Alanoly
Research Scientist.
S. Sankar
Professor and Director.
CONCAVE Research Centre,
Department of Mechanical Engineering,
Concordia University,
Montreal, Canada H3G1M8
1 Introduction
X
Suspensions are an integral part of any ground vehicle.
They perform the important task of isolating the passenger
and cargo from terrain-induced shock and vibration. A vehicle
Active force
) generator
suspension unit usually consists of a spring and damper. Normally they are "passive units" because they do not require any
4-'
external power. By using hydraulic or pneumatic power and
(a)
(b)
<>
sophisticated control devices, it is possible to make "active
suspensions" which could give substantially improved perfor- Fig. 1 Schematic of (a) passive, (b) active, and (c) semi-active isolators
mance over passive ones. But they are more complex, expensive, and less reliable than passive suspensions.
input). For the passive system of equation (1), the
A compromise between the active and passive types is the transmissibility, Tis given by
"semi-active" suspension system. In this system, virtually no
external power is required. Desired forces are generated in a
1+
damper by modulating orifice areas for fluid flow. However,
like an active suspension system, this also requires an in(2)
T=strumentation package and control devices. It has been shown
that a semi-active suspension can provide a performance
superior to that of a passive one without the cost and complexFigure 2 shows this transmissibility plotted for several values
ity of a fully active suspension.
of damping ratio, f. This plot illustrates the fundamental performance characteristics of most passive suspensions, both
2 Passive and Active Isolators
linear and nonlinear.
In Fig. 1, a single-degree-of-freedom vehicle model is shown
Based on this transmissibility plot, it can be seen that lower
with (a) passive, (b) active, and (c) semi-active suspensions. damping gives good isolation at high frequencies but poor
The passive system using linear elements has the equation of resonance characteristics. However, higher damping results in
motion
good resonance isolation at the expense of high frequency perx+2fa(x-y) +o>2(x-y) = 0
(1) formance [1].
When active suspensions are used the suspension force can
where
be generated based on any number of control strategies [2].
u>l = k/m and f=c/2\km
Using optimal control theory and a commonly used quadratic
Its vibration isolation can be characterized by acceleration performance criterion it was shown [3] that an optimum
transmissibility (defined as the ratio of the steady-state peak single-degree-of-freedom isolator must generate suspension
response acceleration to peak input acceleration for sinusoidal force as
Fs/m=-2fanx-o>Hx-y)
(3)
leading to a sprung mass equation of motion as
x + 2frx + u2{x-y)=Q
(4)
Contributed by the Design Automation Committee and presented at the
Design Engineering Technical Conference, Columbus, Ohio, October 5-8,1986,
The
optimum
value
of
damping
ratio,
f,
is
1/V2.
The
of THE AMERICAN SOCIETY OF MECHANICAL ENGINEERS. Manuscript received at
transmissibility of this system is given by
ASME Headquarters, July 9, 1986. Paper No. 86-DET-28.
c
['-)]>
10.0
=ML
-A
1
! \ > -5=0.1
\ \
1.0
1.0
0.25
0.5
0.707( optimal)
1.0
5-1.0
0.7070.1
0.5
0.25
0.1
-|
0.1
IC
K^
Vtt,
^s
J~l\"
0.0
\
0.001
\
0.01
1.0
0.1
10.0
1.0
ai/co
Fig. 2
T=-
Fig. 3
[-)H*3 :
(5)
*-
2fax,
x(x-y)>0
0,
x(x-y)<0
(6)
u)/u
10.0
3 Semi-Active Isolators
F.-
V-
F1,-
2fa(x-y),
x(x-y)>0
(7)
x(x-y)<0
0,
The difference between equations (6) and (7) is that, in the
latter, the force is proportional to the relative velocity across
the damper. Thus this scheme can be implemented using an
on-off damper. Experimental results reported indicate that the
on-off semi-active isolator is superior to passive one.
Rakheja [22] and Rakheja and Sankar [23] have proposed
an on-off semi-active damper based a different control
scheme. The control requires the measurement of relative
velocity and relative displacement, both of which are directly
measurable even in vehicle applications. The scheme is
C2^u{x-y)\x~y\,
(x-y)(x-y)<0
F'd = \
(8)
\2$2u>n(x-y)\x-y\,
(x-y)(x-y)>0
We will refer to the three semi-active control schemes,
represented by equations (6)-(8), as Type 1, Type 2, and Type
3 semi-active suspensions, respectively. In all these cases, the
damper force is determined by one of two expressions depending on the sign of a certain function. This function will be called the "condition function." Types 1 and 2 are based on the
same condition function. The former employs continuous
control of damper forces while the latter uses on-off control.
