension behaves essentially as a passive system with
damping, included in the model with an active suspension, may time-varying dampers, a simplified understanding can be
be disregarded in the current model since, even if such dampers achieved by analyzing the effect of different levels of damping

Journal of Dynamic Systems, Measurement, and Control

SEPTEMBER 1996, Vol. 1 1 8 / 5 1 9

Downloaded From: http://asmedigitalcollection.asme.org/ on 12/16/2015 Terms of Use: http://www.asme.org/about-asme/terms-of-use

(within the range used in automotive suspensions) on these


characteristics.
When the frequency domain responses for body acceleration,
tire deflection and suspension rattle space for a rigid vehicle
with a passive suspension are considered, two major resonance
peaks can be observed that correspond to the natural frequencies
of the sprung mass and the wheels. More detailed analysis of
the frequency response curves for a passive vehicle with various
levels of damping reveals four distinct regions in the relevant
frequency range (Lizell, 1988; Hennecke and Zieglmeier,
1988):
(1) The sprung mass mode (typically 0.5-3 Hz) in this
range the ride comfort, road holding, and suspension deflection
are improved simultaneously by increasing the damping coefficients; hence, firm damping is required to suppress the body
resonances.
(2) The mid-frequency range, i.e., the range between the
resonance peaks of the sprung and unsprung masses (typically
3 - 8 Hz). In this range, reduction of the damping coefficients
improves both ride comfort and road holding with only a marginal effect on packaging requirements. The damping should
be kept low, especially since, in this region, the human body
is the most sensitive to mechanical vibrations (ISO/DIS 2631,
1972).
(3) Thewheel-hopmodetheregionofthe resonance frequency of the wheels ( 8 - 1 3 Hz). Firm damping is required to
improve road holding, which also reduces suspension motions
and results only a slight or no penalty in ride comfort.
(4) Harshness (over 13 Hz)above the wheel resonance
frequency, acceleration of the body increases with increased
damping without significant benefits in road holding ability.
Therefore, low damping is required to improve body isolation.
Some roads have isolated harmonic or periodic excitations
in various frequency bands that can be detected and processed
selectively. On such roads an "adaptive damping" strategy
(Hennecke and Zieglmeier, 1988) may be successful, in which
constant damping that suits particular road conditions is selected
on the basis of the frequency content of the road input, and
updated when the road conditions change. Many roads, however, have a broad frequency spectrum covering several of the
above ranges. On such roads, adaptive damping has limited
appeal since different levels of damping are required in various
frequency ranges. Superior performance can be achieved with
rapidly varying dampers that can closely approach the behavior
of active systems, which are known to be capable of providing
a good balance of all control objectives. However, the price to
be paid for introducing rapid variations of damping is highfrequency harshness, whose consequences cannot be fully predicted when rigid-body models are used.
When the flexibility of the vehicle body is considered in the
model, additional peaks corresponding to the structural resonances will appear in the frequency characteristics for body
accelerations, usually above the wheel hop frequency. Low
damping in this range, required for the rigid-body model, should
reduce forces transmitted to the body and lessen the excitation
of structural resonances. If, however, the structural modes are
excited, their transient vibrations will persist for a long time
because of the very limited structural damping. More important,
for semi-active suspensions, the rapid variations of damping
coefficients, and consequently suspension forces, will provide
persistent excitation of the structural vibrations. Therefore, in
order to obtain satisfactory performance, increased modal
damping of the structural modes is necessary. In summary, a
good semi-active system should provide high damping for low
frequency inputs to achieve good body isolation, low damping
in the mid-frequency range for good comfort, adequate damping
to control the wheel hop, especially under conditions of motion
that require the development of lateral forces, and finally increased damping of structural modes. As shown in Part I these
520 / Vol. 118, SEPTEMBER 1996

