Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter.
Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.
OCR for page 163
CHAPTER 5. STUDY OF SMOOTHNESS-MEASURING
EQUIPMENT AND INDICES
Purpose and Objectives
As more and more agencies are adopting pavement smoothness specifications,
ancT as the monitoring of pavement smoothness becomes a more common activity In
the pavement management area, there is a strong motivation to evaluate the
smoothness-measuring equipment and associated smoothness indices commonly in
use. Currently in the U.S., there is a wide variety of smoo~ness-measur~ng
equipment and smoothness indices employed. Moreover, many agencies employ
different equipment and indices for measuring Initial pavement smoothness and time
1 1 ~ 1
. .
monitored smoothness. Icleally, the same equipment usect to measure crucial
smoothness would be used to measure the smoothness of that pavement over time.
The results of the analyses conducted in chapter 4 strongly indicate the positive
effect of initial smoothness on the future smoothness of the pavement. This is a very
important finding that supports what has been intuitively believed by many within
the pavement community. However, a set of recommended guidelines is now
needec! for measuring and reporting crucial smoothness which, to the extent possible,
represent a consensus or trend of best practices. These guidelines may therefore
serve as a dynamic "nucleus document" around which to add unprov~ng and
clarifying information over a reasonable period of lime, such that State DOTs can
migrate from their current varied practices toward a standardized approach to
measuring and specifying Axial smoothness.
Due to the integrated nature of a smoothness index and smoothness-measur~ng
equipment In an overall test procedure, this project studied each side of the
smoothness issu~quipment and indices to arrive at recommended specifications
for equipment used to measure ~rutial pavement smoothness. This was accomplished
through a review of current smoothness-measuring equipment and Trough an
evaluation of current smoothness summary statistics. Once recommended summary
statistics were selected, specifications for the equipment were developed.
Review of Smoothness-Measuring Equipment
aL 1
Over the last century, a vast array of equipment has been developed to evaluate
the smoothness of pavements. In recent years, the growing interest in pavement
management systems has focused on pavement ride as a major factor in evaluating
the condition of a road, thereby creating He need for economically evaluating the
extensive mileage of a highway system. Electronics and modern technology have
been applied, creating an ever-expanding approach to the subject.
163
OCR for page 164
For this study, a review was conducted of the significant types of equipment that
have been used for evaluating the smoothness of pavement surfaces, the details of
which are provided in appendix E. Those smoothness-measur~ng devices currently
being used by major highway agencies, as well as those believed to provide a
potential contribution in the immediate future, are clescribed in detail.
The questionnaire survey of highway agencies conductecl under this study
revealed that Mere are a variety of types of equipment in use for measuring the
smoothness of newly constructed pavements. Table 40 shows the breakdown of
equipment usage among the 47 State highway and 3 Federal Lancl agencies that
responded to the agency questionnaire.
The profilograph is clearly the most popular device ~ use today for the
measurement of initial pavement smoothness. However, the review in appendix E
indicates Cat there are several limitations to the profilograph and several legitimate
concerns about its accuracy (precision and bias).
Table 40 shows that Mays Meters are still in use by some agencies, but these
devices are known to have accuracy limitations. This is because they are highly
dependent upon characteristics of the host vehicle. As such, they require frequent
calibration.
A few agencies are using noncontact pavement profilers for Me measurement of
initial pavement smoothness. These devices offer We advantage of providing a
pavement profile from which a more accurate representation of Me smoothness
characteristics of Me roadway can be made. The relatively high cost of these
pavement profiling systems and their inability to test young concrete (say, less than 3
days old) have been drawbacks to the use of these systems for smoothness
construction control. However, advancements are currency being made In the area
of lightweight profit - measuring devices Cat cost far less than their full-size
counterparts and yet are light enough that they can be operated over young concrete
pavement surfaces.
Table 40. Usage of various smoothness measuring indices.
Equipment Type
California-type Profilograph (generic)
Rainhart Profilograph
No. of Agencies Using
Mays Meter (Vehicle mounted)
. . . . . .
Mays Meter (Trailer mounted)
-
GM Type Profilometer~
Rolling Straight Edge
.
Straight Edge or Stringline
33a
a Six agencies use this category as the sole method of smoothness measurement, and the remaining use it to
augment other smoothness testing.
164
OCR for page 165
The review of equipment focused on types of devices that could be appropriate
for use on major highway projects, are currently available, and have at least limited
use by a major highway agency. The following important principles can be defined
as a result of this equipment review:
1. A testing procedure should be "profile-based" and have a high correlation
with user response to road roughness.
2. A testing procedure should be refined to a degree of accuracy (precision and
bias) to allow it to be used with adequate def~rution in a smoothness
specification.
3. Pavement surface characteristics should be evaluated ~ a manner Mat is fair to
the agency and construction industry alike.
4. A testing procedure should be understandable by all involved anti defensible
In a court of law.
A testing procedure should have the support of managers and technicians.
Although the review indicated several pieces of equipment Cat could be used In
measurement of initial pavement smoothness, it is apparent that the capabilities and
characteristics of the necessary smoothness-measuring equipment is largely driven by
the smoothness index that it measures. As such, a smoothness Index must first be
selected before suitable equipment (or equipment characteristics) can be identified.
For example, once a recommended index is chosen, the necessary equipment
properties (sample interval, wavelength accuracy range, and dynamic accuracy) can
be defined to complete the equipment specification. Therefore, a review of various
smoothness indices is warranted and is presented in the following section.
Descuplion of Smoothness (Roughness) Indices
Discussed in this section are the more prominent pavement roughness
measurement statistics for assessing pavement smoothness. Because the degree of
index correlation with the response of highway users is critical to the selection of
appropriate indices, available information regarding each index's relationship with
user response is included below. These Indices of pavement smoothness are grouped
according to three categories: subjective ratings, mechanical fiIter-based numerics, and
profile-based numerics (Paterson 1987a).
