Transportation Research Record 1816 ■
Paper No. 02-4039
43
Use of Distress and Ride Quality Data to
Determine Roughness Thresholds for
Smoothing Pavements as a Preventive
Maintenance Action
Doseung Lee, Karim Chatti, and Gilbert Y. Baladi
In this discussion, 462 pavement sections from 37 projects in Michigan
were analyzed to investigate the interaction between pavement surface
roughness and distress. The main hypothesis of this research is that an
increase in roughness leads to higher dynamic axle loads, which in turn
can lead to a tangible acceleration in pavement distress. If this relationship is established, then it will be possible to plan a preventive maintenance (PM) action to smooth the pavement surface. Such a PM action
is bound to extend the service life of the pavement by several years. The
objective of the research presented was to develop such roughness thresholds. The selected projects include 13 rigid, 15 flexible, and 9 composite
pavements. The ride quality index (RQI) and distress index were used
as measures of surface roughness and distress, respectively. To get a good
relationship between distress caused by dynamic loading and surface
roughness, dynamic-load–related distress types were extracted. The
analysis showed very good relationships between distress and roughness
for rigid and composite pavements (R2 = 0.739 and 0.624). However, for
flexible pavements there was significant scatter (R2 = 0.375), reflecting
the higher variability in flexible pavement distresses. A logistic function
was used to fit the data, and roughness thresholds were determined as
the RQI-values corresponding to peak acceleration in distress. These were
determined to be 64 [international roughness index (IRI) = 1.99 m/km or
124 in./mi] for rigid pavements and 51 (IRI = 1.43 m/km or 89 in./mi) for
composite pavements.
Dynamic pavement loading due to the mix of commercial heavy
trucks that make up the national fleet has been steadily increasing in
the United States as the national highway network continues to age.
Although this may be somewhat counteracted by an increase in the
use of “road-friendly” vehicles, truck loading and its dynamic component remain a major cause of road damage. Previous research has
suggested that dynamic wheel loads, caused by surface roughness,
can be up to 40% higher than static loads (1). All road surfaces are
inherently rough even when they are new, and they become increasingly rougher with age depending on pavement type, traffic volume,
environment, and so forth. This process is the result of the interaction between vehicles and pavements. Accordingly, it is reasonable
to assume that there is a threshold value in roughness in which
dynamic load increases sharply leading to acceleration in pavement
damage. If this threshold value exists, it could be a useful preven-
Department of Civil and Environmental Engineering, Michigan State University,
East Lansing, MI 48824.
tive maintenance (PM) tool, whereby a PM action, such as smoothing the pavement surface, is taken to extend the service life of a
given pavement for several years.
Many government and state agencies use roughness as one of the
objective measures for the management of their pavement network.
The Michigan Department of Transportation (DOT) uses the ride
quality index (RQI) to characterize surface roughness. Michigan DOT
has also been surveying surface distresses for the entire pavement
network in a systematic fashion, since 1992. The distress index (DI)
is used to quantify the severity and extent of surface distresses along
a project. In this discussion, measured roughness and DI data from
37 pavement projects (462 sections including rigid, flexible, and
composite pavements) were analyzed for the purpose of determining RQI threshold values aimed at minimizing pavement damage
due to dynamic loading.
METHODOLOGY
The main hypothesis of this research is that an increase in pavement
roughness leads to higher dynamic axle loads in certain portions of
the road. This amplification in the load magnitude can lead to a tangible acceleration in pavement distress. If the relationship between
increased loading and increased distress is established, then it will
be possible to determine the optimal timing for PM action in the form
of smoothing of the pavement surface (by way of milling, grinding,
or a thin overlay). This is bound to extend the service life of the
pavement by several years.
The objectives of this research were as follows:
• Test the above hypothesis,
• Develop a threshold based on roughness and dynamic loading
considerations, and
• Determine the optimal timing of the PM action aimed at extending the pavement service life.
These objectives were achieved by using two parallel approaches:
(a) a mechanistic approach and (b) an empirical approach. In the
mechanistic approach, dynamic loads are predicted using a truck
simulation program and pavement profiles; then pavement damage
is calculated using a damage law such as the fourth power law. In the
empirical approach, pavement distresses are correlated directly to
surface roughness at increasing levels of roughness. This discussion
only summarizes the results from the second approach.
