You are on page 1of 22

> ARR No.

126
:u ';N
<0
:u .,...
z 0
~ Z
~ a::
N a::
0) ct

ISSN 0518 - 0728

Speeds on Curves:
Side Friction Factor Considerations
J.R. McLean

mg _ _ _ ----- ---- - - -

AUSTRALIAN ROAD RESEARCH BOARD


RESEARCH REPORT
AUSTRAUAN ROAD RESEARCH BOARD

REPORT SUMMARY
THE PURPOSE OF THIS REPORT

is to document the development of the design side friction factor values adopted for the NAASRA (1980) Interim
Guide to the Geometric Design of Rural Roads.

THIS REPORT SHOULD INTEREST


those concerned with geometric road design and the relationships between road alignment and driver behaviour.

THE MAJOR CONCLUSIONS OF THE REPORT ARE

On curves with a speed standard less than 90 lan/h, drivers typically develop side friction factors considerably in excess of
the values assumed for traditional geometric design practice. This anomaly can be removed by re-defining curve design speed
so that it represents an estimate of the 85th percentile curve speed, and deriving limiting design values of side friction factor
from observed driver behaviour.

AS A CONSEQUENCE OF THE WORK REPORTED, THE FOLLOWING ACTION IS RECOMMENDED


The side friction factor design values and associated curve speed prediction relations be used as a basis for alignment design
which is consistent with actual driver behaviour.

RELATED ARRB RESEARCH


P359 - Analysis of Rural Road Traffic Operations
ASH - NAASRA Geometric Standards and Policy Review

--------------------------------
McLEAN, J.R. (1983) : SPEEDS ON CURVES: SIDE FRICTION FACTOR CONSIDER-
ATIONS. Australian Road Research Board. ARR No. 126. 16 pages including 11 figures
and 6 tables.
-----------------------------------------
KEYWORDS : Speed/bend (road)/alignment/degree of curvature/design speed/skidding
resistance/friction
ABSTRACT : Observed speeds on rural road curves in the lower speed standard range have
been found to be in excess of the assumed design speed, and it has been suggested that an
alternative design procedure, based on the predicted 85th percentile speed, be developed
for these curves. Within this context, an examination is made of the side friction factor
basis of current curve design standards, and the relationships between current design
values of side friction, available friction, and the values typically utilised by drivers. It is
CUT OUT INFORMATION found that the driver behaviour interpretation of the side friction factor-design speed
relationship cannot be substantiated by empirical evidence, and that the side friction
RETRIEVAL CARD . ........ . values actually utilised are often greatly in excess of the design values. Empirical data
are used to develop an alternative set of design friction factor values, more in keeping with
current driver behaviour, which can be used in conjunction with predicted 85th percentile
curve speeds. It is argued that the proposed values should be treated as maxima, and that
the object of alignment design, in terms of the side friction criterion, should be to produce
alignments based on driver behaviour, such that the 85th percentile driver does not utilise
side friction factors in excess of the design values.

'"'liii~immr
0003990 "Non IRRD Keywords
Austral ian Road Research Board
Research Report ARR No. 1 26

AUSTRALIAN POAD
RESEARCH BOARO
25 MAY 1933

LIB8ARY

SPEEDS ON CURVES: SIDE FRICTION FACTOR


CONSIDERATIONS

by
J.R. McLEAN
Principal Research Scientist
Australian Road Research Board

Australian Road Research Board


500 Burwood Highway, Vermont South
Victoria 3133, Australia
March 1983
Research Report ARR No. 1 26

The ARR series of reports was created by the Board to quickly and economically reproduce the conclusions of
road and road transport research subjects for those associated with roads and the road transport industry. Unlike
other technical publications produced by the Australian Road Research Board, ARR reports are not reviewed by
referees or ARRB Directors.
Although this report is believed to be correct at the time of its publication, the Australian Road Research Board
does not accept responsibility for any consequences arising from the use of the information contained in it. Peo-
ple using the information contained in the report should apply, and rely upon, their own skill and judgment to the
particular issue which they are considering .
Reference to, or reproduction of this report must include a precise reference to the report.

ISBN 0 86910 102 1 Report


ISBN 0 86910 103 X Microfiche
ISSN 0518 - 0728
Research Report ARR No.1 26

CONTENTS.
ABSTRACT

NOTATION

1. INTRODUCTION

2. TRADITIONAL APPLICATION OF SIDE FRICTION


FACTOR VALUES IN CURVE DESIGN STANDARDS
2.1 Basis of Curve Design Standards
2.2 Interpretation of the Side Friction Factor-Design Speed
Relationship 2

3. CRITICISMS OF THE SIDE FRICTION FACTOR


CRITERION 2
3.1 The Design Speed Assumption 2
3.2 Assumptions in the Design Equation 3
3.3 ' Side Friction Factor-Speed Relationsh ip and Driver
Behaviour 3

4. PROPOSED APPROACH TO SIDE FRICTION


DESIGN 3

5. PAVEMENT SKID RESISTANCE 4

6. SIDE FRICTION FACTORS EMPLOYED IN


PRACTICE 4
6.1 Side Friction Factors Computed from Observed Speeds 4
6.2 In-Vehicle Measurements of Lateral Acceleration 6
6.3 Relationships Between Observed Values. Actual Friction
Demand and Design Values 8

7. DEVELOPMENT OF ALTERNATIVE DESIGN VALUES 8


7.1 Rationale of Approach 8
7.2 Consideration of the Side Friction Factors-Speed Plot 9
7.3 Distributions of Utilised Side Friction Values 11
7.4 Consistency with Traditional Standards 11
7.5 Proposed Design Values 11

8. DISCUSSION 11
8.1 Basis of Proposed Values 11
8.2 Minimum Curve Radius 14
8.3 Safety Margin 14

9. CONCLUSIONS 16

REFERENCES 16
NOTATION

lateral acceleration (g) R curve radius (m)


e curve superelevation (m/m) skid number

side friction factor side force coefficient

side friction factor assumed for v speed (m/s)


design
v speed (kmlh)
side friction factor appropriate to 85th percentile speed (km/h)
the 85th percentile speed
g acceleration due to gravity design speed values currently
(m/SfO!c 2 ) applied (kmlh)
ARR No . 126 1

