Slope Variance as a Measure of Roughness

And the CHLOE Profilometer

W. N. C A R E Y , J R . , H . C . H U C K I N S and R. C. L E A T H E R S
Respectively, Chief Engineer for Research, and Chief, Instrumentation
Branch, AASHO Road Test; and Engineer of Special Assignments,
AASHO Road Test, Bureau of Public Roads.

A short discussion of existing pavement roughness measuring devices is

given. The rationale is described by which the statistic, slope variance,
was chosen as a measure of pavement roughness at the AASHO Road Test
for use as an element of serviceability. The choice of the slope rather than
the displacement or elevation profile is discussed.
The C H L O E Profilometer, a device developed at the Road Test for deter-
mination of the slope variance on pavements in the field, is described.
The C H L O E Profilometer is a relatively simple, electronic-mechanical
device. It is towed by a vehicle over any section of pavement. Three
statistics, the number of sample points (slope is sampled at 6-in. intervals
along the pavement wheelpath), the sum of the deviations of the slopes
from an arbitrary reference value, and the sum of the squared deviations
of the slopes from the same reference are detected, computed and dis-
played on the panel of the electronic device that is in the towing vehicle.
These three statistics can be combined to determine the summary statistic,
slope variance.
The device uses solid-state circuitry (transistors and diodes). It is
extremely compact and rugged. It is easy to operate requiring no elec-
tronic technicians. It uses plug-in printed circuit boards to permit rapid
field maintenance by non-technical personnel. Provision is included for
field calibration and testing of all components to insure that everything
is working properly. The cost for labor and materials for constructing
one unit is approximately $6,000 excluding overhead and profit. A disad-
vantage of the unit is that in its present form it cannot be operated while
recording at a speed in excess of about 5 mph. However, pavements under
study are generally sampled in short segments. Between such sample
segments the device can be towed at normal road speeds.

For many years engineers and others have a moving plane established by a series of bogie
been concerned with the need for a device that wheels so arranged that if one small bump or
would indicate the wheelpath profiles of roads irregularity is traversed by one of the wheels
or railroads. Some of the earliest devices for its effect on the reference plane is reduced in
this purpose, then called land planes, were de- proportion to the number of bogie wheels.
veloped in the 19th Century in Germany and Michigan and California use various statis-
France. The first to come to attention in the tics to summarize these analog records. Nota-
United States was the land plane developed by ble is one in which one-half of the total excur-
the Illinois Division of Highways about the sion of the sensing wheel with respect to the
time of the Bates Road Test (Fig. 1). Recently plane established by the bogie wheels is con-
the land plane was brought into the 20th Cen- sidered a roughness index.
tury, modified to include electronic recording, These devices are particularly useful to those
by the California Division of Highways. A investigators who wish to study the changes
rather elaborate unit called the California Pro- in profile of a specific section of pavement over
filograph was constructed and has been used a period of years as they might relate to warp-
extensively in California (Fig. 2). Another ing and frost heave, for example. Changes in
unit modeled after the California device, with profile are clearly presented on the analog
some additional features, was built for the records obtained by these devices. However,
Michigan Highway Department. These devices when one pavement is to be compared with
provide continuous analog records of the dis- another, the devices have an inherent short-
placement profile of a pavement referenced to coming, best illustrated by the case where a
126
 SELECTED SPECIAL STUDIES
127

Figure 1. Illinois Profilograph, 1920 version.

Figure 2. California Profilograph, 1958.