Subjective Rating Indices
Several subjective ratings of pavement roughness have been developeci, most
significantly the present serviceability rating (PSR) concept developed at the AASHO
Road Test (Carey and Trick 1960~. This rating index has received widespread usage
since He AASHO Road Test, as many highway agencies adoptecl the PER or
modified it based on regional information. However, the use of subjective pavement
ratings based on He response of a pane} of pavement raters is not economically
practical for the assessment of initial pavement smoothness. Furthermore, other
factors, such as vehicle size and type, pane] size, extent and type of pane! training,
165
OCR for page 166
and panel regionality, may also affect the repeatability and bias of subjective rating
numerics (Hutchinson 1964; Sayers 1996; Janoff, Nick, Davit, and Hayhoe 1985).
Because of their economic Unpracticality and difficulties In obtaining repeatable
assessments, subjective ratings are not reco~runended as prunary indices for the
assessment of initial pavement smoothness. However, the response of highway users
and perhaps the additional pavement damage from dynamic loads should be
controlling factors In defining acceptable levels of pavement smoothness. Because of
this strong desire in the highway community to correlate roughness indices with user
response, He roughness-user response relationship has been explored in various
studies for both response-type and profile-based numerics Janoff, Nick, Davit, and
Hayhoe 1985; Sayers and Gillespie 1986; lanoff 1988; Al-Omari and Darter 1992~. The
correlation of roughness indices with subjective ratings of user response is used later
as a critical factor in the selection of an index for pavement smoothness.
Mechanical Filter-Based Indices
Mechanical systems for measuring roadway roughness Include response-type road
roughness measuring (RTRRM) systems and rolling straighteclge systems. RTRRM
systems measure the cumulative relative displacement between the axle and He
vehicle body and average Hat value over some distance of roadway; ~us, the Index
is reported In terms of vertical deviation over distance of roadway traveled (e.g.,
in/mi). Summary numerics measured by response-type systems (e.g., Mays Ride
Meter iMRM], PCA Roadmeter, and BPR Roughometer), calibrated to a profile or
over numeric in some cases, are reported to not correlate well win user response to
roadway roughness.
Mays Ride Number
The Mays ride number (MRN) is a measure of accrued vertical axle travel per
length of highway travel. It is obtained from a displacement transducer that detects
small Increments of axle movement relative to the car body, expressed In units of
length/length (generally In terms of in/mi). This summary statistic rectifies each
increment of positive or negative movement of the axle relative to the vehicle body
and then calculates the average rectified axle displacement over the length of the test
section. In essence, the MEN statistic (in the absence of nonlinear effects) is He
average rectified velocity (ARV) of the axle motion, multiplied by the tune needed to
travel ~ mi (~.6 km) at the prescribed test speed (Gillespie, Sayers, and Segal 1980~0
The repeatability and accuracy of this technique is affected by vehicle loading,
placement of He measuring apparatus in the vehicle (preferably in He center to
neutralize the effects of vehicle roll), tire pressure, suspension spring rate, friction,
and shock absorber characteristics.
The MRN is noted to correlate better with user response than other REARM
system outputs, such as the PCA Roadmeter and the BPR Roughometer indices;
therefore, it is considered the best of the available response-type systems (Gillespie,
Sayers, and Segal 1980~. Results of a correlation between the average of a 21-member
166
OCR for page 167
pane] rating and MRN in 1984 yielded a moderate correlation coefficient (R2) of 0.58
danoff and Davit 1984~. However, results of the International Road Roughness
Experiment (IRRE) conducted in 1982 indicated good correlation between MRM
output and the subjective response of a small pane} for AC pavements (Sayers and
Gillespie 19861.
Two NCHRP studies completed in 1985 and 1988 using strict quality control and
large rating panels yielded Tow to moderate R2 values (0.55 and 0.62, respectively)
between MRN and mean pane] ratings (MPR) for all pavement types danoff, Nick,
Davit, and Hayhoe 1985; lanoff 1988~. Those same studies concluded that for PCC
and composite surfaces and for all Tree pavement types (AC, PCC, and composite)
combined, MRN measures do not accurately predict rideability danoff 19881. A 1987
Kansas DOT research project using 108 AC, PCC, and composite test sections resulted
In moderate panel rating R2 values of 0.67 for both AC and PCC pavements (Moore,
Clark, and Plumb 1987~. A 1987 research project in Ohio indicated poor correlation
between MRN and user response for PCC pavements, although the correlation for
AC pavements was fair (Spangler and Kelly 1987~. Finally, a smaller panel rating
study In Colorado yielded high correlation values between MRN and user response
for AC pavements and very low correlation values between MRN and user response
for PCC pavements; tests were conducted at both 35 and 55 mi/hr (56 and 88 km/hr)
(Arterburn and Suprenant 1990~. MRN values for the Colorado studies were
determined from root-mean square acceleration (RMSA) values obtained using a
high-speed profile measuring system win ultrasonic sensors. Table 41 summarizes
the reported correlations between MRN and user response.
The inability of the response-type system indices to correlate well with user
response and with similar measuring systems can be related to both He inability of
response-type systems to measure and sufficiently weight He surface profile
wavelengths that are most related to user response and to overall variability within
these systems. Response-type systems, such as the ~M, respond to spatial
frequencies (i.e., reciprocal of wavelength, spatial frequency of 10 equals wavelength
of 0.~) only from about 0.013 to 0.150 cvcles/ft (0.042 to 0.492 cycles/m) whereas
, , , ~ , . ..
. ~ . ~ ~ ~ ~ . . ~ . . ~ ~ . ~ ~ _ _
panel ratings are highly correlated With spatial frequencies between U.-12d and Mu
cycles/ft (0.410 and 2.067 cycles/m) Qanoff 1988~. As a result, the MRM does not
respond to short wavelength roughness that is felt by highway users. Similarly, the
PCA Roadmeter responds to frequencies between 0.014 and 0.027 cycles/ft (0.046 and
0.089 cycles/m) at 50 mi/hr (80 km/hr), and He BPR Roughometer is excited by
pavement surface frequencies between 0.05 and 0.2 cycles/ft (0.164 and 0.66
cycles/m) at 20 mi/hr (32 km/hr) (Gillespie, Sayers, and Segal 1980~. Consequently,
these devices, especially the PCA Roadmeter, do not respond adequately to the
critical short wavelengths of a pavement profile.