Type 3 is an on-off semi-active damper based on a different
condition function [refer to equation (8)]. A continuous con-
1
v
5.1 Simulation. The equations of motion of the semiactive suspension system described in equation (9) have step
1 discontinuities and hence are nonlinear. A direct and fast way
of solving the system equations is by computer simulation.
1 This is the approach used by previous investigators [4, 22]. In
simulation, the differential equations are solved as an initialvalue problem for a given set of parameters. When a frequency domain performance measure is desired (such as
' transmissibility), the simulation is carried out at each discrete
frequency. The simulation proceeds in time until a steady state
is reached. The response variables are then evaluated and the
process is repeated for the next frequency.
During numerical integration, the discontinuities in the
equation require special treatment. Otherwise the integration
algorithm will be very inefficient and will cause inaccurate
results [24]. Several researchers have studied the problem of
numerical integration of ODE's with discontinuities [24-27],
Their basic approach is to incorporate a zero-finder in the integration procedure. At each integration step, a test is performed to detect the change in sign of the condition function.
If no change occurs, the integration proceeds. Otherwise a
zero-finder determines accurately the time, r*, when the condiFig. 4 Steady-state response of Type 4 system at (a) u/w = 0.5,
tion function is zero. The integration then stops with the old
(b) w/u = 1.0, and (c) u/u = 5.0; z = x - y
set of system functions and restarts with the appropriate new
set. In the present study, a fourth-order Runge-Kutta method
trol scheme based on this condition function is the new semi- was used for numerical integration in conjunction with a
active concept being proposed in this paper. This scheme will bisection algorithm as the zero-finder.
be referred to as a Type 4 system.
5.2 Bifurcation Analysis. Bifurcation, or branching of
solutions, is a phenomenon that occurs in nonlinear systems.
Some engineering problems where this occurs include
4 Type 4 Semi-Active Isolator
nonlinear oscillators, such as, Duffing spring [28], impact
It is observed that the damping force in a passive damper oscillators [29, 30], buckling of structures, etc. [31, 32].
tends to increase the acceleration of the sprung mass during
In the case of a Type 4 semi-active system, it was observed
part of a vibration cycle. This happens when the spring force in simulation runs that the periodic solutions did not exist for
and the damper force have the same direction. Therefore it is certain values of parameters. This suggested a bifurcation
proposed that a semi-active damper be used which gives zero problem. It is not easy to examine bifurcation and stability
(or a very low) damping during this part of the cycle. When problems using simulation techniques. Therefore the study
the spring force and the damper force are in opposite direc- was done with the help of a bifurcation analysis software
tions, the damper can generate a force with the same package called AUTO [33, 34]. AUTO solves the system as a
magnitude as the spring force but opposing it. This way the boundary-layer problem and determines stability of periodic
net force acting on the mass will be zero for this part of the solutions under perturbations. Details regarding the use of this
vibration cycle. This idea is the basis of the Type 4 semi-active software for semi-active suspensions can be found in [35].
scheme.
Thus for a Type 4 system
5.3 Performance Characterization. Since the semi-active
system is nonlinear with step changes in damper force, the ac2
f -aco (x-y),
(x-y)(x-y)<0
celeration response will have discontinuities. Previous reF'd = \
(9) searchers [5] have used displacement transmissibility (ratio of
L
0,
(X-y)(X-y)>Q
the peak displacement response to the input amplitude) to
where a is the gain.
characterize the suspension performance. Since the human
As in the case of the Type 1 system [4], the possibility of body or a suspended mass is sensitive to inertial forces, the
damper lockup exists when (x y) (xy) = 0. In this situa- characterization in terms of acceleration would be more aption, two special cases arise. In the first case (xy) = 0, in propriate. In this study, the ratio of the root-mean-square
which case, the damping force F'd = 0. The second case occurs (rms) value of the response acceleration to the rms-value of the
when (x y) = 0 and (xy) ^ 0. In this case, the system will input acceleration is taken as the transmissibility.
