Vous êtes sur la page 1sur 92

Rev.

SAWE Paper No. 3528


Category Number 31.0

MASS PROPERTIES AND AUTOMOTIVE LATERAL ACCELERATION


By
Brian Paul Wiegand, P.E.

For Presentation at the


70 Annual International Conference
of the
Society of Allied Weight Engineers, Inc.,
Houston, TX, 14-19 May 2011
th

Permission to publish this paper, in full or in part, with


credit to the Author and to the Society, may be obtained by request to:

Society of Allied Weight Engineers, Inc.


P.O. Box 60024, Terminal Annex
Los Angeles, CA 90060

The Society is not responsible for statements or opinions in


papers or discussions at its meetings. This paper meets all regulations for public
information disclosure under ITAR and EAR

Rev. B

SAWE Paper No. 3528


Category Number 31.0

TABLE OF CONTENTS
Chapter:

Page:

Table of Contents ... i


Abstract ......... ii
1 Introduction..... 1
2 Tire Behavior: Lateral Force Generation ..... 3
3 Weight Transfer Along an Axle: A Two-Dimensional Model . .. 5
4 Weight Transfer Between Axles: A Three-Dimensional Model .................... 11
5 Weight Transfer Between Axles: A Sprung Model .. ..14
6 The Transient Condition ...

...... ..... 22

7 The Steady State Condition 30


8 Rollover . 35
9 Tire Behavior: Slip Angles . .40
10 Directional Stability .......... 50
11 Safety ........ 60
12 Conclusions ................

62

References................ 71
Authors Biographical Sketch.......... 73
Appendices.. ....... 74
A Symbolism.. ......75
B Lateral Acceleration Program ...

.....82

C Steering .... .. 84
D Derivation of Equation 3.5 .. 86
E Roll Stiffness Determination ..87

Rev. B

SAWE Paper No. 3528


Category Number 31.0

ABSTRACT
There are a number of automotive performance aspects which are associated with
accelerations in the lateral direction: maneuver (transient and steady state), roll-over, and
directional stability. For each of these automotive performance aspects certain mass property
parameters play significant roles; it is the intent of this paper to make explicit exactly how
those mass property parameters affect each of those automotive performance aspects.
With regard to maneuver, the maximum lateral acceleration which can be attained in
steady-state turning is an important index of performance and safety. The obtaining of high
maximum lateral acceleration levels has inherent vehicle weight and center of gravity
(longitudinal, lateral, and vertical) implications. However, before attaining a steady-state
condition, a turning maneuver must first go through a transient phase. When the transient
phase is included in the full maneuver picture, the previous list of significant vehicle mass
properties parameters acquires two more members: the mass moments of inertia about the
roll and yaw axes.
For modern passenger vehicles, the lateral acceleration point at which roll-over can
occur is generally at a level significantly greater than the maximum lateral acceleration. That
is, a modern car will tend to slide out of control long before there is a possibility of overturn.
Accidents involving rollover generally occur because the vehicle was flipped by obstacles
in the roadway, not because the vehicle traction was great enough to reach the critical lateral
acceleration level. However, the level at which rollover could occur is still an important index
of safety, and the most significant mass property for the determination of that level is the
vertical center of gravity.
Lastly, there is the matter of directional stability, which has to do with the lateral tire
traction force balance front-to-rear, and the front-to-rear drift angle relationship of the
vehicle tires due to those forces. The lateral force/drift angle relationship is dependent upon
normal load, so the most significant mass properties with regard to directional stability are
the vehicle weight and static longitudinal and lateral weight distribution.
However, the static normal loads are dynamically modified in response to lateral
directional disturbance forces. Such disturbances generate lateral initial inertial reactions at
the vehicle c.g.; the consequent roll moment not only causes lateral changes in the normal
load distribution, but also longitudinal changes due to the front-to-rear suspension roll
resistance balance. Such changes readjust the initial lateral force/drift angle relationship frontto-rear, and thereby affect the lateral inertial reaction. If this reaction augments the effect of
the original disturbance, then the vehicle is termed unstable or oversteering; if the reaction
is such as to diminish the effect of the original disturbance, then the vehicle is termed stable
or understeering. Therefore, for directional stability, the primary mass property parameters
are the vehicle weight, and total weight distribution (longitudinal, lateral, and vertical).
ii

Rev. B

SAWE Paper No. 3528


Category Number 31.0

1 - INTRODUCTION
Maneuver, the case of an automobile undergoing directional change, is a situation of
general plane motion: translation plus rotation. Initially, as a vehicle begins a directional
change, the situation is as depicted in Figure 1.1:

Figure 1.1 GENERAL PLANE MOTION: AUTOMOBILE IN TURNING


Note that the radius of the turn is to be considered large enough so that simplification,
by ignoring the angularity which would require resolution of the forces into X and Y
components, is plausible; therefore the forces producing acceleration are to be considered
essentially purely lateral in orientation.
For this case, the principle of dynamic equilibrium requires the following relationships
between forces, moments, and the accelerations produced thereby:

TRANSLATIONAL:
ROTATIONAL:

= +

(EQ. 1.1)
(EQ. 1.2)

Equation 1.2 shows that the turning situation involves mass properties other than just
weight and center of gravity; the yaw rotational moment of inertia (I) will have an
important effect on the turn-producing forces (F f and F r ) whenever there exists some
appreciable angular acceleration (). Such angular acceleration is associated with the
1

Rev. B

SAWE Paper No. 3528


Category Number 31.0

transient phase of initiation or termination of a turn, or with application of the accelerator or


brakes in a turn, or with a turn of varying radius. The transient condition will be dealt with at
length in Chapter 6.
Consider for the moment only the steady-state condition of constant angular velocity;
it is in this steady-state condition (or as close to it as can be reasonably approximated on a
skidpad) that the maximum lateral acceleration level is to be obtained. Therefore, in this
limited case, the matter reduces to just a consideration of how the lateral forces are influenced
by the weight and center of gravity (no I term) 1:

= = +

Substitute W t /g for m, and V2/R for a:


= +

The dynamic equilibrium also requires a moment balance about the CG:

= =
=

Substitute this expression for F r in the force equation and do a little manipulation:

= + = + =
=

Now solve for F f , and F r , then substitute W f for W t l r /l wb , W r for W t


l f /l wb , and a y for V2/gR:

(EQ. 1.3)
(EQ. 1.4)

So, for steady-state turning the lateral tire traction force(s) at each axle need only
equal the weight load at each axle times the lateral acceleration in gs. However, the
generation of those tire forces is a bit more complex than might first be supposed; the matter
is not simply a case of applying Coulombs friction law
1

Reference [1], pp. 199-201.

Rev. B

SAWE Paper No. 3528


Category Number 31.0

2 TIRE BEHAVIOR: LATERAL FORCE GENERATION


Automobiles produce all primary direction controlling forces at the tire/road interface.
As noted, the force generation is not necessarily in accord with Coulombs Friction Law: F =
N. Because of the nature of rubber pneumatic tires, the tractive force (F) and normal load
(N) relation is nonlinear. Empirical studies show that for a tire the lateral traction
coefficient is itself a function of the normal load as per Equation 2.1 2:

(EQ. 2.1)

The coefficients b and m are particular to the type of tire concerned. The b
coefficient is the basic coefficient of traction and is dependent upon the type of tire material
and road surface and is directly proportional to the magnitude of the contact area. The m
coefficient is a measure of the decrease of contact area due to tire distortion under lateral load
(and, therefore, is an inverse measure of tire structural stiffness). Such contact area decrease,
due to distortion, leads to decreased lateral force generation potential from what otherwise
may be expected. The nature of this distortion is depicted in Figure 2.1:

Figure 2.1 CONTACT AREA DECREASE UNDER LATERAL LOAD


2

Reference [11], page 127. Longitudinal traction force potential also decreases under increasing normal load,
but in accord with a somewhat different mechanism; for more info see Reference [7], page 57.

Rev. B

SAWE Paper No. 3528


Category Number 31.0

Combining Equation 2.1 and Coulombs law by substitution for results in the
normal load/potential lateral force relationship of Equation 2.2:

= ( )

(EQ. 2.2)

Graphically, this function may be depicted for a typical set of coefficient values as
shown in Figure 2.2:

Figure 2.2 NORMAL LOAD / LATERAL TRACTION RELATIONSHIP


Knowledge of this function is basic but cannot, by itself, be used to determine, even
roughly, the maximum lateral acceleration potential of a vehicle because there is at least one
very significant modifying factor: weight transfer 3 in a turn. This matter of weight
transfer will bring to the fore the role of the c.g. in determining lateral acceleration.

Weight transfer is a bit of a misnomer; some say a more appropriate term would be load transfer.

Rev. B

SAWE Paper No. 3528


Category Number 31.0

3 WEIGHT TRANSFER ALONG AN AXLE: A TWO-DIMENSIONAL MODEL


The lateral force generation potential for an axle is essentially only a matter of adding
the lateral force generation potentials of the tires. This should be readily discernable from the
axle force Equation 3.1 4:

= ( ) + ( )

(EQ. 3.1)

In the static case the normal loads would be equal, N i = N o . However, it is not the
static case, but that of dynamic equilibrium in a steady-state turning situation, in which we are
interested. In such a case, a weight transfer moment occurs which alters the lateral force
generation potential by decreasing the normal load on the tire closest to the turn center (inner
tire) and increasing, by an equivalent amount, the normal load on the tire furthest from the
turn center (outer tire). This situation is as depicted in Figure 3.1:

Figure 3.1 LATERAL WEIGHT TRANSFER IN TURNING

Reference [11], page 127.

Rev. B

SAWE Paper No. 3528


Category Number 31.0

From the information given in this figure, expressions for the normal loads, including
the effect of weight transfer for the inner and outer tires, may be determined: Equations
3.2 and 3.3, respectively:

(EQ. 3.2)
(EQ. 3.3)

The quantity W a y (h cg /t) is the weight transferred. Substituting N i and N o


from Equations 3.2 and 3.3 into Equation 3.1 for N i and N o , respectively, produces the
two-dimensional model Equation 3.4 which gives us the axle lateral traction force potential
taking weight transfer into account:

+
+

(EQ. 3.4)

From this equation, further equations for the maximum axle lateral acceleration limits
of slide and overturn 5 can be determined, Equations 3.5 and 3.6, respectively 6:

( )

(EQ. 3.5)

(EQ. 3.6)

Reference [15], Section V.A. Reference [7], page 311.


For derivation of Eq. 3.5 see Appendix D. The value for the overturn acceleration is also approximately (given
the simplifications inherent in the analysis) equal numerically to the coefficient of traction at overturn. Reference
[12] therefore suggests that the greatest coefficient of traction anticipated to be encountered in vehicle use (the
coefficient depending on tire and road surface) be used as a design criterion to establish t and/or hcg so
that it would be impossible for the vehicle to rollover from traction forces (comment: this tends to occur
regardless for passenger vehicles). Note, the result of Eq. 3.6 is called the rollover threshold by Reference [7].

Rev. B

SAWE Paper No. 3528


Category Number 31.0

Examination of a plot of Equation 3.4 for typical variable values, such as Figure 3.2,
allows for greater understanding of the mechanism of the lateral traction force phenomena:

Figure 3.2 LATERAL TRACTION POTENTIAL W/ WEIGHT TRANSFER,


SINGLE AXLE
Note that as lateral acceleration levels increase, the lateral traction force potential
decreases. This is due to the effect of weight transfer. Simultaneously, the inertial loading,
or the force required to achieve a y , increases. At some lateral acceleration level, the
potential and inertial forces will be equal, i.e., the function plot lines will intersect (point
A). After this point, dynamic equilibrium can no longer be maintained and slide sets in
(a y = 0.7 gs).
Note that slide would not have set in until a significantly higher acceleration level
(a y = 0.8 gs, point B) had the phenomenon been one of lateral traction force potential
without weight transfer, i.e., if the tire loadings stayed even. Also note that for these typical
variable values, the lateral acceleration limit of overturn is much larger (a y = 1.4 gs, point
C) than that of slide. In such a case, the slide acceleration is truly the maximum lateral
acceleration possible.
7

Rev. B

SAWE Paper No. 3528


Category Number 31.0

The point of intersection of the traction potential and inertial loading functions, i.e.,
the slide acceleration, is given us directly by means of Equation 3.5. Observe what happens if
this equation is plotted for increasing total (axle) weight as in Figure 3.3:

Figure 3.3 SLIDE AND OVERTURN LATERAL ACCELERATION vs. AXLE LOAD
It would seem that not only is it beneficial to keep the tire loadings even (i.e.,
minimize weight transfer), but also to keep the loadings as light as possible in order to
achieve the highest maximum lateral acceleration (slide) levels.
The reason for this last observation is that while increasing weight does produce
increasing lateral traction potential, in accord with the non-linear Equation 3.4, the inertial
loading also increases in accord with the Newtonian law F=ma. In other words, with
increasing weight, inertial forces increase more rapidly than lateral force generation
capacity; this is due to the decreasing traction factor b mN.

Rev. B

SAWE Paper No. 3528


Category Number 31.0

Plotting Equations 3.4 and 3.5 while varying the center of gravity height produces
Figure 3.4:

Figure 3.4 SLIDE & OVERTURN ACCELERATION vs. C.G. HEIGHT


Note that on this figure, not only does the slide lateral acceleration level decrease with
increasing h cg but the overturn acceleration drops also. This is because h cg is the weight
transfer and overturn moment arm.
For the variable values used, the overturn acceleration level drops so precipitously that
at about an h cg value of 60 inches, the overturn acceleration curve intersects with the slide
acceleration curve. For yet higher c.g. values, the overturn acceleration would be the
maximum lateral acceleration limit as it would be lower in value than the slide acceleration
level; this would be a most unsafe condition.
For most vehicles, and certainly any modern passenger car design, such an extreme
situation need not be of concern; the slide acceleration is normally the maximum lateral
acceleration limit. The purpose of Figure 3.4 is to point out the fact that, in order to obtain
maximum lateral acceleration levels, the c.g. height should be kept at a minimum.

Rev. B

SAWE Paper No. 3528


Category Number 31.0

From this consideration of the physics of a single axle, the next step towards the
reality of a conventional four-wheeled vehicle configuration would be to consider the case of
two axles in tandem.

10

Rev. B

SAWE Paper No. 3528


Category Number 31.0

4 WEIGHT TRANSFER BETWEEN AXLES: A THREE-DIMENSIONAL MODEL

A conventional vehicle has two axles in tandem, and its maximum lateral acceleration
level is the acceleration level of whichever axle reaches its slide point first. To determine the
vehicle maximum lateral acceleration level, therefore, an equation for lateral force generation
must be written for each axle. The portion of vehicle weight that can be assigned to each axle
equation is determined from the total vehicle weight and the longitudinal center of gravity.
Using the notation of Figure 1.1, the weights (axle loadings) to be apportioned to the front
and rear axle are to be in accord with Equations 4.1 and 4.2:

= ( )

(EQ. 4.1)
(EQ. 4.2)

Using the same values for weight (4000 lbs. total vehicle weight, 2000 lbs. per axle
initially), c.g. height, track, and tire coefficients as were used to generate all previous figures,
the effect of varying the longitudinal c.g. on the maximum lateral acceleration (slide) may be
seen from Figure 4.1:

Figure 4.1 SLIDE ACCELERATION vs. LONGITUDINAL WEIGHT


DISTRIBUTION
11

Rev. B

SAWE Paper No. 3528


Category Number 31.0

Note in Figure 4.1 that as the c.g. moves rearward from the 50/50 position, the vehicle
maximum lateral acceleration decreases, as the rear axle goes into slide at lower and lower
acceleration levels. For instance, at 35/65 the maximum lateral acceleration level that could be
attained by the vehicle would only be 0.637 gs, the rear axle governing.
If this were the case with an actual vehicle, then the logical thing to do would be to
change the rear axle parameters so as to acquire a higher rear axle slide point. Due to the
limitations inherent in this paper there are various suspension changes that wont be
discussed, but changing the rear axle tire behavior so that m is 0.00035 instead of 0.0004 is
well within the scope of this paper, and could possibly be accomplished by a wider wheel,
higher tire inflation pressure, a different tire size and/or type, etc. Any such changes that
would make m equal 0.0035 would also make the 35/65 vehicle competitive with the
previous 50/50 vehicle, with a maximum lateral acceleration of 0.69 gs for each.
However, if the same modifications leading to an m of 0.00035 were carried out at
both axles for the 50/50 weight distribution then that vehicle configuration would corner at a
maximum lateral level of 0.745 gs! The conclusion drawn from this is that, if all other things
are equal, then the 50/50 weight distribution vehicle will always attain a higher lateral
acceleration than a vehicle of uneven weight distribution. This is the rationale behind the
commonly encountered statement that a 50/50 weight distribution is the ideal for vehicle
design. Note that this is analogous to the findings for a single axle: the maximum lateral
acceleration level is highest when the tire loadings are as even as possible (i.e., without
weight transfer, point B on Figure 3.2).
In the light of the above, some of the results of a comparative testing reported in an
August 1983 issue of Road & Track should not be surprising. The 1984 Corvette and the 1984
Porsche 944, when at driver onboard test condition weights of 3390 lb (1537.7 kg) and 2940
lb (1333.6 kg) respectively, were found to have exactly a 50/50 longitudinal weight
distribution; in this condition on the skidpad these vehicles attained maximum lateral
accelerations of 0.842 and 0.821 gs, respectively 7.
Given such apparent adherence to a 50/50 weight distribution among sports cars, one
would expect all race cars to be of a 50/50 weight distribution, but that is not the case.
Group Seven (a former racing classification sanctioned by the Fdration Internationale de
lAutomobile) vehicles tended to have longitudinal weight distributions of about 45/55 to
40/60 8. This was the result of performance considerations other than lateral acceleration being
factored into the design such as braking and longitudinal acceleration. Also, the mathematical
model from which the 50/50 longitudinal weight distribution ideal was obtained was a very
7

Reference [20], page 58. Presumably the skidpad was of 100 foot (30.5 m) radius. It is also assumed that the
Porsche 944 was a 1984 model as it was not so stated.
8
Reference [13], pages 20-23.