167
OCR for page 168
OCR for page 170
OCR for page 171
OCR for page 172
OCR for page 173
OCR for page 191
OCR for page 192
OCR for page 193
OCR for page 194
OCR for page 195
OCR for page 196
OCR for page 197
OCR for page 198
OCR for page 199
OCR for page 200
Representative terms from entire chapter:
pavement smoothness
Table 41. Correlation of MAN with user response.
No. of PCC No. of AC PCC AC All
Reference ~ Sections ! Sections ~R2 ~R2 ~Types, R2
, anoff and Davit 1984 i l i I 1 0.58
layers and Gillespie 1986 I 01 11 1 _ 1 0.94 1
I anoff, Nick, Davit, and | 17| 18 | 0. 3 | 0.74 | 0.55
| ~ anoff 1988 1 1111 97 1 0' 18 1 0 85 1 0.62
Moore, Clark, and Plumb | 2~301 78-84 | 0. '7 | 0.67 |
|';pangler and Kelly 1987 1 181 17 1 0.36 1 0.74 1 0.56
.\rterburn and Suprenant | 18 | 51 | O ~0.81 ~
Mechanical filters, such as the RTRRM systems, respond to only a small range of
roadway profile wavelengths, amplifying and attenuating Me wavelengths that they
measure according to the lateral properties of each mechanical device. The
estimated gain for a typical RTRRM mechanical system is shown in figure 93. A gain
of 1.0 for a waveband indicates that the surface profile output measured by the
device is the same as the actual surface profile. Increases or amplifications of Me
actual profile elevations are noted by a gain greater than I, whereas decreases or
attenuations of Me true elevations result In a gain of less than 1. It is evident from
this figure that the true pavement surface profile is not measured by RTRRM
systems.
Variability in RTRRM systems has been documented since 1980, when Gillespie et
al. (1980) concluded that a significant amount of random error is associated win the
dependence of RTRRM systems on the dynamics of Weir host vehicle. Factors such
as variations In fire pressure, shock absorber properties, vehicle weight changes, and
testing speed all affect the output index of RTR1
4 ~ - |-St f shock
absorbers
3 ~ ~ 1~\ ~ ~ absorbers
~:~
0.001 0.01 0.1 1 10
Spatial frequency (cycles/ft)
Figure 93. Sensitivity of typical RTRRMS to spatial frequency
(adapted from Gillespie, Sayers, and Segal 1980).
Profile Index (PA from a Profilograph
The profilograph, commonly used for initial pavement smoothness control, applies
a mechanical rolling straightedge filter to a pavement profile, measuring wavelengths
within the range of 1 to 75 ft (0.3 to 23 m) and amplifying or attenuating the
wavelengths Mat are factors of the profilograph length (25 ft [7.6 m]). This is evident
in figure 94, which shows the profile amplified by a factor of 1.85 at wavelengths of
25 ft (7.6 m) and attenuated by a factor of 0.45 at wavelengths of 12.5 ft (3.S m)
(Scofield, Kalevela, Anderson, and Hossa~n 1992).
Figure 94 does not include the weighting effect of the in/ml statistic as noted by
Darl~ngton (1995). Using the ~n/mi statistic as Me unit of a smoothness index
attenuates longer pavement wavelengths and amplifies shorter wavelengths. For
example, if a profilograph trace contains sinusoidal waves with a 0.5-:n (12.7-mm)
amplitude and a wavelength of 0.2 mi (.32 km), there would be 1 in (25.4 mm) of
deviation from the zero trace in each wavelength, and the zero blanking band
profilograph in/ml statistic would be 10. However, if sinusoidal waves of the same
height were spaced at 0.1 mi (0.16 km), there would be the same deviation per
wavelength, and the in/ml statistic would be 20. The relationship between
169
4 -
o ~ -
Or, n 1 n r, 1 rat 1
Spatial frequency (cycles/ft)
1 - 10
Figure 94. Sensitivity of simulated profilograph to spatial frequency
(adapted from Scofield, Kalevela, Anderson, and Hossa~n 1992~.
wavelength and in/ml is linear, resulting In a larger gain for pavements with shorter
wavelengths. In figure 94, use of Me in/ml statistic would result In the simulated
profilograph response gain increasing to the right as spatial frequency increases.
Thus, use of the ~n/mi statistic effectively filters me profile data placing a linearly
increasing emphasis on shorter wavelengths.
Whether by accident or design, the ~n/mi statistic amplifies me shorter
wavelengths mat tend to most affect me users' options of a roadway. This has
helped the statistic to provide smoothness index values that correlate fairly well with
user response. However, the statistic is not based on signal theory, and it applies a
linear weighting to profile data wavelengths that is not based on analytical
investigation of the relationship between user response and pavement wavelengths.
However poorly the profilograph measures the true pavement surface profile, the
PI, as derived from the profilograph output, has served me highway corrununity
fairly well as an easily understood index of ~rutial pavement smoothness.
Nevertheless, recent studies have highlighted cases in which the PI has not correlated
well with user response (Darlington 1995~. In addition, one study showed that the PI
correlated only moderately well with PSI, with R2 values of 0.705 using a 0.2 in (5
mm) blanking band on roadways less than 20 in/ml (0.32 m/km) (Walker and Lin
170
1988). A slightly better correlation (R2 = 0.742) was obtained when a 0.1 In (2.5 mm)
blanking band was used (Walker and kin 1988). This moderate correlation has
resulted In several States experimenting with the reduction or eluntnation of the
blanking band in their profilogram data reduction (Parcells and Hossain 1994).
Research completed In 1989 by the Pennsylvania Transportation Institute
~. ~. ~'.. ~
.
concluded that using a blanking band on profilograph traces from new pavements is
unacceptable for pavements with PI values less than 7 in/ml (~10 mm/km)
(Kulakowski and Wambold 1989~. That study also showed correlations unproved
significantly in linear regression between PI and IR] when the blanking band was
reduced to 0.! and 0.0 in (2.5 and 0.0 mm). Another study conducted for the South
Dakota DOT in 1993 indicates increasing correlation with IRI, PST, and the lanoff ride
number (RN~anoff) when the blanking band is reduced to 0.! and 0.0 in (2.5 and 0.0
mm) (Moser, Hudson, and Hudson 1993~. These studies conclude that, as the move
toward smoother pavements progresses, it will be important to reduce or eliminate
the blanking band from profiIogram evaluation to avoid missing times of pavement
roughness important to highway users.