somewhat conflicting objectives can be fulfilled with a properly


designed active suspension and our purpose here is to investigate whether, by adjusting the forces in the dampers based on
the same control laws, comparable performance can be
achieved.
3.2 Control Laws. The control laws developed for the
active suspension in the Part I of this article may be applied
with some modifications to the semi-active system analyzed
here. Therefore, two classes of control algorithms are considered. In the first one, an extra feedback loop which uses information about the structural modes is added to the rigid-body-based
suspension controller. In the second approach, an active vibration absorber mounted on the vehicle body is used to suppress
structural vibrations, while suspension control is based on a
rigid-body model. In the case of semi-active suspensions, however, the required suspension forces Uf and u,. cannot be arbitrarily manipulated by the designer. They are achieved by varying the damping coefficients of the shock absorbers, Vi{t) and
1)2(0, which are subject to inequahty constraints (1). In this
case damping coefficients may be obtained that minimize at any
given time the difference between the control forces required by
the active system and the damping forces generated by variable
dampers. This approach is sometimes referred to as "cUppedoptimal" control (Margolis, 1983; Hrovat et al., 1988; Butsuen
and Hedrick, 1989) or "model reference" approach (Choek et
al., 1985). Let *, k = \,2, and Uo,, refer to the control forces
obtained for the fully active system using any of the algorithms
developed in Part I. Then the elements of the control vector are
given by
Vkmui

Vkit) =

if

MotAXj. S i;i,j( A X j )

if

UicAXt a

UoJAxt

Uimax(AXt)^

(6a)

otherwise

"(0 = oo(0

(6b)

for fc = 1,2 where Ax,, = e^x denotes the velocities across the
dampers; e^ is the kth row of matrix E introduced in Eq. (2).
According to Eq. (6a) the damping force, %(r)Axt, follows
exacfly the suspension control force that would be produced by
a fully active system unless it requires a rate of damping outside
the inequality constraints set for the damping coefficients. In
that case the damping is switched either to a minimum or to a
maximum level depending on which setting is capable of producing a damping force that is closer to the desired control
force. As a result, the closed-loop system is piecewise linear
and varies among various combinations of a fully active system
and the systems with passive damping. The closed-loop system
can be described by the following equation
% = Ac(v)x + Dw

(7)

where A<,(v) = A(v) + [0, BjJG,, and A(v) is the matrix A


with the damping coefficients v^rim (k = 1, 2) replaced by v^it)
and B2 is obtained by partitioning matrix B according to B =
[Bi, B2]. Matrix Go is a control gain matrix obtained by using
any of the controllers developed in Part I. For the system without
absorber, Ac(v) = A ( v ) .
It should be pointed out that the above problem has a singularity when the velocity across the /:th damper Ax* = 0 for any k
= 1, 2. In that case the force produced by the damper is zero
regardless of the value of the damping coefficient Vi,{t). The
most recent value of Vt(t) can then be maintained to avoid
unnecessary switching.
When two state (on-off) dampers are used instead of continuously variable dampers, the following control law for the variation of the damping ratio is proposed:
Transactions of the ASME

Downloaded From: http://asmedigitalcollection.asme.org/ on 12/16/2015 Terms of Use: http://www.asme.org/about-asme/terms-of-use

Table 1 Components of the performance index for various types of suspensions in soft mode; the passive suspension is
taken as 100 percent
Variances of

Type of suspension
Passive soft
Semi-active, full state feedback
Semi-active, output feedback
Semi-active, rigid-body control.
full state
Semi-active, rigid-body control.
output feedback
Semi-active with absorber.
full state
Semi-active with absorber.
decentrahzed
On-off, full state feedback
On-off, output feedback
On-off, rigid-body control.
full state
On-off rigid-body, output
feedback

Heave
ace.

Pitch
ace.

1-st
modal
ace.

2-nd
modal
ace.

Susp.
deil.

Total
index

m.sq.
power
[kW]

100%
34.8%
33.0%
38.6%

100%
75.3%
74.8%
73.8%

100%
38.8%
47.9%
128.1%

100%
222.3%
189.4%
326.8%

100%
26.8%
27.6%
26.5%

100%
63.9%
63,9%
70.0%

38.2%

73.6%

144.7%

348.9%

27.1%

71.0%

33.0%

73.5%

29.2%

171.0%

27.1%

62.1%

0.037

40.5%

79.3%

36.3%

236.9%

26.7%

66.6%

0.038

51.1%
56.8%
47.7%

84.0%
83.5%
81.3%

74.5%
88.4%
122.9%

345.1%
298.9%
610.5%

44.7%
50.6%
39.6%

71.2%
78.1%
77.1%

47.3%

81.6%

120.6%

633.3%

39.3%

77.4%

Table 2 Components of the performance index for various types of suspensions in firm mode; the passive suspension is
taken as 100 percent
Variances of

