|Research Home | Pavements Home|
|This report is an archived publication and may contain dated technical, contact, and link information|
Publication Number: FHWA-RD-03-049
Date: November 2005
Improving Pavements With Long-Term Pavement Performance: Products for Today and Tomorrow
Paper 2. Transforming LTPP Distress Information for Use In Mtc-Pms
By Shameem A. Dewan1
The severities, types, and definitions of surface distresses used in the Strategic Highway Research Program (SHRP) database for Long-Term Pavement Performance (LTPP) sites are not the same as those used in the Metropolitan Transportation Commission Pavement Management System (MTC-PMS) system. Therefore, to use the LTPP distress data as inputs in the MTC-PMS software, the LTPP data must be transformed to match the MTC-PMS distress definitions. The objective of this paper is to describe a method to complete such transformations. Data conversion and use of converted data as inputs in the MTC-PMS were performed to develop a model for International Roughness Index (IRI) as a function of pavement condition information (the IRI model is intended for use in estimating user costs/benefits in the pavement management system). The condition information includes all MTC distress-severity combinations transformed from LTPP data, and corresponding deducts, percent load related deducts, percent nonload related deducts, and pavement condition index (PCI) values calculated using MTC-PMS software. The paper first presents the differences in definitions of distresses and severities in the two systems. It describes the selection of appropriate LTPP distress types to be transformed to generate required MTC distress data. Then the data transformation techniques for different distress types and severities from the LTPP system to the MTC system are explained. It was found that several types of manipulations were required to conduct the transformation of different distresses. These manipulations were performed based on the differences in definition for distresses and severities in the two systems. An IRI model was eventually developed using the transformed distress data and the output from MTC-PMS software.
1Graduate Research Student, Texas A&M University, College Station, TX 77843 Phone: 979-845-5982, Fax: 979-845-0278, firstname.lastname@example.org
The MTC-PMS of Oakland, CA, does not consider road user costs in producing decision support recommendations (MTC, 1999). Management system capabilities can be improved by incorporating models that estimate road user costs and/or user benefits attributable to different management strategies. Because the road user cost is a function of pavement roughness (Gillespie, 1981), it will be helpful to estimate pavement roughness to incorporate user cost models in the pavement management system. However, the MTC-PMS system uses only distress information to estimate and predict pavement conditions, so it was necessary to establish a correlation between IRI and pavement distresses. The objective was to estimate road user costs for the streets in the cities and counties of the San Francisco Bay area directly from MTC pavement distress information and incorporating additional models for user costs relating IRI. The distress information used all MTC distress-severity combinations, deduct values, and PCI values calculated from all distress-severity combinations using MTC-PMS.
The distresses types, along with their severities, used in MTC-PMS software, are defined in MTC’s distress identification manual (MTC, 1986). The cities and counties of the San Francisco Bay area do not generally have IRI data available for their city and local streets, but the distress and IRI data for State highways and freeways of California’s LTPP sites are available in the SHRP database. In the future, it would be more appropriate to use data for the model from the city streets instead of using LTPP site’s data. However, the current effort to develop an IRI model using the LTPP distress data can be considered as a starting point, and the intended model will need refinements in future using data from city streets.
This study requires the MTC-PMS system to use SHRP’s LTPP distresses as inputs to calculate PCI values and deducts prior to the use of these data in the statistical analysis to develop the intended IRI model. But the severities, types, and definitions of surface distresses used in the SHRP database for LTPP sites are not the same as those used in the MTC-PMS. The definitions of distress types and extents of severities in the two systems differ in several respects. For asphalt concrete surfaces, the LTPP database uses 15 types of distresses (SHRP, 1993), while MTC-PMS uses only 7 distress types to define road conditions. Because of the differences in the two systems for distress types, severities, and definitions, the LTPP distress information had to be transformed to MTC distress information before being used as inputs in the MTC-PMS software.
To achieve the intended IRI model using LTPP distress data, the following major activities were required:
DATA EXTRACTION FROM LTPP DATABASE
The extracted data for distress, transverse profile, and IRI values from the LTPP database were found in IMS Modules: Monitoring and Tables: MON_DIS_AC_REV, MON_T_PROF_PROFILE, and MON_PROFILE_MASTER, respectively (FHWA, 2001). The data were extracted to Microsoft® Excel® spreadsheets for further evaluation, conversion, and analysis.