~, ~
User response correlations are not available for pavements measured using 0.: or
0.0 ~ (2.5 or 0.0 mm) blanking bands. Therefore, it is not possible to determine
what, if any, increase in correlation with user response can be obtained by reducing
the blanking band Further study is necessary to identify the correlation between PI
using a reduced blanking band profilogram and user response, especially in the
smoothness range of newly-constructed pavements.
Computer modeling of rolling straightedge (profiIograph) measurements using
pavement surface profiles obtained from inertial and inclinometer equipment are
becoming more common. This application allows PI values to be more easily
obtained. In addition, it provides a more accurate crucial pavement profile from
which other pavement smoothness indices can be obtained. The flexibility and speed
provided by PI values modeled using high-speed, more accurate pavement profiles
presents the opportunity to maintain Pl-based specifications while simultaneously
evaluating other profile-based smoothness specifications.
Slope Variance (SV)
Another mechanical-type, profile-based statistic developed at the 195~60 AASHO
Road Test is the slope variance (SV) (Carey and Huckins 1962~. This value was
measured by the CHLOE ProfiIometer as the difference in angles between a small
beam with two wheels, 9 in (229 mm) apart, and He 25.5-ft (7.~-m) long trailer towed
at a speed of 2 to 3 mi/hr (3.2 to 4.S km/hr) (Gillespie, Sayers, and Segal 19801. The
slope variance for He AASHO Road Test was estimated by sampling the slope record
at Oft (0.3 m) intervals, finding the mean of the values in each test section, and then
computing the mean squarest deviation of those values from the mean (Carey and
Huckins 1962~.
171
The gain from the CHT=OE system is near 1.0 for spatial frequencies between 0.02
and 30 cycles/ft (0.06 and 98 cycles/m) and is less than 1.0 for higher spatial
frequencies (Gillespie, Sayers, and Segal 1980). A gain of 1.0 indicates perfect
representation of the "true" pavement profile. This trueness to the actual slope
profile is a positive advantage of the SV statistic; however, SV is derived from a band
of profile spatial frequencies much broader than is significant to an automotive
vehicle, and as a result introduces a random error that degrades the agreement
between RTRRM system output and SV (Gillespie, Sayers, and Segal 1980~.
Although the simple geometry of the early profilers implies that SV can be
computed mathematically from more accurate profile measures, the earlier
instrumentation systems had quirks and complexities that have not been well
documented; thus, estimates of SV made from measured profiles are not equivalent
to outputs from old instruments. The SV is sensitive to We choice of profile
measurement and does not describe a standard roughness measure. The "true"
variance of me slope of a road profile is infinite, since Me true profile includes
texture effects (Savers and Gillesnie 1986: Savers Gillesnie and Oueiroz 19861. As a
~, ~, ~, ~, _
, ~ ~ ~ . . . ~ ~ . ~
result, the rev is not an ideal pavement smoothness measurement.
The information presented above leads to Me conclusion Cat subjective ratings
and roughness indices based on mechanical vehicle-response output have
shortcomings Cat render them sometimes inadequate or unpractical for use as
indices of initial pavement smoothness. Roll~ng-straightedge or profiIograph-based
roughness indices are not based on the range of roughness wavelengths that highway
users find offensive. The simplicity of PI measurement and its moclerate correlation
with user response have helped to make it the most widely used index for ~rutial
pavement smoothness quality control. However, advances In the ability to measure
and analyze pavement surface profiles and more frequent reports of drivers finding
pavements with low PI ratings to be offensively rough have made the use of the PI
statistic less ideal. Thus, indices based on measurement of true pavement profile are
investigated for their ability to more completely relate with user response and
provide a practical method of measuring nutial pavement smoothness.
Profile-Based Indices- Mechanical System Simulation
Since Me advent of high-speed profile measuring devices, analog and digital
filtering have been used to remove unnecessary and unproper wavelengths from
measured "true" surface profiles, that is, to amplify or attenuate selected wavebands.
Profil~based pavement smoothness numerics are generally obtained by either
simulating the response of an RTRRM system as it traverses the profile or by
separating (filtering) and weighting the spectra of wavebands that make up the road
surface profile. These approaches are similar In that the profile for each is filtered
with a band-pass filter, transformed to a positive value, and averaged over He length
of the profile. The differences lie mostly in the use of either rectification or squaring
and In He filter bandwidths and weighings (Hayhoe 1992~.
- r---
1 1 - 1
^. ~. .
172
Most roughness indices use either the average rectified slope/velocity (ARS/ARV)
or the root mean squared (RMS) elevation/slope/acceleration methods. The average
rectified method and its deviations uses the average of the positive values of
measured elevation, slope, or velocity. The root mean squared method on the other
~ ~ ~ . . .~ . ~ -~
nanu, cteterm~nes the square root of the squared elevation, slope, or acceleration
profile values. Neither of these methods stands alone as an index for pavement
smoothness. The RMS and AR slope and acceleration approach infiruty as Me profile
surface becomes vertical or as it changes direction abruptly. In addition, a baselength
for filtering must be defined for the RMS to be consistent (Sayers, Gillespie, anct
Paterson 1986~.
The type of filter (mechanical or digital/analog) applied to the pavement surface
profile is the determining factor needed, In addition to the RMS or ARS reduction
method, for defining a profile-based smoothness Index.
RARV and RARS
con
Distal filters for pavement profiles sometimes attempt to simulate the response of
mechanical filters or isolate the profile wavelengths that are well correlated with
some factor, such as user response. The reference average rectified velocity (RARV)
and the reference average rectified slope (RARS) statistics are examples of indices that
filter the pavement profile using quarter-car, four-degree-of-freedom models
(Gillespie, Sayers, and Segal 1980~. These statistics were user! to provide a base for
... .. . . ~%, .
cat~oranng actual I<~
Table 46 lists the statistics and their perceived ranking with regard to the
evaluation criteria. The rankings listed in this table are based on the information
listecI above. Guidelines used In the determination of the ranking of each criterion
are listed In table 47.