44
Paper No. 02- 4039
Transportation Research Record 1816
from 55 to 70 indicate fair ride quality; whereas pavements with
RQI values of more than 70 are considered as having poor ride quality (3). The longitudinal profile for the entire pavement network in
Michigan is measured annually using a Rapid Travel Profilometer.
The data are used to calculate both RQI and IRI, which are reported
to FHWA. Figure 1 shows the correlations between RQI and IRI for
rigid, flexible, and composite pavements.
ROAD SURFACE ROUGHNESS
Roughness is not only an indicator of road surface condition but also
excites the dynamic behavior of trucks, thus increasing damage.
There are several ways of quantifying road roughness. One of the
most widely used parameters is the international roughness index
(IRI). Another way of quantifying road roughness is RQI, which
was developed and is being used by Michigan DOT.
PAVEMENT DAMAGE EVALUATION
RQI
The Michigan DOT collects both functional and structural distress
data to assess the surface condition of the pavement. Distress data
are collected by videotaping 50% of the pavement network every
year. The videotapes are reviewed in the office and each distress on
the pavement surface within each 10-ft (3-m) long section is identified, reviewed, checked, scored, and stored in the pavement
management system (PMS) databank. Hence the data include information on the status of each crack and its location within the 10-ft
(3-m) long section. The distress data are then grouped into surveying unit sections that are 0.1-mi (161-m) long. Thus the PMS databank contains, for each 0.1-mi (161-m) segment of the road, detailed
data for each type of pavement distress and the severity and extent of
the associated distress. The term “associated distress” is used in the
Michigan DOT rehabilitation practice to denote secondary distresses associated with the principal distress. For example, spalling
associated with a transverse crack would be considered as associated distress for the transverse crack.
The Michigan DOT PMS group has developed a rating system
whereby each type of principal distress and its associated distress
level are ranked and assigned distress points (DP) based on their
impact on pavement performance and on experience. For any pave-
As its name suggests, RQI describes the ride quality of the road. In
the early 1970s Michigan DOT conducted a study to determine an
objective measure that would correlate ride quality to the subjective
opinions of highway users. Using psychometric tests, it was found
that some components of a road have a strong effect on user opinion,
whereas others have a significantly lesser effect (2). Through a series
of mathematical and statistical steps, the power spectral density was
found to correlate at 90% with subjective opinions. On the basis
of this, the profile is split into three wavelength bands: 0.6 to 1.5 m
(2 to 5 ft), 1.5 to 7.6 m (5 to 25 ft), and 7.6 to 15.2 m (25 to 50 ft).
Wavelengths shorter than 0.61 m (2 ft) mostly create tire noise and
those longer than 15.2 m (50 ft) fail to disturb the vehicle suspension. RQI is calculated from these three PSD wavelength bands
according to the equation shown below (3):
RQI = 3 ln(Var1 ) + 6 ln(Var2 ) + 9 ln(Var3 )
(1)
where Var1, Var2, and Var3 are variances for 7.6- to 15.2-m, 1.5- to
7.6-m, and 0.6- to 1.5-m wavelengths, respectively.
An RQI value between 0 and 30 indicates excellent ride quality;
RQI values from 31 to 54 indicate good ride quality; and values
120
100
RQI
80
60
Rigid
Composite
40
Flexible
Rigid
20
Composite
Flexible
0
0
0
50
100
1.0
150
2.0
3.0
IRI
FIGURE 1
Relationship between RQI and IRI.
200
250
4.0
300
in/mi
m /k m
Lee et al.
Paper No. 02- 4039
ment section, DI can be calculated as the sum of DP along the section normalized to the section length. The length of the pavement
section is expressed with regard to 161-m (0.1-mi) unit sections. The
equation for DI follows:
DI =
∑ DP
L
(2)
where Σ DP is the sum of DP along the pavement section and L is
the length of the pavement section in 161-m (0.1-mi) unit sections.
