|FHWA > Asset Management > System Management & Monitoring > RWD > Texas|
Rolling Wheel Deflectometer (RWD) Demonstration and Comparison to Other Devices in Texas
TABLE OF CONTENTS
LIST OF TABLES
LIST OF FIGURES
Applied Research Associates, Inc. (ARA) is developing a high-speed Rolling Wheel Deflectometer (RWD) for structural evaluation of highway pavements. The research and development has been sponsored primarily by the Federal Highway Administration (FHWA) and the Small Business Innovation Research (SBIR) program. RWD testing is a nondestructive method for measuring the structural response of highway pavements. It is designed to measure continuous deflection profiles at normal highway speeds, thereby increasing both the safety and productivity of network-level pavement assessment. The RWD is intended to provide structural information to pavement managers for use in the administration of highway networks.
The current RWD design was assembled and made limited field tests in 2002. During the winter of 2002-2003, the project team made several hardware and software modifications. In July 2003, a comprehensive field test was made in College Station, Texas, sponsored by FHWA and the Texas Department of Transportation (TxDOT). The RWD was tested on six pavements representing a range of surface characteristics and deflection levels. A total of 38 individual test sections were designated for evaluation of the RWD’s effectiveness over a wide range of conditions. Falling Weight Deflectometer (FWD) and Rolling Dynamic Deflectometer (RDD) data were collected over the same sections for comparison to the RWD. In addition, two test sections were instrumented with Multi Depth Deflectometers (MDDs) to provide a reference deflection for comparison to the RWD.
Objectives of Study
The specific objectives of this study were the following:
ARA would like to acknowledge the many participants whose tremendous efforts contributed to this study. We would like to recognize TxDOT for sponsoring the project and for providing technical support including FWD testing, site characterization data, and traffic control services. In particular, ARA wishes to thank the TxDOT Pavements Section, headed by Dr. Ken Fults, and the leadership of his staff, including Dr. Dar-Hao Chen and Dr. Michael Murphy.
Also, we would like to recognize the efforts of the Texas Transportation Institute (TTI) for providing RWD storage facilities, MDD instrumentation, and MDD testing services. In particular, ARA wishes to recognize Mr. John Ragsdale for his hard work and efforts, and Dr. Tom Scullion for his contributions to site selection and ground penetrating radar (GPR) testing. Furthermore, ARA acknowledges the University of Texas (UT) at Austin, including Dr. Ken Stokoe and Mr. Jeffrey Lee for providing the RDD testing and analysis on this project.
Furthermore, ARA is grateful to FHWA for their continued support of the development and deployment of the RWD. Specifically, we wish to thank Mr. Frank Botelho, RWD Project Manager, for his long-term efforts into the development of a valuable tool for asset management. Mr. Botelho’s persistence has truly been the primary factor driving the RWD development. In addition, Mr. Max Grogg, RWD Technical Representative, has provided many years of valuable input on the RWD design and testing. In addition, Mr. Grogg and FHWA team member Mr. Tom Van participated extensively throughout this study, and their contributions are much appreciated.
Finally, ARA wishes to recognize Mr. Bob Walker, Alpha to Omega, and Dr. Andy Peekna, Innovative Mechanics, Inc., for their invaluable contributions as consultants to ARA. Their expertise has been essential to the advancement of the RWD.
Figure 1. Participants of the field testing program.
Initially, ARA pursued the use of a scanning laser to measure pavement deflection under an 18-kip, single axle load. While the scanning laser device did not prove successful, the RWD trailer itself provided a sound platform for measuring pavement deflections at highway speeds. In 2002, ARA designed, manufactured, and field tested an alternative design utilizing four displacement lasers mounted on a rigid beam placed beneath the RWD trailer. Initial field tests were encouraging and identified several key hardware and software modifications to make the device more effective. During the winter of 2002-2003, ARA modified the software, positioning of the laser between the tires, and laser calibration procedure. The Texas demonstration was the first opportunity to test the improvements thoroughly and expand the matrix of pavements over which the RWD had been tested.
The RWD utilizes a “spatially coincident” methodology for measuring pavement deflection. Three lasers are used to measure the unloaded pavement surface (i.e., forward of and outside the deflection basin), and a fourth laser, located between the dual tires and just behind the rear axle, measures the deflected pavement surface. Deflection is calculated by comparing spatially coincident scans as the RWD moves forward. In other words, the profile of the undeflected pavement surface is subtracted from the profile of the deflected pavement surface measured at the same exact locations. This method was originally developed by the Transportation and Road Research Laboratory (TRRL)1 and furthered by Dr. Milton Harr at Purdue University2. It was later employed on the Dynatest/Quest prototype RWD.