Type of suspension
Passive firm
Semi-active, full state feedback
Semi-active, output feedback
Semi-active, rigid-body control,
full state
Semi-active, rigid-body control.
output feedback
Semi-active with absorber,
full state
Semi-active with absorber.
decentralized
On-off full state feedback
On-off, output feedback
On-off, rigid-body control,
full state
On-off rigid-body, output
feedback

Heave
ace.

Pitch
ace.

1-st
modal
ace.

2-nd
modal
ace.

Susp.
defl.

Total
index

m.sq.
power
[kW]

100%
91.9%
115.1%
98.7%

100%
96.6%
103.2%
92.3%

100%
64,6%
92.4%
191.2%

100%
110.6%
97.0%
158,3%

100%
59.4%
53.6%
60.4%

100%
95.5%
101.2%
101.8%

123.0%

103.9%

152.5%

134.0%

53.4%

105.4%

82.4%

92.4%

28.8%

77.1%

62,0%

93.1%

0,086

114.5%

104.2%

32.5%

87.6%

58.9%

100.7%

0,127

117.7%
160.2%
129.1%

111.3%
132.8%
105.6%

101.7%
139.8%
273.5%

848.5%
400.6%
1401%

71.1%
62.3%
69.9%

105.9%
117.5%
115.7%

162.7%

132.8%

176.0%

1309%

54.2%

121.2%

second beaming mode increased sharply for semi-active suspensions. In Fig. 2(b) the responses in terms of tire deflections are
shown. These plots are very similar for all the suspensions
discussed below since the variances of tire deflections on the
randomly profiled road were adjusted to the same level. ThereTable 3 Percent increase in the variances of acceleration
at the body center of mass and at point B due to flexibility.
Semi-active and on-off suspensions in soft modes.
Type of suspension
Passive firm
Semi-active, output feedback
Semi-active, rigid-body control.
output feedback
Semi-active with absorber,
decentralized
On-off, output feedback
On-off, rigid-body control.
output feedback

522 / Vol. 118, SEPTEMBER 1996

Center
of mass

Point B

91.4%
52.2%

71.9%
40,6%

88.3%

81.3%

18.4%
38,9%

27.6%
37,2%

57.6%

75.7%

fore, plots of the tire deflections will not be shown in subsequent


figures. When the semi-active suspension with output feedback
is used (Fig. 3(a)) the amplitudes of the structural vibrations
remain small and are quickly suppressed, although the motion
is slightly more irregular than for the corresponding active system. For the semi-active system with the absorber (Fig. 3(b))
the performance is generally similar to that of the semi-active
system with output feedback. When on-off dampers are used
(Fig. 4 ) , significant discontinuities in the acceleration curves
occur which are caused by discontinuities in the suspension
forces. The performances of the systems with output feedback
and with rigid-body control are close, indicating that no significant improvement in controlling the structural modes can be
achieved with these dampers. This conclusion is again consistent with the results obtained for random road input.
In Fig. 5 the control forces are compared for semi-active
suspensions with output feedback and that with proof-mass actuator and decentralized control. For the system with an absorber, the suspension forces are essentially identical to those
obtained for the system with rigid-body control (which are not
shown). In spite of the severity of the bump, the maximum
Transactions of the ASME

Downloaded From: http://asmedigitalcollection.asme.org/ on 12/16/2015 Terms of Use: http://www.asme.org/about-asme/terms-of-use

I I I I I'

elastic response
rigid response
S

-^
N
in

'I

I I I I I I I I I I I

elastic response
rigid response

2.0

B 1.0

1-0

.2 0.0
4-)

4)

-2.0

-2.0

'
'

b)

' '

' t ' ' ' I '

' I ' I ' ' I ' ' I I ' '

elastic response
rigid response

B
0.0

elastic response
rigid response

X 2.0
S

1.0

0.0

ii4) -1.0

I I ' ' I ' I

b)

-2.0
I

0.0

0.2

0.3

. .