The DI scale starts at zero for a perfect pavement and it increases
(without a limit) as the pavement condition worsens. Michigan DOT
categorizes DI into three levels: low (DI < 20), medium (20 < DI
< 40), and high (DI > 40). A pavement with a DI of 50 is considered to have exhausted its service life; hence its remaining service
life (RSL) is zero, and it is a candidate project for rehabilitation. This
DI threshold value was established based on historical pavement
performance data and on experience.
DATA ANALYSIS
The analysis in this research consisted of three phases: The first phase
was aimed at showing how one could use roughness and distress data
for the purpose of discerning between dynamic-load–related distress
and non-load-related distress. The results from the first phase showed
that plotting roughness and distress profiles together along the project
length allows for identifying zones of high distress and high roughness, and high distress and low roughness. Combining the above
information with the trends of distress and roughness indexes from
year to year allows for distinguishing load-related from non-loadrelated distress accumulation. In addition to these plots, dynamic load
and DP profiles along the project length were plotted and the coefficients of correlation (ρ) between dynamic loading and DP were determined. In this analysis, dynamic load profiles for thirty-six 161-m
(0.1-mi) sections were generated using a standard five-axle tractor
semitrailer and the TruckSim truck simulation program. DP were also
calculated at 3-m (10-ft) intervals along each section, using Michigan
DOT’s distress data. These plots and the corresponding coefficients
of correlation also allowed for distinguishing dynamic-load–related
from non-dynamic-load–related distress accumulation. The detailed
results from this phase appear elsewhere (4, 5).
The second phase was aimed at getting a relationship between
roughness and load-related distresses by isolating dynamicload–related distresses from non-load-related ones and deciding on
roughness threshold values. This phase is the focus of this discussion. It should be noted that, ideally, one should use a roughness
indicator that is correlated to truck dynamics as opposed to using
roughness indicators that are based on the response of cars to the
pavement surface. However, there were no such indicator available
at the time. Later, the need for developing a truck-based roughness
indicator became apparent if such an indicator were to be used at
the project level (i.e., for a specific pavement profile). This led to
the development of a new roughness index, called the dynamic load
index, which is based on the truck’s response to a pavement profile.
The details of that research are described elsewhere (5).
The third phase consisted in developing a model for determining
the optimal timing of PM smoothing action. In this model, the con-
45
cept of an RQI “trigger” value was introduced. These trigger values
represent the value at which planning for the smoothing PM action
some years down the road needs to take place. On the basis of actual
RQI growth rates from 112 in-service (rigid) pavements, a reliability table was generated indicating the probabilities that the pavement
will not reach the threshold RQI value before x-years for different
RQItrigger values, with x being the PM planning period. The detailed
results from the third phase are reported elsewhere (5, 6).
Site Selection
The study involves the analysis of roughness and distress data from
a small number of actual pavement projects in Michigan. Ten projects with known performance records and having exhibited some
distress and 27 projects in which PM activities were done during 1997
and 1998 were selected. The 37 sites were analyzed to get relationships between roughness and distress, and to come up with roughness
threshold values. Thirteen of the 37 sites were rigid, 15 were flexible,
and 9 were composite pavements. The length of these pavement
projects varied from 2.4 to 26.4 km (1.5 to 16.5 mi) with an average
project length of 10 km (6.25 mi). Their ages ranged from 1 to 43 years.
The commercial daily traffic volume ranged from 70 to 8,900. The
selected sites for each phase are listed in Table 1.
Data Collection
For the 37 selected sites, DI and RQI data as well as road surface profiles were obtained from the Michigan DOT PMS database. DI values
were available for 1993, 1995, and 1997, whereas RQI values and
road profiles were available for the period between 1992 and 1996.
The DI and RQI data were available for each 161-m long (0.1-mi)
section. Surface profile data were converted to ASCII files containing
surface elevations at 76-mm (3-in.) intervals. Detailed distress data
in the form of distress type, severity, and extent were also available
at 3-m (10-ft) intervals. The data were then converted to DP.