It is important to remember that at 55 mph and a 2 kHz sampling rate a reading is being taken approximately every 0.5 in. The random error associated with the individual deflection readings can be very high, due to factors including equipment limitations and pavement factors. For example, due to pavement texture, one laser may read the top of an aggregate while the next laser to pass over that point may read the valley between two adjacent aggregates. This type of error is random, resulting in an approximately equal distribution of overestimated deflections, as underestimated deflections. If a sufficiently large number of readings are averaged, the random error is reduced to the point that the overall mean is not significantly affected by the random noise. For the Texas demonstration, ARA determined that a 100-ft interval was sufficient to reduce random error. This sample unit length is also suitable for network-level purposes.
Figure 2 shows an overview of the RWD trailer and beam. The 53-ft trailer was custom designed and built specifically for the RWD. Its length minimizes pitching of the reference beam, thereby minimizing the laser range needed to accommodate bouncing of the trailer during normal operation. In addition, its natural frequency of 1.45 to 1.8 Hz is low enough that it does not couple with the high frequency vibration of the 25.5-ft aluminum beam. The beam uses a curved extension to pass under and between the dual tires, placing the rearmost laser approximately 6 inches rear of the axle centerline and 7 inches above the roadway surface. The wheels have been spaced a safe distance from the laser and beam using custom lugs and a spacer. The beam and extensions are covered with foam rubber material for thermal insulation. Figure 3 shows the laser and spacer between the tires.
Figure 2. Overview of the RWD.
Figure 3. Laser placed between the dual tires.
The RWD uses four LMI/Selcom laser-triangulation displacement sensors to measure pavement deflection. The three forward lasers are mounted 12 in. above the road surface, while the lasers between the wheels has a standoff distance of 6 in. All lasers have a 3-in. measurement range.
The lasers come absolutely calibrated from the factory; however, ARA’s experience has shown that further refinement of the factory scale factors was necessary for our purposes. In addition, the spatially coincident methodology assumes that all lasers are mounted in a plane parallel to a flat surface. To mathematically remove small deviations in laser mounting height, ARA developed a calibration procedure that uses a water level to provide a flat surface beneath the lasers. With the trailer level, the laser readings to the flat surface are compared and corrected to make all lasers read the same height. In addition, as water is added and removed, the lasers measure the change in water level. The change in height measured by each laser is compared to all other lasers to relatively calibrate the laser scale factors.
For the Texas demonstration, the laser calibration procedure was performed at TTI’s storage facility prior to and following the field data collection program.
ARA participated in a demonstration planning meeting in Austin on March 19, 2003. The meeting took place at the office of the TxDOT CST-Materials and Pavements Section - Pavements and Materials Systems Branch and included representatives from TxDOT, the University of Texas, FHWA, and ARA. The purpose of the meeting was to define the objectives and scope of the demonstration, the roles of key parties, and to chose the location of the test sites. Based on this discussion, it was decided that the demonstration should include the following key elements:
Parties Involved and Roles
The following parties were identified as participants in the demonstration and assigned the following roles:
Selection of College Station Area
Prior to the meeting, TxDOT had performed preliminary screening, including some FWD testing, on groups of pavements located in three different geographic regions of Texas. After a briefing on the candidate sites, the group decided to make a site visit the following day to the College Station area. On March 20, 2003, representatives from TxDOT, FHWA, and ARA visited the TTI facilities at the Texas A&M University (TAMU) Riverside campus near College Station. We met with Mr. John Ragsdale (TTI) and Dr. Tom Scullion (TAMU) to discuss the project goals and to field review potential test pavements. Based on this visit, the group decided to establish approximately 20 individual test sections on SH 47 between SH 21 and College Station. In addition, several sections were selected on SH 21. Later, sections were added on nearby FM 50 and FM 2154, as well as a 16-mi stretch of Interstate Highway (IH) 45 located about 1.5 hrs north of College Station.
The group selected mid-July for the testing and planned to collect data over a 2-week period. In the meantime, the group proceeded to finalize test site locations for the general study, perform a detailed evaluation for the location of MDDs, conduct MDD installation, and make final modifications to the RWD in preparation for mobilization to Texas.