12

Rev. B

SAWE Paper No. 3528


Category Number 31.0

simplified model that did not include the modifying effects of steering, transients,
aerodynamics, and directional stability. When all of that is factored in, then a design can stray
somewhat from the ideal 50/50 weight distribution, but still not too radically if obtaining
high maximum lateral acceleration levels is of any concern, at least for conventional designs.

13

Rev. B

SAWE Paper No. 3528


Category Number 31.0

5 WEIGHT TRANSFER BETWEEN AXLES: A SPRUNG MODEL

The two dimensional model for weight transfer along an axle as used in Chapter 3
was highly simplified; there was no consideration of the effects resulting from the presence of
a suspension. By design, a suspension allows for deflection under vertical loads, and
consequently there is some deflection under longitudinal loads (dive, squat) and some
deflection under lateral loads (roll). The deflection under lateral loads modifies the weight
transfer results from that obtained using just the previous simple moment balance equations
Eq. 3.2 and Eq. 3.3.
The exact nature of the effects resulting from the roll deflection under lateral load
depends upon the type of suspension, in particular the suspension roll center (RC) height
(h rc ) and roll stiffness (k roll ), and upon the mass property parameters of weight and
center of gravity 9. It is the suspension at the front of the vehicle, and the suspension at the
rear, which acting in concert determines the roll response to a lateral load. To illustrate the
matter, consider Figure 5.1 which depicts the roll axis and c.g. situation of a 1980 Ford Fiesta
S (1.1 liter, European version):

Figure 5.1 1980 FORD FIESTA S (EURO VERSION) ROLL AXIS


The vehicle is in the 4 up condition, which is the curb weight condition of 1609.4
lb (730.0 kg), plus 4 occupants at 165.3 lb (75 kg) each, and a small amount of baggage at
28.7 lb (13.0 kg), resulting in a total vehicle weight of 2299.4 lb (1043.0 kg, GVWR is 1160
9

The effects of damping upon roll are neglected in this analysis. For information on this aspect of roll see
Reference [1], page 83.

14

Rev. B

SAWE Paper No. 3528


Category Number 31.0

kg) at c.g. coordinates of 41.4, 0.0, 22.74 inches (105.2, 0.0, 57.8 cm); the wheelbase is 90 in
(228.6 cm), and in this condition the front/rear longitudinal weight distribution is 54/46.
The total unsprung weight is determined to be 161.0 lb (73.0 kg) at c.g. coordinates of 45.0,
0.0, 9.6 inches (114.3, 0.0, 24.4 cm) 10. From all this some simple weight accounting
establishes the sprung weight and its c.g. coordinates:

The front suspension roll stiffness k froll is 218.3 lb-ft/deg (296.0 Nm/deg), and the
rear suspension roll stiffness k rroll is 154.9 lb-ft/deg (210.0 Nm/deg), which makes for a
total stiffness about the roll axis of 373.2 lb-ft/deg (506 Nm/deg) 11. The front suspension roll
center height h frc is 7.28 inches (18.5 cm), and the rear suspension roll center height h rrc
is 7.52 inches (19.1 cm) 12, which by proportioning makes the roll axis height under the
sprung mass c.g. 7.39 inches (18.8 cm). Looking at the vehicle cross-section through the
sprung c.g., the situation is as depicted:

Figure 5.2 - SPRUNG MASS IN ROLL AT 0.5 gs LATERAL


10

The unsprung mass c.g. is assumed to be at mid-wheelbase longitudinally, at centerline laterally, and at the
rolling radius height of the 155SR12 tires.
11
See Appendix E for roll stiffness determination methodology.
12
Reference [6], pages 191-192.

15

Rev. B

SAWE Paper No. 3528


Category Number 31.0

Given all this information, the roll angle s of the sprung mass under a 0.5 g lateral
acceleration a y can be determined from the fact that the roll resistance (roll angle s times
the roll stiffness k roll ) has to equal the roll moment (lateral force W s a y times the roll
moment arm h r ) at equilibrium 13:

Roll Angle x Roll Stiffness = Lateral Force x Roll Moment Arm


Into this simple equation we can plug all the necessary parameters as drawn from the
previous discussion and solve for s :

s x 373.2 lb-ft/deg = 2138.4 lb x 0.5 g (23.73 in 7.39 in) / (12 in/ft)


s = 3.9 deg
However, it can be seen from Figure 5.3 that there are at least two complications that
makes the roll angle determination not quite so simple. One complication is that the roll
height h r decreases by an amount d z during the rolling action, which would tend to make
the roll angle s less than the 3.9 degrees calculated. Another complication is that the
rolling motion moves the sprung mass c.g. off centerline laterally by an amount d y , which
would tend to make the roll angle s more than the 3.9 degrees 14:

Figure 5.3 ROLL GEOMETRY


13

Reference [3], pages 133 and 136.


Ibid, page 134. The first complication is implicitly neglected, and the second complication is explicitly
neglected.

14

16

Rev. B

SAWE Paper No. 3528


Category Number 31.0

Finally, there is the further complication that the unsprung masses at the front and
rear axles also make their contributions to the sprung roll moment as transmitted through their
linkages (if the suspension is of an independent type, otherwise the unsprung mass moments
are absorbed internally). From this fact, and the geometry of Figure 5.3, a more complex
reality 15 than that initially considered may be expressed 16:

s kroll = W s a y h r Cos s + W s h r Sin s + W usf a y r rf + W usr a y r rr (EQ. 5.1)

Where:

s = The sprung mass roll angle, degrees.


k roll = The total vehicle roll resistance, lb-ft/deg.
W s = The weight of the sprung mass, lb.
a y = The lateral acceleration, gs.
h r = The sprung mass roll moment arm, ft.
W usf = The front axle unsprung mass weight, lb.
r rf = The front axle unsprung mass vertical c.g. (approx. the
rolling radius 17), ft.
W usr = The rear axle unsprung mass weight, lb.
r rr = The rear axle unsprung mass vertical c.g. (approx. the
rolling radius), ft.
Even with all this complication, the resulting value for the roll angle will still just be
approximate as matters such as free surface effect of liquids, lateral shift of the unsprung
mass c.g.(s), and various secondary deflections (including shift of the RCs in roll) are all still
unaccounted for. Still, the roll angle value as determined by Equation 5.1 may have all the
15

The resultant roll angle determination is still approximate, if for no other reason than the fact that the effect of
free surface movement of liquids has not been considered. However, in high performance vehicles sometimes
fuel tank baffles, etc., have been provided to ameliorate free surface effects.
16
Reference [3], page 136. The best (perhaps only) way to solve this equation is iterative.
17
Ibid, page 134.

17

Rev. B

SAWE Paper No. 3528


Category Number 31.0

accuracy that is needed for early design studies. Since this equation does not allow for an easy
analytical solution a numerical solution is recommended. A simple iterative spreadsheet
approach yields 4.6 degrees, which seems reasonable, but probably is a bit low considering
that free surface effects, etc., were not taken into account.
Now that the roll angle s has been reasonably estimated, the weight transfer
effect of roll, the normal loads N i and N o , can be determined at the front and rear axles.
First, lets consider the front IFS suspension which is of the MacPherson strut type. Based on
the sprung weight c.g. location, the front sprung weight W sf is determined to be 1160.2
lb (526.3 kg). That, plus fact that the front suspension roll center height h frc is 7.28 inches
(18.5 cm), the front track is 52.52 inches (133.4 cm), the front unsprung weight W usf is
80.5 lb (36.5 kg), and the front unsprung weight c.g. height r rf is 9.6 inches (24.4 cm),
can all be depicted in the following diagram:

Figure 5.4 1980 FORD FIESTA S FRONT STRUT SUSPENSION IN ROLL


The appropriate equations for N if and N of , taking the sprung mass roll and the
relative roll stiffnesses into account, are now:

(EQ. 5.2)

(EQ. 5.3)

+ /

18

Rev. B

SAWE Paper No. 3528


Category Number 31.0

Plugging the appropriate values into Equations 5.2 and 5.3 produces the following
results:
N if = (1240.7/2) 4.6 (218.3)/(52.52/24) = 161.47 lb, or 73.24 kg
N of = (1240.7/2) + 4.6 (218.3)/(52.52/24) = 1079.23 lb, or 489.48 kg
If Equations 3.2 and 3.5 were used, i.e. if roll stiffness was not considered, the results
for N if and N of would have been:
N if = (1240.7/2) 1240.7 (0.5) (23.73/52.52) = 340.06 lb, or 154.25 kg
N of = (1240.7/2) + 1240.7 (0.5) (23.73/52.52) = 900.64 lb, or 408.52 kg
Now that the front axle has been considered, lets turn our attention to the rear axle,
which is non-independent of the dead beam type. The rear sprung weight W usr is
determined to be 978.2 lb (443.7 kg), the rear suspension roll center height h rrc is 7.52
inches (19.1 cm), the rear track is 52.01 inches (132.1 cm), the rear unsprung weight W usr
is 80.5 lb (36.5 kg), and the rear unsprung weight c.g. height r rr is 9.6 inches (24.4 cm),
as depicted in the following diagram:

Figure 5.5 1980 FORD FIESTA S REAR BEAM SUSPENSION IN ROLL


The appropriate equations for N ir and N or , taking the sprung mass roll and the
relative roll stiffnesses into account, are now:

19

Rev. B

SAWE Paper No. 3528


Category Number 31.0

+ /

(EQ. 5.4)
(EQ. 5.5)

Plugging the appropriate values into Equations 5.4 and 5.5 produces the following
results:
N ir = (1058.7/2) 4.6 (154.9)/(52.01/24) = 200.55 lb, or 90.97 kg
N or = (1058.7/2) + 4.6 (154.9)/(52.01/24) = 858.15 lb, or 389.25 kg
If Equations 3.2 and 3.5 were used, i.e. if roll was not considered, the results for N if
and N of would have been:
N ir = (1058.7/2) 1058.7 (0.5) (23.73/52.01) = 287.83 lb, or 130.56 kg
N or = (1058.7/2) + 1058.7 (0.5) (23.73/52.01) = 770.87 lb, or 349.66 kg
Note that without incurring roll, the little Fiestas normal loads in a 0.5g maneuver
would have been much more even front-to-rear on the heavily loaded outside wheels. The
effect of the Fiestas relative roll stiffnesses, which is the result of the spring rates and
moment arms, is to have much more of the roll moment resisted by the front suspension than
by the rear. Consequently, when roll is taken into account the Fiestas normal loads are much
more skewed toward the front outer wheel. This is by design; the suspension roll centers (roll
axis slanting down towards the front) and spring rates (greater roll stiffness at the front) were
carefully chosen to create this effect.
This may seem peculiar in light of the previous chapters admonition that for
maximum lateral acceleration the normal loads should be kept as even all around as possible.
However, the Fiesta was not a high performance vehicle; what the designers were intent upon
was the obtaining of a high degree of stability for a rather prosaic little grocery getter to be
driven by the semi-conscious general public. The absorption of most of the weight transfer
under lateral load by the front suspension is conducive to maintaining directional stability for
reasons to be dealt with in Chapter 10. Race car drivers, as opposed to the general public, are
assumed to be highly skilled and alert, at least when in competition, and so the compromise
point in design between stability and performance is shifted for racing vehicles toward
performance 18.

18

Race car design may even include a fair amount of directional instability for sharp low speed maneuvers, and a
changeover to increased directional stability for high speed sweeping maneuvers; this has to do with the fastest
way to get around a road race course.

20

Rev. B

SAWE Paper No. 3528


Category Number 31.0

Speaking of performance in the current context of roll is a natural lead-in to the


subject of roll gain. Roll gain is the steady state equilibrium amount of roll, usually in
degrees, per lateral acceleration, usually in gs, as illustrated by Figure 5.6 19; note that the
little Fiesta S has more than double the roll gain (by this authors calculation) of any of the
high performance sports cars shown:

Figure 5.6 ROLL GAIN: DEGREES ROLL PER Gs LATERAL ACCELERATION


The lower the roll gain the less weight transfer due to lateral c.g. shift and the
harder a car can corner without incurring adverse suspension camber angles; the result is
flatter and faster cornering. This can be achieved by increasing roll stiffness, preferably
through the use of anti-roll bars, but general suspension stiffening will also suffice at the
expense of ride quality. However, a better means to this end may be by the reduction of the
roll moment arm by roll center manipulation (good) and/or c.g. height reduction (best) 20.

19
20

Reference [8], pages 37-38.


Reference [1], page 49.

21

Rev. B

SAWE Paper No. 3528


Category Number 31.0

6 THE TRANSIENT CONDITION

The subject of roll gain is closely inter-related with the topic of transient response in
maneuver; there are a number of aspects to the subject of transient response of a vehicle in
maneuver, and roll gain is just one such aspect. As noted in Chapter 1, there is an angular
acceleration associated with the transient condition at the initiation or termination of a
turn, or with the application of the accelerator or brakes in a turn, or with a turn of varying
radius. The yaw inertia of the vehicle I times this angular acceleration represents an
inertial moment which must be equaled by moments generated by forces at the tires in order
for the transition from straight ahead to turning to occur:

The above, where F f l f > F r l r , is the case when the vehicle is initiating a maneuver
as was depicted in Figure 1.1. When the vehicle comes out of the turning maneuver there will
be a less intense shorter lived reversal of this situation, i.e., F f l f < F r l r . Then, as is often the
case in slaloms, chicanes, or lane changing maneuvers, a turning action in the opposite
direction may commence causing a situation of F f l f > F r l r once again, only now the forces
will be pointing in the direction opposite from before!
With all this fluxing of forces and moments inflicted upon a damped spring-mass
system it should not be surprising that there would be some oscillatory behavior observable.
Figure 6.1, which is a plot of vehicle yaw velocity versus time for the transient phase at
the commencement of a turn leading up to a steady state condition, illustrates such yaw
oscillation behavior and its two phases, rise and decay, which sum to the total vehicle
response time:

Figure 6.1 TRANSIENT YAW RESPONSE TO TURN INITIATION

22

Rev. B

SAWE Paper No. 3528


Category Number 31.0

For the same specific vehicles as noted in Figure 6.1, the over plot of transient
responses is illustrated by Figure 6.2:

Figure 6.2 COMPARATIVE TRANSIENT YAW RESPONSE FOR FOUR


VEHICLES
Of course, along with this yaw oscillation there would be some corresponding roll
oscillation 21, and maybe even a little pitch oscillation. All these oscillations result in
fluctuating demands upon the tire contact patches, shifts in suspension linkage, and confusing
sensory inputs to the driver: oscillationis critical to the feel of the car. a high decay
rate might...be described as twitchinessa vehicle that would wiggle around as the driver
turned into a corner 22. It is best that this transient phase be as short in duration (response
time) as possible.
Transient behavior is so important to the feel and capability of a vehicle that General
Motors conducted a special effort during the development of the fourth generation Corvette
(C4) to acquire superior transient behavior with respect to various foreign sports cars
renowned for their handling prowess. Exhaustive testing of a Ferrari 308, a Porsche 928, and
a Datsun 280ZX established the performance targets that GM was determined to beat 23; how
well they did with respect to roll and yaw transients is illustrated in Figures 5.6, and 6.1/2,
respectively.

21

Reference [7], page 319. Scale estimation of the example roll oscillation plot indicates a roll response time of
0.324 seconds for an unspecified vehicle..
22
Reference [8], pages 38-39.
23
Ibid, page 37.

23

Rev. B

SAWE Paper No. 3528


Category Number 31.0

The data for all the sports car figures was obtained by instrumented runs on a skidpad.
For the roll gain measurement of the roll angle and corresponding lateral acceleration (in
gs) was taken after all transient effects had died out and the vehicles were circling the pad
in a steady state condition. For all transient data appropriate steering angle was quickly dialed
in and held constant along with velocity until all fluctuation died out and steady state attained.
Such skidpad testing allowed for the obtaining of data useful in making engineering
determinations. Less rigorous in producing useful data, but more common for vehicle
comparisons with regard to transient behavior, is the slalom. Slalom consists of a course of
evenly spaced traffic cones, or other such items, laid out in a straight line. Such a course
doesnt allow for a vehicle settling in to a steady state condition; running the course keeps a
vehicle in a series of alternating transient conditions. A vehicle is taken on a timed run though
the obstacle course as quickly as possible, often for multiple times in alternating directions in
order to average out any directional (wind, gradient) or driver anomalies. The resultant
average time and/or speed for completion of the course is taken as indicative of how
beneficent or malevolent the vehicles transient behavior is.
As with all test results care must be taken to assure apples-to-apples comparison.
General Motors, Motor Trend, and Hot Rod Magazine all favor a 600 foot (182.9 m) slalom
course, while MIRA and Road & Track 24 favor a 700 foot (213.4 m) course. And, of course,
even when the length is the same there are many other possible variances, like cone
spacing. There just is no universal standard slalom course, although the course depicted in
Figure 6.3 comes closer to that designation than any other.

Figure 6.3 STANDARD SLALOM COURSE


To get the transient characteristics of response time (rise + decay) and roll gain
(degrees/g) for the four sports cars presented earlier GM utilized a skidpad for two reasons:
24

Reference [16], page 16.