Relationship of Distress, Roughness, and
Dynamic Load
The relationship between surface roughness and dynamic axle loads
is well established, and the models for simulating the response of
heavy vehicles to road roughness are well developed and validated
(7). As mentioned above, the TruckSim program was used to generate dynamic axle loads for a standard five-axle tractor semitrailer
moving at 96 km/h (60 mph). This program was developed at the University of Michigan Transportation Research Institute. The program
is used to predict vehicle vibration due to pavement surface roughness
and the dynamic axle loads resulting from these vibrations. It uses
detailed nonlinear tire and spring models and includes major kinematic and compliance effects in the suspensions and steering systems.
The use of the five-axle tractor semitrailer is based on earlier analysis
which showed that dynamic loads from the three main truck types
traveling on U.S. highways (two-axle and three-axle single unit trucks
and the five-axle tractor semitrailer) were repeatable in space, with a
coefficient of correlation higher than 0.7. In this analysis, actual pavement surface profiles from 333 sections evenly divided among the different pavement types were used. The sections are 161-m (0.1-mi)
46
Paper No. 02- 4039
TABLE 1
Transportation Research Record 1816
List of Projects Investigated
Route
EB I-94
EB US-10
WB US-10
NB I-69
EB I-69
WB US-2
WB US-2
EB US-2
WB US-2
EB US-2
NB US-24
NB US-31
WB I-196
EB US-10
EB I-96
NB US-23
M-88
M-95
M-57
M-57
M-30
M-57
M-28
M-65
M-36
M-36
NB US-31
NB US-27
EB I-94
NB US-127
US-12
M-115
WB US-2
M-95
M-52
M-21
SB US-131
Control
Section
11017
18024
18024
23063
76024
21022
21022
21022
21025
21025
63031
70014
70023
09101
33084
47014
05031
22013
25102
25102
26032
29022
31021
35012
47041
47041
61074
72014
38101
38131
11011
18011
21022
22012
76012
76062
78012
Pavement Type
Rigid
Rigid
Rigid
Rigid
Rigid
Rigid
Rigid
Rigid
Rigid
Rigid
Rigid
Rigid
Rigid
Flexible*
Flexible*
Flexible*
Flexible
Flexible
Flexible
Flexible
Flexible
Flexible
Flexible
Flexible
Flexible
Flexible
Flexible
Flexible
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Composite
Mile Post
1.015-5.875
0.0-7.598
0.0-7.598
0.0-5.0
0.0-3.827
4.1-5.650
5.650-8.414
6.3-8.429
0.0-6.191
0.0-6.191
0.0-2.5
0.0-5.5
0.0-5.5
0.924-7.356
8.979-17.491
0.0-7.165
19.313-20.804
0.0-13.012
0.0-5.780
5.780-9.841
0.0-16.591
0.0-9.819
0.0-9.605
7.040-16.027
13.448-20.630
20.630-23.252
0.0-3.084
0.0-8.890
0.0-8.7
0.0-5.2
0.0-3.160
0.0-12.230
0.0-1.981
10.170-16.230
2.282-9.210
3.278-12.583
0.092-2.463
Construction
Year
1959
1976
1976
1991
1991
1957
1962
1962
1971
1971
1953
1954
1956
1990
1993
1992
1978
1982
1975
1978
1984
1984
1984
1973
1971
1986
1978
1975
1985
1987
1989
1985
1991
1987
1987
1987
1990
NOTE: 1 mi = 1.6 km. EB = eastbound; WB = westbound; NB = northbound; SB = southbound; M = Michigan.
* rubblized PCC base
long. Figure 2 shows plots of 95th percentile dynamic load (normalized relative to the static axle load) versus RQI. The figure shows a
clear increase in dynamic load with increasing RQI.
The relationship between load and distress, however, is complex
and not well understood. The variability in pavement materials and
environmental and traffic factors, coupled with errors in field measurements and the subjectivity of the field distress surveys, make it
hard to relate roughness and dynamic loading to the accumulation
of pavement distress in in-service pavements. Furthermore, some
distresses are mainly material or environment related. Therefore, any
interpretation of field data must be made with caution. Analysis conducted as part of this research showed that it was possible to use
roughness, dynamic load, and distress data profiles along project
length and unit [161-m (0.1-mi)] sections to distinguish between cases
of dynamic-load–related distress and non-load-related distress (4).
