Study of Long-Term Pavement Performance (LTPP): Pavement Deflections

Chapter 1. Introduction

BACKGROUND

The Long-Term Pavement Performance (LTPP) program, which began as part of the Strategic Highway Research Program (SHRP) and is now administered by the Federal Highway Administration (FHWA), has been gathering falling weight deflectometer (FWD) load-deflection data since late 1988. The FWD database is large—by the fall of 1998, there were already more than four million records (or lines) of load-deflection data, representing FWD tests conducted throughout the United States, Canada, and Puerto Rico. In addition, a considerable volume of ancillary information—such as sensor calibrations, pavement temperatures, sensor positions, and FWD operator observations and comments—exists as well.

The FWD database was used for comparison to screen the pre-autumn 1998 FWD level E data for errors or anomalies. The term level E refers to those data elements that have undergone a screening process already, and have been uploaded to the LTPP database for public dissemination and use. One common source of these data, and other LTPP data elements as well, is data from the various versions of DataPave.

STUDY OBJECTIVES

FWD load-deflection data are generally used to characterize the tested pavement by an analysis of the applied load and the magnitudes (or shape) of the measured deflection basin. Often, these data are used to backcalculate layered elastic stiffnesses or moduli. The results give the pavement researcher a measure of the pavement’s bearing capacity, which can in turn be linked to future pavement performance.

The primary objective of this study was to identify data errors or anomalies in the LTPP loaddeflection database that were not identified during routine data screening required to reach level E. Routine screening applies more general procedures, such as broad range checks, to the data. The intent of this study was to review the level E deflection data and ancillary information, looking for data discrepancies and errors that routine screening may not have identified. The overall objective of the postscreening, final data check was to assure that good quality load-deflection and ancillary data are available for researchers and highway engineers.

The majority of the errors and anomalies found during this quality assurance (QA) screening of the level E FWD database either have been, or are in the process of being, corrected. The resulting database will, in turn, be much more useful for pavement analysis and design engineers who wish to understand and properly evaluate new or rehabilitated highway pavements, based on the FWD load-deflection data in the level E database, generally using an up-to-date version of DataPave. This report documents the screening methodologies employed and the findings of this study, along with the extent of the various categories of errors and anomalies identified and reported. Specific examples of these FWD-associated data categories are also presented in this report.

SCOPE AND ORGANIZATION OF REPORT

It was initially expected that the primary thrust of this study would be a straightforward screening for anomalies in the level E load-deflection data, possibly accompanied by parallel anomalies in the peak load readings. However, it was found that the majority of the questionable FWD2 associated data in the database in fact did not involve the deflection or load readings directly, but rather a variety of other manual data entry errors or oversights. The number and magnitude of direct equipment errors found were in fact surprisingly small, with the vast majority of the data (well over 99.5 percent) appearing to be of very good quality—highly accurate and very repeatable.

The following list is a breakdown of the various categories of errors and anomalies found, together with the approximate percentage of data affected by the questionable load-deflection data in the pre-autumn 1998 database:

• Inconsistent deflection basin anomalies: ~0.1 percent.
• Systematic load-deflection anomalies: <0.1 percent.
• Load-deflection calibration anomalies: None.
• Long-term sensor positioning errors: ~7 percent.
• Single section sensor positioning errors: 0.4 percent.
• Lane designation errors: <0.1 percent.
• Date- or time-stamp errors: 0.1 percent.
• Drop height designation errors: 0.3 percent.
• Site errors (tested at wrong test section): 0.1 percent.
• Stationing errors: 0.1 percent.

As can be seen in the list of errors and anomalies, only the first three categories are directly related to the FWD load and deflection values present in the level E database. The remaining categories have little or nothing to do with the quality of the deflection data gathered; these anomalies are generally due to inadvertent data entry errors, where manual keyboard input to the field data collection program(s) is required. Moreover, the approximate percentage of the pre-autumn 1998 data directly affected by anomalous load-deflection readings is probably less than 0.2 percent (by any reasonable measure, a very small percentage of the FWD data), while the corresponding percentage affected by other types of data errors may be greater than 8 percent. In fact, of all the error types identified, one category of error dominates all other errors combined: Incorrectly placed sensors along the FWD’s raise-lower bar over extended periods of time. Still, the quality of the FWD data in the database has to be regarded as excellent overall. As previously noted, most of the FWD data anomalies identified by this study can be (or already have been) either corrected or flagged.

Two other categories of anomalies or potential errors were also identified in the pre-autumn 1998 FWD load-deflection data, as follows:

• Unbound layer anomalies (all data): Not screened (0.8 percent).
• Unchanged (noted only) data discrepancies: ~ 1 percent.

These two categories of data were not recommended for alteration or flagging in the database for several reasons. For the category “unbound layer anomalies,” there is considerable variation in the data for most FWD test points (even drop-to-drop at the same drop height). Thus it was not possible to find any automated and reasonable criteria to sort out the “good” from the “bad” data. With respect to the general category “unchanged data discrepancies,” it was deemed adequate to merely note the nature and extent of each discrepancy found; no defendable changes, deletions, or flags in the database could be justified based on available information (see also chapter 6). Since it is possible that the FWD test results (or at least many of these) are correct (or nearly so) in both categories, no changes or flags are recommended in the level E dataset for these categories of data anomalies. 3 The information presented in the following chapters is organized as follows. Chapter 2 describes the data obtained from the LTPP level E database, along with an overview of how these data have been organized. Load-deflection errors and anomalies are discussed in chapters 3 and 4. Categories of nondeflection associated manual data entry errors are covered in chapter 5. Chapter 6 deals with other data anomalies that have been noted but not recommended for changes or flags in the database, due to lack of definitive information. Chapter 7 presents suggested computed parameters and new FWD test procedures. A summary and conclusions are presented in chapter 8. Appendices A through M are found at the end of this report.