The desired IRI model was a proof-of-concept model for the cities and counties of the San Francisco Bay area. Because of the lack of IRI data for Bay area city and county streets, SHRP data for California LTPP sites were used in this pilot study. The LTPP data for pavements with asphalt concrete on granular base (general pavement study GPS-1) and asphalt concrete on bound base (GPS-2) were extracted form LTPP DataPave 3.0 released in September 2001 (Federal Highway Administration (FHWA), 2001). Only 39 sets of data were available in the database September 2001 in the specific categories GPS-1 and GPS-2 in California and for which the profile dates of IRI data match survey dates of distress and transverse profile data. The profile dates of IRI were not exactly the same as the survey dates of distresses because the roughness measurements and the distress measurements were made on different dates. Reasonably close dates were considered in selecting these 39 data sets for conducting further analysis.
LTPP DISTRESSES VERSUS MTC DISTRESSES
MTC uses seven types of distresses with three severity levels that are slight modifications of the PAVER distress definitions (Shahin and Walters, 1990). The MTC distress types are (MTC, 1986):
The three severity levels of the distresses are “Low,” “Medium,” and “ High.” Considering the similarities between the MTC distress types and the LTPP distress types, the set of LTPP distresses used in this study includes:
The MTC definitions of distress types and severities are different, in several cases, from the LTPP definitions of distress types and severities. Table 1 briefly describes the differences in definitions for severities between the two systems for the MTC distress types (MTC, 1986; FHWA, 2001). One difference between the LTPP system and MTC system not included in table 1 is that LTPP longitudinal cracking in the wheel path of any severity level is considered as a part of the low severity alligator cracking in the MTC system.
Table 1. Differences in definitions between MTC and LTPP for MTC distresses-severities (MTC, 1986; SHRP, 1993)
1 mm = 0.039 inches
DATA TRANSFORMATION TECHNIQUES
It was necessary to transform several of the LTPP distress data types and severities to equivalent MTC distress types and severities. Table 2 shows which distresses in LTPP system were used to calculate which distresses in MTC system. First, the data in the LTPP database are stored in metric units (mm, etc.) were converted to English units (inch, etc.) to match the units system used in the MTC PMS software. Moreover, several manipulations were required to calculate MTC distress quantities in three severity levels from LTPP distress types and severities.
Table 2. Distress quantities in LTPP system used to obtain distress quantities in MTC system
LTPP TRANSVERSE PROFILE DATA TO MTC RUTTING DATA
Transverse profile data from the LTPP database were used to calculate MTC rutting quantities. MTC measures rut depths and quantities by laying a 3-meter (m) (10-feet (ft)) straightedge across the rut. Rut depths are the maximum depths found in the wheel paths and rutted widths are the parts of the widths in the wheel paths rutted more than 13 mm (0.5 inches). The rutted widths were divided into three severity levels according to the definitions given in table 1. A rutting quantity (area) in a specific severity level was calculated multiplying the rutted width portion in that severity level by the length of the test section between the transverse profile measurements.
In the LTPP database, transverse profile data are stored for 152.4 m (500 ft) long by 3.66 m (12 ft) wide; sections are collected at an interval of 15.2 m (50 ft). For the current study, all transverse profile data sets that were selected for the study were plotted. Rut depths and rutted widths in different severity levels for both wheel paths were recorded from the plots. An Excel-Visual Basic® macro was written and used to facilitate this. Figure 1 shows a typical transverse profile plotted with LTPP transverse profile data and the measurement of rutted widths and rut depths for calculating MTC rutting quantities and severities.
Figure 1. A schematic diagram for the measurement of rutted widths and rut depths from LTPP transverse profile data
LTPP BLOCK CRACKING TO MTC BLOCK CRACKING
According to the definitions of severities for block cracking, the ranges defining the boundaries of severity levels in the LTPP system are different from those in MTC system (see table 1). Figure 2 provides a graphic comparison of the definitions in two systems. This figure was used to develop ratios of quantities based on the width of the cracks to convert LTPP block cracking quantities in different severity levels to MTC block cracking quantities. The following equations were developed based on these relationships and used for the conversions.
The high severity distress quantity in LTPP system was distributed into 70 percent and 30 percent of the medium and high severity, respectively, in MTC system. This distribution was selected because of the high coverage of MTC medium severity (57 mm in figure 2) on the LTPP high severity region.