pavement has a sinusoidal undulation longi- ometer, towed by any vehicle at relatively high
tudinally w i t h a wave length equal to one-half speeds (20 to 30 mph) over a pavement (Fig.
the distance between the forward and rear 3) . Attached to a single wheel trailer is what
bogie wheels. Here the forward bogie wheels, is essentially a seismic mass. The total excur-
the sensing wheel at the center, and the rear sion i n one direction of a wheel running on the
bogie wheels are all traversing the ascending pavement with respect to the seismic mass is
part of the undulation at the same time. They accumulated through a mechanical integrator.
all reach the peak and descend at the same The resultant excursion in inches per mile of
time. Consequently, the analog record shows a pavement is termed the road roughness. This
straight line indicating a smooth pavement. device is exceedingly simple to operate. I t too,
Obviously this kind of a pure sine wave is however, has shortcomings in that i t is highly
never found i n an actual highway, but the fact susceptible to changes in temperature and to
remains that components of periodic waves the condition of its bearings and mechanical
exist in all wave forms so that there can be components and i t also has a resonant fre-
an appreciable error associated with the output quency which, i f excited by a large component
of these devices when they are used on certain of a corresponding wave length in the pave-
highways. A n extensive investigation of the ment, produces erroneous results.
mathematics involved i n these errors associ- Another device was developed f o r the A i r
ated with the profilographs (and also with the Force by the Midwest Research Institute (Fig.
AASHO Road Test Profilometers) was made 4) . This is the first semi-automatic device that
by Thomas W. DeVries (Ph. D. dissertation, has come to the attention of the authors that
Purdue University, 1959). Another disadvan- utilizes a fixed reference plane. Consequently,
tage of the profilographs is that they must the output is in absolute units of displacement
operate at slow speed (5 mph) when recording. profile. The reference plane is provided by a
Another device developed f o r the purpose of collimated beam of light established by fixing
studying pavement roughness is a simple trail- a light source on a tripod at one end of the
er, called the Bureau of Public Roads Rough- section of pavement under investigation. A
 C O N F E R E N C K ON T H E AASHO ROAD TEST
128

Figure 3. Bureau of Public Koads roads roughness indicatorRoughometer.

Figure 4. Midwest Research Institute Profile equipment.

traveling device employing a servomechanism mentation Panel f o r the AASHO Road Test.
follows the light beam and records the profile A large number of ideas were considered, some
of the pavement. The most serious disadvan- discarded immediately, some tried with models
tage of this device, aside f r o m its rather cum- and later discarded and some tried as f a r as
bersome and complicated makeup, is that i t the prototype stage and, unfortunately, also
must be operated leap-frog fashion over short discarded. The most difficult problem associ-
stretches of pavement at each set-up. ated with this work was that of providing a
Because each of the existing devices f o r the reference plane f r o m which measurements
determination of roughness had what appeared could be made, the difficulty being that the
to be f a i r l y serious shortcomings and because reference plane was required to move along
pavement roughness was considered to be a the pavement with the balance of the instru-
very important element of pavement perform- ment system. Gyroscopes were immediately
ance, it was decided early in the deliberations discarded because they precess due to rotation
concerning the AASHO Road Test to attempt of the earth and due to the inevitable accelera-
to develop a new device that would overcome tions and decelerations associated with travel
these shortcomings. The advice of a highly over irregular surfaces. Some highly ingeni-
qualified group of instrumentation and physi- ous devices suggested by George W. Cook of
cal research men was utilized toward this end. the Road Test instrumentation contractor's
These men were the members of the Instru- staff were tried. One of these involved a highly-
 SELECTED SPECIAL STUDIES 129

Figure 5. A A S H O Road Test Profilometer.

sensitive angular accelerometer; another, a

spinning pendulum with a period of about 84 SLOPE RECOm

min. Each of these was partly successful but

none was entirely so. A horizontal reference
utilizing a free rotor spherical gyroscope with
a highly complicated and sensitive erection sys- OtSTSNCE RECORD^
tem was considered. I t is believed that this
system would have performed satisfactorily
but its cost (in the neighborhood of $200,000) H H- ONE fOOT ON PAVEMENt'

and the fact that many of its components had ' TR4NSITI0N -)- TE8T SECTIOW

high military security classification precluded

its use. For Rough Pavement

Ultimately the Road Test Profilometer with-

out horizontal reference was used f o r the bulk
of roughness determinations upon which the
analyses of pavement performance at the Road
Test were based.
I t is important to recognize that the Road
Test Profilometer has an inherent shortcoming
identical in cause to that of the California
Profilograph, that is, the output of the device
varies f r o m the t r u t h when i t is operated over
undulating pavement of certain frequencies.
However, the magnitude of the error and the
chance f o r the appearance of error (over nor-
mal pavements) are considerably lower than For Smooth Pavement