24

Rev. B

SAWE Paper No. 3528


Category Number 31.0

1. Skidpad results tend to be less driver dependent.


2. It was necessary to reach a steady-state termination.
However, with proper instrumentation hard data can also be obtained from slalom
course testing. Colin Campbell relates how roll steer and steering lag characteristics can be
obtained from empirical tracking of steering angle and rear wheel slip angle; a plot of such
tracking would look as follows in Figure 6.4 25. Because the front slip angles closely mimic
the front steering angles, and the angles are similar on left and right sides, only the front
steering angle and the consequent rear slip angle are shown for clarity:

Figure 6.4 ROLL STEER EFFECTS


As directional input is made at the steering wheel, the front wheels create steering
angles with respect to the velocity vector; the exact nature of those angles being dependent
on the steering geometry of the vehicle (see Appendix C). The tires, being elastic, distort as
the wheels turn and run at some slip angles to the new direction vector. The slip angle
25

Reference [4], page 58. Reference [7], page 323 shows a similar plot for left side front and rear wheels (tires).

25

Rev. B

SAWE Paper No. 3528


Category Number 31.0

function is proportional to, but has some small lag time to, the steering angle function 26. Slip
angle is a misnomer, like the term shock absorber when used to refer to a damper. The
tires dont actually slip but, through a process of continual deformation at the road contact
patch with rotation, give the impression of slippage; the only real slippage occurs in a small
portion of the total tire/road contact area (as shown). It is for this reason Donald Bastow
suggests the term drift angle in lieu of slip angle 27, which is a practice which this paper
will now adopt:

Figure 6.5 DRIFT ANGLE AND LATERAL FORCE


As the driver turns the steering wheel the front wheels begin to turn, forming a
steering angle with the vehicle velocity and a drift angle builds up as the tire at the
ground contact area distorts. Associated with this distortion is a lateral force vector F y
which begins to push the front of the vehicle away from the original velocity direction toward
the new direction of the front wheels.
This incipient turning action generates a translational lateral inertial reaction through
the sprung mass c.g. and a rotational reaction about the vertical axis through the c.g. It is
these inertial reactions which prevent the vehicle response to steering input from being
26

A reasonable estimate for this time lag might be the time it takes for the wheels to go through a half rotation or
so depending on the type of tire, inflation pressure, and other factors.
27
Reference [1], page 73.

26

Rev. B

SAWE Paper No. 3528


Category Number 31.0

instantaneous; there are a number of small lag time contributions involved: play in the
steering, deflection of the tires, compliance of suspension bushings, etc.; but the vehicle
weight, yaw inertia, and sprung mass inertia are the biggest factors.
Note that it is Case (A) of Figure 6.4 that gives the most precise response to steering
input. In Case (A) the time lag between steering input and rear drift angle is at a minimum,
and the drift angle closely follows the steering angle with only slightly less magnitude (which
is the essence of understeering behavior). The worst response to steering input is Case (D)
which is very imprecise with high lag and a general overreaction; in this case even after the
driver has reversed course and is turning right the rear end is still trying to go left! This is a
case guaranteed to take out quite a few cones on the slalom!
The mass properties palliatives for adverse steering behavior such as typified by Case
(D) is to reduce roll and yaw inertias (and perhaps move the vehicle c.g. forward). The
sprung mass roll inertia reduction would reduce roll gain, keeping suspension linkages
closer to optimum positions and improving ground contact. Reducing vehicle yaw inertia
would decrease response time, both rise and decay portions, and decrease lateral force
demand/fluctuation at the tires. Such reductions and proper balance would do much to ensure
a minimum of drama in the transient phase from steady state straight-ahead to the steady state
turning condition.
It is important at this point to take special note of the aligning moment, or the lateral
force F y time the pneumatic trail d. The effect of this torque is create an automatic
reversal of any driver induced course change 28; when the driver relinquishes the wheel after a
turn it is the aligning moment which causes the wheel to self-center and return course to
straight ahead (more or less). Fluctuating forces at the tires during the transient stage of a
maneuver will not only affect the motion of the vehicle directly, but will cause fluctuations in
aligning moment and consequent behavior of the steering wheel, resulting in greater difficulty
for a driver to control the vehicle 29.
The relationship of the rotational inertia, represented by the radius of gyration K
(I = K2M), to the location of the vehicle longitudinal center of gravity as given by l f
and l r , is neatly encapsulated by the quantity K2/(l f x l r ) which is known as the Dynamic
Index (DI) when K is the pitch radius of gyration 30. In SAE J670e standard symbolism
this quantity is depicted as k2/ab. The values of the yaw radius of gyration and the pitch
28

Some additional aligning moment arm may be added by that suspension characteristic known as castor;
which is the angle of the kingpin to the vertical.
29
At extreme maneuver, i.e. racing conditions, the aligning moment sometimes reverses itself, causing an effect
something like the reversal of controls in WW II era aircraft when inadvertently entering the transonic flight
realm. This is another reason why race car drivers, like fighter pilots, are a special breed.
30
Reference [19], page 55.

27

Rev. B

SAWE Paper No. 3528


Category Number 31.0

radius of gyration are highly interrelated as both derive from the common basis of the
vehicle mass distribution, but K z (yaw radius of gyration) is always greater than K y
(pitch radius of gyration) for two reasons:
1. The vehicle plan view about the yaw axis is always greater
than the vehicle elevation view about the pitch axis.
2. The yaw inertia is generally the inertia of the entire vehicle
mass, but the pitch inertia most commonly referred to is the
inertia of just the vehicle sprung mass.
Just as the DI or K y 2/(l f x l r ) is an important factor in the pitch motion of the sprung
mass, the quantity K z 2/( l f x l r ) (or X) is an important factor in the yaw motion of the
entire vehicle as it determines an oscillation center (OC) about which the vehicle will
initially tend to pivot. For the value of this important yaw motion factor there are three
possibilities 31:
1. K z 2/( l f x l r ) = X < 1
2. K z 2/( l f x l r ) = X = 1
3. K z 2/( l f x l r ) = X > 1
Each of these possibilities corresponds to a certain physical reality with a particular
location for the oscillation center or pivot point:

Figure 6.6 TRANSIENT EFFECT OF YAW INERTIA & LONGITUDINAL CG


31

Reference [1], pages 32-33.

28

Rev. B

SAWE Paper No. 3528


Category Number 31.0

For Case (1) the yaw inertia is relatively small, equivalent to W f (l f X0.5)2 + W r (l r
X0.5)2, and the oscillation center is at distance l r X aft of the c.g. As soon as the front
wheels begin to steer the tendency to pivot about the OC will generate lateral reaction forces
at the rear tires in the same direction as the forces will be when the steady state condition is
reached; there will be no reversal of forces from transient to steady state. This case would
represent a short and very smooth transient period.
For Case (2) the yaw inertia is as if the front and rear axle loads represented
equivalent masses actually concentrated at the respective axle lines. The tendency to pivot
about the rear axle means that steering angle input at the front wheels will not immediately
generate lateral reaction forces at the rear tires; the will be some small lag time.
For Case (3) the yaw inertia is relatively large, equivalent to W f (l f X0.5)2 + W r (l r
X0.5)2, and the oscillation center is at distance l r X aft of the c.g. As the front wheels begin
to steer the tendency to pivot about the OC will generate lateral reaction forces at the rear
tires in the opposite direction as the forces will be when the steady state condition is reached;
there will be a reversal of forces from transient to steady state. This case would represent a
long and fluctuating transient period; this has been described as creating an uneasy sensation
of floating at the rear of the vehicle.

29

Rev. B

SAWE Paper No. 3528


Category Number 31.0

7 THE STEADY STATE CONDITION

The transient condition has terminated and the steady state has begun when the
sprung mass has attained some constant roll angle, the vehicle has reached some constant
yaw velocity, and the associate fluctuations of traction forces at the tires have ceased. This is
the condition generally sought after while circling the circumference of that test course known
as the skidpad.
Like the slalom, there is no official standard course. Road & Track uses a 200 ft (61
m) diameter skidpad 32, and General Motors used a 216 ft (66 m) diameter skidpad to obtain
the lateral acceleration results of Figure 7.2 33. The size of the skidpad is critical, because the
larger the radius the less angularity is involved in the resolution of the forces at the tires, and
thus the higher the maximum acceleration possible. Furthermore, if the vehicle in question is
a race vehicle designed to generate aerodynamic down force, then the higher velocities
required on the larger radius skidpads (a = V2/r) brings those aerodynamic qualities into
greater account, again making a higher maximum acceleration possible.
The aero effect is such that Car and Driver magazine recommended testing race cars
such as the 1981 Lola T600, an IMSA GTP race car, on a skidpad of 372 ft (113.4 m) radius,
as opposed to a more prosaic 141 ft (43 m) radius pad. The Lola, which came replete with an
aerodynamic quality called ground effects, pulled 1.42 gs on the large pad, versus 1.23 gs
on the small one. How well the Lola would have done on the large pad without ground
effects is open to question, but it is very likely it still would have bested its figure from the
141 ft (43 m) pad. The point here is that, whenever confronted by empirically obtained lateral
acceleration figures, it is wise to ascertain how they were obtained, and the size of the pad is
the most basic consideration. 34
Although it has been stated there is no officially recognized standard skidpad, the 100
ft (30.5 m) radius pad as depicted in Figure 7.1 comes close to being a standard by virtue of
its popularity:

32

Reference [16], page 16.


Reference [8], page 40.
34
Information regarding all relevant considerations is seldom given, but still very important. For instance, what
was the nature of the skidpad surface (asphalt, concrete, etc.) and whether that surface was clean or dusty, damp
or dry. Also, exactly how was the skidpad test conducted; are the results an average of multiple readings taken in
runs going in both clockwise (CW) and counter-clockwise (CCW) directions? And, of course, what was the
condition of the vehicle at the time of test; was the vehicle in some standard and reproducible configuration?
33

30

Rev. B

SAWE Paper No. 3528


Category Number 31.0

Figure 7.1 STANDARD SKIDPAD


Figure 7.1 reveals the nature of another interesting result that can be obtained from
skidpad testing other that the maximum lateral acceleration that a vehicle is capable of. That
other result is the basic nature of the directional stability of the vehicle tested; that basic
nature is characterized as understeering, neutral, and oversteering behavior. These are
terms having to do with directional stability, and a full discussion of directional stability and
what is involved therein is reserved for Chapter 10, but a quick discussion of the behavior
patterns depicted in Figure 7.1 is in order at this point.
When a vehicle is circling around the skidpad along the 100 ft (30.5 m) radius in a
steady state condition, after the maximum lateral acceleration has been recorded, then another
type of test may be carried out. Holding the steering angle input constant, the vehicle velocity
may be slowly increased and the vehicle response noted. If the vehicle attitude changes,
taking a nose in/tail out position with respect to a tangent to the circle, and changes course so
as to spiral in toward the center of the circle (- - - path), then the vehicle is said to
oversteer.
If, however, the vehicle adopts a nose out/tail in attitude with respect to a tangent to
the circle, and changes course so as to spiral out away from the center of the circle (oooo
path), then the vehicle is said to understeer. If the vehicle tends to neither oversteer nor
understeer but instead continues to circle the pad at constant 100 ft (30.5 m) radius (
path), then the vehicle is said to be neutral.
31

Rev. B

SAWE Paper No. 3528


Category Number 31.0

As should now be apparent, these terms oversteer, understeer, and neutral; are
characterizations of a basic vehicle handling tendency. This handling character is determined
by a number of factors, but the most basic is the vehicle longitudinal mass distribution. The
oversteering character is associated with an aft weight bias. The understeering character is
associated with a forward weight bias. And, of course, the neutral handling character is
associated with a vehicle c.g. near the midpoint of the wheelbase.
The trouble with this last condition is that neutral handling tends to not be constant;
as a vehicle is operated fuel is consumed, loading varies, and onboard objects can shift
positions. All of this variation in condition results in a shifting longitudinal c.g. location
during operation, which in turn can cause the vehicle handling character to shift, perhaps from
neutral to understeer then back to neutral on the way to oversteer again Such change in a
vehicles basic handling character can be disconcerting to the driver and is therefore
inherently dangerous. For reasons better dealt with in Chapter 10, it is generally best if a
vehicle maintains a degree of understeering character throughout its operational envelope.
It was on a skidpad somewhat larger than the standard skidpad of Figure 7.1 that
GM obtained its data from the target vehicles for its design of the 1984 Corvette. That
skidpad was of 108 ft (33 m) radius and, along with the roll gain info of Figure 5.6 and the
transient response data of Figures 6.1/2, the lateral acceleration info of Figure 7.2 was
obtained 35:

Figure 7.2 MAXIMUM LATERAL ACCELERATION RESULTS


35

Reference [8], page 40.

32

Rev. B

SAWE Paper No. 3528


Category Number 31.0

Note that once again the Corvette Z51 severely trounced its target opposition, as it did
in almost all objectively measurable categories of performance. However, the numerical test
advantage the Corvette Z51 demonstrated perhaps may have been obtained by too narrow a
focus on certain performance aspects to the neglect of other less quantifiable aspects. In contrast to
the highly enthusiastic review by Motor Trend, Car and Driver had this to say about the Corvette
Z51 36 (which was corroborated by a number of other sources) :

bad pavement sent its wheels boundingminor bumps or


irregularities threw the car off on a momentary tangentride
was annoyingly harshstraight-line stability was practically
nonexistent.
The problem seems to have been that GM had narrowly focused on beating the
targeted competition with regard to a number of performance criteria, at which it succeeded,
but thereby neglected the refinement for which some of the competition is famous. It probably
also didnt help that GM was undoubtedly designing to much lower cost targets than most of
the competition.
Car and Driver obtained a maximum lateral acceleration figure on a 141 ft (43 m)
radius skidpad of 0.84 gs for the Z51, and 0.85 gs for the base Corvette 37! This contrasts
sharply with the 0.95 gs GM supposedly attained with the Z51 on a 108 ft (33 m) radius
skidpad 38. In the very same issue that Motor Trend reported the GM figure, Motor Trend
reported the result of their test of the Z51 as 0.92 gs (skidpad radius not stated, but M/T
currently uses a 100 ft/30.5 m radius pad) 39. Road & Track, on a 100 ft (30.5 m) radius pad,
put the 1984 base Corvette at 0.842 gs 40. All of this underscores the variability of max lateral
acceleration testing possible due to skidpad and methodology variations.
All of the above indicates that if the 84 Corvette was not as overwhelmingly superior
in maximum lateral acceleration as GM and M/T indicated, it was still better than most of its
competition. The Corvette performance had been achieved by GM in large part by the mass
properties engineering tactics of weight and c.g. height reduction. Basic Corvette
construction consisted of the traditional (for Corvette) fiberglass body and steel frame, but
much weight and c.g. reduction was obtained through material substitution: suspension
control arms and uprights were forged in 6061 T-6 aluminum (saving 36 lb/16.3 kg over the
previous configuration), the brake calipers were also aluminum, and the rear transverse leaf

36

Reference [17], page 64. This papers author recognizes such criticism as being reminiscent of his own
experience with his 1999 Firebird.
37
Ibid, pages 65 and 68.
38
Reference [8], page 40.
39
Reference [8], page 36.
40
Reference [20], pages 57 and 58.

33

Rev. B

SAWE Paper No. 3528


Category Number 31.0

spring was fiberglass 41. Cross-members were reduced on the steel frame by tying the
transmission and the differential together via an aluminum C-section beam; this saved weight
and allowed for the entire driveline to be installed with only four bolts 42. Completing the
materials picture was the use of magnesium for the air cleaner housing and valve covers;
urethane sheet molding compound was used for the immense hood 43.
Yet, despite all this intense focus on weight reduction, GM missed the 3000 lb
(1360.8 kg) Corvette target curb weight by 150 lb (68 kg)! The 84 Corvette weight was
somewhat lighter than most of the vehicles used to formulate its design specification goals,
but still disappointing in view of all the aluminum, magnesium, plastics, and high strength
steels that had gone into its construction; there would seem to have been ample scope left for
further weight reduction 44.

41

Reference [18], page 30.


Ibid, page 31.
43
Ibid, page 29.
44
Ibid, page 30.
42

34

Rev. B

SAWE Paper No. 3528


Category Number 31.0

8 ROLLOVER

As noted in Chapter 3, rollover (overturn) seldom occurs for modern passenger cars as
the result of lateral acceleration resulting from tire traction forces because the slide
acceleration point is generally reached first (page 8, top). However, this does not mean that
vehicle rollover is not a concern. Although rollover was involved in less than 3% of passenger
vehicle accidents in the US for 2009, rollover was involved in about 35% of all fatalities
(23,437 total fatalities, 8,296 roll-over fatalities). Of the 8,296 fatalities about 66% failed to
wear seatbelts, with many resultant ejections from the vehicle; it can be assumed that some
would have survived had seatbelts been utilized. However, that leaves 34% who were
properly belted in, yet who fail to survive anyway, which still represents a disconcertingly
large proportion of all fatalities at almost 12% 45.
Therefore, even though rollover is unlikely, it still merits serious concern. The
NHTSA used to (NCAP 2001-2003) rate vehicles for rollover resistance based solely on a
mathematically derived figure of merit called the Static Stability Factor (SSF) 46. The SSF is
exactly the same as as calculated by the rigid model Equation 3.6 in Chapter 3:

Where:

(EQ. 8.1)

SSF = A figure numerically equal to the lateral acceleration for overturn (in
gs) as calculated by Eq. 3.6.
t = The average vehicle track width, front plus rear divided by two.
h cg = The vehicle center of gravity height above the ground plane.
Of course, the accuracy of using the SSF as a means of comparison between vehicles
depends on the accuracy of the track measurements and c.g. measurement, and for the latter it
is crucial that all vehicles be in the same weight condition when measured. Even with all due
care in measurement, the SSF cant be a totally accurate means of comparison between
vehicles for reasons touched on in Chapter 5 and illustrated in Figure 5.3, under lateral inertial
load a rolling movement of the sprung mass will occur through some angle s which will
reduce the h cg by some amount d z and cause the sprung weight to shift laterally by some
45

Reference [9], page 1. Per the NHTSA, roll-over accidents have on average accounted for 10,000 deaths per
year over the decade from 2000 through 2010.
46
Reference [2], page 1.