Based on the results shown In table 46, five statistics stand out as good candidates
for use as an Crucial pavement smoothness index. Top scoring indices are the TRI and
PI statistics, followed closely by the RN developed by Sayers (RNsayers) (Sayers and
Karamihas 1996), the Michigan DOT RQI (Darlington 1995), and the RN developed
by Janoff (RNJa7,off) Janoff 1988).
Correlation (R2) of TR! win user response for AC pavements is excellent and for
PCC pavements is fair, based on several reputable shldies using a range of pavement
roughnesses. Although for all pavement types combined, {RI does not correlate as
well win user response ~ R2 ~ 0.65 to 0.70), its correlation with panel ratings using
data from Minnesota for all pavement types was reported as 0.89 (Sayers and
Karamihas 1996~. The IRI is currently used as an index of initial pavement
smoothness In Minnesota.
For PCC surfaces, the reported correlation of PI with PST was moderate, and
correlation for AC pavements is estimated to be slightly beUer, because Me PI
responds to a long-wavelength range similar to the Ad.
Table 46. Evaluation of pavement roughness measuring Indices.
Criterion (see table 47) || A | B | C | D | Weighted ||
Weighted Value of Each Criterion 5 5 4 4 Score Rank
Index Description l
Sayers Ride Number (RN-Sayers)
r
Michigan DOT Ride Quality Index (RQ1)
.
Janoff Ride Number (RN-Janoff)
.
Mays Ride Number (MRN)
.
Mays Meter RMSVA Output (MO)
.
Spangler/Kelly Ride Number (RN-S/K)
Quarter Car Simulation (Qlrj
Telescoped Rolling Straightedge (TRS)
Slope Variance (SV)
Table 47. Guiclel~nes used for smoothness index rating.
l Basis of ll I I l l
Critenon | rating || Rating 1 | Rating 2 | Rating 3 | Rating 4 | Rating 5 |
A. Correlation with rider 2 . 0.55 to 0.65 to 0.75 to
l | respor~se for AC I R of best studies l | ~| 0.65 | 0.75 | 0.85 | 0.85
B. Correlahon with rider 05 0.55 o 0.65 o 0.75 o > 5
l | respo e for PCC | R2 of best studies | ~| 065 | 075 | 085 | 0.8 |
l | C. Correlation with other | Ability: PI, MRM, 11 one I two I Free I four | > five |.
roughness statistics IRI, RN, RQI,
D. Information Info: vanability,
availability | spec. limits ~poor ~fair ~ moderate ~good ~ excellent
~1~ [~- She ~
~_Wll~l"~LlVlLO 1V1 L1L~ 1~sayer~ RQI' and RNJ=off are excellent for AC pavements
(A 0.86) and good for PCC surfaces (A 0.81), based on reputable studies completed
using NCHRP-furnished data. Of these three, Me RQl is the only one currently used
in smoothness specifications for new PCC and AC pavements.
Indices Cat are currently used for initial pavement smoothness control, the PI and
the MRN, fell short in correlation with user response. Because of the long-stand~ng
use of these indices, there are a large number of correlations developed between
these indices and other roughness statistics. The correlations between PI and MRN
with IRI are reportedly good, with R2 values ranging from 0.92 to 0.97. However,
correlations of PI with MRN are not as good, with reported correlation coefficients
(R2) between 0.57 to 0.94 (Scofield 1993~.
Based on available information, no index of pavement smoothness stands out as
the single statistic of choice. The TR] is well known In the pavement management
arena but is a newcomer to the crucial smoothness control area. PI has been used for
many years for initial smoothness quality control, resulting in improved control of
pavement rideability. Reducing or eliminating the blanking band should make the PI
more sensitive to the short wavelength oscillations that can occur In Me paving
operation, thus resulting in better relation with user response. Since the PI statistic
can be computed using inertial-based, higher-speed profile measurements, data
collection, repeatability, and correlation with other smoothness indices can also be
greatly improved by the new technology and appropriate models.
171~T 17~T VIA 12(~)T ~11 hears mr~t~nti~1 tic On ini4;~1 c~mnr~thn~ce ct:~tictir hilt
Janoff, Sayers' ~` Len ~ art ~` ~ &~ ~ ~ ~ ~ ~ ~ ~ ~ - ~ ~ ~ &^ ~` - ~A LV~ ~ ~ &~ - ~ ~ ~ - ~` ~ ~ L
have seen limited use. Because of Ignited data about He relationship of these indices
win user response for new pavements and about He index levels required to
provide smoothness levels equal to or better than current specifications, it will be
difficult at this time to develop crucial smoothness specifications based on these
statistics. A carefully planned correlation study of rider response to the roughness of
new pavements measured by these indices would help to define appropriate
specifications smoothness limits. In the absence of such a study, assistance from
191
I]MTRI, the Michigan DOT, and Jim Research (developers of the RNSaye~;, RQI, and
RNJar~off, respectively) would be required to define these limits.
Because of the lack of a proven replacement, the PI, notwithstanding its
limitations previously described, should remain the current index of choice for
PCC and AC pavements. However, it is recommended that the blanking band be
eliminated and PI statistics be computed from surface profiles measured using
inertial-based systems. Since the PI amplifies wavelengths that are not related to user
response, and profiIograms do not provide accurate pavement profiles, measuring
"true" profiles will permit State Agencies to concurrently study other indices that
relate better with user response.
The IRI provides another opportunity to develop a profile-basecl smoothness
specification, although its correlation with user response is reportedly not better than
the PI statistic. Correlations are available for developing required levels of
roughness. To address He limited ability of the ~ statistic to differentiate between
short and long wavelengths, it may be possible, with information from I:MTRI, to
develop a mov~ng-average IR] specification. However, the window used for filtering
these profiles must be chosen carefully, and rectangular filters should be avoided.
Thus, the currently available information suggests that, for the immediate future,
the PI fusing "true" profile modeling and a zero blaring band) should remam the
primary index of pavement smoothness. An IRI-based smoothness specification will
provide a cradl~to-grave statistic but is not expected to unprove the current PI
correlation win user response.
Additional study of the RN~;U,off, RNSayers' and RQl statistics is recommended to
further improve the quality of initial pavement smoothness specifications.