f o r the case of the land plane device. This is Figure 6. Sample Profilometer records.
demonstrated by DeVries.
The Road Test Profilometer was used suc-
cessfully throughout the AASHO Road Test $1,500 per month. Maintenance of the equip-
(Fig. 5 ) . Its principal disadvantages are the ment required highly qualified electronic tech-
inherent error mentioned above, its extreme nicians. Consequently, although this instru-
complexity and its slow operating speed (about ment was highly useful f o r the special purpose
5 mph). The device and its associated elec- of a large highway research project, i t would
tronic equipment cost about $100,000 (includ- be out of the question f o r routine highway
ing development cost). Its output is in the evaluation.
f o r m of an analog record in ink on paper tape The Road Test Profilometer was designed to
(Fig. 6 ) . This record required the develop- provide continuous analog records f r o m both
ment of an automatic chart reader at a devel- wheelpaths of the slope of the pavement with
opment cost of about $20,000 (Fig. 7 ) . A n elec- reference to a horizontal plane as the device
tronic digital computer also was required to moved along the pavement. The slope of the
digest and summarize the output of the chart pavement at a point was taken to be the tan-
reader. The cost of the computer was about gent of the angle f r o m horizontal of two points
 CONFERENCE ON T H E AASHO R O A D T E S T
130

Pavement
Surface

Figure 8. Sclietnatic of A A S H O Profilometer.

(serviceability). Obviously, these last two uses

are closely allied since low serviceability leads
to the need f o r maintenance. I t was clear that
roughness was related to serviceability at the
outset of the Road Test efforts to define pave-
ment performance. The specific relationship
between roughness and serviceability, however,
was not at all obvious.
A rather large number of pavements in vari-
ous conditions of repair in three States were
Figure 7. A A S H O Road Test chart reader. rated subjectively as to their ability at the time
of rating to serve high-speed, high-volume
spaced 9 in. apart along the wheelpath. The mixed truck and passenger traffic. A 14-man
continuous analog was recorded since, at the rating panel was appointed f o r this purpose.
time that the device was under development, A t the same time, objective measurements
no decision had been made as to the proper were made of characteristics of the pavement
means f o r summarizing the roughness infor- surfaces. Among the objective measurements
mation into workable statistics. Figure 8 is a were Bureau of Public Roads Roughometer
schematic of the profilometer. Without a suc- runs and Road Test Profilometer runs. The
cessful horizontal reference device i t was correlation between the subjective ratings of
necessary to assume that the angle between serviceability of these pavements and the
the pavement and the long trailer tongue was roughometer values f o r those where the rough-
equal to the angle between the pavement and ometer was run is shown in Figure 9. The
the horizontal f o r practical purposes. Inas- scatter is associated, in part at least, with the
much as long wave lengths such as those asso- shortcomings in the measuring instrument.
ciated with change i n nominal vertical align- The output of the Road Test Profilometer f o r
ment (grades of the pavement) were not of a given test section is a continuous analog
interest in the studies of roughness, errors in- record (Fig. 6) of the slope of the pavement
troduced by this assumption were seldom surface i n each wheelpath. Since the slope is
greater than 0.005 in slope at a point. This is essentially the first derivative of the displace-
equivalent to less than 20 min of arc. Further- ment profile, continuous integration of the
more, i t is believed that the occurrence of such slope record should produce the elevation or
error was random i n nature and, therefore, no displacement profile. Similarly, differentiation
systematic bias was introduced f r o m one pave- of the slope record should produce a profile
ment to another. that is called the acceleration profile. This
profile is really the rate of change of slope with
distance.
USE OF SLOPE V A R I A N C E TO REPRE- Summaries of these three profiles, the dis-
SENT P A V E M E N T ROUGHNESS placement profile, the slope profile and the
Some measure of pavement roughness is acceleration profile, were compared with the
desired f o r several reasons. One is to check on subjective ratings of the sections. Fairly high
the work of a contractor constructing a pave- correlation was found between the slope profile
ment. Specifications have been written around and the subjective ratings and between the
certain requirements f o r pavement roughness acceleration profile and the subjective ratings
(or smoothness). Detailed profiles of pave- but the correlation was considerably lower
ment are of interest to those studying warp- when i t was attempted to compare the eleva-
ing, frost heave, etc. Another use f o r such tion profile with the subjective ratings of.
measurements is to determine needed main- serviceability.
tenance. Still another, is to aid in the deter- Consequently, f o r Road Test use, since the
mination of the ability at the time of the slope profile was already on the record, i t was
measurement of the pavement to serve traffic decided to confine further studies to considera-
 S E L E C T E D SPECIAL STUDIES 131

50
1 1
X Flexible
4.0
Rigid
X

? 3.0 tM
X*'
o
01 2.0
X X
> f

X
e X
X
X
e
1.0

100 200 300 400

Roughometer
Figure 9. Serviceability rating vs output of modified BPR Roughometer.