35

Rev. B

SAWE Paper No. 3528


Category Number 31.0

amount d y , all of which is among a number of suspension dependent changes which will
render the real somewhat different (undoubtedly less! 47) than that predicted by the
SSF.
Still, while the difference between the actual and the SSF is significant,
the SSF is simple to determine and exhibits a strong statistical correlation with the incidence
of roll-over accidents, as determined by NHTSA statistical analysis 48. As part of the
NHTSAs New Car Assessment Program (NCAP), the NHTSA began providing SSF
based rollover resistance ratings for new cars in 2001.
Other measures of rollover resistance include passive tests such as the Tilt Table
Ratio (TTR), which is depicted in Figure 8.1. Clinometers are utilized to measure the degree
of tilt of the table and of the vehicle sprung mass; the degree of table tilt corresponds to a
certain amount of lateral gs:

Figure 8.1- TILT TABLE RATIO TEST

47

As discovered in Chapter 3 the drop in c.g. height dz due to roll is probably negligible, but the lateral shift in
c.g. dy significantly diminishes the track dimension t/2 significantly; essentially or the SSF

becomes ( )/( ).
48

Reference [15], page 37 (Section V.A.). Reference [2], page 1, states that the SSF was chosen as the basis for
vehicle rollover ratings because it highly correlated with actual crash statistics; it can be measured accurately
and inexpensively and explained to consumers, and changes to vehicle design to improve SSF are unlikely to
degrade other safety attributes. Note: the degrade other safety attributes is a reference to possible c.g. height
reduction by manufacturers through reduction of roof and roof support structure, thereby reducing crush
resistance. However, increased rigor and wider applicability of the FMVSS 216 roof crush resistance standard
has since made such reduction unlikely.

36

Rev. B

SAWE Paper No. 3528


Category Number 31.0

TTR is the ratio of the lateral force (parallel to the table top) at onset of high side
wheel lift off to vehicle weight (and therefore equivalent to gs); for a rigid vehicle model the
TTR should be the same as the SSF calculation. For a real vehicle the TTR results should be
more realistic (i.e., less) than the SSF as most of the complications of roll, deflections, and
liquid free surfaces will be present. However, TTR is regarded as likely to be a little different
from a real rollover because the whole vehicle is tilted; the physical situation is not the same
as if the lateral g level were attained by forces at the tires while the vehicle was on level
ground 49. Reportedly, real vehicle TTR measurements are about 10% to 15% less than SSF
calculations 50.
There is also another empirical determination, the Side Pull Ratio (SPR), which
involves winching in of a cable acting through the vehicle c.g. in order to measure the force
required to cause two wheel lift on a side. That force is divided by the vehicle weight to
produce the SPR metric in gs. Again, the result should be near identical to the SSF for a
solid body vehicle, and very close to the TTR for any real vehicle (and therefore also about
10 to 15% less than SSF) 51. However, the SPR is regarded as being somewhat more realistic
than the TTR as the vehicle weight vector stays perpendicular to the ground plane.

Figure 8.2 SIDE PULL RATIO TEST


All the roll-over resistance metrics mentioned so far are essentially static in nature. However,
there are at least two dynamic methods for rollover resistance determination, the J-turn and the

Fishhook. Since 2004 the NHTSA has been using the results of the SSF calculation plus the
Fishhook test to arrive at rollover resistance ratings for new vehicles. The Fishhook and
49

Reference [7], page 317.


Reference [15], Section V.A., page 2.
51
Ibid.
50

37

Rev. B

SAWE Paper No. 3528


Category Number 31.0

the J-turn are both open loop maneuvers that dont really provide a direct measure of roll
resistance, but do give an indication of vehicle dynamic response under high lateral loads. The
NHTSA grades vehicles on a pass or fail basis for the Fishhook maneuver, and combines that
result in a complex formulation that favors the SSF results to produce a rating of one to five
stars.
The Fishhook name comes from the shape of the path taken by the vehicle during
the test. The Fishhook invokes the rollover tendency of a vehicle by approaching as close as
possible to actual rollover through a rather harsh maneuver. The Fishhook uses steering inputs
that approximate the steering a panicked driver might use to regain lane position after
dropping two wheels off the roadway onto the shoulder, but is performed on a level pavement
with a rapid initial steering input followed by an over correction. The original version of this
test was developed by Toyota, and variations of it were adopted by Nissan and Honda 52.

Figure 8.3 FISHHOOK TEST GROUND PATH


NHTSAs test version includes roll rate measurement in order to time the countersteer to coincide with the maximum roll angle each vehicle takes in response to the initial
steering input. The test utilizes an automated steering system programmed with inputs
intended to compensate for differences in vehicle steering gear ratio, wheelbase, and stability
properties. To begin, the vehicle is driver controlled in a straight line. The driver releases the
throttle, coasts to the target speed (which starts around 35 mph/56 kph and increased in 5
mph/8 kph increments for each run until termination 53 is achieved), and then activates the
auto-pilot which commences the maneuver. The test runs conclude when a termination
condition is achieved involving two inch or greater lift of the vehicles inside tires (fail), or
52
53

Reference [2], page 4.


Hopefully not the drivers.

38

Rev. B

SAWE Paper No. 3528


Category Number 31.0

if the vehicle completes the final run at maximum speed of 50 mph/80 kph without lift (pass).
If needed, further testing is undertaken to confirm the exact speed at which lift occurs, and
that the lift point is repeatable 54.

54

Reference [2], page 4.

39

Rev. B

SAWE Paper No. 3528


Category Number 31.0

9 TIRE BEHAVIOR: SLIP ANGLES

As noted and diagramed in Chapter 6, the lateral traction force generated at the
tire/road contact area is associated with a certain slip (drift) angle. Of course, the contact
patch is also often called upon to generate longitudinal forces (acceleration and braking) as
well, and all these forces interact just as stresses acting in different orthogonal directions in a
material sample have an interacting effect expressed in a quantity called Poissons Ratio 55.
The relationship between lateral and longitudinal forces (analogous to stresses) and
drift angle (analogous to strain) is not as clean-cut as the situation expressed in Poissons
Ratio This is to be expected, as most materials are isotropic 56 and tires are anisotropic. The
relationship with respect to tires can be illustrated by the following diagram:

Figure 9.1 TIRE TRACTION CIRCLE (ELIPSE)


Any combination of lateral and longitudinal traction forces is possible as long as the
resultant total traction force does not exceed the maximum traction possible represented by
the bounds of the outer circle. In the situation depicted in the first quadrant of the traction
circle, the acceleration force of 539.8 lb (244.8 kg) allows for a lateral left hand traction force
of up to 762.6 lb (345.9 kg). If the centrifugal force needed at this time to complete some
maneuver actually goes up to that level, then the resultant force will be 934.3 lb (423.8 kg),
and no further increases in acceleration forward or to the left will be possible; the tire is at
55

Simon Denis Poisson, 1781-1840.


This would be true regarding traditional engineering materials, excluding the new burgeoning field of
composites which have a directional structure, as do tires (though on different scales).
56

40

Rev. B

SAWE Paper No. 3528


Category Number 31.0

force saturation. If the driver hits the gas to go faster, the tire will spin helplessly making
huge clouds of white smoke (burnout!) and the vehicle will slide off the road to the right.
If a lateral traction force of 762.6 lb (345.9 kg) is actually required of the tire in this
situation of also supplying an acceleration force of 539.8 lb (244.8 kg), then the tire will be
running at a drift angle of 8 degrees. If the driver lifts off the gas, then the drift angle will
snap back to about 5.5 degrees. Conversely, if the driver were taking a turn which caused a
lateral force load at his rear wheel(s) of 762.6 lb (345.9 kg), and he stepped on the gas so as
to acquire an acceleration force of 539.8 lb (244.8 kg) at the rear wheel(s), the rear drift angle
would go to 8 degrees; the rear end would step out adopting a new attitude and smaller
turning radius. This effect is called throttle steering, and in some automotive sporting events
(gymkhana, autocross, drifting) a driver may make as much use of the throttle to steer as the
steering wheel.
Simplifying the traction field as a circle instead of the more realistic ellipse 57 allows
for an easier determination of how much traction potential is available for lateral
acceleration. The Pythagorean relationship between lateral and longitudinal traction forces
can be used to revise the equations for the potential lateral traction force at the axles.
Essentially Equation 3.1 becomes for the front axle:

= +

(Eq. 9.1a)

Or
= +

(Eq. 9.1b)

And for the rear axle:

= [( ) ] + [( ) ]

(Eq. 9.2a)

Or
= +
57

Reference [11], page 129. Even the ellipse is a simplified model of a more complex reality.

41

(Eq. 9.2b)

Rev. B

SAWE Paper No. 3528


Category Number 31.0

This can be made even more representative of reality by explicitly taking weight
transfer into account in determining the lateral force potential; Equations 5.2/3 (front) and
5.4/5 (rear) for the determination of Ni and N o front and rear can be substituted into
Equations 9.1 and 9.2. However, the result is cumbersome and there is little point in doing so
other than to create a computer simulation of automotive maneuvering. For such a simulation
it would also be important to also know how the drift angle varies as a function of lateral and
normal loads, which is a topic only touched on so far.
As is apparent from Figure 9.1, as lateral force increases the tire deformation 58, i.e. the
drift angle, increases until the deformation reaches a limit and any further demands upon the
tire will tend to result in 100% slippage; the vehicle will tend to go into an uncontrolled skid.
Thus the limit of tire deformation determines the maximum lateral resistance possible and
restricts the range of the drift angles; at present for passenger car tires the maximum b and
drift angle seems to be around 1.0 g and 12 degrees. However, most driving on public roads
rarely exceeds 5 degrees or 0.3 gs lateral acceleration 59. For racing vehicle tires (like those
used in Formula 1) the present limits (without aerodynamic loading) seem to be about 8
degrees and 2.0 gs, but the softer compounds that allow for such greater traction do not
give the sort of mileage expected of road tires (race tires are designed to last only a few
hundred miles or less).
Since the drift angle/lateral force relationship is dependent upon quite a few
parameters, it is common to look at functions which constitute only a partial differential of the
total relationship (for which no one has yet established a complete definitive formulation
based on physics 60) in order to achieve a degree of understanding. If the tire drift angle/lateral
force partial differential function is plotted the result looks like Figure 9.2.
The lateral force increases with drift angle (or vice versa) on a shallow curve,
reminiscent of stress-strain curves, up to a deformation limit. At that point the situation is very
unstable and will either quickly return to the safety of the useful traction region or result in
100% slip (a slide) at the tire/road contact patch.

58

Conversely, in the literature one often reads of increasing deformation (drift angle) causing an increase in the
lateral force. It is a matter of perspective; if a lateral force is applied to a vehicle running in a straight line then
the tires will develop drift angles and the vehicle will deviate from the straight line, but if a driver twists the
steering wheel of a vehicle running in a straight line then the resultant drift angles will result in a lateral force
and again the vehicle will deviate from the straight line. The line between cause and effect is often difficult to
discern.
59
Reference [1], page 78.
60
Hans Bastiaan Pacjeka, Professor Emeritus at Delft University of Technology in the Netherlands, has
developed a tire model called the Magic Formula because it is based on relatively little underlying physics;
instead it is mainly based on a regression analysis of reams of empirical tire data.

42

Rev. B

SAWE Paper No. 3528


Category Number 31.0

If the appropriate portion of the useful region is linearized by the fitting of a straight
line to the original shallow curve, then a simplified lateral force/drift angle relationship (the
slope of the fitted line) can be obtained. This quantity is commonly called the Cornering
Power of the tire, which is yet another misnomer like shock absorber (damper) or slip
angle (drift angle). It would seem more appropriate to call this quantity the Cornering
Stiffness. Such linearized tire relations are often used for simulations/studies of automotive
dynamic behavior. Note that a little earlier it was demonstrated by use of the traction circle
how the addition of longitudinal traction force for acceleration could decrease the cornering
stiffness F y / at a tire (762.6 lb/5.5 deg vs. 762.6 lb/8.0 deg); braking force additions
likewise tend to decrease the cornering stiffness (C s ) of the affected tire(s).
The problem with simulations/studies based on linear models is that tire behavior
beyond the proportional limit is not taken into account. Although road going passenger
vehicles seldom venture into the nonlinear region between proportional limit and deformation
limit, racing vehicles frequently do. A simulation/study of race car behavior would seem to
require a non-linear curve fit to the entire (both linear and transition portions) useful
traction region 61.

Figure 9.2 - CORNERING STIFFNESS: LINEARIZED


LATERAL FORCE / DRIFT ANGLE FUNCTION
61

Reference [14], page 126.

43

Rev. B

SAWE Paper No. 3528


Category Number 31.0

If matters were just as simple as Figure 9.2 then understanding of automotive dynamic
behavior, and how mass properties influences that behavior, would be very easy. The
potential or maximum lateral force that a tire can supply, and the drift angle associated with
that force, is dependent on many parameters. That the lateral traction force potential is
influenced by longitudinal forces is a subject already touched on, but the lateral force
potential is also influence by normal load, camber angle (which can change with roll), roll
steer (which can be the result of normal load and camber change with roll, but toe in/out can
also change with roll), tire type (size, carcass type and material, rubber type, tread design,
aspect ratio), inflation pressure, wheel rim width, road material and surface (smooth, rough,
dusty, etc.), weather (rain, snow, ice), temperature (both ambient and of the tire itself), and the
speed of the vehicle (all basic tire coefficients of traction are slightly speed dependent; the
same lateral force will produce a slightly smaller drift angle at high speed than at low
speed 62). While all these factors are significant, only normal load is fundamental and germane
to the topic of this paper.
To illustrate the effect of normal load on lateral resistance and drift angle a plot such
as Figure 9.3 is often used. Note that it is essentially like Figure 9.2 except that there are now
a large set of F y , functions which serve to represent an infinite variation; any change in
normal load alters the F y , relation, but these five example curves may suffice as the
intermediate possibilities can be approximated by interpolation:

Figure 9.3 LATERAL TRACTION vs. DRIFT


ANGLE FUNCTIONS AT NORMAL LOAD
62

Reference [5], page 165.

44

Rev. B

SAWE Paper No. 3528


Category Number 31.0

Only the useful traction region is shown; 13 degrees is about the maximum drift
angle/deformation limit. Note that as normal load increases the lateral traction force necessary
for a certain drift angle increases as well, but at a decreasing rate just as was the case for
Equation 2.2/Figure 2.2. Actually, Figure 9.3 is a poor way to illustrate this behavior, which is
better shown by the following real data plot of lateral force vs. normal load for an actual
6.00x16 tire inflated to 28 psi (193 kPa) 63:

Figure 9.4 LATERAL TRACTION vs. NORMAL LOAD


AT DRIFT ANGLE
Figure 9.4 clearly shows how increasing normal load will increase the lateral force
necessary to cause the same amount of deformation (drift angle), but at a decreasing rate and
only up to a point. That point is the deformation limit, and traction demands beyond that point
are likely to result in an out-of-control slide of the vehicle.
This figure provides the ability to demonstrate the directional effect of weight
transfer/roll under lateral acceleration. Recall that the 1980 Ford Fiesta S in the 4 up
condition as modeled in Chapter 5 had static axle loads of 1240.7 lb (562.8 kg) front and
1058.7 lb (480.2 kg) rear, as determined from the total vehicle weight of 2299.4 lb (1043.0
63

Reference [5], page 165.

45

Rev. B

SAWE Paper No. 3528


Category Number 31.0

kg) and weight distribution of 54/46. This means that at 0.5 gs lateral acceleration in steadystate equilibrium the lateral traction forces at the axles must be 620.35 lb (281.4 kg) front and
529.35 lb (240.1 kg) rear (assuming a turn radius large enough that the effects of angularity
requiring the resolution of forces into X & Y components can be neglected). From
Equation 5.2 it was determined that for the Fiesta the lateral acceleration of 0.5 gs translated
into a roll angle of 4.6 degrees. From Equations 5.2 through 5.5 it was determined that for a
roll angle of 4.6 degrees the normal loads (now dynamic) become N fi = 161.47 lb, N fo =
1079.23 lb, N ri = 200.55 lb, N ro = 858.15 lb (73.2, 489.5, 91.0, 389.3 kg respectively).
Assuming tire characteristics of b = 1.2 and m = 0.0004 and using Equation 2.2
these dynamic normal loads correspond to potential lateral traction forces of F yfi = 183.33 lb,
F yfo = 829.18 lb, F yri = 229.57 lb, F yro = 735.21 lb (83.2, 376.1, 104.1, 333.5 kg), or on a
potential force per axle basis F yf = 1012.5 lb (459.3 kg), F yr = 964.78 lb (437.6 kg). These
per axle potential traction forces easily exceed the required lateral forces of 620.35 lb (281.4
kg) front and 529.35 lb (240.1 kg) rear for dynamic equilibrium, and there is no need to resort
to Equations 9.1a through 9.2b to make a determination whether the combined lateral and
longitudinal forces would exceed available traction because, for simplicitys sake, there are no
longitudinal forces in this example (although in reality there would always be some
longitudinal forces due to rolling resistance and the need to counter that resistance with some
driving forces to keep the vehicle going around the skidpad at constant velocity).
Assuming the actual forces generated at each tire will be in proportion to the potential
forces possible at that tire, the actual lateral forces work out to be F yfi = 112.33 lb, F yfo =
508.02 lb, F yri = 123.86, F yro = 405.49 (51.0, 230.5, 56.2, 183.9 kg). Assuming for this
example that the tires of the Fiesta were those of Figure 9.4, by interpolation of that figure
(and not one like Figure 9.1 because again there is no X and Y forces requiring
combination into resultant forces via use of Equations 9.1a through 9.2b) the corresponding
drift angles are fi = 4.0 deg, fo = 5.8 deg, ri = 4.0 deg, ro = 4.5 deg. Note that the
inner and outer drift angles are different, which they would tend not be if there were no
weight transfer. Also note that the average drift angle at the front axle (4.9 deg) is greater
than the average drift angle at the rear axle (4.24 deg), which is indicative of an understeering
vehicle (and the front to rear drift angle comparison is even more so indicative if limited to
just the outer wheels; with increasing lateral acceleration the outer wheels become ever more
dominant), which the Fiesta is known to be.