1. INTRODUCTION Safe vehicle operation on alignment features is


very speed dependent, and the design speed concept
The Australian Road Research Board (ARRB) Project has evolved as a basis for alignment design. Design
200 (Speeds on Curves) was directed at investigat- speed was originally defined as 'the maximum
ing the relationships between vehicle operating reasonably uniform speed which would be adopted by
speeds and the geometric properties of curves on the faster driving group of vehicle operators, once
two-lane roads. Data collected during the course of clear of urban areas ' (Barnett 1936) , but is now used
the study included vehicle spot speeds (curve and more as a means of 'matching ' alignment elements
approach) and road geometry measured on 72 curves (McLean 1978) . However, an implicit assumption of a
with design speed standards nominally in the range design vehicle travelling at the design speed is re-
40 to 120 km/h . Where possible, separate speed tained in the application of the concept. Values of
studies were made for each direction of travel with design speed ranging from 40 to 130 km/h, in 10 km/h
each direction being considered as a separate study increments, are typically used in the derivation of
site. A total of 120 sites resulted . curve design standards from eqn (2) .
McLean (1981) described how the data were The National Association of Australian State
analysed to produce a family of curve speed predic- Road Authorities (NAASRA) (1970 and 1973) give
tion relations for 85th percentile car speeds, and sug- design values of (as a decreasing function of design
gested that these be used as a basis for design stan- speed as shown in Fig. 1. These are typical of the
dards for low-speed al ignments (less than 90 km/h). design values found in most design manuals and
This report exam ines the side friction factor (f) guides, and are referred to in this report as the 'tradi-
basis of curve design standards and the relationships tional ' design values . The basis of these functions is
between design values of (, available friction, and the discussed in the next sub-section.
(values typically uti lised by drivers. Side friction fac - Maximum values of superelevation are based on a
tors computed from the measured curve speeds and number of considerations including the stability of
geometric parameters are used as the basis for a set slow moving, high centre of gravity vehicles, the
of design values, consistent with current driver lengths required for superelevation transitions, and
behaviour, which can be used in conjunction with the appearance. A target maximum value of 0.06 m/m is
predicted 85th percentile speeds. advised in NAASRA (1973), with values as high as
The curve speed prediction relations and the 0.10 in difficult situations and 0.12 in severe terrain.
associated set of side friction factor design values With the controlling values of design speed, side
have been adopted for the NAASRA (1980) Interim friction factor and superelevation fixed as described
Guide to the Geometric Design of Rural Roads . above, eqn (2) can be used to derive minimum curve
radii as a function of design speed. The values so
derived in NAASRA (1973) are shown in Table I. It
2. TRADITIONAL APPLICATION OF should be noted that there is a certain arbitrariness in
the (-design speed relationship, so that it would be
SIDE FRICTION FACTOR VALUES IN equally valid to arbitrarily assign design values of
CURVE DESIGN STANDARDS radius and superelevation without reference to eqn
(2) or the relationships shown in Fig. 1.
2.1 BASIS OF CURVE DESIGN STANDARDS
When curves are provided with radii greater than
Minimum curve design standards are based on two the general minimum pertaining to the design speed,
criteria : the normal practice is to reduce superelevation
(a) ensuring that the side friction demand is not ex- below the general maximum, but not to an extent cor-
cessive for the design speed; and responding to the increase in radius. Thus , vehicles
travelling at the design speed on above minimum
(b) ensuring that the available sight distance is radius curves will be operating with f values below
adequate for the design speed. those shown in Fig. 1.
This report is concerned solely with the first criterion. 30 _ _ _ Intersection

The (criterion is based on the force balance for a __ Rural Road

pOint mass on a superelevated surface travelling at \


\
constant speed along a circular arc. With small angle \
approximations, the force balance yields the equa-
tion :
\ ,,
"
(1)
~
en 10
or

e + f = V2 / 127 R (2)

oL-__-L__~____~--~---=~~~--~
This is the design equation used for the derivation of 02040
Design Speed (km/ h)
minimum curve standards satisfying the ( criterion .
However, the values of certain parameters need to be Fig. 1 - Relationship between design values of side friction
set for application of the equation. factor and design speed (NAASRA 1970 and 1973)
2 ARR No. 126

TABLE I
MINIMUM RADIUS OF HORIZONTAL CURVES BASED ON
SUPERELEVATION AND SIDE FRICTION REQUIREMENTS
(from NAASRA 1973)

Minimum Radius (m)


Design
Speed Mountainous General Flat
(kmlh) Terrain Maximum Terrain
e = 0.12 e = 0.10 e = 0.06
40 40 45 50
50 70 75 90
60 105 110 130
70 145 155 185
80 195 210 (g55)
90 255 280 340
100 330 360 440
110 400 435 530
120 495 540 670
130 785

2.2 INTERPRETATION OF THE SIDE FRICTION 3. CRITICISMS OF THE SIDE


FACTOR-DESIGN SPEED RELATIONSHIP
FRICTION FACTOR CRITERION
There appears to be confusion in the standard design
texts regarding the '-speed design relationship in that
it is regarded, alternatively, as a design criterion , or Criticisms of the 'criterion as applied to curve design
as an explanation of driver behaviour. standards fall into three categories :
(a) the design speed assumption.
The design criterion interpretation is based both (b) the constant radius assumption and the point
on providing a factor of safety over actual pavement mass equation , and
friction and on providing comfortable travel condi-
tions for drivers (AASHO 1965). The skid resistance (c) the '-speed relationship as a representation of
of pavements tends to decrease with increasing driver behaviour.
speed in a manner similar to the design relationship,
but is consistently in excess of the design' values so
that a safety margin is provided. The provis ion of
comfort is based on the claim that drivers or 3.1 THE DESIGN SPEED ASSUMPTION
passengers will experience discomfort from 'values
greater than the design values. Application of eqn (2) and the '-speed relationship
(Fig. 1) is only truly valid for a vehicle travelling at, or
The driver behaviour interpretation assumes that below, the design speed. While this may be a
drivers adjust their speed according to the comfort reasonable approach in railway engineering, where
criterion represented by the '-speed relati onsh ip . the maximum speed for each curve is stipulated, there
This reasoning is implicit in the NAASRA (1973) is no guarantee that drivers are either aware of, or
statement that: 'The maximum value of , which most drive at, the design speed.
drivers will tolerate in negotiating a curve range from
0.19 at 40 km/h to 0.11 at 130 km/h' . Similar reason-
ing is used in NAASRA (1 970) to justify the higher The results of Project 200 have shown that actual
values of ' for intersection design, viz. 'it is found that vehicle operating speeds tend to vary along the road ,
drivers turning on curves of small radius at intersec- and are often quite inconsistent with the design speed
tions will accept a lower degree of comfort'. (McLean 1981) . In particular, curves based on design
speeds below about 90 km/h are consistently driven
at speeds in excess of the design speed.
The emphasis on a comfort criterion appears to
have been adopted from railway engineering practice,
where passenger comfort was an important con- The latter point is recognised in AASHO (1965)
sideration in the setting of curves. It was proposed as to justify the use of very conservative design values
a basis for road curve design by Barnett (1936) of 'for the lower range of design speeds: 'The lower
following field tests on volunteer subjects, and was rates at the low speeds are desirable since drivers
subsequently adopted by AASHO. However, the sup- tend to overdrive low design speed highways'. The
posedly 'comfortable' design values were not NAASRA (1973) design va lues at low design speeds
checked against actual driver behaviour. are less conservative.
ARR No. 126 3