Figure 2. A comparison between SHRP and MTC definitions for block cracking severities, and conversion of LTPP quantities to MTC quantities
LTPP ALLIGATOR CRACKING TO MTC ALLIGATOR CRACKING
The LTPP system records the portions for wheel path and non-wheel path of the longitudinal cracking in three severity levels in two separate columns. Since LTPP longitudinal cracking in the wheel path at any severity level is considered as a part of the low severity alligator cracking in the MTC system, MTC low severity alligator cracking incorporates both LTPP low severity alligator cracking and LTPP longitudinal cracking in the wheel path of any severity. LTPP longitudinal cracking quantity (in length) in the wheel path was converted to MTC low severity alligator cracking quantity (in area) by multiplying the crack length by a unit width of 1 foot (0.305 m). The medium and high severity alligator cracking figures in the MTC system were obtained from the medium and high severity alligator cracking, respectively, in the LTPP system.
LTPP LONGITUDINAL AND TRANSVERSE CRACKING TO MTC LONGITUDINAL AND TRANSVERSE CRACKING
The differences in definitions of the three severity levels of longitudinal and transverse cracking in the two systems are the same as the differences for severity levels in block cracking. The conversion technique used to convert LTPP longitudinal and transverse cracking quantities in three severity levels to MTC longitudinal and transverse cracking quantities was similar to that used for converting (equation (1)) block cracking quantities, except that MTC longitudinal cracking includes LTPP longitudinal cracking only from the non-wheel path (because MTC assigns the portion in the wheel path to low severity alligator cracking, according to the definition). Moreover, the quantities in LTPP longitudinal cracking in non-wheel path and LTPP transverse cracking are added together after the conversions to MTC system to obtain MTC longitudinal and transverse cracking quantities in different severities.
LTPP PATCHING, SHOVING, AND RAVELING TO MTC PATCHING, DISTORTIONS, WEATHERING, AND RAVELING
During conversions of these distresses, it was assumed, based on their definitions, that there is no difference between LTPP shoving and MTC distortion. Similarly, LTPP patch/patch deterioration was considered the same as MTC patching and utility cut patch, and LTPP raveling was considered the same as MTC weathering and raveling (see table 2).
PCI AND DEDUCTS FROM MTC-PMS
When appropriate data conversions were complete, the converted data were used as inputs in the MTC-PMS software version 7.5 (MTC, 1999) to calculate PCI values, deducts associated with each of the distress type-severity combinations, percent-load related deducts, and percent-nonload related deducts. Each of the selected 39 data sets was considered as if it were from an individual inspection unit. A database was first developed for the inspection units using the MTC-PMS software’s “Road Inventory” and “Section Description” modules, and distress information was provided as input distresses using its “Inspection Units” module. The “Calculations” module was used to calculate “PCI from Inspection Units.” All necessary data for statistical analysis were extracted from the report “PCI Calculation-Deduct Values” produced by the software (MTC, 1999). The software calculates deduct values for all distress type-severity combinations based on the pavement surface type. It combines all deducts to a single value, corrects for multiple occurrences, and subtracts the corrected value from 100 to determine PCI value (MTC, 1999). This process is a modification of the PAVER PCI calculation process (Shahin and Walters, 1990). The percent-load related and nonload related deducts are calculated based on the severity levels of distress types related to the cause of deterioration. Table 3 shows the severity levels of distress types related to cause of deterioration (Smith, 1999). The report titled “PCI Calculation-Deduct Values” provides deducts for each of the distress-severity combinations, total deduct amount, percent load related deduct, percent-nonload related deduct, and PCI for each inspection unit.
Table 3. Severity levels of distress types related to cause of deterioration for asphalt and surface treatment pavements (Smith, 1999)
Partial results are shown in table 4 with the IRI values associated with each data set.
Table 4. Partial results from MTC-PMS and corresponding IRI values
STATISTICAL ANALYSIS AND ESTABLISHING IRI MODEL
A statistical analysis was conducted in an effort to establish a model of IRI as a function of PCI, all distress-severity combinations, deducts from each distress-severity combination, percent-load related deducts, and percent-nonload related deducts. A total of 45 predictor variables was considered in the statistical analysis, which came from 21 distress-severity combinations (7 distress types times 3 severity levels), 21 deducts from 21 distress type-severity combinations, 1 PCI, 1 percent load deduct, and 1 percent nonload deduct. These 45 predictor variables, data for 16 variables were zero for all 39 data sets. These 16 predictors thus were removed from further consideration in the analysis. The removed predictor variables were: medium and high severity rutting and their deducts; high severity block cracking and its deduct; low, medium, and high severity distortions and their deducts; high severity patching and its deduct; and high severity weathering and raveling and its deduct.