_9ft

. . I . . .

0.2

0.4

Time (sec)
Fig. 4 Response of the vehicle model with suspension with two-state
dampers and output feedback (soft mode) to the impact bump input: (a)
Rigid-body control; (b) output feedback control

Arst
second
beaming beanung
(17.2 Hz) (30.x H)

passive fiim
.. passive soft
- semi'active output feedback
TTI I

2 10-'
800 p r r "Tr-|"-r"i"'f'Tni'T "I' i i|'t' t-y-r^r-i^t-ry

0.4

"^'S- Response of the vehicle model to the impact bump input for the
semi-active soft suspension with output feedback control and: (a) A time
oel^V * 2 ms; (b) a time delay of 4 ms

present. Therefore, the characteristics for semi-active systems


presented here cover only this frequency range.
Figures 7, 8, and 9 show the frequency responses of the semiactive systems with output feedback, with rigid-body control,
and with an active absorber, respectively. The characteristics
for the passive firm and passive soft systems are also provided
for reference. The characteristics for semi-active systems exhibit fluctuations at frequencies above 20 Hz, which have been
observed in earlier studies (Butsuen and Hedrick, 1989). In the
frequency range 3-20 Hz the performance of the semi-active
systems is essentially the same as that of the corresponding
active systems. Around the (rigid) body resonance and for high
frequencies, their performance is worse. All the semi-active

a)

0.3

Time (sec)

10

' " I

10'

Irr-nrri

10'

Frequency IHz)

i 'irr-

Fig. 7 Frequency response in terms of the acceleration at point B to


the front wheel velocity input for the passive suspensions and the semiactive soft suspension with output feedback

flit
second
beaming beaming
{17.2 H J ) (30.1 Hi)
L\

-A.! I

800 IIII I I I "1 "t "? j - T I I "II' 1 -I " ' i T ' i - I -r I I )

- passive Arm
passive soft
- semi-active rigid body control
-TTTTl
2H10''

Fig. 5 Control forces generated while traveling over the impact bump
for semi-active soft suspensions: (a) With output feedback; (b) with
proof-mass actuator and output feedback

524 / Vol. 118, SEPTEMBER 1996

10'

10'
Frequency (Hz)

10'

Fig. 8 Frequency response in terms of the acceleration at point B to


the front wheel velocity input for the passive suspensions and the semiactive soft suspension with rigid-body control

Transactions of the ASME

Downloaded From: http://asmedigitalcollection.asme.org/ on 12/16/2015 Terms of Use: http://www.asme.org/about-asme/terms-of-use

Table 4 The effect of pure time delay on the performances of semi-active soft suspension with output feedback. The
system without delay corresponds to 100 percent.
Time delay [ms]
Variances of

2.5

5.0

7.5

10

15

Heave acceleration
Pitch acceleration
1-st modal acceleration
2-nd modal acceleration
Acceleration of center of mass
Acceleration of point B
Jerk
Suspension deflection
Tire deflection
Total

107.5%
105.6%
88.6%
126.3%
99.5%
108.0%
312.8%
97.7%
99.0%
102.2%

124.2%
117.3%
97.7%
314.3%
111.3%
150.1%
707.4%
93.8%
95.1%
108.3%

154.5%
134.8%
116.3%
833.4%
136.2%
243.9%
1254%
91.2%
90.9%
120.5%

197.7%
166.5%
167.3%
2385%
192.6%
490.3%
2171%
83.9%
83.5%
145.3%

256.7%
190.0%
328.0%
3881%
298.1%
779.2%
2656%
83.2%
69.9%
170.9%

control forces do not exceed 800 N, and the peak value is


smaller for the system with output feedback that includes feedback from flexible modes. The forces generated by the proofmass actuator are nearly an order of magnitude smaller than
the suspension forces.
In Fig. 6 the effects of pure time delays on the performance
of the semi-active system with output feedback are shown. It
can be seen that the presence of time delays introduces jerks in
the system response that become more pronounced as the time
delay is increased. These discontinuities in accelerations are
similar to those observed in real systems. The difficulties in
controlUng the second beaming mode with a time delay of 4
ms can be observed.
4.3 Frequency Domain Characteristics. In this section,
the frequency response characteristics of the vehicle model with
semi-active soft suspensions are analyzed. As in the case of the
system with active suspension, the input and output were the
ground velocity under the front wheel and the acceleration at
point B, respectively.
For the semi-active system described by the bilinear state
equation, obtaining the frequency domain characteristics is
more difficult than for a fully active, linear system because the
system is not linear, but piecewise linear. In this study, the
frequency domain responses were obtained by subjecting the
front wheel to a random road input (with the same stochastic