Table 1 presents a summary of the location and length, specific characteristics of interest, and typical pavement structure for each of the roads selected for testing.
Table 2 summarizes the exact location and feature of interest of the 38 individual 528-ft sections. Note that section R21 was eventually eliminated from study.
Note. Section R21 was eliminated from study.
DATA COLLECTION PROGRAM
RWD Setup and Calibration
The field data collection program began on July 11, 2003, with the arrival of the RWD to TTI’s facilities outside of College Station. Upon arrival, the RWD lasers and data acquisition systems were assembled, and the lasers were calibrated using the water calibration procedure. A preliminary field test of the system was conducted the next day, and on July 13 the aluminum beam was insulated with foam rubber sheeting and the laser calibration was performed one last time prior to testing.
Main Data Collection Program (RWD, FWD, RDD, and MDD)
RWD testing began on the SH 47 instrumented sections (R5 and R6) on July 14 and continued through July 16. During this time, the RWD and MDDs collected a total of 62 useable test runs. RWD passes were made primarily at 50 mph; however, several tests were made at 5, 10, 15, 30, 60, and 65 mph. MDD data were collected by TTI crews, and a lane closure was provided over the length of the instrumented sections. Following completion of the instrumented sections, the RWD proceeded with testing of the 6 roads and 38 test sections included in the general study. ARA tested these roads on July 17-21. Overall, as part of the general study, the RWD tested approximately 264 miles of roadway over a four day period.
As the RWD was testing the instrumented sections, the FWD and RDD proceeded with the 38 sections included in the general study. FWD testing was performed by TxDOT with a Dynatest Model 8001 FWD, while the University of Texas conducted RDD testing using their vibratory-type device. The majority of FWD and RDD data collection was completed between July 14 and July 17, with the last sections being tested on August 14.
During the course of the field program, a large number of people participated either as members of the data collection teams or visitors to the demonstration. TxDOT assembled a register of all participants in the demonstration, included for reference in appendix A.
Recalibration of RWD
To verify the validity of laser calibration factors obtained prior to data collection, ARA recalibrated the RWD near the end of data collection on July 19. The factors obtained were very similar to those determined prior to the start of data collection, and therefore, the decision was made to use the July 13 calibration factors for analysis. ARA staff members performed the calibration indoors at the TTI facilities and were accompanied by Mr. John Ragsdale (TTI) for both the pre- and post-data collection calibrations. Following completion of data collection on July 21, the RWD lasers were removed from the beam and stored for transport. The RWD departed College Station on July 22.
A table summarizing the locations and dates for RWD, MDD, FWD, and RDD testing is included in appendix B.
Site Characterization Data Collection (Rut, Ride, Texture)
Once the main field program was completed, TxDOT continued with site characterization data collection using the TxDOT Modular Vehicle (TMV). The TxDOT Modular Vehicle includes:
The data from the TMV is processed in real time and stored for later upload to the Texas Pavement Management Information System.
IRI was measured for all 38 test sections included in the general study. Rutting was measured for all asphalt concrete-surfaced pavements sections, which included everything except the five continuously reinforced concrete pavement (CRCP) sites on IH 45. Texture was measured on the 20 test sections on SH 47. Data collection was completed by September 2003, and all data were processed by TxDOT and forwarded to ARA for use and inclusion in this study.
The mean IRI, rutting, and texture results for each section are included in a table in appendix C.
ARA post-processed the raw RWD data using the spatially coincident method coded in a proprietary program. The software calculates a deflection for every set of laser readings, taken at the rate of 2,000 samples per second. At 55 mph, this is equivalent to approximately 1 deflection reading per 0.5 in, or 2,400 deflections per 100-ft sample unit. Based on ARA’s previous experience, we chose a 100-ft sample unit length for averaging of RWD deflections to sufficiently reduce the random error associated with the mean RWD deflection. In addition, a sample unit length of 100 ft is considered acceptable for pavement management purposes.
To facilitate processing of long data files, ARA analyzed the data assuming that the truck speed was constant over the test section. The software calculates the speed of the truck as it enters the test section and uses this speed to locate spatially coincident laser readings in the data file. Experience has shown that variations in truck speed over the section length will slightly increase the random error associated with individual deflection readings, but do not significantly influence the mean deflection calculated over the 100-ft sample unit length.