3.2 ASSUMPTIONS IN THE DESIGN EQUATION direct bearing on driver curve speed selection
behaviour.
Equation (2) assumes that a vehicle acts as a point
mass travelling at constant speed on a constant Good and Sweatman (1976) proposed a number
radius . In fact , the driverlvehicle is a complex of alternative models for driver speed behaviour on
dynamic system which will seldom be in a fixed, or curves, but lacked the necessary empirical data to
steady-state , condition. test these models. They considered that an (-speed
interpretation of driver speeds on curves could lead
The ability , or inability, of the driver/vehicle to erroneous conclusions, and argued that (be tre-
system to maintain lateral positioning within the lane ated as an outcome of driver behaviour rather than as
is of specific relevance here. It is not possible for a an explanation for it.
driver to steer a vehicle perfectly. Even if it were, en-
vironmental disturbances such as road surface ir-
regularities constantly act on a road vehicle , causing
it to deviate from its path. While traversing a curve, a
driver must also adjust the position of the vehicle 4. PROPOSED APPROACH TO SIDE
within the lane, so that deviations from a constant path
radius are required. FRICTION DESIGN
This problem has been investigated experimen- Despite the criticisms of the ( approach to curve
tally by Glennon and Weaver (1971) . They found that design standards, designers still consider that ( is
vehicle paths on curves with radii ranging from 250 to both a necessary and valid design criterion. They
875 m were such that, at some point along the curve, argue that the difference between (and the actual
vehicle path curvature was more severe than that of tyre/pavement friction available is the most important
the curve, and that the discrepancy between minimum factor affecting safe vehicle operations on curves.
path radius and curve radius increased with increas- Furthermore, as maintenance of pavement skid resis-
ing curve radius. Thus, the actual maximum ( ex- tance is outside their realm, geometric design must
perienced by a vehicle traversing curve will generally focus on the friction demand, or side friction factor,
be greater than that indicated by eqn (2), and the dis- contribution to the difference. In effect, the criticisms
crepancy will tend to be greater for larger radius, or given in the previous section do not invalidate the
higher speed curves . Good and Joubert (1977) found safety margin interpretation of the design f-speed
that, for substantial deviation angles (more than 90 0 ) relationship .
and strong constraints on drivers' lateral positioning, If (is to remain as a fundamental curve design cri-
this relationship extends to curves of lower radius (18 terion, recent driver behaviour studies suggest that it
to 116 m). However, for low radius curves with should be based on a different conceptual framework
smaller deviation angles, or with room for lateral from that traditionally employed. In particular, (de-
manoeuvring, drivers tend to 'cut the corner' such that mand should be seen as an outcome of driver
the actual maximum (is smaller than that given by eqn behaviour rather than as a representation of it. The
(2) . objective of alignment design for the (criterion is,
McLean (1974) suggested that , when ( is com- then, to ensure that the design driver (some chosen
puted from observed speeds and eqn (2), the percentile of the driver population) does not exceed
difference between minimum vehicle path radius and the design values of (.
curve radius may account for part of the observed Given the complexities of driver/vehicle
trend of decreasing (with increasing speed. behaviour on curves, it is not feasible to base design (
values (for substitution in eqn (2)) on measured pave-
3.3 SIDE FRICTION FACTOR-SPEED RELATIONSHIP ment skid resistance and an arbitrary safety factor.
AND DRIVER BEHAVIOUR However, through experience, drivers have learnt to
As a consequence of its driver behaviour studies, the operate with their own safety margins. While utilised (
Human Factors Research Group at the University of does not appear to have a direct bearing on driver
Melbourne has been critical of the (-speed relation- speed selection, it can be assumed that learnt driver
ship as an interpretation of driver curve speed behaviour includes an appreciation of the bounds on
behaviour (Good, Rolls and Joubert 1969; Good the friction forces which can reasonably be de -
1975; Good and Joubert 1977; Good and Sweatman manded. Thus, the upper range of ( values typically
1976) . There would appear to be two major objec- utilised in practice can form the basis of design (
tions. values which include a subjective safety margin as
applied. by drivers.
(a) There is no empirical evidence that drivers res-
pond to actual, or subjectively predicted (in the Implementation of such an approach requires that
selection of curve speed, rather than some the design (values be based on a realistic assess-
other parameter. ment of the behaviour and comfort tolerance of
modern drivers, and the pavement skid resistance that
(b) Owing to the inter-relationship between speed, can be anticipated. Attempts to introduce additional
curve geometry and ( (eqn(2)), attempts to conservatism by designing to low ( values (as per
represent driver behaviour as an (-speed rela- AASHO 1965) are quite unrealistic when the resultant
tionship may cloud the more fundamental issue alignments are driven faster than the design speed
of driver speed behaviour and road conditions such that higher (values are used in practice.
(including curve geometry) .
McLean (1981) showed that, for design speeds
The failure of superelevation to appear as a greater than about 100 kmlh, traditional curve design
statistically significant variable in the regression practice results in conservative deSigns relative to
analysis of the Project 200 curve speed data (McLean the behaviour of the 85th percentile driver. The lower
1981) supports the argument that (does not have a range of deSign speeds (less than about 90 kmlh)
4 ARR No . 126