tion of this profile. A slope profile (as shown power in the pavement is great at the resonant
in Fig. 6, for example) is simply an aperiodic frequency of the vehicle, the vehicle is going to
wave form with frequencies identical to those get an extraordinarily rough ride. Whereas
in the displacement profile but 90 deg out of for the same vehicle resonant frequency, if the
phase. Little can be learned from a study of a pavement has a small power component, the
complicated wave form of this type in itself. vehicle will get a smooth ride. The analysis,
Perhaps the best transformation for many pur- however, is much too cumbersome for ordinary
poses involves a highly complex analysis called pavement roughness determinations.
power spectral density analysis. This analysis Nevertheless, the total area under the power
transforms the original wave form into a plot spectral density curve, and therefore, the total
similar to the one in Figure 10 where the power in a particular pavement record, is equal
abscissa is frequency in cycles per foot and mathematically to the variance of the original
the ordinate is the power associated with each profile record. It is this line of reasoning that
band of frequencies. Power is a term borrowed led the engineers at the Road Test to choose
from communication engineers but it seems to variance as the most meaningful single-number
fit very well in this context, in that it may be summary statistic for the aperiodic wave form
said that the power at any frequency is that that is the record of pavement slope along its
available to excite a vehicle and, therefore, wheelpaths. Variance, fortunately, can be esti-
cause discomfort in the ride. This analysis is mated by rather simple mathematical tech-
exceedingly useful for research studies of the niques. Variance is defined as the average
pavement-vehicle relationship. Since, if the squared deviation of a function from its mean.
The slope variance for Road Test usage was
estimated by sampling the slope record at 1-ft
intervals on the pavement, finding the mean
of the values in each test section and then find-
ing the mean squared deviation of those values
from the mean. Computing formulas are given
in the Appendix. Figure 11 shows the relation-
ship between the slope variance and the sub-
jective ratings of the same pavements used in
Power
the illustration of roughometer-rating correla-
tion (Fig. 9). It is clear that the correlation
is higher than that for the roughometer. The
relationship in Figure 9 is curvilinear. The
logarithm of the slope variance serves as a
satisfactory linearizing transformation. For
Frequency, c y c l e s per f t use in the present serviceability index, one was
added to the slope variance before taking the
Figure 10. Typical power spectral density curve. logarithm so that the minimum value of the
132 CONFERENCE ON T H E AASHO ROAD TEST

5.0

4.0 * X Flex ibie

X Rigi d
X (
3.0 (
V
c
( X
o
cn 2.0 i
>
X

1.0

0 10 20 30 40 50 60 70 80 90 100
Slope Variance
Figure 11. Serviceability rating vs slope variance from AASHO Profilometer.

transformation is zero. Figure 12 plots sub- serviceability indexes include consideration of

jective ratings of a larger number of pave- surface slipperiness and perhaps faulting of
ments against log (1 -I- SV), the statistic final- rigid pavements at cracks and joints.
ly chosen to repi-esent longitudinal profile in
the Road Test serviceability indexes. Other THE CHLOE PROFILOMETER
elements of serviceability were summarizations
of depth of rut in the wheelpaths, and crack- When it had been determined that the slope
variance represented satisfactorily the rough-
ing and patching of the pavement surfacing
ness or longitudinal pavement profile, it was
material. It is recommended that further clear that the elaborate equipment in use at
efforts towards derivation of more universal the Road Test was a great deal more sophisti-
cated than it need be to do the job at hand.
50 As previously mentioned, the Road Test Pro-
filometer will be highly useful for research in
o which the power spectral density analysis is
ee
0 desired. For evaluation of the serviceability
4.0 o s of the pavements in the various State highway
0 o systems, however, a much simpler device was
9 indicated. Such a device, the C H L O E Pro-
IT 3.0 filometer, was developed at the Road Test. To
O
V) e ' JO
date three C H L O E Profilometers have been
Q. constructed and used extensively over the high-
way systems of the country. The first, Model
2.0 eoo 1, was used by the New York Thruway in a
comprehensive study of the serviceability of
o 1
e
the entire Thruway in the interest of predic-
I 0
e
\ e
tion of future maintenance needs. The other
two profilometers. Model 2, were used along
4 9 R igic Pa\ rem jnts
with the New York Thruway device in a sur-
vey of pavements that were scheduled for early
overlay in about 40 States. This survey was
0.4 08 I2 1.6 20 conducted by the Bureau of Public Roads for
log (I + SV )
the American Association of State Highway
Officials in an attempt to determine the level
Figure 12. Serviceability rating vs log of one plus of serviceability at which the State highway
slope variance. engineers felt it necessary to overlay pave-
 SELECTED SPECIAL STUDIES 133
Roller Contactor