Information necessary for unplementation of such a specification Includes the
correlation of each index win user response for each pavement type in the
smoothness range typical of new pavements. Variability associated with profile
measuring equipment, operators, cI~matological factors, profile filters, and sampling
Intervals must all be evaluated for the Index most related to user response, to allow
development of statistically based, performance-related smoothness specifications.
To allow introduction of specifications using the Ad, RQI, or RN statistics,
recommendations for equipment capabilities have been developed that are broad-
based. This will provide measured pavement profiles that can be used to provide
sufficient accuracy and breadth for use win all promising smoothness statistics.
Profile Measuring Equipment Requirements Based on Selected Indices
The primary Indices reconnnnended for potential use in initial s m oothness
elm; ~ ~a +; ~ ~ PIT MAT Teal
O~_l`~Ct~l=~, ~ A, 6~Janoff, $~saye=, RQI, and I~J have inherent characteristics that
help to define He requirements of He equipment necessary to measure the pavement
profiles needed for index computation. These characteristics include the pavement
192
wavelengths to which the index responds and the variability associated with the
index. Using these characteristics, the requirements of construction quality control
pavement profile measuring equipment can be defined; these requirements are
described below.
Measured Profile Wavelengths
Because each primary index responds to slightly different pavement surface
profile wavelengths, it is critical Cat equipment for measuring construction pavement
smoothness be capable of accurately measuring surface profile ~ these wavelengths.
These critical wavelengths reported In the literature are listed In table 48 for the
primary indices Janoff 1986; Paterson 1987a; Scofield 1992; Darlington 1995; Sayers
and Karamihas 1996~. To provide additional accuracy and capacity, these wavelength
ranges have been modified slightly, accounting for gains greater than 0.l, and both
are shown in table 48.
It is evident from table 48 that equipment capable of accurately providing the
primary smoothness indices must be able to accurately measure pavement surface
wavelengths from 0.9 to Il0 It (0.27 to 33.5 m). Current equipment is capable of
measuring wavelengths from 3 in (76.2 ~run) to 300 ft (91.4 m).
,Sampl~n~ Interval
To measure the shorter wavelengths accurately. an anDroDriate sampling interval
must be selected.
-- -r - - -A
The required sampling interval is defined by the minimum
wavelength of interest. Shannon's Sampling Theorem indicates that the sampling
interval needs to be at least two times the minimum wavelength of interest. The
actual sampling rate required to recover the necessary signal frequencies is called Me
Nyquist rate (Stanley, Dougherty, and Dougherty 1984~. If the Nyquist sampling rate
is not attuned, a phenomenon known as "aliasing" results. This causes frequencies
-a ~ - - - --r r - - r
Table 48. Wavelengths measured by primary indices.
~.
Index ~Source ~ Reported bandwidth ~ Moclified bandwidth
Paterson 1987a 3.0 to 80 It
Scof~eld 1992 I.0 to 75 ft
RNsayers Sayers 1996 I.7 to 36 It
RQl Darlington 1995 2.0 to 50 ft
~anoff ~JaIloff 1986 ~1.6 to ~ ft
2.! to I10 It
0.9 to 85 It
l.: to 77 It
1.4 to 88 it
I.2 to 10 ft
ft = 0.305 m
193
to be mistaken for entirely differently frequencies upon signal reconstruction or
recovery. A practical example of this phenomenon is the wagon wheel effect often
noted In western movies. Because of an insufficient rate of film collection, long
wavelengths are added to Me visible picture, resulting in the speed of the wheels
appearing to slow and sometimes reverse as the wagon wheel rotation speed
Increases (Scofield 1993~.
The reason for requiring smaller sample intervals is to help to avoid abasing of
the profile. Aliasing occurs when phantom long wavelengths are adcled to the profile
because of an Insufficient sampling interval. DarI~ngton reports Cat if the phantom
wavelengths are digitized before they are filtered from the profile, they will remain in
the final profile (Darlington 1995~. As a result, the Michigan DOT collects eight
samples per minimum wavelength of interest and uses short wavelength analog
antialias~ng fitters on its profiles prior to digitizing it in its computers. The PRORUT
system collects four samples per m~rumum necessary wavelength anct uses a low-pass
(smoothing) filter set at about Oft (0.3 m) wavelength to provide antialiasing (Sayers
1990~. However, according to DarI~ngton and DeFrain at the Michigan DOT, Me
moving average, smoothing filter does not sufficiently reduce the aliased effect.
:~creasing Me sampling rate reportedly helps to reduce the aliasing phantom
wavelengths.
Table 48 indicates that the smallest wavelength of interest for the primary indices
is 0.9 It or 11 in (0.27 m) in length. If an analog antialiasing filter is used bra the
profile measuring equipment, a sampling interval of 2 In (50.S ~run) is reco~runended.
If digital antialiasing filters are employed, a sampling interval of ~ in (25.4 man)
shouIc} be used.
It is interesting to note that profiIograph sampling rates are very near this range.
For example, Me sampling rate for Me Cox CSS200 computerized profiIograph is 1.3
In (33 mm), and the sampling rate for the IPMC-McCracken computerized
profiIograph is 1.22 in (Cox and Sons 1994; Noonan 1994~. Because current profile
measuring equipment can measure pavement profile at intervals of ~ In (25.4 mm) or
less, it is recommended that the sampling interval for the initial smoothness control
equipment specification be I.0 in (25.4 mm), unless analog antialiasing filters are
used.
Distance Accuracy
Longitudinal accuracy is necessary for defining must-grind locations and for
correlating output from repeat passes over a pavement surface. A typical range of
distance accuracy varies from 0.02 to 0.! percent for most high-speed profile
measuring devices (Evans 1993~. The new FHWA LTPP profile measuring devices
can measure longitudinal distances within 0.05 percent of the actual length, 0.5
ft/l,OOO It (0.15 m/305 m) (FHWA 1994b). Currently used profiIographs measure
distances with about 0.10 percent error, that is, ~ It (0.3 m) per 1,000 It (305 m)
(Noonan 1994~. Because He level of longitudinal measurement accuracy of
profilograph equipment has not been noted as insufficient, and because current
194
technology is available to measure longitudinal profile lengths at high speeds with
the same accuracy, a distance accuracy requirement of 0.! percent (1 ft/l,OOO it [0.3
m/305 m]) is recommended for the initial smoothness control equipment
specification. This is consistent with the accuracy requirements for Class ~ equipment
in ASTM E950-94.