Development of RQI Threshold Values
The above-referenced analysis on the relationship between distress
and roughness showed that distresses in some projects were dynamic
load related whereas they were not in others. This is due to the fact
that load is not the only factor that causes distress in pavements; there
are many other factors that contribute to distress, such as material
problems, environmental factors, and so forth. Therefore, to get roughness threshold values in which distress starts to pick up due to dynamic
loading, it is necessary to isolate dynamic-load–related distresses,
and then relate them to pavement surface roughness.
Extraction of Load-Related Distress Types
To isolate dynamic-load–related distress types, several subsections
having the same range of DI level but different RQI levels were
selected. Table 2 shows the criteria of subsection selection and the
number of subsections selected for each DI range. For this analysis,
805-m (0.5-mi) subsections were used instead of 161-m (0.1-mi)
subsections to minimize any shifts between roughness and distress
data. For each selected subsection, DPs for each distress type were
calculated. In this calculation, transverse and longitudinal cracks
were divided into two types: (a) a crack having no associated distress,
and (b) a crack with associated distress such as spalling. Associated
distress is a distress occurring around a main distress and is used to
define subsequent deterioration of the main distress in the Michigan
DOT PMS. It should be noted that the Michigan DOT distress list
does not include fatigue cracking as a separate distress type. Fatigue
Paper No. 02- 4039
95th Percentile
Dynamic/Static Load Ratio
Lee et al.
1.6
1.5
1.4
1.3
1.2
1.1
1
0
20
40
60
80
100
120
80
100
120
80
100
120
RQ I
95th Percentile
Dynamic/Static Load Ratio
(a)
1.6
1.5
1.4
1.3
1.2
1.1
1
0
20
40
60
RQ I
95th Percentile
Dynamic/Static Load Ratio
(b)
1.4
1.3
1.2
1.1
1
0
20
40
60
RQ I
(c)
FIGURE 2 The 95th percentile dynamic load (normalized to static axle load) versus RQI:
(a) rigid, (b), composite, and (c) flexible pavements.
47
48
Paper No. 02- 4039
Transportation Research Record 1816
TABLE 2
Criteria of Section Selection and Number of Samples
Rigid Pavements
Relative
Roughness
0-10
10-20
20-30
30-40
Smooth
Rough
RQI<50 (11)
RQI>65 (10)
RQI<55 (17)
RQI>70 (17)
RQI<65 (4)
RQI>80 (4)
RQI<60 (4)
RQI>75 (4)
DI Level
Flexible Pavements
Relative
Roughness
10-20
Smooth
Rough
RQI<35 (12)
RQI>55 (8)
20-30
DI Level
30-40
40-50
50-60
RQI<40 (5)
RQI>60 (4)
RQI<40 (9)
RQI>60 (4)
RQI<40 (6)
RQI>60 (2)
RQI<40 (2)
RQI>60 (6)
Composite Pavements
Relative
Roughness
10-20
Smooth
Rough
RQI<40 (10)
RQI>55 (8)
DI Level
20-30
30-40
40-50
RQI<40 (5)
RQI>60 (4)
RQI<45 (5)
RQI>60 (7)
RQI<55 (4)
RQI>62(4)
NOTE: Number of samples is in parentheses.
cracking is implicitly described by an increased number of transverse and longitudinal cracks. In Figures 3, 4, and 5, proportions of
DP of each distress type are plotted for relatively rough and relatively smooth subsections at each DI level. These figures show that
transverse cracks with associated distress are clearly dominant in
relatively rough subsections at most DI levels, whereas transverse
cracks without associated distress are dominant in relatively smooth
subsections at most DI levels. These plots allow for discerning distresses that are dynamic load related. On the basis of Figure 3,
“transverse cracks with associated distress,” “transverse joint deterioration,” “delamination,” and “patch deterioration” were selected
as dynamic-load–related distress types for rigid pavements. On the
basis of Figures 4 and 5, only “transverse cracks with associated distress” were selected as dynamic-load–related distress for flexible
and composite pavements.