so

1.0

characteristics as the input used in Section 4.1), and then by


taking Fourier transforms of the input and the output to generate
the spectral estimates of both signals. The magnitudes of the
transfer function at the desired frequencies were obtained as
the square roots of the ratios of the spectral estimates of the
output and the input at those frequencies. For linear systems,
this method should yield the same result as direct calculation
of the magnitude of the frequency domain transfer function
obtained from the equation of motion (Newland, 1975). We
used the IMSL subroutine FTFPS, which applies fast Fourier
transform (FFT) to prefiltered time sequences to generate the
spectral estimates. This method is inherently burdened with
random errors, that can be reduced by using a sufficiently long
data record and applying final smoothing by calculating the
average values of adjacent spectral estimates (Newland, 1975).
In the results presented here, over 65,000 data points were used
with a sampling interval of 5 ms; final smoothing was performed
by averaging every m adjacent estimates where m varied from
5 to 40 depending on the frequency (for high frequencies the
data points were more closely spaced than at low frequencies).
To test the reliability of this method, the frequency domain
characteristics for the passive system were obtained and compared with those calculated from the transfer function. The results were virtually identical in the frequency range of 0.5-50
Hz; beyond this range discrepancies between both curves were

- elastic response
rigid response

-4 '0

elastic response
rigid response

<n

\
S

10

.2 0.0
0.0

2
3-10

'

'

'

1 '

'

'

I I '

'

'

'

'

'

front wheel
rear wheel

_aft

b)

elastic response
rigid response

\
_g 1.0
a

nn

o -o.

^
-1.6

0.2

0.3

0.4

Time (sec)
Fig. 2 Response of the vehicle model with semi-active soft suspension
and rigid-body control to the Impact bump input: (a) Acceleration at
point S; (b) tire deflection

Journal of Dynamic Systems, Measurement, and Control

OZ

0.3

0.4

Time (sec)
Fig. 3 Response of the vehicle model with semi-active soft suspension
to the impact bump Input with (a) output feedbaci< control; (b) with
proof-mass actuator and decentralized control

SEPTEMBER 1996, Vol. 1 1 8 / 5 2 3

Downloaded From: http://asmedigitalcollection.asme.org/ on 12/16/2015 Terms of Use: http://www.asme.org/about-asme/terms-of-use

I I I I I'

elastic response
rigid response
S

-^
N
in

'I

I I I I I I I I I I I

elastic response
rigid response

2.0

B 1.0

1-0

.2 0.0
4-)

4)

-2.0

-2.0

'
'

b)

' '

' t ' ' ' I '

' I ' I ' ' I ' ' I I ' '

elastic response
rigid response

B
0.0

elastic response
rigid response

X 2.0
S

1.0

0.0

ii4) -1.0

I I ' ' I ' I

b)

-2.0
I

0.0

0.2

0.3

. .

_9ft

. . I . . .

0.2

0.4

Time (sec)
Fig. 4 Response of the vehicle model with suspension with two-state
dampers and output feedback (soft mode) to the impact bump input: (a)
Rigid-body control; (b) output feedback control

Arst
second
beaming beanung
(17.2 Hz) (30.x H)

passive fiim
.. passive soft
- semi'active output feedback
TTI I

2 10-'
800 p r r "Tr-|"-r"i"'f'Tni'T "I' i i|'t' t-y-r^r-i^t-ry

0.4

"^'S- Response of the vehicle model to the impact bump input for the
semi-active soft suspension with output feedback control and: (a) A time
oel^V * 2 ms; (b) a time delay of 4 ms

present. Therefore, the characteristics for semi-active systems


presented here cover only this frequency range.
Figures 7, 8, and 9 show the frequency responses of the semiactive systems with output feedback, with rigid-body control,
and with an active absorber, respectively. The characteristics
for the passive firm and passive soft systems are also provided
for reference. The characteristics for semi-active systems exhibit fluctuations at frequencies above 20 Hz, which have been
observed in earlier studies (Butsuen and Hedrick, 1989). In the
frequency range 3-20 Hz the performance of the semi-active
systems is essentially the same as that of the corresponding
active systems. Around the (rigid) body resonance and for high
frequencies, their performance is worse. All the semi-active

a)

