Reformulated Pavement Remaining Service Life Framework

Chapter 7. Collection of Units

INTRODUCTION

Collection of data on the condition state of pavements under an agency's jurisdiction should be based on the same construction triggers that form the basis for local decisions on corrective construction needs.

Field data collection on present pavement conditions should properly be used to determine the impact of the data element on future construction requirements based on the current agencies construction triggers. This effort is complicated with the need to adapt past and new pavement condition measurement practices.

The current challenge to SHAs is integrating, adapting, and adopting advancements in measurement of the physical features of pavement assets to legacy management systems. The development of datasets to establish performance curves of the long-term performance of a pavement requires a uniform set of data based on uniform measurements. Since common past pavement design practice was based on a 20-year life span, developing datasets with consistent data over this type of timeframe is difficult at best. This challenge will continue since emerging practice is to design pavements with even longer life spans (i.e., 50 years).

PAVEMENT ROUGHNESS

The use of high-speed longitudinal pavement profile equipment has become an accepted industry standard used to measure pavement roughness. The technique uses an inertial profiler that measures the change in longitudinal profile in the wheel paths at or near the speed limit. Roughness indices are computed from this profile and summarized at user-defined intervals. IRI is one of the most commonly computed pavement roughness measures. Other pavement condition feature measures and indices can be computed from this type of profile data such as the roughness index, half-car roughness index, ride quality index, ride number, profilograph index, rolling straight edge simulation, "bump/dip" detection, fault heights on jointed PCC pavements, slab curvature on jointed PCC pavements, heavy truck dynamic loading index, etc.

If pavement roughness measurements are performed using inertial longitudinal profiler devices, both network- and project-level data requirements can be satisfied using this common set of data.

PAVEMENT DISTRESS

Pavement distress ratings can be performed by human raters driving the pavement network using a manual process, a semi-automated process where field collected images are interpreted by human raters in an office, or fully automated systems. Manual pavement distress measurements that require raters to drive each route typically cost the most, put raters at risk (raters must drive and get down on the road to note any necessary information), and generally have the greatest variability in the ratings. In the semi-automated process, pavement distress ratings are performed using field video images, which reduces the risk to the raters and provides a historical archive of images for use in project-level investigations. Fully automated pavement distress ratings systems uses computer algorithms to interpret distresses obtained from field images and/or three-dimensional measurements. It offers the potential to reduce the cost of collecting data by eliminating the need for human interpretation. The newer three-dimensional imaging technology, which assigns a depth to image pixels, provides a more robust dataset that can be used for pavement cracking, rutting, and possibly pavement roughness measurements.

PAVEMENT STRUCTURAL RESPONSE

Deflection measurements are used to measure the response of a pavement structure to a known applied load. Interpretation of the resulting deflection data ranges from the identification of weak spots to advanced non-linear characterization of the engineering properties of pavement material layers.

Pavement deflection measurement techniques vary by the types of applied load, load magnitude, deflection measurement characteristics, and traffic control requirements. The types of applied loads range between nearly static creep, sinusoidal dynamic loads, impact loads, and wheel loads. Many modern deflection measurement devices allow the application of different applied load magnitudes that approach the legal limit of truck axle loads. Devices that measure the resulting deflection basin as a function of distance from the applied load allow the application of engineering algorithms to characterize the material properties of pavement layers. Pavement deflection measurement devices currently used in practice must stop at each measurement location, which requires traffic control.

FHWA has supported the development of a rolling wheel deflection measurement device that does not require traffic control. While this device can measure the maximum deflection response under the moving wheel load, it is not yet capable of measuring the shape of the deflection basin around the moving wheel load. While some maximum deflection pavement performance models exist based on the old Benkelman beam measurement technology, the shape of the deflection basin is currently required for pavement deflection analysis tools.

To ensure accuracy of data interpretation using pavement deflection response devices, it is important to know the pavement thickness and types of near-surface material layers. Manual measurement techniques of pavement thickness are being replaced by automated ground penetrating radar technology. Network-level pavement deflection measurements and estimation of pavement structural condition from those measurements can aid in identifying more cost effective future construction actions by incorporating structural information in addition to surface distresses in to the decisionmaking process. The cost associated with the network-level deflection measurements can be justified by the overall economic benefit obtained from optimum decisions that consider structural condition in addition to surface conditions.