46

Rev. B

SAWE Paper No. 3528


Category Number 31.0

The drift angles at front and rear determine the turn center of the vehicle and the
attitude or yaw angle () the vehicle adopts with respect to the tangent line to the turn
circle. If the vehicle is oversteering, as is the case with the Fiesta, then f > r and the
vehicle adopts a nose out attitude with respect to the tangent line as depicted in Figure

9.5 64.
Figure 9.5 UNDERSTEERING TURNING GEOMETRY
As noted in Chapter 7, an understeering vehicle on the skidpad will not hold to the
constant radius of the pad as velocity is slowly (minimizing longitudinal force effect)
increased; it will nose out and take a path of an ever increasing spiral outward. Since the
Fiesta is an understeering vehicle it should display such behavior via the models presented in
this paper. For the purpose of demonstrating the fidelity of those models, lets say that the
Fiestas speed is slowly increased to a lateral acceleration of 0.7 gs.
For 0.7 gs lateral acceleration in steady-state equilibrium the lateral traction forces
required at the axles are 868.49 lb (393.9 kg) front and 741.09 lb (336.2 kg) rear. From
Equation 5.2 the lateral acceleration of 0.7 gs corresponds to a roll angle of 6.4 degrees.

64

For Ackermann (Darwin) turning geometry information see Appendix C.

47

Rev. B

SAWE Paper No. 3528


Category Number 31.0

From Equations 5.2 through 5.5 for a roll angle of 6.4 degrees the normal loads become N fi
= 0.00 lb, N fo = 1260.85 lb, N ri = 70.41 lb, N ro = 988.29 lb (0.0, 571.9, 31.9, 448.3 kg).
Using Equation 2.2 these dynamic normal loads correspond to potential lateral traction
forces of F yfi = 0.00 lb, F yfo = 877.12 lb, F yri = 82.51 lb, F yro = 795.26 lb (0.0, 397.9, 37.4,
360.7 kg), or on a potential force per axle basis F yf = 877.12 lb, F yr = 877.77 lb (397.9,
398.1 kg). These per axle potential traction forces exceed the required for dynamic
equilibrium @ 0.7 gs lateral forces of 868.49 lb (393.9 kg) front and 741.09 lb (336.1 kg)
rear.
Again assuming the actual forces generated at each tire will be in proportion to the
potential forces possible at that tire, the actual lateral forces work out to be F yfi = 0.00 lb,
F yfo = 868.49 lb, F yri = 69.67 lb, F yro = 671.42 lb (0.0, 393.9, 31.6, 304.6 kg). Also again
assuming that the tires of the Fiesta were those of Figure 9.4, by interpolation of that figure
the corresponding drift angles are determined to be fi = 0.0 deg, fo = 14.5 deg 65, ri = 3.7
deg, ro = 9.0 deg. Note that the average drift angle at the front axle (7.25 deg) is greater
than the average drift angle at the rear axle (6.4 deg), but comparing just the dominant outer
wheels results in an even more disparate 14.5 deg front vs. 9.0 deg rear, all of which is
understeering indicative.
Note that at 0.7 gs that the front inner wheel has lifted off the ground and, judging
from the traction forces and drift angles, the vehicle is just about ready to slide nose first off
the skidpad. This is severe understeer, and the Fiestas designers seem to have done
everything in their power to make it so. The most basic thing was the weight load bias to the
front, but they also had a MacPherson independent suspension at the front with a nonindependent beam axle suspension at the rear (an independent suspension system generally
allows roll camber changes that have an adverse effect on lateral traction). Of course, a lot of
the oversteering factors are the natural result of a FWD/front engine configuration, but to
complete the picture the track at the front is wider than at the rear (more understeer), the roll
axis is nose down to the front (yet more understeer), the front roll stiffness is greater at the
front than the rear (even more understeer) 66.
The vehicle concept explains this; the Fiesta was just a little front wheel drive pulling
a relatively big box behind. In anticipated usage that box could be filled to the brim with
people and cargo, with possibly more cargo strapped to the roof and towed along behind.
Note that this is off the chart of Figure 9.4 and in treacherous territory. The total traction coefficient at
this point is only about 0.69, but 15 degrees drift is extreme. However, this is probably the result of not having a
real set of tire characteristics; the b and m coefficients came from Reference [11] for an unspecified tire,
while Figure 9.4 came from Reference [5] and is for a 6.00x16 tire at 28 psi. The Ford Fiesta S ran on 155SR12
tires, possibly with a large pressure differential front to rear (another over/under steer factor).
66
Notably, with all that bias toward understeer, this author calculates an initial cornering stiffness differential of
only about 2% front to rear (front bias) growing to about 27% at 0.7 gs (slide).
65

48

Rev. B

SAWE Paper No. 3528


Category Number 31.0

With such loadings, possibly right up to the 5622 lb (2550 kg) GVWR (and maybe beyond!),
the c.g. could migrate quite a distance aftward from its normal location. Apparently the design
intent was to ensure consistent handling characteristics no matter how the little grocery
getter was loaded, and the consequent reduction of the maximum lateral acceleration level
was not of much significance.
Now that it seems that the Ford Fiesta reaches its lateral acceleration limit near 0.7
gs, sliding off the skidpad nose first, lets see how well Equation 3.5 predicted this
occurrence. First, the equation for ay slide from Chapter 3:

( )

Into this equation the front end values b (1.2), m (0.0004), W f (1240.7 lb),
h cg (22.74 in), and t f (52.52 in) are input:

(. ).
(. )(. ).
+

.

.
((. )(. ). . )

= 0.8252 gs

So, if the Fiesta actually did slide off line at around 0.7 gs then the rigid model
Equation 3.5 would be definitely optimistic, as expected. However, at 0.7 gs the lateral
traction force potential at the front axle (877.12 lb) still exceeded the required traction force
(868.49 lb) to maintain heading, although by the slightest of margins; the slide point would
actually be just a little more than 0.7 gs, say 0.71 gs.

49

Rev. B

SAWE Paper No. 3528


Category Number 31.0

10 DIRECTIONAL STABILITY
Vehicle directional stability is a subject that was initially developed for, and by, the
aircraft industry starting from that industrys earliest days 67. Although the advent of the
automobile preceded the airplane, it would be more than 35 years before the science of
stability would even start to be applied to ground vehicles 68. Stability as it pertains to a
vehicle can have many aspects, mainly yaw (directional) stability, roll stability, and pitch
stability. The emphasis of this paper on lateral acceleration narrows the focus to just
directional stability and roll stability, and to a limited extent the latter has been dealt with
previously in Chapters 6 (transient effects) and 8 (rollover).
What held back the development of directional stability theory as applied to
automobiles was the lack of information regarding tire behavior. The behavior of an aircraft in
flight is determined by aerodynamic forces, and the effort to understand such forces was
present at the very dawn of the aircraft; Orville and Wilber Wright built a wind tunnel to
study aero effects long before their first aircraft ever left the ground 69. The behavior of
automobiles is predominantly determined by tire behavior, and the first tire testing machine
was possibly the rotating steel drum tire tester of Becker, Fromm, and Maruhn circa 1930 70.
These German researchers generated tire data as a prerequisite for their investigation of the
great automotive problem of the time: steering shimmy. In this they were following up on
the French researcher George Broulheit who had identified the tire characteristic of slip
angle in his investigation relating to shimmy 71. A notable follower in the footsteps of these
Europeans was R.D. Evans of the Goodyear Tire and Rubber Company who continued the
investigation of the physical properties of tires via another drum tire tester 72.
The war years of 1939-1945 brought most research into tire and automotive behavior
to a halt. However, in Germany at Junkers Aircraft the researchers Von Schlippe and Dietrich
developed a simple structural model of the pneumatic tire which after the war would tie in
with the research of Dr. A.W. Bull and (later) S. A. Lippmann at US Rubber. Essentially this
had to do with the lag time of drift angle formation after application of a side force being a
67

Orville and Wilber Wright thought a certain amount of instability was unavoidable; they regarded part of the
pilots role was to serve as a stability augmentation system. This design approach became less and less viable as
aircraft performance increased. F.W. Lanchester was one of the first to investigate airplane stability analytically;
he published his book on the subject, Aerodonetics, in 1908.
68
Reference [14], page 123.
69
Ibid, page 101.
70
Becker, G.; H. Fromm, and H. Maruhn, Schwingungen in Automobillenkungen, Krayn, Berlin, 1931.
71
Broulheit, G.; La Suspension de la Direction de la Voiture Automobile: Shimmy et Danadinement, Socit
des Ingnieurs Civils de France, Bulletin 78, 1925.
72
Evans, R.D.; Properties of Tires Affecting Riding, Steering, and Handling, Journal of the Society of
Automotive Engineers, 36(2):41, 1935.

50

Rev. B

SAWE Paper No. 3528


Category Number 31.0

function of the distance rolled, and hence of velocity and time. This meant that, for highway
speeds and small drift angles, the transient phase of the tires adjusting to the lateral forces
could be neglected; modeling of vehicle behavior for just the steady state condition would still
have broad validity. This opened up the avenue to all the steady state modeling investigations
into vehicle directional stability which would subsequently occur.
One of the first to attempt a mathematical analysis of automotive stability was Prof.
Yves Rocard of the Sorbonne in 1954. Prof. Rocards model was 4-wheeled but rigid; no
weight transfer or roll effects were included. In 1955 M.A. Julien and G.J. Arnet presented a
paper based on an analysis of a less simplified model in that it separated the vehicle into a
sprung and an unsprung mass. Maurice Olley at General Motors had begun research into
the shimmy problem in the early 1930s, and by the 1950s he and his GM compatriots had
progressed to major revelations regarding automotive stability, benefiting greatly from the tire
work of Evans, Bull, Gough 73, and others. Also around this 50s period William F. Milliken,
David W. Whitcomb, and Leonard Segel of the Cornell Aeronautical Laboratory 74 would do a
great deal to advance the understanding of tire behavior and automotive directional stability,
including the construction of the AF-CAL on-road tire tester 75. The stability model utilized
by the CAL investigators appears to have been limited to the bicycle model, so the effects
of lateral weight transfer and roll were again not inherently accounted for, but they can be
integrated into the results after the fact (somewhat).
Since the heady period of the 1950s many new approaches to understanding
automotive directional stability have been taken. Mathematical vehicle modeling and
empirical tire research, utilizing newer and ever more realistic test machines, would continue,
but computer simulations would come to play an ever more prominent role. At first the
computer simulations were largely through the use of analog devices intended to serve in
validation of theory, and vice versa. However, analog computers were limited in capability,
and were to be replaced by more versatile digital machines. Since early digital machines were
typically slower than analogs, for a while in the early 1970s hybrid computers found favor.
As the computational speed and other characteristics of digital computers rapidly improved
73

V.E. Gough was a major researcher into tire behavior at the Dunlop Tire Research Center in England during
the 1950s/early 1970s; he is best known for the Gough Diagram which brings all the major characteristics of
a tire together into a single convenient diagram (see W. Steeds Mechanics of Road Vehicles, page 223, for
further info).
74
Originally founded in 1943 as the Curtiss-Wright Airplane Division Research Laboratory, the lab was donated
to Cornell University in 1946. In 1972 the laboratory became the Calspan Corporation, a publically traded
company.
75
This device was CAL constructed under contract to the US Air Force to study the behavior of aircraft nose
wheel tires; later it was put to more general use. Testing tires on the road eliminated the artificialness of the tiredrum contact, but was limited to the study of free rolling tires at speeds up to 40-50 mph (64-80 kph). It was later
superseded by the Calspan TIRF (TIre Research Facility) in 1973, an advanced flat-bed tire tester which is in
use up to the present day.

51

Rev. B

SAWE Paper No. 3528


Category Number 31.0

they came to dominate the simulation world, initially as just mainframes but ultimately also as
personal computers. The change in platforms also coincided with corresponding changes in
simulation software; specific purpose built software was to a large extent supplanted by
general multipurpose dynamic simulation and finite element analysis (FEA) codes.
The result has been a modern plethora of highly detailed and complex modeling
attempts. A full understanding would involve a long exposition involving a good deal of
higher math and complex algorithms which are beyond the scope of this paper (and beyond
this authors capability as well). Therefore, this paper will address the subject through a
simple analysis that will cover most automotive cases, especially passenger cars, and arrive
at a limited result with the limitations as fully identified as possible.
The greatest single complicating factor with regard to automotive directional stability
is that an automobile must be dynamically stable, not just statically stable. The defining
difference between those two conditions is that a statically stable system may show the
accepted stable state characteristic of returning to equilibrium position after the application of
a disturbance input, but in returning may overshoot the equilibrium position and then
execute a reverse correction back to the equilibrium point once again, which results in an even
greater overshoot, and so on. Although the system initially exhibits what would be regarded
as stable behavior, that is not the full story; that initial behavior degenerates into an ever
increasing series of oscillations ultimately leading to a loss of control. Such a system is
statically stable, but dynamically unstable.
The cause of this curious dichotomy in behavior is the presence of an energy source. A
spring displaced from its equilibrium position will return to that position when released,
although it may take a large number of oscillations to do so if not critically damped. However,
if the same spring is subject to energy input from a vibrating ground plane, then it will never
settle back to its equilibrium position; it will vibrate about that position indefinitely as long as
the energy input continues (or, if the energy input is at the spring systems resonance
frequency, the oscillations may grow in amplitude until the spring breaks). To illustrate the
condition of a vehicle being statically stable, but dynamically unstable, Figure 10.1 is
presented:

52

Rev. B

SAWE Paper No. 3528


Category Number 31.0

Figure 10.1 STATICALLY STABLE, DYNAMICALLY UNSTABLE


So, whether a system is fully, i.e. dynamically, stable or not has to do with the
presence of an energy source. The traditional high school physics example of a stable system,
such as the one consisting of a marble in a bowl, is limited as that example does not involve a
source of energy input; the stability exhibited by the displaced marble in ultimately coming
back to rest in the center of the bowl is only a static stability. Vehicles are quite different; as
a system they all have a source of energy input in the form of an engine or motor. The energy
input from that engine or motor is in large part reflected in the kinetic energy of the moving

vehicle, which is essentially equal to 76, so it should not be surprising that a full
automotive directional stability analysis should hinge on the parameter of speed (velocity).

That is, a vehicle may be dynamically stable in a certain speed range, and unstable in
another, usually higher, range. For this paper the involvement of speed in the stability
question will only be lightly dealt with in the course of the following simple exposition77. The
matter of automotive directional stability will be approached initially from the standpoint of
an old and incomplete analysis, a static stability analysis, which has to do with the lateral
force balance front-to-rear and the consequent drift angle relationship of the vehicle tires
front-to-rear. The lateral force/drift angle relationship is dependent upon normal loads and the
lateral inertial reaction forces through the c.g., so the most significant vehicle mass property
with regard to directional stability is the weight distribution.
Consider the case of a vehicle traveling in a straight line on a level (slightly crowned)
road just before encountering a disturbance such as a sudden change in the road crown
76

To be truly equal to all the kinetic energy inherent in the vehicle, not just the translational energy but also the
energy of all the rotational components, the mass symbol is the effective mass me.
77
For a more complete treatment see Reference [14], pages 123-229. Speed not only indicates the presence of
kinetic energy for oscillations to feed on, but also affects traction directly. However, one would think that the
slight dependency of the traction coefficient on speed would have little effect on directional stability as it
depends on a front to rear balance, and such traction dependency would affect front and rear equally.

53

Rev. B

SAWE Paper No. 3528


Category Number 31.0

(highly exaggerated) resulting in a lateral (with respect to the vehicle) component of the
weight vector acting through the c.g. 78 (disturbance force F d ):

Figure 10.2 LATERAL DISTURBANCE FORCE F d


Most disturbance forces act through the c.g., or generate a reaction force acting
through the c.g. Aerodynamic disturbance forces tend to act through the center of pressure
(CP), which for most modern automobiles is located just forward of the c.g., so the effect is
much the same as if the aero force had been at the c.g. At any rate, since the disturbance force
is taken as acting laterally though the c.g., the location of the c.g. is very significant to the
nature of the vehicle reaction. To illustrate the possibilities of that reaction, consider the case
of three vehicles of varying longitudinal weight distribution, with suspension and tire
characteristics equal all around, encountering the same lateral disturbance force F d 79.
The forward c.g. bias case (l f < l r ) develops larger drift angles at the front tires
then at the rear in response to the disturbance force F d ; this action results in the forward
c.g. vehicle steering away from F d generating an inertial reaction ma that tends to
counteract the effect of the disturbance. Therefore this understeering action is considered to
be stable. The mid c.g. case (l f = l r ) develops equal drift angles all around in response to
F d ; this results in no turning action, just a sideways shift, no rotational inertial reaction is
generated. The rear c.g. bias case (l f > l r ) develops larger drift angles at the rear tires than
at the front in response to F d ; this results in a steering into the disturbance force F d
generating an inertial reaction ma that tends to augment the disturbance. This in turn
causes an even sharper turning, generating a larger inertial reaction. This sort of
oversteering action is therefore considered to be unstable as it tends to feed upon itself.
78
79

Reference [14], page 134 & 143.