would appear to be the crit ical region where amend- exist ing practices within U.S. State Highway Depart-
ment to traditional practice may be necessary. For ments. The recommended values are generally lower
this region , it may be more appropriate to define than the target values of skid resistance rec om-
design speed as the likely 85th percentile speed, and mended in the U.K . by the Ministry of Transport
to base the design criteria on the behaviour of the (1970) .
85th percentile driver. (The 85th percentile approxi-
mates a point of inflection in the normal distribution It can be seen from Fig. 2 that the Kummer and
curve, and, as speeds tend to be normally distributed, Meyer (1967) recommended minimum skid resistance
is likely to repr~sent the point of diminishing returns values are consistently below the likely range of SFC
when designing according to a percentile speed values as measured by the CRB. This can be partly at-
value .) tibuted to the difference in measurement techn iques
for SN and SFC. Skid number is derived from the fric-
The remainder of this report exam ines data on t ion force generated by a fully skidding tyre , and, par-
pavement skid resistance and the values of ( ob - ticularly at high speeds, this tends to be lower than
served in practice. The data from Project 200, being the side force generated by a tyre operating at the
comprehensive and up-to -date observations of the crit ical slip angle as used for the measurement of
curve speed behaviour of Australian drivers, are used SFC.
to derive a design (-speed relationship which is both
consistent with actual driver behaviour and safe with The main purpose of Fig. 2 is to illustrate the
respect to likely pavement skid resistance. range of pavement friction which can be anticipateo
for well maintained rural roads. As the Kummer and
Meyer recommended minimum values appear to
represent a conservative estimate , and as (app roxi-
5. PAVEMENT SKID RESISTANCE mates the fri ction actually demanded by the
The foremost requirement of design values for ( is that driverlveh icle , the Kummer and Meyer values might
they not be excessive in relation to the friction wh ich reasonably be regarded as an upper bound fo r design
a pavement is likely to provide. This section of the values of (.
report considers the results of two studies which
have examined available pavement skid resistance.
Pavement skid resistance is usually expressed 6. SIDE FRICTION FACTORS
as a skid number, friction number, or side force coeffi- EMPLOYED IN PRACTICE
cient. The values quoted, and the term used, are de-
pendent on the method of measurement. However, for This sect ion of the report examines the results of a
number of studies which have looked at the (values
the purpose of the current exercise , these numbers
can be regarded as being approximately equal to the actually used by drivers. There is considerable
wet tyre/pavement coeffic ient of friction multiplied by literature on this topic , and only the more recent in-
ve~tigations carried out on rural roads are con-
100.
sidered here.
Country Roads Board (CRB) (1976) reported an
6.1 SIDE FRICTION FACTORS COMPUTED FROM
investigation of side friction values for the NAASRA
OBSERVED SPEEDS
Review of Road DeSign Standards. As part of this in-
vestigation , an examination was made of curve site Taragin (1954) measured speeds of passenger cars,
side force coefficient (SFC) data collected from at the point of minimllm sight distance, for thc in3ide
routine SCRIM testing carried out by the CRB on the and outside lanes of 35 curves on rural , two-lane
arterial road network of Victoria (see Bethune and highways in the U.S. He presented values for the
Read (1974) for details of SCRIM testing) . Test mean , 90th percentile and 95th percentile for each set
results taken at, or near, the centres of curves were of speed measurements, and the (values correspond-
extracted and grouped according to test speeds at ing to these speeds as computed from eqn (2) . The
the same points, with each group covering a 5 km/h New South Wales Department of Main roads (DMR,
range of test speed. A mean and standard deviation N.SW.) collected similar data during the 1960s, some
were computed for the SFCs within each speed of which appeared in DMR, N.SW. (1969) .
group. The speed of SCRIM on each curve would be
determined by the nominal maximum testing speed McLean (1974) presented plots of 90th percen-
(80 km/h) , road gradients, traffic conditions, or the tile f against 90th percentile speed for the Taragin
safe speed of travel for the vehicle. As the results data, and average ( against average speed for the
were mostly taken from the more lightly-trafficked DMR data. These are reproduced here as Fig. 3.
rural road network, the test speed would often ap- Although it was possible to obtain statistically signifi-
proximate the safe travel speed of the curve . The cant linear regressions of (against speed (p < 0.05) ,
results of the CRB study for speed groups containing there was a great deal of scatter in f values . Only
a significant number of curves (more than 100) are small proportions of the variances could be explained
summarised in Fig. 2. The pOints shown for each group by the regressions.
are the mean SFC values, and the values one and two The 85th percentile (values for the Project 200
standard deviations below the mean. data, as computed by eqn (2) , are shown plotted
Kummer and Meyer (1 967) produced a set of against the speeds from wh ich they were derived in
'minimum recommended skid numbers' for American Fig. 4. The line shown as curve 1 on Fig. 4 represents
highways as a function of mean traffic speed. As the linear regression of ( against speed, which was
shown in Fig. 2, these range from SN = 50 at 16 km/h statistically significant at p < 0.01 . This regression
to SN = 31 at 130 kmlh . Their recommendations were had an ~ value of 0.58, which is considerably higher
derived from a consideration of the technical and than that obtained for either the Taragin or DMR data.
economic feasibility of achieving levels of skid resis- Curves 2 and 3 are explained in Section 7 of the
tance , the likely frictional demands of traffic , and the report.
ARR No. 126 5

80

CRB Observed Values

Mean

60
Mean - 1 s.d.

+'
c:
Q) Mean - 2 s.d.
'u
fQ)
o
u
Q)

...
u
o
u..
Q)
'0
40
en...
...o
Q)
.0
E
::J
Z
:s!
~
en
L Kumme'& Meve,

20

o ~ ____-L______L -____- L______ ~ ____ ~ ______ ~ ____ ~

o 20 40 60 80 100 120 140


Speed (km/h)

Fig. 2 - A comparison of the eRB (1978) side force coefficient


measurements on curves with the Kummer and Meyer (1987)
recommended minimum values for skid number on rural highways
6 ARR No. 126

• Taragin (1954) 90%ile

& --- DMR, N.S.W. Average

.. • •
0.30
, •
r2 = 0.16

••
••
~0.20
... •
...u0
cu
u..
c:
, ..
•• •
...u
.
.2
.;:
u..
Q)

. ..
.. • I .t~
"C

en 0.10
•• ,
..•.. ,..•"
. •
••

0.00
o 20 40 60 80 100 120
Speed (km/h)