Vertical Elevation Accuracy
The vertical elevation accuracy of a profiling system can be measured both
statically and dynamically. Static accuracy determination is measured when the
device is stationary and is Me accuracy method used for current profilograph
systems. Accuracy requirements are generally made up of two
components precision and bias (ASTM El77~. Precision is a measure of Me profile
repeatability of one or more pieces of equipment using one or more operators. Bias
is the deviation of the device from the true profile, as measured by the best available
profile measuring method-typically either rod and level or Face Dipstick.
The static accuracy method is used for the current profilographs with a required
vertical measurement precision and bias of +0.02 in (0.51 mm) on 0.5-in (12.7-mm)
and 1.5-in (38-mm) gage blocks (ASTM El274~. For the new FHWA LTPP profiling
equipment, the required static precision and bias is +0.005 in (0.125 mm) at 95
percent reliability. ASTM E950 requires a vertical measurement resolution of <0.005
In (0.1 mm) for Class ~ measurement. Among commonly available high-speed profile
measurement devices, the static precision and bias ranges from 0.002 to 0.012 In (0.05
to 0.3 mm).
Because Me static accuracy of a profile measuring device is not as critical as the
dynamic accuracy, it is recommended that the static accuracy required In the crucial
smoothness control equipment specification be better than the required dynamic
accuracy) but not unduly restrictive. This is the logic used In defining Me
specifications for the FHWA LTPP profiling equipment. Therefore, a static accuracy
(precision and bias) of 0.005 in (0.125 mm) is recommended.
A measure of the dynamic accuracy of a pavement profile measuring device
widen a specified range of wavelengths is much more difficult than that of static
accuracy. However, it is necessary to identify inaccuracies in the hardware and
software that may not be evident in static accuracy measurements. To quantify
dynamic accuracy, repeat measurements must be made of pavement profiles through
a range of new construction roughness, using care to eliminate the effects of
horizontal wander, longitudinal positioning deviations, filtering variations, and data
nror~.~in~ divergence. Such noint-bv-noint comparisons were performed bv
rim o ~ ~ -I- A rim rat --- A--------- - A
Sp angler, et al. (1990) for the Federal Aviation Administration (FAA). The reported
average standard deviations for AC pavements from that study are listed In table 49
(Sp angler, Gerardi, and Yager 1990).
195
Table 49. FAA Bluegrass Parkway Profilometer~ evaluation results
(Spangler, Gerardi, and Yager 1990~.
. .
Average Std. Deviation Average Absolute Bias
Profiler (18 runs) (~S runs, 1,057 points)
North-Central region 0.012 In (0.31 mm) 0.016 In (0.40 mm)
ProfiIometer~
AZ ProfiIometer@-Laser 0.010 In (0.26 mm) 0.054 In (~.36 mm)
AZ ProfiIometer~-Infrared ~0.019 in (0.49 mm) ~0.035 In (0.89 mm)
Sp angler, Gerardi, and Yager (1990) tested Me K. I. Law mode! 690DNC
ProfiIometer~ used by the FHWA Norm Central Region contractor on a 0.2-mi (0.3-
km) section of the Bluegrass Parkway in Kentucky. That pavement was a relatively
smooth AC surface and should be considered as a good site for obtaining repeatable
profile measurements. Rod and level measurements were recorded at I-ft (0.305-m)
Intervals.
The method of profile analysis used for this equipment included moving averages
and filtering of profile data from IS runs (Sp angler, Gerardi, and Yager 1990~. Prior
to digitizing, the pavement profile was measured on An (25.4-mm) Intervals,
averaged over ~ It (0.3 m), and recorded at 0.5-ft (0.152-m) intervals. During profile
measurement, the profile was high-pass filtered using a third-order filter set at 300 ft
(92 m). To adjust for phase shifting, the profile was later reverse filtered using Me
same filter, resulting In a s~xth-order high-pass filter.
The rod and level profile was "tipped" to remove the very long wavelengths.
Then it was subjected to the same filtering as the ProfiIometer~ data received.
ProfiIometer~ precision was determined using the standard deviation of Me
elevations for IS runs at each ~ ft (0.3 m) interval.
The average standard deviation was 0.012 in (0.31 mm), and the mean absolute
bias between the average ProfiIometer~ elevations ant! the rod and level elevations
was 0.016 In (0.40 mm) (Spangler, Gerardi, and Yager 1990~. Two other K. I. Law
mode! 690DNC systems were also tested at this site, and me average standard
deviations for those systems were about the same as the North Central Region
ProfiIometer@. The average absolute bias for these systems was larger, as shown In
table 49. Po~nt-by-point comparison statistics of repeated profiIograph traces are not
available but are expected to be In the range of Me data reported by Sp angler,
Gerardi, and Yager (1990~.
ASTM E950-94 requires a maximum one-standard deviation precision of 0.015 in
(0.38 mm) from at least 10 repeat runs to achieve Class ~ measurement status. In
addition, to achieve Class ~ status, the average of the difference between the
196
measured elevation and a baseline elevation for 1,057 points spaced at I-ft (0.305-m)
Intervals must not exceed 0.050 In (~.25 mm).
Although the ASTM precision and bias requirements for Class ~ measurement are
slightly less stringent than the Kentucky data indicate is feasible, it is recommended
that the Class ~ ASTM E950 dynamic precision and bias requirements be followed in
the development of specifications for initial smoothness measuring equipment.
Other Considerations
Several other attributes of the Initial smoothness measuring equipment that may
not be specifically recommended In the specification, but that will nonetheless be
useful, include minimal weight, higher speeds, ease of use, must-grind location
capability, horizontal location accuracy control, rapid availability of survey data
results, and automated calibration, data collection, filtering, and reporting. An
understanding by the construction contractor of the newly constructed pavement
smoothness level widen days or hours instead of widen weeks can improve road
surface quality and result in savings to both the contractor and the owner agency.
Lightweight surface measuring systems, like hose developed in Michigan and Texas,
can assist Me contractor and Me State In early detection of profile roughness.