Ibid, page 169. Reference [6], page 56.

54

Rev. B

SAWE Paper No. 3528


Category Number 31.0

Figure 10.3 THE THREE CASES: STABLE, NEUTRAL, AND UNSTABLE


When forces such as F d and ma act at the c.g. some weight transfer and roll is
created, especially in the oversteering case. The static normal loads are dynamically modified
in response to the weight transfer and roll, which generally tends to amplify the vehicle
response, as does camber changes and other effects. However, for directional stability, it can
be seen from the above illustration that the primary parameters are the vehicle mass
properties of weight, weight distribution (longitudinal and lateral), and the yaw inertia.
To this list may also be added the vertical c.g. and roll inertia, as it has already been shown
how those mass properties affect the normal loads at the tires, and thereby the drift angles.
The essence of automotive directional stability may be gleaned from the illustrative
example just given of the three vehicles of varying weight distribution. However, such an
explanation falls far short of an authentic analysis. Such an analysis begins with an
assessment of the geometry of a vehicle undergoing a slight turning action, either as the result
of a lateral disturbance force or an input from the steering wheel.
55

Rev. B

SAWE Paper No. 3528


Category Number 31.0

From the geometry of the situation the differential equations of motion are written as
if the vehicle were a free body moving in space, but generally in only 2-DOF (1 translational,
1 rotational); these equations consist of the summation of forces in the lateral direction, and
the summation of moments about the vehicle c.g., in accord with the concept of Dynamic
Equilibrium 80. The analysis leading to the following discussion was essentially (other
modeling results may have been blended in) based upon the bicycle model; the geometry of
which is presented in Figure 10.4:

Figure 10.4 BICYCLE MODEL TURNING GEOMETRY


The resulting differential equations are solved in accord with the appropriate classical
math methodology. Depending on the complexity of the model, the resulting expressions may
include:
80

Although Newtons (1643-1727) Second Law is the basis for the solution of the problem, the concept of
Dynamic Equilibrium results from the work of Jean le Rond dAlembert (1717-1783), and is sometimes
referred to as DAlemberts Principle.

56

Rev. B

SAWE Paper No. 3528


Category Number 31.0

1) Understeer Coefficient, symbolized as K us is expressed in


radians or degrees. This presents a more general way (than comparing
specific drift angle values f and r ) of classifying a vehicle
condition as stable (understeering): K us > 0, neutral: K us = 0, or
unstable (oversteering): K us < 0, and can take various forms 81, one
of which is:

(Eq. 10.1)

2) Characteristic Speed provides a means of comparing designs with


respect to the degree of understeer present 82. Mathematically, it is the
speed at which the steering angle to make a turn of radius r 83 is
equal to 2l wb /r (that is, the steering angle is twice that determined by
the low speed Darwin steering geometry l wb /r, see Figure C.2):

(Eq. 10.2)

3) Critical Speed is a speed at which an oversteering vehicle can


switch from initially controllable (slight oversteer) to very unstable
(high oversteer) behavior:

(Eq. 10.3)

Note that the system of units conversion factor 57.3 is now negative.
This is in recognition of the fact that an oversteering vehicle will have
a negative K us .
4)

Steering Angle, as required at the front wheels to make a steadystate turn of radius R (see Figure 10.4) is another measure of
under/over steer:

81

Reference [14], pages 161-164. The formulation used is different from that of Eq. 10.1, but the essence is the
same.
82
Ergo, both vehicles must be essentially understeering.
83
The turn radius R is measured from the turn center (TC) to the vehicle c.g.

57

Rev. B

SAWE Paper No. 3528


Category Number 31.0

(Eq. 10.4)

5) Static Margin is a determinant like K us for classifying a


vehicle condition as stable (understeering): SM > 0, neutral: SM
= 0, or unstable (oversteering): SM < 0; but as a dimensionless
fraction with the advantage of relating the longitudinal c.g.
placement to the Neutral Steer Point for that condition (LCG
NSP = SM l wb ) 84:

(Eq. 10.5)

There are a number of other significant relationships, as well as variations on the


above, which result from the bicycle model 85 analysis or are obtained from similar analyses
that differ in the complexity of the model and/or the details of the approach. The advantages
of obtaining such relationships from an analysis of the equations of dynamic equilibrium are:

1) Insights are obtained, such as the existence of a critical speed (Eq.


10.3) or the exact role played by the longitudinal c.g. location (Eq.
10.5) in obtaining stability for a particular design possessed of unequal
tire characteristics (cornering stiffnesses) front to rear, which would be
much more difficult to obtain otherwise. Obviously, the greater the
magnitude of a positive K us or SM the more stability is present.
However, if the goal is to obtain the maximum amount of lateral
acceleration, then the closer these values are to zero then the closer the
design is to its maximum lateral acceleration capability given a
particular set of tire characteristics (which is a lot more sophisticated
judgment than the 50/50 longitudinal weight distribution conclusion
of Chapter 4).
2) Whole areas within the realm of linear behavior can be determined as
stable or unstable without detail calculations of the sort carried out in
84

This is all very similar to aircraft stability theory wherein the relationship between the center of gravity and
the center of pressure is expressed in terms of percent of the Mean Aerodynamic Chord (MAC). Typical
passenger car SM values may range between +0.03 to +0.07.
85
The utilization of the bicycle model means that lateral weight transfer and roll effects are neglected.
Also, the analysis is based on linearized tire functions, so the results are limited in applicability to the lower
levels of the performance spectrum under 0.4 gs (i.e., the linear portion of the Fy, N, curves).

58

Rev. B

SAWE Paper No. 3528


Category Number 31.0

Chapter 9 where considerable work went into the investigation of only


two points in the Ford Fiestas operational envelope: the 4-up
condition at 0.5 gs and at 0.7 gs. Given enough computational
resources, it would be possible to investigate a vast number of such
points so as to gain an understanding of the lateral acceleration/
directional stability performance for a specific vehicle throughout its
operational envelope, regardless of whether the region is linear or not.
However, that is appropriate for detail design; in the conceptual design
stage the sort of theoretical relationships obtained herein are the most
useful.
In conclusion, it is good idea to review the limitations of the theoretical
relationships obtained. Remember that the model did not inherently include the effects
of weight transfer or roll 86, that the model was based on low speed (less than 0.4 gs)
large radius (greater than 24.0 ft or 7.3 m) turns in the steady-state condition which
means small angles (less than 10 deg or 0.175 rad); all of which means that whatever
value the relations have is restricted to the linear tire behavior zone. Given all those
restrictions and the various simplifications inherent in the theory, its amazing that the
results are as useful (at least for passenger car design) as they are.

86

Therefore there is also no accounting for the effects of roll steer or roll camber, in addition to the effects
covered in Chapter 5. These and other effects (aligning torque variation with lateral load, etc.) constitute a large
area of design not covered in this paper.

59

Rev. B

SAWE Paper No. 3528


Category Number 31.0

11 SAFETY

Just what safety is, and what is to be considered safe, is ultimately a matter of
philosophy. However, once a safety standard has been established, then the matter comes into
sharper focus; either the standard is complied with or it is not, and a question regarding
compliance can be settled by empirical means. However, there still remains the question as to
whether the standard is adequate.
That automotive lateral acceleration, rollover, transient behavior, and directional
stability all have to do with safety is understood, but no standards of performance in these
areas have established, at least not as a federal regulation with all the force of law. That leaves
possible only some general statements regarding those automotive characteristics and safety.
So, in general, the higher the maximum lateral acceleration the safer a vehicle should
be, although there will always be some possible occurrence of reckless driving and/or adverse
road situations resulting in an accident no matter how many gs the vehicle may record on the
skidpad. It should also be obvious with regard to safety that transient behavior in maneuver
should be as smooth and as short-lived a phase as possible and that, with regard to directional
stability, road going vehicles sold to the general public should maintain a healthy dose of
understeer throughout the operational envelope.
Regarding rollover, however, there is something more to be said with regard to safety
than just some vague statement urging a high level of performance. Just as with lateral
acceleration, raising the level at which out-of-control behavior ensues is well and good, but
again no matter how high the lateral acceleration threshold there will always be cases of
exceedence. When a vehicle slides off course into guard rails, other vehicles, etc., the
passengers have the protection afforded them by carefully engineered crush zones and
restraints such as air bags and seat belts; when a vehicle rolls over the passengers essentially
have only that protection as may be afforded by divine providence.
As noted in Chapter 8, rollover accidents in the US for 2009 accounted for a
disproportionately large percentage of vehicle accident fatalities even when properly seatbelted in, about 12% or 2821 deaths 87. If the unbelted 5475 that died in roll-over accidents in
2009 had been belted in, then a fair percentage of them would have died anyway, adding
substantially to that 12% figure. This is a situation that has been notable for a long period of
time, and the obvious expedient of strengthening the roof structure, thus completing the

87

Reference [9], page 1.

60

Rev. B

SAWE Paper No. 3528


Category Number 31.0

supposedly inviolate nature of the passenger crash cage, has been consciously and
deliberately avoided by the US automotive manufacturers 88 for four decades.
In 1973 Federal Motor Vehicle Safety Standard (FMVSS) 216, Roof Crush
Resistance, went into effect. This was a very limited test of roof strength as it involved a
quasi-static 89 pushing at constant (slow) speed of a large area steel plate against the most
robust portion of the roof in a test with a total force equal to 1.5 times vehicle empty
weight 90. That standard was so obviously inadequate that finally in 2009 (36 years later) an
improved standard was forced through, which involved applying a force of 3 times the vehicle
weight. This new standard will begin to be phased in by 2013 with 100% compliance required
of each manufacturer by 2017. Even then, vehicles of GVWR exceeding 10,000 lb will be
exempt.
Of course, a great many people have died (not to mention those suffering quadriplegia,
paraplegia, and various other spine and brain injuries) during the struggle to obtain a roof
crush standard, and it is certain many more will do so despite the revised standard as such
limited quasi-static loadings in no way reflect the magnitude of the dynamic loadings actually
experienced in real rollover accidents. This has been a matter of common knowledge for as
long as the matter has been in discussion; this is evidenced by the fact every motor sport
governing body in existence that sanctions, or did at one time sanction, racing of production
motor vehicles required additional structure to resist roof crush.
NHRA, NASCAR, SCCA, and many other such organizations have a long history of
requiring that roll bars of a very definite quality (often integrated into complete roll
cages) be installed on vehicles used in competition. The effectiveness of such reinforcement
has been observed innumerable times in crashes which involved high speed dynamic impacts
to the roof structure but resulted in nothing more than a shaken driver. That private
organizations, and very often even private individuals, have long achieved that which seems
impossible for the federal government and major corporations to do gives cause for concern.

88

Volvo has been traditionally an outstanding exception with regard to roof crush standards, not only in
comparison to US manufacturers, but in comparison with automotive manufacturers in general. Three years after
buying Volvo in 1999, Ford told Volvo executives that their view on one of the most contentious areas of
automotive safety was out of step with Ford's and had to change. Ford executives, certain e-mail messages
suggest, were concerned that Volvo's view on roof strength would be used against Ford in rollover cases, a
potentially expensive liability concern.
89
This ignored the dynamic nature of most rollover accidents. SAE J996 is a dynamic drop test, but the
automotive industry fought against the application of such a test as a federal standard.
90
The wording was Empty weight, or 5,000 lb, whichever is less, not GVWR. Vehicles in excess of 6,000 lb
GVWR were excluded altogether from need to comply with the standard.

61

Rev. B

SAWE Paper No. 3528


Category Number 31.0

12 CONCLUSIONS

This paper has concentrated on the influence of mass properties on certain aspects of
automotive performance associated with acceleration in the lateral direction, and in so doing it
was inevitable that various non-mass properties parameters would be encountered, if only to
note their influence and then exclude them from further discussion. Therefore, if the reader
wishes to obtain a more complete and thorough understanding of automotive performance in
this regard, then he is referred to the many fine texts listed in the References section of this
paper (References [1], [7], and [14] are highly recommended).
The mass properties conclusions (plus a few non-mass properties conclusions) to be
drawn from this paper are, in chapter by chapter order:
Chapter 1 Consideration of a turning vehicle free body in plan view (2D) highlights
the significance of translational (mass) and rotational (yaw) inertias in determining the
magnitudes of the lateral traction forces at the tire/road contact patches necessary to
accomplish the maneuver.
Chapter 2 Vehicle weight and its distribution determines the magnitude of the
static normal loads at tire/road contact patches; the traction potential that can be generated at
those patches varies in accord with those normal loads. The coefficient of traction is itself a
function of normal load, thereby resulting in a highly nonlinear relationship between normal
load and traction.
Chapter 3 - A simple two dimensional study of the weight transfer along an axle
due to lateral acceleration produces two equations: one for the approximate lateral
acceleration at which slide will occur, and one for the approximate lateral acceleration at
which rollover will occur. These accelerations, in gs, are numerically equal to the coefficient
of traction necessary to reach those accelerations. The most significant mass property
parameters for these lateral acceleration limits are vehicle weight and c.g. height.

( )

=
62

(EQ. 3.5)

(EQ. 3.6)

Rev. B

SAWE Paper No. 3528


Category Number 31.0

Chapter 4 - Studying maximum lateral acceleration from just the simple models

utilized up to this point places limits on the conclusions drawn. These limits were
encountered, some perhaps implicitly, throughout the development of the theory. To keep
things straight in mind it is best to explicitly enumerate the conclusions and the limits on
those conclusions now. The limitations are:
1)
2)
3)
4)
5)

Steady-state turning: constant angular velocity, no braking


or drive torque at wheels, no change in turn radius.
Conventional automotive configuration: four wheels, two
axles in tandem.
No roll; suspension characteristics not taken into account
(front/rear roll stiffness, compliance, camber change, etc.).
The front and rear track dimensions are equal.
The tire characteristics (b, m, etc.) are equal all around.

These limitations circumscribe the utility of the conclusions. However, the


conclusions are still of value as the weight and c.g. are the most basic (but not sole)
determining factors of maximum lateral acceleration levels. High lateral acceleration levels
can be obtained, even in apparent violation of the following conclusions, by tailoring of
suspension characteristics, tire characteristics, and track relationship front to rear. Violation of
the conclusions for maximizing lateral acceleration is often done because there are other
important design criteria (directional stability, longitudinal acceleration, braking, ride) the
pursuit of which requires compromise. However, design cannot stray too far from the
following conclusions if lateral acceleration is of any priority:
1)
Minimize weight, the lighter the load on the tires the
better the relative traction.
2)
Locate the vehicle c.g. so that there is an even static
weight/area load on all tires (except in case of special
purpose vehicles like some race cars that make left or right
hand turns exclusively).
3)
Minimize c.g. height; the lower the c.g. the less
weight transfer for a given lateral acceleration, and so the
more even the dynamic load on the tires, laterally and
longitudinally, which results in greater traction.
4)
Maximize the track, which reduces the effect of
weight transfer, again tending to even the dynamic loads
and thereby increase traction.
5)
Maximize b (increase tire contact area and tread
compound stickyness), and minimize m (brace the tread,
keep it flat against the ground).
63

Rev. B

SAWE Paper No. 3528


Category Number 31.0

The mass properties conclusions may be more succinctly stated: for maximum lateral
acceleration a minimum weight and a 50/50 weight distribution is the ideal. However, there
are other performance criteria which militate against this perfect balance; thus necessitating
the manipulation of other parameters to attain an acceptably high lateral acceleration level
without perfect balance. In such a case, it should be remembered that, no matter how high
the lateral acceleration level, it will never be as good as what could have been achieved had the
weight balance been perfect.
Chapter 5 In Chapter 3 a rigid (no roll) model was used to develop the equations for
the limiting lateral accelerations" and ; in Chapter 5 the inclusion of roll
effects in the vehicle model revealed that roll adversely affects (reduces) these lateral
acceleration limits due to a lateral shift in sprung mass c.g. toward the outer wheels.
Therefore, to the five factors previously identified as necessary for maximizing lateral
acceleration levels can now be added a sixth:
6)
Minimize roll, which causes an adverse lateral shift in
the sprung weight (and often a similar shift in the unsprung
weight if the suspension is of an independent type).
The measure of sprung mass roll under lateral load is called roll gain. The roll gain
for the 1980 Fiesta, which seems limited in lateral acceleration to about 0.71 gs, calculated
out to about 9.2 deg/g, while the 1984 Corvette Z51 had an empirically determined lateral
acceleration of 0.95 gs and a roll gain of 2.0 deg/g 91; this tends to illustrate the association of
maximum lateral acceleration with minimum roll.
Chapter 6 The transient condition represents a period of transition (response time)
from one steady state condition to another steady state condition, and the design objective is
to minimize the duration and smooth the character of this transition as much as possible. To
this end the mass properties objectives are the minimization of the roll (I x ) and yaw (I z )
inertias, plus the minimization of the Oscillation Center location factor X which involves
the yaw radius of gyration K z and the longitudinal c.g. locating lengths l f and l r :

X = Kz2/( lf x lr )
In its simplest form, this is yet another argument for minimizing the yaw radius of
gyration K z (the numerator) while getting the weight distribution as close to 50/50 as
possible (maximizing the denominator l f x l r ).

91

Reference [8], pages 39-40. The reported figures were GM test results.