Determination of Roughness Threshold Values
Dynamic-load–related distress types were extracted from distress
data, as described above, and DP counting only these distress types
were calculated for each subsection and plotted against the corresponding RQI values. The curves relating DI and dynamic-load–
related DP to RQI values are shown in Figures 6, 7, and 8. The data
from all pavement sections (for each pavement type) were included
in these plots. This was necessary because it was not possible to get
DP–RQI plots for individual projects separately, given that distress data were available for only 3 years (1993, 1995, and 1997). The
logistic model having the following form was used for the regression
analysis (8):
DI (or DP ) = a
exp(b + cRQI )
1 + exp(b + cRQI )
(3)
where a, b, and c are regression constants.
The logistic function was chosen because it fits the data best and
allows for calculating critical points of distress accumulation as
described below. The results shown in Figures 6 through 8 show that
the curves relating load-related DP to RQI have higher R2 values
than those relating the all-inclusive DI versus RQI curves. When
dynamic-load–related distress types were considered, R2 increased
from 0.488 to 0.739 for rigid pavements, and from 0.522 to 0.624
for composite pavements. For flexible pavements, there is no good
trend and the scatter in the data is very large, with an R2 value of
0.311 for DI versus RQI. This probably reflects the higher variability in flexible pavements, indicating that weak spots in the pavement
will tend to “attract” load-related damage as opposed to rougher spots
inducing higher dynamic axle loads. On considering only load-related
distress types, the R2 value increased from 0.311 to 0.375. The results
of the regression analysis are summarized in Table 3.
Relationships between RQI and dynamic-load–related DP show
that the increased rate in distress is not constant, with DP sharply
increasing at a critical RQI level. This RQI value can therefore be
taken as a roughness threshold value. It corresponds to where the
acceleration in pavement distress is maximal. Mathematically, it is
where the second derivative of RQI–DP function is maximal. Acceleration in the accumulation of DP versus RQI values for each pavement type is shown in Figures 6, 7, and 8. The RQI threshold value
for rigid pavements was determined to be 64. This corresponds to
an IRI of 1.99 m / km (124 in./mi). For composite pavements, the
RQI threshold value was equal to 51. This corresponds to an IRI of
1.43 m/km (89 in./mi). For flexible pavements, it was deemed difficult to determine a meaningful threshold value from this plot because
of the large scatter in the data. These threshold values represent
the overall behavior of pavements at the network level and may not
be applicable for a particular pavement project. Therefore, such
thresholds can be used for network-level pavement management,
and not necessarily at the project level. It is also interesting to note
that the roughness threshold for rigid pavements corresponds to a DI
of 13, as opposed to a DI of 27 for composite pavements. This may
imply that the optimal time window for PM actions corresponds to
a lower distress level (higher RSL) for rigid pavements than for
composite pavements.
CONCLUSIONS
In this discussion, distress and roughness data from the Michigan
DOT PMS database were used to develop roughness thresholds, for
all three pavement types, aimed at minimizing pavement damage
Lee et al.
Paper No. 02- 4039
1
0. 8
0. 6
Low-RQ I
High-RQI
0. 4
0. 2
0
TC-N
TC-A
TJ
LC
LJ
DL
P atch
(a)
1
0. 8
0. 6
Low-RQ I
High-RQI
0. 4
0. 2
0
TC-N
TC-A
TJ
LC
LJ
DL
P atch
(b)
1
0. 8
0. 6
Low-RQ I
High-RQI
0. 4
0. 2
0
TC-N
TC-A
TJ
LC
LJ
DL
P atch
(c)
1
0. 8
0. 6
Low-RQ I
High-RQI
0. 4
0. 2
0
TC-N
TC-A
TJ
LC
LJ
DL
P atch
(d)
FIGURE 3 Proportions of distress types in DI for rigid pavements: (a) 0 < DI < 10;
(b) 10 < DI < 20; (c) 20 < DI < 30; and (d) 30 < DI < 40. (TC = transverse crack;
TJ = transverse joint deterioration; DL = delamination; −N = with no associated distress;
−A = with associated distress.)