Fig. 3 - Plots of side friction factor against the speeds from


which they were derived

6.2 IN-VEHICLE MEASUREMENTS OF LATERAL Ritchie, McCoy and Weide (1968) recorded
ACCELERATION speeds and lateral accelerations for 50 subjects who
Accelerometers mounted in vehicles have been used drove a 180 km rural course containing 227 curves .
in a number of studies to gain direct measures of the The subjects were instructed to drive 'normally'.
lateral acceleration levels employed by test subjects Measurements of lateral acceleration and speed were
when negotiating curves. For steady-state condi- grouped into speed cells of 8 km/h . The drivers were
tions, lateral acceleration expressed in units of g can ranked by their overall mean speed for the course and
be approximately related to f via: divided into five groups of ten, with the ten fastest in
group one and the ten slowest in group five . Mean
lateral accelerations were computed for the data in
each speed cell for each driver group. These were
a (1 + K)f (3) plotted against cell speed, and showed an 'inverse '
y
relationship. The plots for the faster (group one) and
intermediate (group three) driver groups are shown in
Fig. S.
where K = the static roll sensitivity factor of the vehi- Herrin and Neuhardt (1974) recorded speed and
cle , which is typically about 0.2 rad/g . lateral acceleration for two groups of ten subjects
ARR No . 126 7

driving over pre-defined routes on two- lane ru ral each curve within each data group, across driver
roads. The first group repeated a 14 km course con- means were computed for maximum lateral accelera-
taining ten curves (route one) four times . The second tion and speed . These results are shown plotted in
group made two runs over a 22 km course contain ing Fig. 6.
15 curves (route two) . On different runs the subjects
were instructed to drive according to either of two
scenarios: At the anecdotal level , in 1973 the author and a
colleague were required to take a vehicle fitted with
Scenario A - drive as though you are late for an im- advisory speed data logging equipment (including in-
portant interview, and strumentation for measuring lateral acceleration) on
Scenario 8 - drive as though you are on a leisurely an interstate journey. On the return trip , the oppor-
Sunday afternoon drive . tunity was taken to test the performance of the instru-
ments over a protracted period. As the recording
Data were grouped according to the scenario and, for system had been based on a maximum lateral ac-
the first group of drivers, familiarity with the route . For celeration of 0 .5 g, the la!eral acceleration level was

®
0.50

0.40




_ 0 .30

!:!=.

...o •
U
<0
u..
c:
o£o
.;:
u..
I'
Q)

"en 0.20

••
• •

0.10

·1·

o L -____ ~ ______- L______ ~ ______ ~ ____ ~~ ____ ~ ______ ~

o 20 40 60 80 100 120 140


Speed (km/h)
Fig. 4 - Plot of 85th percentile f values against the
corresponding 85th percentile curve speed for the ARRB Project
200 data
8 ARR No. 126

Driver Grou p 1 Despite these limitations, the results considered


v= 73.6 km / h definitely show that the faster group of drivers often
utilise (values considerably in excess of those used
Driver Group 3 as a design criterion . This is particularly so for curves
0.30
v=64.8 km / h of sufficient severity that the faster drivers negotiate
them at speeds less than about 90 km/h. At higher
speeds, the utilised (values are generally below the
design values.

0.20

.g" 7. DEVELOPMENT OF
~
Q) ALTERNATIVE DESIGN VALUES
tiu
<I:
~
7 .1 RATIONALE OF APPROACH
~ 0.10
..J
Tile Project 200 data provide comprehensive infor-
mation on the ( values actually utilised by current
Australian drivers on rural road curves. These might
reasonably form the basis of a set of design (values
for the anticipated actual curve operating speeds
rather than the traditional design speeds.

0~----~----~----~-----7~--~~- As discussed earlier, the critical situation ap-


o 20 40
Speed (km / h)
pears to be curves of speed standard below about 90
kmlh , where the 85th percentile driver tends to
negotiate the curve at speeds in excess of the tradi-
tional design speed and hence, employs ( values
Fig. 5 - Mean lateral acceleration v. speed for the faster and
greater than the traditional design values. The effort
intermediate driver groups of Ritchie et al. (1 968)
has been concentrated on this region, with an attempt
to develop an 85th percentile (value ((ss) appropriate
for the 85th percentile curve speed.
monitored relative to this value . With the 'steady' driv-
ing required for advisory speed survey data collec-
tion, this value was not exceeded. However, with nor- The approaches used have implicitly assumed
mal (for the author and colleague) fast driving , lateral that the observed (ss values are an outcome of driver
accelerations in excess of 0.5 g were not infrequent. curve speed behaviour, rather than a determinant of it.
These observed values have been examined with the
object of determining what should be a reasonable
6 .3 RELATIONSHIPS BETWEEN OBSERVED VALUES, upper limit of , for design purposes at the various
ACTUAL FRICTION DEMAND AND DESIGN VALUES levels of curve operating speed. The view is taken
The values of ( computed from eqn (2) can only pro- that the upper range of 'ss values encountered for the
vide an estimate of the actual tyre/pavement friction Project 200 data are probably a result of poor align-
being utilised by the vehicle. As discussed pre- ment design, and that, with good design, it should be
viously, drivers tend to 'cut the corner' on small radius possible to keep actual 'ss values to a lower level.
curves so that the actual friction utilised is less than However, it is obviously not possible to keep actual
that indicated by eqn (2) . On the other hand, for larger subscript (' ss ) within the very conservative values
radius curves , maximum vehicle path curvature is traditionally assumed for design.
generally in excess of the road curvature, so that the
actual maximum friction util ised is greater than that in-
dicated by eqn (2). As a consequence, the results There are a number of ways in which the Project
shown in Figs 3 and 4 would tend to give an over-esti- 200 data might be used to derive a set of 'ss design
mate of the actual friction utilised for the low speeds, values, and three of these were given detailed con-
and an underestimate for the high speeds. These sideration:
biases should not affect comparisons between com-
(a) a consideration of the conventional plot of 'ss
puted values of (and the design criterion . (As eqn (2)
and Vss data computed for each curve study
is used both to compute the (values, and to apply the site ;
design criterion, the biases will apply equally in both
cases .)
(b) the distribution of utilised 'values for curves
Provided maximum readings are used, direct grouped according to either traditional curve
readings of lateral acceleration from vehicle mounted speed standards, or observed curve operating
accelerometers will take account of variations in speed; and
vehicle path curvature. However, an additional com-
ponent is introduced into the acceleration reading by (c) the relationship between obse r ved curve
the vehicle roll dynamics. Lateral acceleration read- operating speeds, the curve speed standard
ings will typically over-estimate the actual friction based on the current 'criterion, and traditional
demand by about 20 per cent. standards for radius and superelevation.
ARR No . 126 9