64

Rev. B

SAWE Paper No. 3528


Category Number 31.0

Chapter 7 The terms oversteer, neutral steer, and understeer are introduced and
defined as characteristics of a basic vehicle handling tendency which is observable in a
skidpad test. Also measurable in a skidpad test is the maximum lateral acceleration capability
of a vehicle. Because of the possibility of lateral asymmetry in vehicle construction, the most
important of which would be an offset lateral (not on the vehicle centerline) c.g., the skidpad
test must be run at least twice, once CW and once CCW, with the results averaged. The
skidpad size (radius, diameter) is very significant, as the larger the skidpad the higher the
maximum lateral acceleration obtained due to decreased angularity and possibly greater
aerodynamic download (race cars). A 100 ft (30.5 m) radius is the most common, and
probably constitutes the minimum acceptable size, for passenger car testing; much larger sizes
are appropriate for vehicles designed to operate while generating significant down force (race
cars).
It should go without saying that skidpad, and all other performance testing, should only be
carried out for a vehicle in a standard, reproducible weight condition (curb weight plus
driver, etc.). However, the obvious is often the most frequently overlooked, which would tend
to account for the wide discrepancies in some reported skidpad test results for the same
vehicle 92.
Chapter 8 There are numerous means to obtain an empirical rating for a vehicles
rollover resistance. However, the most significant rating used is calculated, the SSF (Static
Stability Factor), which is numerically equal to the calculation in Chapter 5:

(EQ. 8.1)

The SSF is used by the NHTSA as part of the NCAP for rollover resistance ratings
because there is a strong correlation between the SSF and accidents involving vehicle
rollover 93. However, that correlation is such that it can be improved. For instance, it has been
noted that the correlation improves when certain other parameters are added to the regression
analysis, such as vehicle wheelbase. However, because the link between such other
parameters and rollover cant be incontrovertibly demonstrated, as it can for c.g. height and
track width, those other parameters are not used in the determination of rollover rating of one

92

Reference [8], pg. 40: 84 Corvette Z51 rated at 0.95 gs (GM test). Reference [17], pg. 65; 84 Corvette Z51
rated at 0.84 gs (C/D test). Test weights are not reported, but even standard curb weights can vary by 100 lbs
or more.
93
Reference [2], page 1.

65

Rev. B

SAWE Paper No. 3528


Category Number 31.0

to five stars 94. So, use of the SSF results in certain anomalies, such as the Ford Pinto and the
Chevrolet Vega obtaining the same SSF, yet one was significantly more inclined to be
involved in rollover accidents than the other 95.
It is largely because of this indication of something lacking in rollover rating
methodology, and manufacturers pressure supporting an empirical dynamic method, that
the NHTSA has come to include the results of the Fishhook test along with the SSF
(heavily biased toward the SSF) in establishing NCAP rollover ratings. Yet this is hardly an
ideal situation; the Fishhook test is time consuming, arduous, and expensive. Consequently
the NHTSA has taken the expedient of not actually subjecting every vehicle to the test; some
are rated according to similarity with other vehicles which have undergone the test, based
on vehicle configuration, suspension type, etc. This is hardly an ideal state of affairs, and it is
hard to see where a significant improvement in methodology is obtained by including a
dynamic test when implemented in such a fashion.
In the interest of improving the NHTSA NCAP rollover resistance rating system, while
reducing effort and expense, the following modest proposal is made. The SSF is only of value
as a first order estimate of a vehicles resistance to rollover, and is very conservative in
nature (actual roll over occurs much more easily) 96. It doesnt take into account the basic
suspension characteristics that so strongly influence roll as demonstrated in Chapter 5. That
situation would be largely corrected if the formula for SSF as used for rollover resistance
rating were revised to:

SSF =

(EQ. 12.1)

The symbols d y and d z represent the dimensional changes due to roll as was
illustrated in Figure 5.3 (d y = h r cos s , d z = h r sin s ):

94

Such precise calibration is typical of all NCAP ratings.


The Ford Pinto and the Chevrolet Vega are US automotive products of the 1970s era; the NHTSA rated
vehicles for rollover resistance solely on the basis of the SSF during the years 2001-2003. However, the Pinto
and Vega were included in studies leading to the adoption of the SSF as a rating tool. NHTSA began to use the
combination of the SSF and dynamic maneuver for establishing rollover resistance ratings in 2004.
96
Reference [7], page 312.
95

66

Rev. B

SAWE Paper No. 3528


Category Number 31.0

Figure 5.3 ROLL GEOMETRY


As those dimension changes are dependent upon the roll angle s that angle must be
determined using Equation 5.1. That leaves only the question of what lateral acceleration level
a y to use in the Equation 5.1 computation of s ; this author suggests use of that value of
a y obtained by the original first order approximation. However, caution must be taken to
ensure that the s value obtained is not beyond the physical capability of the automobile in
question to achieve. The limiting value of s at the point the vehicle sprung mass rolls into
contact with the suspension snubbers can easily be determined by a quick study of the
suspension system geometry, or by simple empirical means.
The result of this modest proposal should be a significant increase in correlation 97 and a
considerable improvement in productivity and expenditure if the Fishhook test is
consequently abandoned for rating purposes. To counter manufacturer objections that to rely
on even the modified SSF alone for establishing rollover resistance ratings is still not fully
taking all pertinent factors into account, the presence of supposed mitigating factors such as
the installation of Electronic Stability Control (ESC) could be given the same emphasis as the
Fishhook tests pass or fail had in previously influencing the overall combined rating.
Chapter 9 An automobiles lateral behavior is almost totally dependent upon its tires
drift angles, and a tires drift angle is totally dependent upon the magnitude of the normal,
lateral, and longitudinal forces incumbent upon the tire; all these forces are weight and mass
properties (c.g., roll/pitch/yaw inertia) driven to varying degree. The most important
distinction to keep clear in ones mind when dealing with tire forces is the difference between
97

The correlation can possibly be further improved in that the dy, dz dimensional changes apply to the sprung
mass c.g. and are here applied to the vehicle c.g. as a good approximation. This was done to keep the matter
simple, as simplicity is one the advantages of SSF usage. However, at the expense of somewhat more
calculation, the corresponding total vehicle dimensional changes can be easily determined.

67

Rev. B

SAWE Paper No. 3528


Category Number 31.0

a potential traction force and a required traction force. Up to the point of slide the required
traction forces will be less than the potential, and the required force distribution among the
tires is assumed to reflect the potential force distribution.
The drift angle distribution will reflect the required force distribution, or vice versa,
depending upon the perceived order of cause and effect. In accord with the drift angles at the
wheels the resultant geometry will determine the vehicle path (maneuver radius R) and the
vehicle attitude (yaw angle ). There are many factors which influence the tire traction
force/drift angle relationship, but the most fundamental is normal load which relates back to
vehicle weight and weight distribution (mainly c.g. location).
Chapter 10 - The terms oversteer, neutral steer, and understeer, previously
introduced in Chapter 7 as relating to observable handling behavior, acquire additional
significance in Chapter 10 as the terms are now related to vehicle directional stability. The
subject of vehicle stability is immense and complicated beyond the scope of this paper; only
an overview of ground vehicle stability, to the extent that it furthers basic understanding of
the role of mass properties, is attempted.
The essence of vehicle directional stability is a matter of obtaining sufficiently
understeering behavior by loading the front wheels, both statically and dynamically, so as to
ensure that the front wheels will always be operating at greater drift angles than the rear
wheels. Unfortunately, this also ensures that the front axle will have a lower " than
would be the case otherwise. So, to the six factors necessary for maximizing lateral
acceleration levels can now be added a seventh:
7) Minimize stability to the bare minimum necessary to maintain control on
the skidpad. The means used to enhance stability (nose down roll axis
inclination, etc.) do so by ensuring a weight bias to the front (especially
the front outer wheel in the dynamic case) thus violating the 11th
Commandment Thou shall keep thy contact patch pressures as even as
possible all around if thou wish for thy vehicle to hath maximum lateral
acceleration.
One of the key relationships resulting from a directional stability analysis and
highlighted in Chapter 10 is the Understeer Coefficient. This coefficient is
symbolized as K us and expressed in radians or degrees; the formulation for this
coefficient is:

68

(Eq. 10.1)

Rev. B

SAWE Paper No. 3528


Category Number 31.0

Since the coefficient is based on a front (subscript f) to rear (subscript s) comparison


of the lateral force (assumed to be in proportion to the front W f and rear W r vehicle
weight distribution) to cornering stiffness ratios, stability (understeering) can readily be
determined (K us > 0) for whole areas within the linear tire region.
Another key relationship is the Static Margin, which is symbolized as SM and
expressed as a portion of the vehicle wheelbase:

(Eq. 10.5)

The greater the magnitude of a positive SM the more stability is present. However, if the
goal is to obtain the maximum amount of lateral acceleration, then the closer SM is to zero the
closer the design is to its maximum given a particular set of tire characteristics. This a more
sophisticated judgment than the Chapter 4 conclusion that a 50/50 longitudinal weight
distribution is the means to maximum lateral acceleration; the SM takes weight distribution
and tire characteristics into account.
Chapter 11 This chapter, termed Safety, was not originally envisioned as a part of
this paper; although safety issues are inherent in all aspects of vehicle performance. This
papers topic is the role mass properties play in regard to lateral acceleration performance;
any associated safety issues were to be given just passing mention. However, it was
impossible to study the matter of vehicle rollover without becoming aware of the issue of roof
crush resistance, and the horrific role the automobile industry has played for decades in
denying the obvious 98. Hence the creation of this chapter as a sounding board for the authors
opinion, even though the issue really has little to do with mass properties, but much to do
with human cupidity.
The essence of the problem is that once you dance with the Devil its hard to get free. The
auto industry drifted into a situation of inadequate roof strength during a period of a general
unawareness of safety issues, and to forthrightly admit now the truth about roof strength
would open the door to thousands, if not millions, of lawsuits by the injured and/or the
relatives of the deceased. So the industry continues to stonewall against the inevitable while
thousands more suffer the adverse effects.
The solution to all this is a clean slate; legislation establishing a mandate for the NHTSA
to rigorously enforce new realistic dynamically based roof crush standards, while limiting the
98

Denying the obvious has long been a tradition in the automotive industry. Maurice Olley relates in his
memoirs how Harold Wilson, head of the Automobile Manufacturers Association, in an address to the Surgeon
Generals Conference in November of 1959, denied that automobile exhaust was in any way responsible for air
pollution (Milliken, Chassis Design, Principles and Analysis, page 577.)

69

Rev. B

SAWE Paper No. 3528


Category Number 31.0

liability incurred by the industry for all previously incurred claims, is the obvious and rational
solution. However, since it is obvious and rational, it probably has little chance of occurring
anytime soon.

70

Rev. B

SAWE Paper No. 3528


Category Number 31.0

REFERENCES

[1] Bastow, Donald; Car Suspension and Handling, Plymouth, Devon; Pentech Press Ltd.,
1980.
[2] Boyd, Patrick L.; NHTSAS NCAP Rollover Resistance Rating System,
Washington, DC; NHTSA Paper 05-0450, June 2005.
[3] Costin, Michael; and David Phipps, Racing and Sports Car Chassis Design,
Cambridge, Mass.; Robert Bentley, 1967.
[4] Campbell, Colin; Design of Racing Sports Cars, Cambridge, Mass.; Robert Bentley,
1976.
[5] Campbell, Colin; The Sports Car, Cambridge, Mass.; Robert Bentley, 1969.
[6] Campbell, Colin; New Directions in Suspension Design, Cambridge, Mass.; Robert
Bentley, 1981.
[7] Gillespie, Thomas D.; Fundamentals of Vehicle Dynamics, Warrendale, PA; SAE R114, 1992.
[8] Grable, Ron; Formula One Chevy: A Technical Overview of the Worlds Best
Handling Production Car, Motor Trend, pages 37-40, March 1983.
[9] Highway Loss Data Institute, Insurance Institute for Highway Safety, Rollover and
Roof Strength, www.iihs.org, March 2011.
[10] King-Hele, Desmond; Erasmus Darwins Improved Design for Steering Carriages
and Cars, Notes Record Royal Society 56, London, pages 41-62, 2002.
[11] Lamar, Paul; More About Cornering Power, Road & Track, pp. 127-131, October
1969.
[12] Martinez, Rosendo; Vehicle Cornering Stability and C.G. Limits, Los Angeles,
CA; SAWE #1930, 1990.

71

Rev. B

SAWE Paper No. 3528


Category Number 31.0

[13] McKibben, Jon; Group 7 Competition, Two Seat Racing with No Holds Barred,
Car Life, pages 20-23, January 1968.
[14] Milliken, William F.; and Douglas L. Milliken, Race Car Vehicle Dynamics,
Warrendale, PA; SAE R-146, 1995.
[15] NHTSA, Rollover Prevention Docket No. NHTSA-2000-6859 RIN 2127-AC64,
2000.
[16] Simanaitis, Dennis; Road Testing at R&T, R&T Road Test Annual, pp. 12-16,
1982.
[17] Sherman, Don; Chevrolet Corvette, Four C/D Apostles Bring Home Seven
Revelations, Car and Driver, pages 63-68, October 1983.
[18] Smith, Kevin; The New Corvette, Road & Track, pages 25-39, March 1983.
[19] Wiegand, B.P.; Mass Properties and Automotive Vertical Acceleration, Los
Angeles, CA; SAWE #3521, 2011.
[20] No Byline, Corvette vs. Ferrari 308GTBi Quattrovalvole, vs. Porsche 928S, vs.
Porsche 944, Road & Track, pages 53-59, August 1983.

72

Rev. B

SAWE Paper No. 3528


Category Number 31.0

AUTHORS BIOGRAPHICAL SKETCH

Brian Paul Wiegand, now retired, was a Senior Weight Engineer and Mass Properties
Handling Specialist for the Mass Properties Analysis and Control Group of Northrop
Grumman Corporation, Bethpage, NY. He is a 1972 graduate of Pratt Institute, Brooklyn, NY,
and a licensed Professional Engineer registered in the State of New York (# 58470). He
continues to be an active member of the Society of Allied Weight Engineers and of the
Society of Automotive Engineers. He has presented three SAWE papers: Mass Properties
and Automotive Longitudinal Acceleration (SAWE #1634, 1984), The Basic Algorithms of
Mass Properties Analysis & Control (SAWE #2067, 1992), and Automotive Mass
Properties Estimation (SAWE #3490, 2010). He has also published two articles: The
Weight and C.G. Implications of Obtaining Maximum Automotive Lateral Acceleration
Levels (SAWE Journal Weight Engineering, Winter 1982-83), and The Mystery of
Automotive POI Values (SAWE Journal Weight Engineering, Spring 2011). Recently he
reaffirmed his long-standing interest in automotive engineering by attending the SAE Seminar
Vehicle Dynamics for Passenger Cars and Light Trucks (Troy, MI; August 11-13, 2009)

73

Rev. B

SAWE Paper No. 3528


Category Number 31.0

APPENDICES

74

Rev. B

SAWE Paper No. 3528


Category Number 31.0

APPENDIX A SYMBOLISM

A translational acceleration (in/sec2, ft/sec2, m/sec2), also can mean the same as
l f per SAE J670e (in, cm).

ay

Lateral translational acceleration, usually in gravity units.

a yslide

Lateral translational acceleration, usually in gravity units, sufficient to cause


loss of traction and consequent sliding.

a yoverturn

Lateral translational acceleration, usually in gravity units, sufficient to cause


vehicle overturn (rollover).

The Greek lower case letter alpha meaning rotational acceleration, usually
in radians/second2.

The basic tire coefficient of traction (dimensionless), also can mean the same
as l r per SAE J670e (in, cm).

BASIC

Beginners All-purpose Symbolic Instruction Code is a computer


programming language dating back to 1964; many variations have appeared
since.

The Greek lower case letter beta used to represent the yaw angle (angle
between the vehicle centerline and the tangent line to the turning circle) of a
maneuvering vehicle.

Cs

Cornering stiffness, or F y /, in lb/deg, kg/rad, or kN/rad. See Figure 9.2.

C sf

The cornering stiffness of the tires on the front axle (lb/deg, kg/rad, or kN/rad).

C sr

The cornering stiffness of the tires on the rear axle (lb/deg, kg/rad, or kN/rad).

C/D

Car and Driver magazine is a major US reviewer and tester of new production
automobiles; it was founded as Sports Car Illustrated in 1955.

CW

The clockwise direction of rotation.

CCW

The counter-clockwise direction of rotation.

CG

Center of gravity.

CP

Aerodynamic center of pressure.

75

Rev. B

SAWE Paper No. 3528


Category Number 31.0

C4

The fourth generation of the General Motors, Chevrolet Division, Corvette


sports car, which is now in its sixth generation (C6).

DI

The Dynamic Index, equal to the sprung mass pitch radius of gyration
squared divided by l f l r (dimensionless).

The Greek lower case letter delta used to symbolize the steer angle (degrees,
radians)

dy

The lateral shift of the c.g. of the sprung mass in roll (in, cm).

dz

The vertical decrease of the height of the sprung mass c.g. in roll (in, cm).

A force, as in F = ma (lb, kg).

Fd

A lateral disturbance force (lb, kg).

Ff

A lateral force in the ground plane acting at the front axle (lb, kg).

Fr

A lateral force in the ground plane acting at the rear axle (lb, kg).

Fx

A tire traction force acting in the longitudinal direction (lb, kg).

Fy

A tire traction force acting in the lateral direction (lb, kg).

F axle

A force in the ground plane acting at an axle (lb, kg).

F xf

The longitudinal force in the ground plane acting at a front axle (lb, kg).

F xfi

The longitudinal force in the ground plane acting at a front inner wheel (lb,
kg).

F xfo

The longitudinal force in the ground plane acting at a front outer wheel (lb,
kg).

F xr

The longitudinal force in the ground plane acting at a rear axle (lb, kg).

F xri

The longitudinal force in the ground plane acting at a rear inner wheel (lb,
kg).

F xro

The longitudinal force in the ground plane acting at a rear outer wheel (lb,
kg).