49
50
Paper No. 02- 4039
Transportation Research Record 1816
0.8
0.6
Low-RQI
0.4
High-RQI
0.2
0
TT
TC-N
TC-A
LC-c -N
LC-c -A
LC-e-N
LC-e-A
LC-w-N
LC-w-A
(a)
0.8
0.6
Low-RQI
0.4
High-RQI
0.2
0
TT
TC-N
TC-A
LC-c -N
LC-c -A
LC-e-N
LC-e-A
LC-w-N
LC-w-A
(b)
0.8
0.6
Low-RQI
0.4
High-RQI
0.2
0
TT
TC-N
TC-A
LC-c -N
LC-c -A
LC-e-N
LC-e-A
LC-w-N
LC-w-A
(c)
0.8
0.6
Low-RQI
0.4
High-RQI
0.2
0
TT
TC-N
TC-A
LC-c -N
LC-c -A
LC-e-N
LC-e-A
LC-w-N
LC-w-A
(d)
FIGURE 4 Proportions of distress types in DI for composite pavements: (a) 10 < DI < 20;
(b) 20 < DI < 30; (c) 30 < DI < 40; and (d) 40 < DI < 50. (LC = longitudinal crack; LJ = longitudinal
joint; TT = transverse tear; −c = at center lane; −e = at edge; and −w = at wheelpath.)
Lee et al.
Paper No. 02- 4039
0.8
0.6
Low-RQI
0.4
High-RQI
0.2
0
TT
TC-N
TC-A
LC-c -N
LC-c -A
LC-e-N
LC-e-A
LC-w-N
LC-w-A
(a)
0.8
0.6
Low-RQI
0.4
High-RQI
0.2
0
TT
TC-N
TC-A
LC-c -N
LC-c -A
LC-e-N
LC-e-A
LC-w-N
LC-w-A
(b)
0.8
0.6
Low-RQI
0.4
High-RQI
0.2
0
TT
TC-N
TC-A
LC-c -N
LC-c -A
LC-e-N
LC-e-A
LC-w-N
LC-w-A
(c)
0.8
0.6
Low-RQI
0.4
High-RQI
0.2
0
TT
TC-N
TC-A
LC-c -N
LC-c -A
LC-e-N
LC-e-A
LC-w-N
LC-w-A
(d)
0.8
0.6
Low-RQI
0.4
High-RQI
0.2
0
TT
TC-N
TC-A
LC-c -N
LC-c -A
LC-e-N
LC-e-A
LC-w-N
LC-w-A
(e)
FIGURE 5 Proportions of distress types in DI for flexible pavements: (a) 10 < DI < 20;
(b) 20 < DI < 30; (c) 30 < DI < 40; (d) 40 < DI < 50; and (e) 50 < DI < 60.
51
Paper No. 02- 4039
Transportation Research Record 1816
50
40
DI
30
20
10
0
0
20
40
60
80
100
60
80
100
RQI
(a)
50
Load Related DP
52
40
30
20
10
0
0
20
40
RQI
(b)
0. 0 4
0. 0 2
0
-0 . 0 2
-0 . 0 4
0
20
40
60
80
100
RQI
(c)
FIGURE 6 Relationship between distress and RQI for rigid pavements: (a) DI versus RQI (R 2 =
0.488); (b) load-related DP versus RQI (R 2 = 0.739); and (c) acceleration in DP accumulation (RQI
threshold = 64).
120
Lee et al.
Paper No. 02- 4039
100
80
DI
60
40
20
0
0
20
40
60
80
100
60
80
100
60
80
100
RQI
(a)
100
Load Related DP
80
60
40
20
0
0
20
40
RQI
(b)
0.06
0.04
0.02
0
-0.02
-0.04
-0.06
0
20
40
RQI
(c)
FIGURE 7 Relationship between distress and RQI for composite pavements: (a) DI versus RQI (R 2 =
0.524); (b) load-related DP versus RQI (R 2 = 0.624); and (c) acceleration in DP accumulation (RQI
threshold = 51).
53
54
Paper No. 02- 4039
Transportation Research Record 1816
200
DI
150
100
50
0
0
20
40
60
80
100
60
80
100
RQ I
(a)
Load Related DP
200
150
100
50
0
0
20
40
RQ I
(b)
FIGURE 8 Relationship between distress and RQI for flexible pavements: (a) DI versus
RQI (R 2 = 0.311) and (b) load-related DP versus RQI (R 2 = 0.369).
due to dynamic loading. These thresholds can be used as trigger values for smoothing pavements in time for delaying damage from
dynamic loading. Such action can play an important role in any state
highway administration’s pavement PM program. Some specific
conclusions are listed below:
• To get a good relationship between distress caused by dynamic
loading and surface roughness, it was necessary to extract dynamicload–related distress types.