Finally, a variant of the above program allows for calculation of a single mean deflection value for points delineated in the raw data file by means of a the RWD photo cell, triggered by reflective cones placed at the beginning and ending of the 528-ft sample units. This routine allows for rapid calculation of mean deflection values for the 528-ft test sections and ignores the remainder of the data collected outside the zones of interest (i.e., outside of the cones).
The results discussed in the following sections were produced using a combination of the two programs described above.
The general study included continuous measurement of all roads at normal highway speeds (e.g., 50 to 55 mph), as well as determination of mean deflections for 528-ft test sections designated along the roadway through the use of reflective cones placed at the start and end of each test section. A total of 6 roads containing 38 test sections were tested. For each road, the RWD made five repeat runs, with the exception of SH 47, for which nine passes were made.
ARA analyzed the general study data in several ways. First, the RWD deflections were averaged over 100-ft intervals and plotted vs. station for the entire road to provide a deflection profile. An example of a raw deflection profile for the first pass made on FM 50 is presented in figure 4. A discontinuity in the profile can be seen at station 4,400 ft, due to excessive bouncing of the RWD trailer passing over a culvert at this location. Excessive bouncing leads to the lasers reading beyond their valid measurement range, which in turn produces erroneous deflections. To provide a more representative profile, the data were filtered by eliminating deflections that plotted either too high or too low with respect to the remainder of the deflection profile. Figure 4 shows an example of the filtered profile for FM 50. In addition, the mean deflections for test sections R28-R30 were calculated using the trigger program and superimposed on the deflection profile data.
Finally, all remaining runs for the road were filtered and plotted simultaneously to evaluate the RWD’s repeatability. Figure 5 shows the filtered deflection profiles for five runs on FM 50. With the exception of run no. 1, the remaining passes show good repeatability of RWD deflection profiles.
Figure 6 illustrates the detrimental effect of excessive roughness on SH 21 caused by swelling soils under the asphalt concrete (AC) pavement. Significant bouncing of the trailer due to the undulating pavement profile resulted in obvious spikes in the deflection profile. Figure 6 also shows the filtered profile once the erroneous data were removed, as well as the mean values for test sections R22-24. Figure 7 displays the filtered profiles for all five passes, again showing good repeatability.
The filtered deflection profiles and examples of unfiltered profiles for all six roads are included in appendix D. The mean values for each test section are used later in this report for comparison to FWD and RDD results.
Figure 4. Filtered RWD deflection profile for FM50 @ 55 mph.
Figure 5. Representative RWD Deflection Profiles for FM50 taken @ 55 mph.
Figure 6. Filtered RWD deflection profile for SH21 – bumpy @ 55 mph.
Figure 7. Representative RWD deflection profiles for SH21 - bumpy @ 55 mph.
The RWD performed 62 passes over a 3-day period on the instrumented sections R5 and R6 on SH 47. These sections were chosen for their distinct pavement structures (i.e., cement-treated vs. granular bases), and each was instrumented with two MDDs to provide reference deflection measurements for comparison to the RWD. The RWD was operated primarily at 50 mph; however, passes were made at speeds ranging from 5 to 65 mph. The RWD driver attempted to operate the truck in such as way that the passenger side, dual tires of the RWD straddled the MDD embedded in the pavement as closely as possible, while minimizing abrupt steering corrections. Both theory and prior experience indicate that abrupt steering movements are detrimental to good results. The MDD ground crews recorded actual transverse placement of the RWD tires and data collected at offsets greater than 4 in. left or right of dead center were discarded in this analysis. Ground crews also recorded the AC mix temperature throughout the day by means of a temperature probe placed mid-depth in the AC layer.
Figure 8 shows RWD deflection profiles collected for five consecutive passes over the instrumented sections. The first 528 ft correspond to R5 (cement-treated base), while the final 528-ft represent R6 (granular base). Overall, the profiles demonstrate good repeatability and the ability to distinguish between the stiffer and weaker pavement sections. However, there is also a slight drift in the profiles, where each consecutive pass produces slightly lower deflections than the previous run. When viewed over the course of the day, this trend is much more evident and a reason for concern. In general, AC deflections should be increasing throughout the day, due to softening of the AC layer with increase in temperature.
Figure 8. Five consecutive passes on the instrumented sections @ 50 mph.