F yf

The lateral force in the ground plane acting at a front axle (lb, kg).

F yfi

The lateral force in the ground plane acting at a front inner wheel (lb, kg).
76

Rev. B

SAWE Paper No. 3528


Category Number 31.0

F yfo

The lateral force in the ground plane acting at a front outer wheel (lb, kg).

F yr

The lateral force in the ground plane acting at a rear axle (lb, kg).

F yri

The lateral force in the ground plane acting at a rear inner wheel (lb, kg).

F yro

The lateral force in the ground plane acting at a rear outer wheel (lb, kg).

GM

General Motors Corporation, once the largest automotive manufacturer in the


world, and one of the worlds largest manufacturing enterprises, was founded
in 1908. On 10 July 2009 a new General Motors Corporation was created
from the bankrupt remains of the original company.

GTP

Grand Touring Prototype was an automotive racing category created under the
jurisdiction of the IMSA.

HRM

Hot Rod Magazine is the oldest existing magazine (US) dedicated to the
modification and customization of production vehicles; it was founded in 1948.

h cg

The height of the center of gravity above the ground plane (in, ft, cm, m).

hr

The distance between the RC and the CG (in, ft, cm, m).

h rc

The height of a roll center above the ground plane (in, ft, cm, m).

h rrc

The height of the rear roll center above the ground plane (in, ft, cm, m).

h frc

The height of the front roll center above the ground plane (in, ft, cm, m).

A vehicle rotational inertia, units lb/in2, lb/ft2, kg/cm2, kg/m2.

Ix

The vehicle roll rotational inertia, units lb/in2, lb/ft2, kg/cm2, kg/m2.

Iz

The vehicle yaw rotational inertia, units lb/in2, lb/ft2, kg/cm2, kg/m2.

IMSA

The International Motor Sports Association is a major sanctioning body for


automotive competition, and was founded by John Bishop (formerly of the
SCCA) and his wife Peggy in 1969.

A radius of gyration, usually pitch per SAE J670e (in, ft, cm, m).

K us

The understeer coefficient , allows for a determination of stability

in the linear region of the vehicle operational envelope. Units: degrees or


radians. See Equation 10.1.

77

Rev. B

SAWE Paper No. 3528


Category Number 31.0

Ky

The sprung mass pitch radius of gyration (in, ft, cm, m).

Kz

The vehicle yaw radius of gyration (in, ft, cm, m).

k roll

The total vehicle roll stiffness (lb-ft/deg, Nm/deg).

k froll

The roll stiffness at the front axle (lb-ft/deg, Nm/deg).

k rroll

The roll stiffness at the rear axle (lb-ft/deg, Nm/deg).

lf

The distance from the front axle to the c.g. (in, cm).

lr

The distance from the front axle to the c.g. (in, cm).

l wb

The wheelbase distance from the front axle to the rear axle (in, cm).

MIRA

The Motor Industry Research Association (Ltd) was founded 1949 in the UK
and is now a major automotive test facility with aerospace, rail, and other
branches.

M/T

Motor Trend magazine is a major automotive publication (US) founded in


1949.

Mass, as in F = ma, or the rate of tire traction coefficient decrease with


normal load.

The normal load at a tire/road contact patch (lb, kg).

NASCAR

The National Association for Stock Car Auto Racing was founded by Bill
France Sr. in 1947. It sanctions and governs various auto racing events, and is
the largest sanctioning body for stock car racing in the United States.

NCAP

The NHTSAs New Car Assessment Program was created in 1979 in response
to Title II of the Motor Vehicle Information and Cost Savings Act of 1972.

NHRA

The National Hot Rod Association was founded by Wally Parks in 1951. It is
one of the worlds largest auto sports governing organizations, specializing in
drag racing.

NHTSA

The National Highway and Traffic Safety Agency is a subdivision of the US


Department of Transportation (DOT). Its mission statement is Save Lives,
Prevent Injuries, Reduce Vehicle Related Crashes.

Ni

The normal load at an inside wheel tire/road contact patch (lb, kg).

78

Rev. B

SAWE Paper No. 3528


Category Number 31.0

No

The normal load at an outer tire/road contact patch (lb, kg).

N fi

The normal load at an inside front wheel tire/road contact patch (lb, kg).

N fo

The normal load at an outer front tire/road contact patch (lb, kg).

N ri

The normal load at an inside rear wheel tire/road contact patch (lb, kg).

N ro

The normal load at an outer rear tire/road contact patch (lb, kg).

OC

The Oscillation Center is a point in vehicle plan view about which the
vehicle initially tends to rotate during the transient phase of a maneuver.

The Greek lower case letter theta with subscript s meaning roll angle of
the sprung mass (weight), units in degrees or radians.

PET

The Personal Electronic Transactor was a personal computer introduced


by Commodore International in 1977.

RC

Roll center, which is a point about which the sprung mass will rotate under
lateral acceleration due to suspension kinematics.

R&T

Road & Track magazine is an automobile enthusiasts magazine founded at


Hempstead, NY, USA in 1947; it was perhaps the most technically orientated
of all major automotive publications.

The radius of a turn (ft, m).

r rf

The rolling radius at the front axle (in, cm).

r rr

The rolling radius at the rear axle (in, cm).

SAE

The Society of Automotive Engineers (now SAE International).

SAWE

The Society of Allied Weight Engineers, which is the premier professional


society for engineers involved in mass properties analysis and control.

SCCA

The Sports Car Club of America is a descendent of the Automobile Racing


Club of America founded in 1933 by brothers Miles and Sam Collier, which
was dissolved in 1941 at the entry of the US into WW II. In 1944 it was
reconstituted as the SCCA, and began sanctioning road racing events in 1948.

SPR

Side Pull Ratio, a test to determine the lateral acceleration rollover point.

79

Rev. B

SAWE Paper No. 3528


Category Number 31.0

SSF

This represents the Static Stability Factor, which is calculated as t/2h cg


(dimensionless). The SSF is numerically equal to a yslide , and also
numerically equal to the coefficient of traction necessary for overturn to occur
without the aid of tripping, etc.

SM

The Static Margin is not only a measure of stability, but allows for
determination of the location of the longitudinal c.g. in order to achieve that
degree of stability. See Equation 10.5.

TC

The Turn Center which is the origin of a vehicle maneuver path radius.

TTR

The Tilt Table Ratio, a test to determine the lateral acceleration rollover
point.

Vehicle track width, from tire tread center to tire tread center (in, cm).

tf

Vehicle front track width, from tire tread center to tire tread center (in, cm).

tr

Vehicle rear track width, from tire tread center to tire tread center (in, cm).

The Greek lower case letter mu meaning the tire coefficient of traction =
b mN (dimensionless).

Vehicle velocity, sometimes called speed (mph, ft/sec, kph, m/sec).

V char

The characteristic velocity provides a means of comparing the degree of


understeer present in one design to the degree of understeer present in another
design; see Equation 10.1. Units: ft/sec, m/sec, mph, or kph.

V crit

The critical velocity is a speed at which a vehicle can switch from initial
mild oversteer (controllable) to wild oversteer behavior; see Equation 10.3.
Units: ft/sec, m/sec, mph, or kph.

A vehicle weight, usually curb weight plus driver (lb, kg).

Wf

The portion of vehicle weight borne by the front axle (lb, kg).

Wr

The portion of vehicle weight borne by the rear axle (lb, kg).

Ws

The weight of the vehicle sprung mass (lb, kg).

W sf

The weight of the front axle vehicle sprung mass (lb, kg).

W sr

The weight of the rear axle vehicle sprung mass (lb, kg).

80

Rev. B

SAWE Paper No. 3528


Category Number 31.0

W us

The weight of the vehicle unsprung mass (lb, kg).

W usf

The weight of the vehicle unsprung mass at the front axle (lb, kg).

W usr

The weight of the vehicle unsprung mass at the rear axle (lb, kg).

A quantity similar to DI, but calculated for the total vehicle from K z 2/( l f
l r ), and is dimensionless.

X,Y,Z

Center of gravity coordinates (in, ft, cm, m).

The Greek lower case letter psi used to symbolize the drift (slip) angle
(degrees, radians).

fi

The Greek lower case letter psi used to symbolize the drift (slip) angle at
the front inner (fi subscript) tire (degrees, radians).

fo

The Greek lower case letter psi used to symbolize the drift (slip) angle at
the front outer (fo subscript) tire (degrees, radians).

ri

The Greek lower case letter psi used to symbolize the drift (slip) angle at
the rear inner (ri subscript) tire (degrees, radians).

ro

The Greek lower case letter psi used to symbolize the drift (slip) angle at
the front outer (fo subscript) tire (degrees, radians).

81

Rev. B

SAWE Paper No. 3528


Category Number 31.0

APPENDIX B LATERAL ACCELERATION PROGRAM

The following computer program was used to generate the data for all the graphs in
the first few chapters of this paper. It was written for the Commodore 2001 PET
minicomputer 99 and, with slight modifications, may be used with any other computer utilizing
BASIC. If the single axle focus of this program is changed to reflect two axles separated
longitudinally by some wheelbase length l wb , and the normal loads N i and N o are
determined for each axle on the basis of Eq. 5.2/5.3 (front) and Eq. 5.4/5.5 (rear), and if
provision is made to calculate the necessary lateral forces at each axle to generate lateral
acceleration a y and rotational acceleration along with the associate slip angles, then the
basis for a lateral acceleration (maneuvering) simulation program will be obtained. However,
the devil is in the details!

MAXGLAT.BAS

1 PRINT PRG FOR MAX LAT ACCEL: TIRE FORCES @ AY


2 PRINT WIEGHT & CG EFFECTS PRIMARY
3 PRINT UNITS: LBS, INCHES
4 PRINT INPUT AXLE WT LOAD, TIRE COEFFS, CG HEIGHT, TRACK
5 INPUT W, B, M, HCG, T
6 AA=2*M*W
7 AB=HCG^2
8 AC=T^2
9 AD=(AA*AB)/AC
10 AE=(AA/4)-B
11 AF=SQR(1-(4*AD*AE))
12 AS=(AF-1)/(AD*2)
13 AOT=T/(2*HCG)
14 PRINT MAX LAT ACCEL (SLIDE), MAX LAT ACCEL (OVERTURN), IN GS
15 PRINT AS,AOT
16 PRINT DECISION: INPUT 1 FOR MAX ACCEL ITERATION, 0 FOR FORCE @AY
17 INPUT DEC
99

This represented the state of the art in personal computers circa 1979.

82

Rev. B

SAWE Paper No. 3528


Category Number 31.0
18 IF DEC>0 THEN GOTO 3
19 PRINT INPUT AY IN GS
20 INPUT AY
21 WTF=(W*HCG*AY)/T
22 NS=W/2
23 NI=NS-WTF
24 N0=NS+WTF
25 FI=(B-M*NI)*NI
26 FO=(B-M*NO)*NO
27 FAXLE=FI+FO
28 PRINT AY, INNER FORCES (NORM & LAT), OUTER FORCES (NORM & LAT) , FAXEL
29 PRINT AY,NI,FI,N0,F0,FAXEL
30 PRINT DECISION: INPUT 1 FOR FORCE ITERATION, 0 FOR END
31 INPUT DEC
32 IF DEC>0 THEN GOTO 19
33 END

83

Rev. B

SAWE Paper No. 3528


Category Number 31.0

APPENDIX C - STEERING

In 1758 Erasmus Darwin (1731-1802), future grandfather of Charles Darwin (18091882), invented the so-called Jeantaud-Ackermann steering system (Figure C.1d) in order to
rectify some of the shortcomings of the axle steering system (Figure C.1b) then in use for
horse drawn carriages, coaches, and wagons 100. The main shortcoming of the axle steering
system was that the steered wheels had to be significantly smaller in radius than the nonsteered wheels in order to clear the body/chassis at full lock; this incurred significantly greater
road shock transmission as such shock varies inversely with the size of the wheel 101. Another
shortcoming was that axle steering reduced the vehicle resistance to overturn when cornering.
A third shortcoming, which would not be readily apparent until the birth of the automobile,
was that the encountering of a bump in the road by one of the steered wheels would generate a
large impact moment about the axle pivot point. The drivers of some of the earliest
automobiles would experience these moments as severe kick-back through the tiller.

Figure C.1 STEERING GEOMETRY SYSTEMS

100
101

Reference [10], page 41.


Reference [19], pp. 20-21 & 118-130.

84

Rev. B

SAWE Paper No. 3528


Category Number 31.0

At around 1816, Darwins steering innovation was apparently independently


reinvented by a German carriage manufacturer named Georg Lankensperger. A certain Rudolf
Ackermann became acquainted with Lankensperger and his invention and, acting obstensibly
as Lankenspergers agent, obtained an English patent (number 4212) in his own name in
1818. However, the geometry of the steering arrangement in Ackermanns patent illustration
was both extreme (control arm inclination with respect to the longitudinal axis was only about
6.5 degrees, while Darwins arrangement involved angles of 23 to 30 degrees) and difficult to
discern. Consequently, Ackermann Steering became associated with the vastly inferior
geometrical layout of Figure C.1c 102 (zero degrees inclination of the steering arms with the
longitudinal axis).
In 1878 French carriage maker Charles Jeantaud corrected the Ackermann Steering
back to the original (Darwinian) concept of Figure C.1d. However, to this day the steering
geometry in question is still known as Ackermann Steering, although Rudolf Ackermann,
despite all his scheming perfidy, never profited from his patent. Life is not only unfair, but
often ironic.
The following diagram is of Darwin Steering geometry in a low speed (no drift
angles) turn, which is the only kind of turn for which such geometry is truly valid:

Figure C.2 DARWIN STEERING GEOMETRY (LOW SPEED TURNS)

102

An even earlier (1714) claimant for this flawed but significant alternative to axle steering may be a certain
M. Du Quet who possibly indicated such steering on his wind-driven carriage; see Reference [10], page 61.

85

Rev. B

SAWE Paper No. 3528


Category Number 31.0

APPENDIX D DERIVATION OF EQUATION 3.5


By inspection of Figure 3.2, it can be seen that the lateral acceleration level for an axle
at which slide will be incipient is the point A where the lateral inertial load W a y is just
equal to the lateral potential traction force (with weight transfer) of Equation 3.4:
=

+
+

Implicit in this is the assumption that the tires will have the same lateral traction
parameter values on both the inside and the outside ends of an axle in a turn (b = b i =
b o , m = m i = m o ), which is a reasonable assumption except for some circle track race
vehicles. With that in mind, the above simplifies into:

W =

This can be divided through by W and rearranged to become:

This is a quadratic equation of the form aX2 + bX + c = 0 and can be solved by


using the standard equation for the quadratic roots:

So, by substituting a y for X, etc., and ignoring the negative root which has no
physical meaning here, the result is obtained:

( )
Q.E.D.
86

Rev. B

SAWE Paper No. 3528


Category Number 31.0

APPENDIX E ROLL STIFFNESS DETERMINATION


For an independent suspension, the roll stiffness at that particular axle line is 103:

Where:

(EQ. E.1)

k roll = Independent suspension roll stiffness, lb-ft/degree.


k cs = Combined stiffness of suspension spring at wheel + tire, lb/in.
t

= The track width at the axle line, in.

Exempli gratia, the 1980 Ford Fiesta S had a MacPherson strut front suspension with
effective suspension spring and tire stiffness at the wheel of 123.9 lb/in, and a front track of
52.5 in, so the front roll stiffness is 104:
k roll = 123.9 (52.5)2 / 1375 = 248.4 lb-ft/degree 105
For a non-independent suspension, the roll stiffness at that axle line is 106:

Where:

(EQ. E.2)

k roll = Non-independent suspension roll stiffness, lb-ft/degree.


k s = Effective suspension spring stiffness at the mount to axle, lb/in.
t sb = Spring base distance between spring mounts along an axle, in.
k t = The spring stiffness of a tire on the axle, lb/in.
t

= The track width at the axle, in.

Continuing with the example, the Fiesta had a dead beam/coil spring rear suspension
with effective suspension spring and tire stiffness k cs of 142.7 lb/in (assuming a k t of

103

Reference [14], page 603. The combined stiffness kcs would be obtained using the formula for springs in
series: kcs = ks kt / ks + kt.
104
Reference [6], page 191.
105
Ibid, page 192 quotes 218.3 lb-ft/deg (296 Nm/deg), which is used throughout Chapters 5 and 9 of this paper.
106
Reference [14], page 604.

87

Rev. B

SAWE Paper No. 3528


Category Number 31.0

1800 lb/in, by reverse calculation the k s is 155.0 lb/in). The rear track was 52.01 in 107, and
the rear spring base is 36.0 inches (as estimated by this author), so the rear roll stiffness is:
k roll = (155.0 (36.0)2)(1800(52.01)2) / (1375(155.0(36.0)2 + 1800(52.01)2)) = 140.3 lb-ft/deg
However, the Fiesta also had an anti-roll bar (a.k.a. stabilizer bar) at the rear, and its
resistance to roll of 37.0 lb-ft/deg must be added (per springs in parallel) in to the quantity
just calculated:
k roll = 140.3 lb-ft/deg + 37.0 lb-ft/deg = 177.3 lb-ft/deg 108
It should be noted that both the roll resistances calculated herein for the 1980 Ford
Fiesta S (1.1 liter, European version) are both about 14% greater than figures quoted by
Reference [6]; the reason for this is unknown, but for the sake of comparability of results the
quoted figures were used for calculations within the body of this paper.

107
108

Reference [6], pp. 191-192.


Ibid, page 192 quotes 154.9 lb-ft/deg (210 Nm/deg), which is used throughout Chapters 5 and 9 of this paper.

88

Rev. B

SAWE Paper No. 3528


Category Number 31.0

89

Vous aimerez peut-être aussi