I \ Slope Switch
/ 29 Contacts

0.0029089
Radians
/ B - D = l9in.
/ A - C = 9in.
Switch Segment \ / A-B=4.5in.
Enlarged \
\

AO
Figure 13. Rear of C H L O E Profilometer.

ments in primary highway systems. The prime Figure 14. Schematic of mechanical
diflference between Models 1 and 2 is in their elements of C H L O E .
system f o r data display. I n Model 1 the dis-
play is in binary form, whereas in Model 2 the switch plates are made by printed circuit tech-
display is in decimal form. A t h i r d model is niques and provision is made f o r easy replace-
now under consideration incorporating simpli- ment of the switch plate when it wears out.
fication and further advances. Pulses at 6-in. intervals along the pavement
The CHLOE Profilometer employs basically are initiated by a device attached to one of the
the same principles as the Road Test Profilom- trailer wheels. This device is a slotted disc
eter without a horizontal reference. A long with a light on one side and a photoelectric cell
trailer is towed behind any vehicle. Toward on the other. As the wheel advances the slots
the rear a set of slope detecting wheels 9 i n . pass by the photocell permitting the light to
apart are so arranged that the angle between energize the cell. The output of the cell is
the long trailer tongue and the slope wheels is recognized at the computer and triggers a
detected. Figure 13 shows the working end of pulse which is designated a command-to-sample
the trailer unit. pulse.
Whereas the Road Test Profilometer meas- The electronic heart of the CHLOE Pro-
ured slope in both wheelpaths simultaneously, filometer is contained in a cabinet that is nor-
the CHLOE Profilometer operates in only one mally carried inside the towing vehicle. Figure
wheelpath at a time. I t has been standard 15 shows the interior of this cabinet, called
operation to run in the outer wheelpath and to the computer f o r the Model 2 CHLOE units,
consider the slope variance so obtained as an
estimate of the average slope variance in both
wheelpaths. Slopes measured by the CHLOE
Profilometer are not recorded continually as i n
the case of the Road Test device, rather they
are sampled at 6-in. intervals along the pave-
ment. With the Road Test device although
slopes were recorded continually, data were
sampled f r o m the record at 1-ft intervals on
the pavement. Therefore, the CHLOE device
actually uses twice as many values in its com-
putations than were used with the Road Test
Profilometer f o r a given length of pavement.
However, the speed of travel of the new unit
is still 5 mph. Where the Road Test Profilo-
meter utilized a highly complex mechanical-
electronic system f o r detection of the slope
angle, the CHLOE Profilometer uses a simple
mechanical system involving a switch with 29
segments each pair of which subtends an arc
of 10 min. Figure 14 shows this mechanical
switching arrangement schematically. The Figure 15. Interior of C H L O E computer.
134 C O N F E R E N C E ON T H E AASHO ROAD T E S T

Figure 16. Front display panel, C H L O E computer.

with the plug-in printed circuit boards. Figure the slope variance measured by the AASHO
16 shows the f r o n t of the cabinet, the display Profilometer. Ideally, the usefulness of any
panel. other instrument would be tested by similar
The operation of the device is relatively sim- correlation studies where its output is com-
ple. As the unit moves along the road at about pared directly with subjective ratings. A t the
5 mph the slope wheels following the pavement Road Test i t was not feasible to accomplish
profile, an arm attached to the slope wheels such a direct study with the CHLOE Profilo-
continually sweeps back and f o r t h over the meter. The most likely available alternative
switch plate which is attached to the trailer plan was to test the CHLOE device against the
frame. Therefore, at any instant the location already proven AASHO profilometer. This was
of the arm indicates the angle between done over pavements with a wide range of
the slope wheels and the trailer frame. variances or roughnesses. The resultant corre-
The pulses formed by the distance sens- lation is shown in Figure 17. U n t i l direct test-
ing device at 6-in. intervals are used to ing of the CHLOE device against subjective
interrogate the switching device. The com- ratings can be accomplished, this curve (Fig.
puter determines the switch position that exists
at that instant. This position is considered to
represent the slope at that point. (The angle i n
radians is used to represent the slope which is
actually the tangent of the angle. The maxi-
mum possible error introduced by this simpli-
fication is f o r switch positions 1 or 29 where o 20
the angle is 0.04072 and the slope is 0.04075.)
The slope values thus obtained are accumulated
in the computer over a sample length of pave-
ment, 500 f t , f o r example. Simultaneously, the
slope values are squared and the sum of
squares also computed as the machine moves
over the sample pavement. The console of the
computer displays the number of samples, the
sum of the slope values and the sum of the
squared slope values at all times so that when
a sample run is completed, these three num-
bers can be read f r o m the console.