Increased speed and automation In the data collection and analysis processes can also
unprove cost-effectiveness of a quality control process. Because of the reported
problems with PI variations along different wheelpaths, a device to ensure horizontal
positioning could eliminate this random variability and unp rove repeatability.
Finally, it remains unperative that a must-gr~nd location identification method be
included in Me software of any system for measuring Initial construction surface
profiles. None of the primary indices provide this capability inherently, and it will
be necessary, as with the PI, to conduct a supplementary simulated straightedge
analysis to identify locations that require grinding. The Michigan DOT reports that
must-gr~nd areas determined using the 0.3 In (7.6 mm) criteria on profiles measured
with their inertial profiler were about 33 percent of the areas determined based on
profiIograms from simulated and actual profiIograph measurements. If must-grind
specifications are based on inertial-based profiles, modifications of the grind [units
may be necessary to maintain a high quality of Initial pavement smoothness.
Summary
This chapter has presented information on both smoothness-measur~ng equipment
and smoothness indices. The various types of smoothness-measur~ng equipment are
discussed, with a detailed review provided in appendix E. Various primary
pavement smoothness indices were presented and reviewed In terms of their ability
to correlate with user response, Weir ability to correlate with other smoothness
statistics, and the availability of information for development of a specification.
Ultimately, two smoothness indices were ranked highest according to those criteria:
IRI and Pl. They were followed closely by three indices that show much better
197
correlation with user response but for which insufficient information is available to
develop a working specification at this time (Michigan DOT RQI, Janoff Ride Number
[~moff] and Sayers ode Number Sayer
Based on the five indices, properties of the equipment recommended for use in
modern smoothness construction specifications were identified. These properties
include the ability to accurately measure pavement surface wavelengths from 0.9 to
110 ft (0.27 to 33.5 m) with a sampling intermural of ~ in (25.4 mm). The static accuracy
(precision and bias) of this equipment should be +0.005 in (0.125 mm) at 95 percent
reliability. The 10-sample, single operator, dynamic precision for the equipment
should be 0.015 In (0.38 mm) at a confidence level of 95 percent within the
wavelengths of 0.9 and Il0 it (0.27 to 33.5 m). The dynamic bias of the equipment,
as compared to high-precision rod anc! level or Dipstick baseline profiles, should be
no more than 0.05 In (~.25 mm), according to the procedures described In ASTM
E950-94. A must-gr~nd location feature should also be Included In the software.
Over attributes, such as m~nunizing weight, Increasing speed, controlling
horizontal wander, and automating data collection, calibration, analysis, and
reporting should also be encouraged. Table 50 summarizes these recommended
properties of smoothness-measuring equipment.
The movement toward adopting a new pavement profiling system will not
happen overnight. Many of the current smoothness-measuring practices are deeply
Ingrained within Me highway agencies and paving industries that make it Impractical
to quickly move toward the adoption of new equipment and smoothness indices.
Highway agencies and paving contractors have invested significant resources In the
current systems, which in many cases are providing satisfactory results. However, in
order to make significant improvements to the way that smoothness data are
collected and reported, a slow and gradual movement toward profile-based systems
appears warranted, and effective training programs will be necessary to fully
mdoctr~nate such systems into the mainstream of data collection activities.
. ~- ~.
198
Table 50. Summary of recommended properties for smoothness
measuring equipment.
11
Property
| Recommended Requirement
. .
Measured Profile Wavelengths
1.2 to 100 ft (0.37 to 30.5 m) up to 55 mi/hr
(88 km/hr)
1 In (25.4 mm) if digital antialiasing used
2 in (51.8 mm) if analog antialiasing used
Sampling Interval
l .
Distance Accuracy
Vertical Elevation Accuracy
Static Precision/Bias
Dynamic Precision/Bias
Over Considerations
0.1 percent
+0.005 in/+0.005 in (0.125 mm/0.125 mm)
+0.015 in/+0.05 in (0.38 mm/1.25 mm)
Lightweight
Rapid sampling speed
Ease of use
Identify must-grinds
Horizontal positioriing
Rapid availability of survey data results
Automated calibration, analysis, reporting
199
MO = 20 + 23(RMSVA12) + 58(RMSVA49)
Unfortunately, no information is available regarding the relationship of MO with
user response. However, its fair correlation with Mays Meter response (R2 = 0.82)
suggests that its correlation with user response may be similar to that of the MEN
(Hudson, Unpin, and Elkins 1987~. The reported correlation of MO with PI is 0.94,
indicating that its user response may be more similar to that of a California
profiIograph (Uddin, Hudson, and Elkins 1990~.
Evaluation of Smoothness Indices
(19)
This section presents an evaluation of the smoothness indices previously
discussed. The purpose of this evaluation is to ultimately identify the smoothness
statistics that are most suitable for use in a smoothness specification. Identification of
these statistics then provides a basis for defining the equipment characteristics and
properties needed to measure initial pavement smoothness.
Review of Kev Selection Factors
To select a smoothness index from the large number of available statistics
described previously, it is necessary to first define the factors necessary for a good
initial pavement smoothness statistic. Foremost among these factors is correlation
win user response for AC and PCC pavements, but He ability to correlate the
statistic with other statistics and the ability to correlate win current smoothness
statistics so that specification Innits can be easily developed are also considered
critical In He selection process. In addition, He information available from past
experience will make an Index more desirable. The Importance of these factors is
discussed below.
Correlation With User Response
Although a large number of pavement roughness indices have been developed,
those that have remained In use over He years have had one common
characteristic" Hey were thought to relate well win the option of roadway users
regarding He rideability of He roadway surface. For example, the BPR and PCA
statistics fall into the category of roughness measures that did not correlate well with
user response, and these statistics are now rarely used. The Mays Ride Number is in
the process of being replaced by statistics that correlate better with user response;
some reported weakness In He correlation of the PI statistic with user response is
weakening its position as the index of choice for new pavement smoothness control.
In determining He level of correlation win rider response for each of the statistics
evaluated In this study, the R2 values noted in the previous section were reviewed.
Because PCC pavements generally contain shorter wavelength roughness components
Han AC pavements, and because different indices react differently to different
wavelengths, He correlation of indices with user response varies with pavement type.
~6