lockup if the desired force Fd, (F'd = aoi2, (x y)) is greater
than the lockup force, which is
6 Results and Discussion
Fd=-y-o>2(x-y)
(10)
Figure 4 shows the steady-state response of a Type 4 system
It is possible to implement this scheme using a servoactuator at 3 frequencies for a = 1. At oi/o> = 0.5, Fig. 4(a), the
to control the damper orifice. This would be similar to the im- damper operates in three phases: (a) exactly opposing the
plementation of a Type 1 semi-active damper reported by spring force, (b) zero force and (c) lockup force. The lockup
Boonchanta [20]. From the control point of view, the obvious condition occurs only at low frequencies. This type of system
244/Vol. 109, JUNE 1987
Table 1
10.0
0.5
1.366
1.02
1.474
2.55
1.857
3.96
1.86
6.384
1.88
10.0
2.00
. Acct leration
r1.0
Vel(icitj
-]?*
Displacement-
\\
"A>\
0.1
10.0
0.01
0.1
1.0
10.0
m/m
/^Acceleration
-s
H>?
'"^.-Disp , V e l .
<s
>
^\
t\
L \_
y>
1.0
-~
\\
\\
1
i
t r P a ss i v e',
\ V Pas s i v e ,
\\
\\\
c=0.1
i i I 1
c=1.0
1- Type 4,a=l .0
0.1
\' 4 xs.
>-V
^ Av s N
e=0. 707 - P / A v^
\ \
\s\ ^-
s.
Active,
V"
\J r^ \
^
1.0
0.1
\ -\l \
\
\x
*
S;
v^-I~
\\
\ \ \
0.01
0.1
1.0
to/to
0.01
10.0
has a unique behavior in that for nearly half of the time the
mass acceleration is zero. This, in turn, leads to good isolation
characteristics.
Figure 5 shows the rms transmissibility of a Type 4 system.
The acceleration transmissibility is higher than the displacement and velocity transmissibilities. This will in fact be the
case for most semi-active systems because of the discon-
Fig. 7
1.0
w/w
10.0
1
i
.. 1
1
1
f\
i
/
/
a=0.
__
1.0"
1.3
\J>
S,^5
-L-NNA
" " T r
V
X1.
X\
L
V"T
r3,\
U-V
0.1
1.0
to/id
10.0
bound on the parameter, a, for stable periodic solutions to exist. For higher values of a, the solutions would not persist
under small perturbations. Table 1 shows the upper bound on
a for some frequencies.
The system is stable throughout the frequency range for a
= 1.3. Simulation was carried out with a = 1.3. Figure 7
shows the transmissibility plot. It is obvious that this value of
a gives a better performance than the unity gain system, Fig.
5. The rms acceleration transmissibility has a peak of 1.19, but
outperforms even active suspensions at high frequencies.
Figure 8 shows the transmissibility for various values of the
gain a. As a is increased, the resonant peak decreases and the
isolation performance improves. However, there is a limit on
maximum a for stability as established earlier.
Figure 9 shows the rms acceleration transmissibility of a
Type 1 system for various values of the damping ratio, *. It
can be seen that higher values of f improves isolation at high
frequencies. But this also leads to deterioration at the very low
frequency end due to the sharper discontinuities in the damper
force. It was noted in [5] that as f -~ oo, the high frequency
performance approaches that of an active isolator with u =
0.6 -4kJrn and f = 1.0. This performance would be superior
to that of a Type 4 isolator with a = 1 . 3 . However, for more
reasonable, finite values of f, a Type 4 system performs better
than a Type 1 system. It is also important to point out that the
main advantage of a Type 4 system over Type 1, or its fully active counterpart, is that Type 4 control logic only involves
directly measurable variables such as relative velocity and
relative displacement.
7
10.0
Conclusion
Active suspensions can achieve the kind of performance not
possible with passive ones. However, their cost and complexity, together with requirements on power and increased weight,
limits their applications. This is especially true in the area of
vehicle suspensions. Semi-active suspensions have been proposed as a compromise between active and passive suspensions. They do not require much power and are less complex
than active suspensions. The new concept in semi-active vibration control proposed in this paper is shown to give performance comparable to that of a fully active isolator. The control scheme's main attraction is that it uses only relative velocity and relative displacement measurements, both of which are
easily measured in vehicle suspensions.
_.
- C=0.25
==s^
v
C=l.
0.5
0.707
-'v7^~
A^ ~S\\
\\ \
N
^L
References
. .-
-"~
VT
X
0- 1
1.0
u/ai
10.0