CORRELATION STUDIES 4 8 12 16 20 24
Slope V a r i a n c e , Chloe Profilometer
The Road Test serviceability indexes were_
based on correlation between the subjective Figure 17. Correlation of C H L O E with
serviceability ratings of the Rating Panel and A A S H O Profilometer.
 S E L E C T E D SPECIAL STUDIES 135

17) will be used to translate C H L O E readings eral Motors device, the Midwest Research
into slope variance for use in the serviceability Institute device, the Purdue accelerometer
index formulas. device, the AASHO Profilometer and the
It is strongly recommended that new experi- C H L O E Profilometer. Only in this way can
ments be conducted in which pavements with the effectiveness of one device as compared
a wide range in condition are rated as to their with another for each specific type of rough-
serviceability by one or more rating groups in ness or profile measuring job be demonstrated.
several parts of the country. Simultaneously, This survey or group of simple experiments
measurements should be made on these pave- can provide the much needed universal present
ments of cracking, patching, rut depth, serviceability index equations. The Highway
faulting, slipperiness, and most important, Research Board is ready to assist in any way
roughness by each of the several available in- possible in the design and conduct of these
struments for roughness determination, notably experiments and in the analyses, interpreta-
the Profilograph, the Roughometer, the Gen- tion and reporting of their results.
 Appendix
D E R I V A T I O N OF COMPUTING FORMULA FOR SLOPE V A R I A N C E

Given a pavement wheelpath profile (Figure and since

18a), the slope, y, varies with distance as i{y)
(Figure 18b). The variance of the slope, SVt,
is the average squared deviation of f { y ) from
its mean. The pavement is sampled from point
A to point B at iV points i (6 in. apart with the N \_ N N
C H L O E profilometer).
(2c)
Elevation (a) and

AT m (2d)

is the computing formula used with the C H L O E

profilometer.
The transducer in the C H L O E profilometer,
-Distance used to sense slope, is a switch with 29 posi-
tions, each representing 10 minutes of arc.
Switch positions are sampled at 6-in. intervals
along the pavement. These positions, F are
(b) entered into the computer as integers, 1 to 29.
Because the angle between switch positions is
10 min, the difference between two adjacent
integers is 10 min, or 0.0029089 radians.
At its console C H L O E presents for each
pavement section 2 Y^-, 2 Y, and A^.
However, the true* slope, y.,, is related to
y . by
-Distance
y, = 0.0029089 F , (3)
Figure 18. Elevation and slope profiles.
Substituting in E q . 2d, true slope variance is
The mean, y, of i{y) from A to B is esti- SVt = (0.0029089) = 2
mated by N
(0.0029089)M2 Y)-
(4a)
v = ^ i y ^ (1)
IS 1=1
Thus, using the C H L O E data.
The deviation of a point on i{y) from the
mean is y^y. The sum of squared deviations SiV _/2ij^
N SVt = 0.0000084616 (4&)
is ^{y%y)^ and the average squared devia- IN \ N ) .
tion, the variance, is In the present serviceability index equation,
for ease of handling, the slope variance esti-
mated from C H L O E is multiplied by 10". This
(2a) value is
N
or
SVt = '2 i y . ' - 2 y y , + y') The angle in radians is used to represent the slope, j
N which I S actually the tangent of the angle. The maxi-
mum possible error introduced by this simplification
2y'S.yi +Ny^'^ (2b) is for switch position 1 or 29 where the angle is
0.04072 and the real slope is 0.04075.
136
 S E L E C T E D SPECIAL STUDIES 137

In the original equation for serviceability the with C H L O E were about 3 x 10-'' higher than
average of both wheelpaths was used (thus the those obtained with the Road Test system,
bar over SV). C H L O E estimates the average Thus the final computing formula for C H L O E
value from observations on one wheelpath only. is
Correlation studies between the first model of ryv - / "5" Y \ - l
C H L O E and the Road Test Profilometer indi- sV = 8.46 =^ ( ^ ) 3.0 (5)
Gated that, for some reason, variances obtained L N \ N / J