A multilinear regression analysis was conducted using the remaining 29 predictor variables. SAS statistical software was used for all statistical validation and modeling (SAS, 2000). The final model for IRI from this statistical analysis was:
where, IRI is in m/km
The model in equation (2) has a correlation coefficient (R2) of 0.53, coefficient of variation (CV) of 28 percent, and a root mean square error of 0.39. Figure 3 compares graphically the actual and the predicted values of IRI from correlation with PCI (equation (2)) and gives a graphical view of the dispersion of data that leads to the R2 value. Equation (2) can be used to estimate pavement roughness for the highways in California, and it needs only the quantities of distress-severity combinations as inputs. Because of the lack of roughness data for the local streets, this correlation should only be used as a starting point, and should be refined using data on city streets. The Bay area agencies need established correlations between user costs and pavement roughness, valid for Bay area cities and counties, to couple with equation (2) to estimate user costs.
A major problem encountered working with the SHRP distress data from LTTP sites was that some distress type-severities were all zeros because those distress-severities were not present on the road surface during distress surveys. It is now evident from the previous discussion that most of these absent distress-severities were associated with high severities and with distortions of all severities. All of these absent distress-severity combinations and distortions should have a significant influence on the values of IRI. The reason for the absence of these distress-severity combinations is that LTPP data are collected from the State highways, and these high distresses or distress-severity combinations are generally quickly repaired on State highways, but common on city streets. There was no way to incorporate these highly influential distress-severity combinations and associated deducts in the statistical analysis because of the use of LTPP data. City streets are generally found with more distress quantities and higher severities than are highways. An IRI model developed using data from city streets would have more applicability.
Some agencies in the cities and counties of the San Francisco Bay area need to collect, at least once, IRI data, along with the distress data from a representative sample of their city streets, and then establish a more appropriate IRI model for their streets by using the data transformations similar to the ones described above. The initial model developed should be refined at reasonable time intervals, for example once in every 5 years, using the most current data from the city streets to maintain the IRI model’s reliability.
1 m/km = 63.36 inches/mi
Figure 3. Actual versus predicted values of IRI
Using the LTPP distress data as inputs in the MTC-PMS system requires transformations to match the data with MTC-PMS definitions because of the differences between the distress definitions and severities used in the LTPP database and in the MTC-PMS system. A model for IRI as a function of pavement condition information was developed using the MTC data, and this IRI model is intended for use in calculating user costs/benefits in the management system. Several manipulations were required to conduct the transformation of different LTPP distresses and severities to obtain MTC distress quantities in three severity levels. Manipulations were performed based on the differences in definition for distresses and severities in the two systems.
The following can also be summarized from the study:
FHWA (2001). DataPave 3.0, CD-ROM. U.S. Department of Transportation, Washington DC, Federal Highway Administration.
Gillespie, T.D. (1981). Technical Considerations in the Worldwide standardization of Road Roughness Measurement-A Report to the World Bank. Highway Safety Research Institute, Ann Arbor, MI, Report UM-HRSI-81-28.
MTC (1986). Pavement Condition Index Distress Identification Manual for Asphalt and Surface Treatment Pavements, 2nd Edition. Metropolitan Transportation Commission, Oakland, CA.
MTC (1999). Pavement Management System, version 7.5. CD-ROM. Metropolitan Transportation Commission, Association of Oregon Counties and Marion Counties, Oregon.
SAS (2000). The SAS System for Windows, Release 8.01, SAS Institute, Inc., Cary, NC.
Shahin, M. Y., and Walters, J. A. (1990). Pavement Maintenance Management for Roads and Streets Using the PAVER System. Technical Report No. M90/05, U.S. Army Construction Engineering Research Laboratory, Champaign, IL.
SHRP (1993). Distress Identification Manual for the Long-Term Pavement Performance Project. Washington, DC, Strategic Highway Research Program, National Research Council, SHRP-P-338.
Smith, R.E. (1999). Conceptual Description of Automated PCI Calculation by MTC-PMS Software. Texas A&M University, College Station, TX, Texas Transportation Institute.
Topics: research, infrastructure, pavements and materials
Keywords: research, infrastructure, pavements and materials,LTPP, pavement performance, DataPave contest, DataPave
TRT Terms: Pavements--Performance--United States, Pavements--United States--Design and construction, Long-Term Pavement Performance Program (U.S.), Pavement performance, Pavement distress, Bituminous overlays, Climate