In most situations, pavement deflection measurements are performed at the project level as a basis to develop design specifications for the most appropriate construction treatment.

TRAFFIC LOADS

Pavement structures are designed to endure heavy truck loads. Since pavement damage has been shown to follow a fourth power exponential relationship to the magnitude of wheel/axle load, some measure of the upper 20 percent of applied heavy axle loads can account for up to 80 percent of the predicted damage potential of those loads. The truck load damage exponent can also be as low as 2.5 depending on the distress type. It is also possible that one truck overload can destroy a pavement structure in one pass. This is why most highway agencies have imposed truck axle load restrictions and use full-time load measurement scales to enforce regulations.

In order to accurately predict future pavement performance, engineers need to know actual heavy loads being applied to a pavement. Access to permitted overloads granted by other divisions within an SHA is important in determining the need for future construction intervention events.

At the network level, research has shown that measuring the type of trucks on a route is more important than the weight of the trucks. This is based on the observation that truck weights are determined by the type of truck. Within a localized region, a typical heavy load truck profile can be applied to trucks in the same classification with an acceptable level of uncertainty.

CLIMATE

Climate data are at best a second-order consideration in the planning of future construction needs. These data measure potential climate-related "loads" such as freeze-thaw cycles and temperature-induced stresses that a pavement may need to resist based on material properties (as affected by climate) and constructed drainage features. Predicting future climate change and accounting for it is not currently incorporated in most pavement management models.

The best source of historical raw climate data in the United States is available from the National Climate Data Center (NCDC). In the past, NCDC data users needed to conduct quality data checks on the NCDC data to avoid using erroneous data; however, NCDC now provides data that have been subjected to quality checks.

MISSING DATA

All infrastructure management systems require a mechanism to handle missing measurement data. One approach is based on the truth-in-data concept in which all data are labeled as observed or imputed. Imputation of missing data based on defined and documented methods is acceptable, provided imputed quantities are appropriately identified.

The best practice is to associate a measure of variability with both measured and imputed condition state parameters. It is expected that the variability of imputed parameters will be greater than those based on measurements.

MEASUREMENT VARIABILITY

A requirement for all input measurement methods is to provide a measure of variability related to their use in prediction models. These measures of variability should be propagated through the models used to predict future pavement condition states to create a posterior probabilistic-based performance curve of future condition state. Variability measures should include repeatability of the measurement method, be partitioned by spatial variation in the pavement response, and reflect the sensitivity of the prediction model to resulting damage estimates.

SAMPLING INTERVALS AND FREQUENCY

Collecting pavement condition data on a partial sample of a network is used to reduce data collection costs. While sensor data such as pavement roughness and rutting are collected continuously, distress data are most often obtained on a sample basis. For manual condition surveys, field crews travel between selected portions of the roadway on which the distress surveys are performed. For semi-automated distress surveys where interpretations of distress are made from video images, only a portion of the video image can be interpreted on a sample basis. The condition ratings from these samples are then used to represent the condition of a larger pavement segment. Findings from recent research on the effect of pavement sampling on errors compared to a continuous measurement include the following:

• For a distress data sample size of 10 percent, errors have been found to have a typical maximum of 400 percent as compared to continuous data. An example of a 10 percent sample size would be rating 0.0621 lane-mi (0.1 lane-km) for every lane-mi (lane-km) of road.
  
  
• A 60 percent sample size reduces the error to approximately ± 25 percent.⁽⁹⁾
  
  
• The error due to sample size increases for pavements with higher variability.

Most agencies use monitoring frequencies of 1, 2, or 3 years between pavement condition measurements. For cracking, 1-year frequencies appear to be the best so that the first appearance of cracking can be detected in order to fit an appropriate model for cracking prediction. Longer intervals for crack monitoring may cause an overestimation of agency costs for pavement repair at the network level. Monitoring IRI using automated sensor readings could be on a 1- to 2-year interval, although longer intervals will result in an underestimation of repair costs at the network level.⁽⁹⁾