0.4 Scenar io A 0.4


Low Famil iar ity
~

c;
s::
.;:0
• •
.~~
~.
~
~
s::
.....
.2
IV
~

~
~
~~
~
~

'*
u 0 .2
U

..
IV
c;;
•••
~ . ~
CII
'ii 0.2
u
u
IV

Scenario A
~

...
CII
IV • ...
CII
IV
High Fam iliarity
...J
...J • Route 2
~ Route 1
~
0 0
0 40 80 120 0 40 80 120
Speed (km/h) Speed (km/h)

0 .4 Scenario B
Low Familiarity

c;
s::
o
.;:
IV
• s::
0
.;:
~

.~ IV
.i 0.2 • ~
~ 0.2
CII ~~ 'ii
u u

·:t.
U u

.....
IV IV
~
~
IV
~
CII
...CII
Scenario B
IV
...J •
~
Route 2 ~
...J
IV

High Famil iarity


Route 1 ~
o ~ __L -_ _ ~_ _~_ _-U~~ __-...J 0
o 40 80 120 0 40 80 120
Speed (km/h) Speed (km/h)

Fig. 6 - Lateral acceleration v. speed as determined by Herrin and Neuhardt (1974)

7 .2 CONSIDER ATI ON OF THE SIDE FRICTION design. As such, it was decided to reject them from
FACTORS- SPEED PLOT considerati on, and recompute the regression equa-
The '85 values corresponding to the observed 85th ti on. This yielded the relationship:
percentile car speeds are plotted in Fig. 4. A linear
regression of , against speed yielded the regression
equation : f85 = 0.584 - 0. 0047 V 85 (5 )

( 2 = 0.60

f 85 = 0.602 - 0.0049 V85 . (4)

which is shown as curve 2 in Fig. 4.


( 2 = 0.58
A design criterion based on curve 2 would result
in approximately half of the observed '85 values being
considered excessive, and it is unreasonable to ex-
which is shown as curve 1. pect alignment design practice to modify driver
behaviour to the extent that all observed values
An examination of the data revealed that two of would fall below the regression line. Raising the line
the '85 values (shown circled in Fig. 4) were ex- by an amount equal to the residual standard devi ation
cessively high , and would have been introducing bias (~O . 0 5), with adjustments for the obvious variation in
into the regression equation. Such excessive values residual variance with speed, would seem to provide
were considered to be well beyond the range of a more reasonable criterion. This is the line shown as
plausible design values , and representative of driver curve 3, and the values corresponding to it are given
behaviour which would be avoided by good alignment in Table II.
10 ARR No. 126

TABLE II
POSSIBLE DESIGN VALUES OF f
INDICATED BY A CONSIDERATION OFTHE
PLOT OF OBSERVED f v. OBSERVED SPEED
85th Percentile
Car Curve Speed f ".;
(kmlh)
50 0.425
60 0 .372
70 0 .319
80 0 .265
90 0.212
100 0 .158
110 0.105
120 0 .051

-0.1 o 0.1 0.2 0.3 0.4 0.5 0.6 0.7


Side Friction Factor (f)

Fig. 7 - Distribution of side friction factors computed from car


curve speeds observed in ARRB Project 200
ARR No. 126 11

7.3 DISTRIBUTIONS OF UTILISED SIDE FRICTION 7.4 CONSISTENCY WITH TRADITIONAL STANDARDS
VALUES
The Project 200 data have displayed a reasonably
The utilised' values were computed for all observed consistent relationship between observed 85th per-
car speeds, and the distribution of these is shown in centile car curve speed and the curve speed standard
Fig. 7. The wide range of , values actually utilised is as given by the traditional' design criterion . This is
indicated by the difference between the 15th pe rc en- shown in Fig. 10.
tile value of 0.030 and the 85th percentile value of
0 .248. If Fig. 10 were regarded as providing a one-to -
one relationship between current curve speed stan -
The curve study sites were subsequently dard and 85th percentile speed, then there would be a
grouped into two sets of 10 kmlh cells, accord ing to corresponding relationship between current minimum
either the traditional curve speed standard, or the ob- curve radius standards (as given in Table I and 85th
served 85th pe rcent ile car curve speed, and separate
, distributions were determined for each group. These
percentile speed . The value of '85
appropriate to the
85th percentile speed for maintenance of this consis-
distributions are shown in Figs 8 and 9 respectively. tency is, from eqn (2) :
The traditional design 'values marked on the distribu-
tions for curve speed standard groupings (Fig. 8) pro-
vide yet further evidence that , on low speed standard
curves, the majorit y of drivers use' values in excess - e (6)
of those traditionally assumed for alignment design.
If the anticipated 85th percentile car speed is to
be regarded as the speed appropriate for design pur-
poses, then the 85th percentile' values for the dis-
tributions based on observed 85th percent ile curve where e represents the specified maximu m
speed groupings (Fig. 9) could provide a reasonable superelevation.
design criterion. The 85th percenti Ie , values for each
As a superelevation of 0 .06 is the target maximum
observed speed grouping are given in Table III.
value employed by most Australian State Road
Authorities, this has been used to compute the '85
values given in Table IV.
TABLE III 7.5 PROPOSED DESIGN VALUES
85th PERCENTILE f VALUES FROM THE The '85 values for the above three interpretations of
UTILISED f DISTRIBUTIONS FOR CURVES Project 200 data (Tables 1/ to IV) are plotted against
85th percentile speed in Fig. 11. The Kummer and
GROUPED ACCORDING TO OBSERVED Meyer (1967) recommended minimum friction values
85th PERCENTILE CAR SPEED are also indicated. As discussed in Section 5, these
Speed Grouping f H'-,
represent a lower bound estimate of the minimum
(kmlh) pavement friction likely to be encountered and, as
such , are regarded as the upper limit for plausible
50 0 .267 design , values .
60 0 .336
70 0.317 Considering the plotted points and the Kummer
80 0 .278 and Meyer recommended values, the 'reverse S'
90 0 .168 shaped curve shown is proposed as a plausible
100 0 .103 design' criterion appropriate for predicted 85th per-
110 0 .075
centile curve speeds. This proposed criterion is
tabulated for 85th percentile speeds up to 90 km/h in
Table V.
It can be seen that for the lowest observed speed For curve speeds, or standards, above 90 kmlh ,
group (50 km/h) the '85 value is inconsistent with the concept of designing for a predicted 85th percen-
overall trends . This is an artifact of the interaction tile speed is not in keeping with the purpose of high
between curve geometry, speed and 'value raised by standard alignments . Here, the objective is to provide
Good and Sweatman (1 976 ), and the limits on the a high level of safety, comfort, and convenience for a
curve speed standards covered by the data. The Pro- wide range of road users. This is evidenced by curves
ject 200 curve sites had a lower bound on curve based on traditional criteria for speeds of 1 20 kmlh or
speed standard (40 km/h) . For a number of reasons, greater, on which few drivers attempt to match the
some of the low standard curves were driven at lower curve speed standard. It is therefore suggested that
speeds than others (e.g. low standard approaches, the traditional design' values and design speed stan-
leisurely trip purposes) , giving rise to lower com- dards be retained for design speeds of 100 km/h or
puted , values. The 50 km/h observed speed group greater.
will contain an over-representation of curves driven
slower than the norm, and the distribution of , values
associated with this group will therefore be below 8. DISCUSSION
that which could be expected for 50 km/h observed
speed curves based on a wider range of curve stan- 8.1 BASIS OF PROPOSED VALUES
dards.
The approach used in arriving at the proposed design
Given the above , it would be prudent to disregard 'values recognised that there is no absolute criterion
the , distribution for the 50 km/h observed speed in terms of driver behaviour or comfort . They are
group in the selection of 'values for design purposes. based on evaluations of observed driver behaviour,
12 ARR No. 126