0.3

Time (sec)

10

' " I

10'

Irr-nrri

10'

Frequency IHz)

i 'irr-

Fig. 7 Frequency response in terms of the acceleration at point B to


the front wheel velocity input for the passive suspensions and the semiactive soft suspension with output feedback

flit
second
beaming beaming
{17.2 H J ) (30.1 Hi)
L\

-A.! I

800 IIII I I I "1 "t "? j - T I I "II' 1 -I " ' i T ' i - I -r I I )

- passive Arm
passive soft
- semi-active rigid body control
-TTTTl
2H10''

Fig. 5 Control forces generated while traveling over the impact bump
for semi-active soft suspensions: (a) With output feedback; (b) with
proof-mass actuator and output feedback

524 / Vol. 118, SEPTEMBER 1996

10'

10'
Frequency (Hz)

10'

Fig. 8 Frequency response in terms of the acceleration at point B to


the front wheel velocity input for the passive suspensions and the semiactive soft suspension with rigid-body control

Transactions of the ASME

Downloaded From: http://asmedigitalcollection.asme.org/ on 12/16/2015 Terms of Use: http://www.asme.org/about-asme/terms-of-use

press structural vibrations. The demands on the speed of response placed on hardware are, however, very high. The
performance of any semi-active system deteriorates rapidly
when time delays in the control loop increase. This is particularly pertinent to the control of structural vibrations of the body.
When more emphasis is placed on road holding in the synthesis of suspension controllers, as required for cornering or braking maneuvers, a satisfactory trade-off between ride comfort
and wheel tracking performance is difficult to achieve without
feedback of the tire deflection. Consequently, the performance
of semi-active suspensions deteriorates to the level of a good
passive system. A reasonable solution is therefore to rnaintain
constant firm damping during an entire maneuver instead of
varying the damping coefficients in real time.

iiist
second
beamiag beuning
(17.2 Hr) (30,1 Hz)

-is, I

pRBsive firm
.. passive soft
- semi-active with absorber
"I I I T I

2HI0'

10

1II

1 1 r 11

10'
Frequency tHzl

10'

Fig. 9 Frequency response in terms of the acceleration at point B to


the front wheel velocity input for the passive suspensions and the semiactive soft suspension with absorber

Acknowledgments
The authors wish to thank Alex Alexandridis of Vehicle Systems Research Department for valuable help and suggestions
and to acknowledge Alan G. Lynch of Engineering Center for
providing the modal data. Financial support for this work by
the National Science Foundation under grant MSS-9213539 and
by General Motors Research Laboratories is gratefully acknowledged.

systems are characterized by better control of the sprung mass


mode than both passive suspensions and provide low damping
in the intermediate range ( 3 - 8 Hz) for good ride comfort.
The semi-active system with an absorber and that with output References
feedback have very similar characteristics, except for the resoAhmadian, M., and Maiijoram, 1989, "On the Development of a Simulation
nant frequency for the second beaming, where the controller Model for Tractor Semitrailer with Semi-Active Suspension," Advanced Automowith an absorber is slightly more effective. Both systems reduce tive Technologies 1989, Karmel, A. M., Law E. H., and Velinsky, S. R., eds.,
the first beaming resonance significantly, but the resonance cor- ASME Winter Annual Meeting, San FrancLsco, CA, Dec. 10-15, pp. 281-289.
Allen, R. R., and Karnopp, D, C , 1975, "Semi-Active Control of Ground
responding to the second beaming is much more pronounced
Structural Dynamics," Proc. AIAA/ASME/SAE 16th Structures, Structhan for the corresponding active systems. This observation is Vehicle
tural Dynamics and Materials Conference, Denver, CO, pp. 1-11.
consistent with the time domain results shown in Table 1. For
Butsuen, T., and Hedrick, J. K., 1989, "Optimal Semi-Active Suspensions for
the semi-active system with rigid-body control the amplitude Automotive Vehicles: The \ Car Model,'' Advanced Automotive Technologies,
of the response at the natural frequency of the second beaming A. M. Karmel, E. H. Law and S. R. Velinsky, eds., pp. 305-319.
Choek, K. C , Loh, N.-K, McGee, H. D., and Petit, T. P., 1985, "Optimal
mode is larger than for the passive firm suspension.
5