• Dynamic-load–related DP showed much better relationships
with RQI than with DI; R2 values increased from 0.488 to 0.739 for
rigid pavements and from 0.522 to 0.624 for composite pavements.
• For flexible pavements, the scatter in the data was very large,
with an R2 value of 0.311 for DI versus RQI and 0.375 for DP ver-
TABLE 3
sus RQI. This probably reflects the higher variability in flexible
pavements.
• For all three types of pavements, transverse cracks with associated distress were shown to be related to dynamic loading.
• Based on relationships of dynamic-load–related distress (DP)
and roughness (RQI), the following roughness (RQI) threshold values
were determined: RQI = 64 for rigid pavements and RQI = 51 for
composite pavements. These threshold values represent the overall
behavior of pavements at the network level and may not be applicable for a particular pavement project. Therefore, they are useful for
network-level pavement management decisions.
• The above relationships also indicate that the optimal time
window for PM action corresponds to a lower distress level (higher
RSL) for rigid pavements than for composite pavements.
Regression Parameters and R 2 Values of Empirical Curves
Rigid
Pavements
Flexible
Pavements
Composite
Pavements
Distress Types Used
All Distress
Load-Related Distress
All Distress
Load-Related Distress
All Distress
Load-Related Distress
a
35
32
100
60
90
60
b
-6.3252
-7.8459
-4.1465
-10.613
-4.4684
-5.8131
c
0.0897
0.1043
0.0643
0.1732
0.0706
0.0908
R2
0.488
0.739
0.311
0.375
0.522
0.624
Lee et al.
Paper No. 02- 4039
ACKNOWLEDGMENTS
The authors would like to thank the Michigan DOT and the Pavement
Research Center of Excellence for their financial support. The comments made by members of the Technical Advisory Group are highly
appreciated; special thanks are due to Dave Smiley, Larry Galehouse,
and Wen-Huo Kuo for their input and support.
5.
6.
REFERENCES
7.
1. Gillespie, T. D., S. M. Karamihas, M. W. Sayers, M. A. Nasim, W. Hansen,
N. Ehsan, and D. Cebon. NCHRP Report 353: Effects of Heavy-Vehicle
Characteristics on Pavement Response and Performance. TRB, National
Research Council, Washington, D.C., 1993.
2. Michigan Department of Transportation. Evaluating Pavement Surfaces: LISA and RQI. Material and Technology Research Record, No. 79,
June 1996.
3. Darlington, J. The Michigan Ride Quality Index. Michigan Department
of Transportation, Ann Arbor, Dec. 1995.
4. Chatti, K., and D. Lee. Investigation of the Relationship Between Surface
Roughness, Truck Dynamic Loading, and Pavement Distress Using Field
8.
55
Data from In-Service Pavements. Proc., 6th International Symposium on
Heavy Vehicle Weights and Dimensions, Saskatoon, Saskatchewan,
Canada, June 18–22, 2000.
Chatti, K., D. Lee, and G. Y. Baladi. Development of Roughness Thresholds for the Preventive Maintenance of Pavements Based on Dynamic
Loading Considerations and Damage Analysis. Final report submitted to the Michigan Department of Transportation, Ann Arbor, June
2001.
Chatti, K., and D. Lee. Development of Preventive Maintenance Strategy. Presented at 80th Annual Meeting of the Transportation Research
Board, Washington, D.C., 2001.
Cebon, D. Handbook of Vehicle-Road Interaction. Swets & Zeitlinger,
Lisse, Netherlands, 1999.
Neter, J., and W. Wasserman. Applied Linear Statistical Models. Richard
D. Irwin, Inc., Homewood, Ill., 1974.
The contents of this paper reflect the views and opinions of the authors, who are
responsible for the accuracy of the information presented. The contents do not
necessarily reflect the views of Michigan DOT and do not constitute a department
standard, specification, or regulation.
Publication of this paper sponsored by Committee on Pavement Management
Systems.