TABLE IV
85th PERCENTILE f VALUES BASED ON THE RELATIONSHIP
BETWEEN OBSERVED 85th PERCENTILE SPEED, CURVE SPEED
STANDARD AND CURRENT RADIUS STANDARDS
Current
Speed fd Vc (85) fHo
Standard (kmlh)
(kmlh)
40 0 .19 51 0 .3 46
50 0 .17 62 0 .294
60 0 .16 72 0.257
70 0 .15 82 0.228
80 0 .14 89 0 .188
90 0 .13 94 0 .147
100 0 .12 97 0 .109
11 0 0 .1 2 100 0 .08 9
1 20 0. 11 103 0 .065

TABLE V
PROPOSED DESIGN f VALUES
APPROPRIATE FOR T HE PREDICTED 85th
PERCENTILE CURVE SPEED
Speed for Ma ximum Side
Design Fric tion
V ... Factor
(kmlh) f "c,
50 0.35
60 0 .33
70 0 .31
80 0 .26
90 0 .18

TABLE VI
MINIMUM RADIUS OF HORIZ ONTAL CURVES BASED ON
PROPOSED MAXIMUM SIDE FRICTION VALUES AND PREDICTED
85th PERCENTILE CURVE SPEEDS
Speed for Minimum Radius (m)
Design
V ~:; e = 0.12 e = 0.1 0 e = 0.06
(kmlh)
50 40 45 50
60 65 65 75
70 90 95 105
80 135 140 155
90 215 230 265
ARR No. 126 13

100

80
~ ,/~ _ _ _ _ __ Speed Standard Grouping
c
0
';:;
"
~
0
>
.."
c

."
0-

u:.
>
.~
-;
E
u"

-0.1 o 0.1 0.2 0.3 0.4 0.5 0.6 0.7


Side Frictior:o Factor (f)

Fig. 8 - Distribution of side friction factors computed from observed car curve speeds for curves
grouped according to their design speed standard

-0.1 o 0.1 0.2 0.3 0.4 0.5 0.6 0.7


Side Friction Factor (f)

Fig. g - Distribution of side friction factors computed from observed car speeds for curves
grouped according to the observed 85th percentile car speed
14 ARR No. 126

140

120

i:ti •
~
tJ
>
:;:
--E 100
= ..
"tl
3lc.
CI)

...
. . •/
/
./

...'"c 80
/
.,c.~ /
0

U
GO
~
:::l
. /
/
//

../ .
~c 60
/
/
~ /
GO
Q.
III
IX) /
/
40 //

40 60 80 100 120 140


Curve Speed Standard (km/h) V d

Fig. 1 0 - Relationship between observed 85th percentile car


curve speeds and the curve speed standard

with a view to arriving at what should be reasonable 8.3 SAFETY MARGIN


target maximum values for deSign purposes. While a At a superficial level, it might appear that the pro-
number of 'rational ' computations were made as as posed f85 values would produce alignments which are
aid to interpreting the empirical data, there was an ap- less safe than those based on traditional standards,
preciable amount of subjective judgement applied in due to the apparent reduction in friction safety
arriving at the proposed values. Given this , the author margins. This is not the case, as the proposed values
acknowledges that further adjustments may prove are based on the f values which are actually being
necessary. utilised on current alignments. In the lower speed
8.2 MINIMUM CURVE RADIUS ranges, attempts to increase safety by nominating
both lower speeds and lower design f values (as per
The proposed f85 design criterion can be used to AASHO 1965) are quite fruitless, as drivers have been
compute minimum curve radii for the predicted 85th shown to travel the resulting alignments at a speed
percentile curve speed. These are shown in Table VI greater than the nominated speed standard.
for speeds up to 90 km/h. Because of the difference
between the proposed fvalues and the current design The reality of driver behaviour (for Australian dri-
values , the radii given for a specified 85th percentile vers at least) is such that many drivers do operate
speed are less than those quoted for a traditional with a very small friction safety marg in on low stan-
design speed of the same value (Table I) . However, dard curves. The proposed approach of using f85
after allowing for the differences between 85th per- values for a predicted 85th percentile curve speed
centile curve speed and traditional speed standard should lead to safer deSigns, as it would make the
shown in Fig 10, these minimum radii do not differ designer more aware of the implications of design
greatly from the traditional standards given in Table I. deCisions, particularly for low standard alignments. It
This is consistent with the approach taken in the also provides quantitative evidence of the need to
development of the proposed f85 values. maintain pavement skid resistance on such roads.
ARR No. 126 15