Conclusions
The purpose of this study was to develop and evaluate control
methods that are capable of reducing high frequency harshness
and the resulting structural vibrations of the vehicle body that
are observed in vehicles with semi-active suspensions. It was
shown that for vehicles with semi-active suspensions and rigidbody based controllers, structural vibrations may severely degrade ride comfort as expressed by the acceleration levels of
vehicle body. Consequently, contrary to the predictions based
on the rigid-body models, it is difficult to approach the performance limits of active suspensions with semi-active systems.
For a semi-active system with on-off dampers, application of
control laws based on a flexible vehicle model yields only small
reductions in high frequency harshness; hence it is unlikely
that structural vibration can be significantly reduced (without
sacrificing other aspects of performance) with this type of hardware. When a semi-active suspension with continuously modulated dampers is used, control of structural vibrations can be
improved either by incorporating the feedback of modal variables corresponding to flexible body modes into existing suspension controllers, or by adding proof-mass actuators to sup-

Journal of Dynamic Systems, Measurement, and Control

Model-Following Suspension with Microcomputerized Damping," IEEE Transactions on Industrial Electronics, Vol. IE 32, No. 4, pp. 364-371.
Hennecke, D., and Zieglmeier, F. J., 1988, "Frequency Dependent Variable
Suspension DampingTheoretical Background and Practical Success," Proc.
International Conference on Advanced Suspensions, pp. 101-112.
Hrovat, D., 1993, "Application of Optimal Control to Advanced Automotive
Suspension Design," ASME JOURNAL OF DYNAMIC SYSTEMS, MEASUREMENT,
AND CONTROL, Vol. 115, No. 2 ( B ) , pp. 328-342.

Hrovat, D., MargoUs, D. L., and Hubbard, M., 1988, "An Approach Toward
the Optimal Semi-Active Suspension," ASME JOURNAL OF DYNAMIC SYSTEMS,
MEASUREMENT, AND CONTROL, Vol. 110, No. 3, pp. 288-296.

ISO/DIS 2631, 1972, "A Guide to the Evaluation of Human Exposure to


Whole Body Vibration," International Standard Organization, New York.
Khulief, Y. A., and Sun, S. P., 1989, "Finite Element Modeling and SemiActive Control of Vibration in Road Vehicles," ASME JOURNAL OP DYNAMIC
SYSTEMS, MEASUREMENT, AND CONTROL, Vol. I l l , No. 3, pp. 521-527.

Lizell, M., 1988, "Semi-Active Damping," Proc. International Conference on


Advanced Suspensions, pp. 83-92.
Margolis, D., 1983, "The Response of Active and Semi-Active Suspensions
to Realistic Feedback Signals," Vehicle Systems Dynamics, Vol. 12, pp. 3 1 7 330.
Miller, L. R., and Nobles, C. M., 1990, "Methods for Eliminating Jerk and
Noise in Semi-Active Suspensions," SAE Transaction,'!, paper 902284, pp. 9 4 3 951.
Newland, D. E., 1975, An Introduction to Random Vibrations and Spectral
Analysis, Longman Group Ltd., London.
Rajamani, R., and Hedrick, J. K., 1991, "Semi-Active SuspensionsA Comparison Between Theory and Experiments,'' The Dynamics of Vehicles on Roads
and Tracks; Proc. 12-th lAVSD Symposium, G. Sauvage, ed,, pp. 504-518.

SEPTEMBER 1996, Vol. 118/525

Downloaded From: http://asmedigitalcollection.asme.org/ on 12/16/2015 Terms of Use: http://www.asme.org/about-asme/terms-of-use

Das könnte Ihnen auch gefallen