The proximity of the proposed '85 values at low centile speed considered in this report. Adjust-
speeds to the Kummer and Meyer recommended ments for this would cause the Kummer and
minimum pavement friction values (Fig. 11) must Meyer curve shown in Fig. 11 to be displaced to
cause concern . However, there are a number of fac- the right.
tors which serve to increase the actual safety margin
over that shown.
(c) On low standard curves with reasonable pave-
(a) Fig. 2 shows that the Kummer and Meyer values ment width, drivers tend to 'cut the corner', so
are more than two standard deviations below that the actual friction utilised is less than the'
the mean friction values on curves measured by value computed from eqn (2).
CRB on routine SCRIM surveys.
At the high speed end, the safety margin is more
(b) The Kummer and Meyer values are based on than adequate to allow for the difference between
mean traffic speed, as opposed to the 85th per- minimum vehicle path curvature and road curvature .


0.40

• Kummer & Meyer


recommended minimum
{ pavement friction

0.30
A''\

-
It)
co
AV,+

,
Proposed
design
values

0.20
'\ ,

• From Section 7.2 (Table II)


\. ,
0.10 + From Section 7.3 (Table III)

From Section 7.4 (Table IV)


\. ,

A "'.

o
o 20 40 60 80 100 120 140
85th Percentile Curve Speed (km/h)

Fig. 11 - Relationships between f", and 85th percentile speed


as a basis for a proposed design criterion
16 ARR No. 126

9. CONCLUSIONS The ARRB speeds on curves data showed that ,


for curves with speed standards less than about 90
A review of design practices relating horizontal curve km /h, 85th percentile speeds and the assoc iated'
design to side friction has found some confus ion in values utilised tend to be in excess of those typically
the interpretation of the traditional ' -design speed assumed for design. For curves of higher standard,
relationship . This appears to have arisen because the driver behaviour tends to be conservative relative to
relationsh ip is treated both as a design criterion and the design assumptions. In keeping with this finding ,
as an explanation of driver speed behav iour. Studies the ARRB data have been used to develop a set of
of driver behav iour have found that this latter in - design 'values appropriate for the predicted 85th
terpretation cannot be substantiated and that , for percentile speeds of curves with speed standards of
design purposes, 'values should be seen as an out- 90 km / h or less. It is suggested that the traditional
come of driver speed behaviour, rather than as a criteria be retained for design speed standards in ex-
determinant of it. cess of 90 km/h .
It is argued that the proposed values be treated
Observational studies have found that drivers fre- as maxima, and that the object of alignment design, in
quently utilise , values much greater than those terms of the 'criterion, should be to produce align-
assumed for design, particularly on low speed stan- ments such that driver behaviour will not lead to the
dard curves . Values in excess of 0.4 are not uncom- 85th percentile driver utilising r values in excess of
mon. the design criterion .

REFERENCES

AMERICAN ASSOCIATION OF STATE HIGHWAY GOOD, M.C., ROLLS , K.J. and JOUBERT, P.N. (1969) .
OFFICIALS (1965). A policy on geometric design of Driver-vehicle behaviour in free-path turns. Transp.
rural highways. (AASHO Washington .) Res. 3, pp. 43-68.
BARNETT, J. (1936). Safe side friction factors and HERRIN, G.D. and NEUHARD1 , J.B. (1974) . An empiri-
superelevation design Highway neseart:h Board. cal model for automobile driver horizontal curve
Proc . Annu . Meet., 16, pp. 69-80. negotiation. Human Factors 16(2), pp. 129-33.
BETHUNE, J.D. and READ , K. (1974) . Road surface KUMMER , HW. and MEYER, W.A . (1967). Tentative
and skid resistance testing vehicle . Proc . 7th ARRB sk id-resistance requirements for main rural highways.
Conf. 7(7) , pp. 380-90. Nat. Coop . Highw. Res . Program. Rep . 37.
COUNTRY ROADS BOARD (1976). The investigation McLEAN , J.R. (1974) . Driver behaviour on curves - a
of f:>ide-fricliun . Report to the NAASRA Review of review. Proc . 7th ARRB Conc. 7(5), pp . 129-43.
Road Design Standards.
---------- (1981). Driver speed behav iour and rural
DEPARTMENT OF MAIN ROADS , NEW SOUTH WALES road alignment design . Traffic Eng. Control 22(4), pp.
(1969) . The behaviour of drivers on horizontal curves . 208-11 .
Main Roads 34(4) , pp. 127-8.
NATIONAL ASSOCIATION OF AUSTRALIAN STATE
GLENNON, J.A. and WEAVER , G.D. (1971) . The rela- ROAD AUTHORITIES . (1970) . Guide to traffic
tionship of vehicle paths to highway curve design. eng ineering practice . (NAASRA: Sydney.)
Texas Transp. Inst., Texas A&M Univ., Research Rep . ---------- (1973) . Policy for geometric design of rural
134-5. roads (Metric Units) . (NAASRA: Sydney.)
GOOD, M.C. (1975) . Road curve design : An historical ---------- (1980) . Interim guide to the geometric
review of the development of design standards for deSign of rural roads. (NAASRA: Sydney.)
horizontal alignment. Rep. HF 25. Human Factors Res.
Group . Dept. Mech. Eng ., Univ. Melbourne . Also RITCHIE, M.L., McCOY, W.K . and WELDE, W.L.
Australian Road Research Board. Special Report , SR (1968) . A study of the relat ion between forward
No. 15, 1978. ve locity and lateral acceleration in curves during nor-
mal driving . Human Factors 10(3), pp . 255-8.
---------- and JOUBERT, P.N. (1977) . Driver-vehicle
behaviour in restricted-path turns. Ergonomics 20(3) , TARAGIN, A. (1954) . Driver performance on horizontal
pp. 217-48. curves . Proc. Highw. Res. Board 33, pp. 446-66.
GOOD, M.C. and SWEATMAN, P.F. (1976) . Driver UNITED KINGDOM MINISTRY OF TRANSPORT
strategies on road curves. Proc. 16th FISITA Cong ., (1970). Report of the Committee on Highway Mainte-
pp . 6-87 to 6-96. nance . (HMSO: London .)

You might also like