The mean deflections measured for R5 and R6 were calculated for all 63 passes and plotted vs. time of day and temperature. Figures 9 and 10 present the mean deflections vs. time for R5 and R6, respectively. It can be seen that deflections are significantly decreasing with time, contrary to expected. This is also apparent in figures 11 and 12, which display the mean deflections vs. temperature for R5 and R6, respectively.
Discussion of Results
The RWD results reveal several findings regarding accuracy, repeatability, and the ability of the RWD to identify variations in pavement stiffness over the length of the road. The first finding is that data averaged over 100-ft intervals and plotted vs. station provide a very detailed profile for pavement management purposes. Significant variations in pavement stiffness, or a change in pavement stiffness that occurs over a significant length of roadway, are apparent. In addition, multiple runs over the same roadway show the RWD is very repeatable, once initial transient effects are overcome. Finally, the profiles on the SH 47 instrumented sections R5 and R6 display the RWD’s ability to distinguish between pavement sections with different stiffnesses, in this case, due to different base types.
Figure 9. Mean RWD deflection vs. time for R5.
Figure 10. Mean RWD deflection vs. time for R6.
Figure 11. Mean RWD deflection vs. temperature for R5.
Figure 12. Mean RWD deflection vs. temperature for R6
The data also show several practical limitations of the RWD. For example, testing identified several conditions that cause the RWD lasers to read beyond their valid measurement range, including excessive bumps due to swelling soils, differential rutting, and localized anomalies, such as culverts and bridge joints. Although these effects result in spikes in the deflection profile, their presence is not seen as a detriment, since they generally occur over a short distance. The erroneous data are easily filtered without disrupting the overall integrity of the deflection profile.
Of much greater concern are the initial shifts and in a few cases, long-term drifts seen in deflection profile for consecutive runs made on the same roadway throughout the day. In several cases, the data show noticeable changes in deflection between the first and second runs of the day, indicating a warming-up effect of the equipment. Several thermal sources may be introducing systematic changes in the RWD throughout the day, including warming up of the lasers themselves, ambient temperature changes, and heat from the RWD tires and brakes. These effects and the research approach recommended to minimize them are discussed in detail in a report prepared by Dr. Andres Peekna, Innovative Mechanics, Inc.3, and included as appendix E of this report.
The data collected on the instrumented test sections R5 and R6 were most significantly affected by thermal effects. In his paper, Peekna points out that testing was performed on the instrumented sections in a significantly different manner than the other roadways, which were more representative of production-type testing. Specifically, the hard braking of the RWD at the end of the instrumented sections resulted in increased heat transfer through the RWD wheels to the area around the laser placed between the dual tires (i.e., laser D). In addition, the short loop driven by the RWD to allow for rapid, repeat testing of the instrumented pavements did not allow sufficient time for the increased heat to dissipate between episodes of hard-braking. Peekna presents arguments that these conditions lead to a systematic lowering of calculated deflections with each consecutive run and further points out that these conditions would not normally be present during production-type testing. In other words, hard, frequent braking is not expected during routine testing, and the RWD wheels would be constantly cooled by the air stream.
Note that figures 9 and 10 show the slopes of the RWD data for all three days testing were very similar, and that similar trends were seen in both sections R5 and R6. This is exactly what is predicted from effects of absence of thermal equilibrium (i.e., if thermal equilibrium had been achieved, the slopes would have approached zero). In any given run, the sensors and their supports were subjected to essentially the same thermal environment on both sections, R5 and R6. Thus, changes in thermal environments between succeeding runs on each of these sections were the same. This results in the similar trends in RWD data for each day and for each section, as seen in the figures.
Device Description and Testing Plan
TxDOT performed nondestructive deflection testing with a Dynatest Model 8001 FWD. Figure 13 shows the FWD testing an MDD location on SH 47. The FWD delivers an impact to the pavement surface by dropping a weight on a set of buffers that transmit load to the pavement by means of a circular load plate. For this project, TxDOT configured the FWD with sensors positioned at 0, 12, 24, 36, 48,60, and 72 in. from the load center. At each test point, a series of drops at target loads of 6, 9, 12, and 16 kips were performed. Each 528-ft test section was tested at 50-ft intervals, giving a total of 11 tests per site. Testing was conducted in the outer wheel path, corresponding to the RWD laser location. The instrumented sections on SH 47 (R5 and R6) were each tested twice, once in the morning and once in the afternoon, to determine any deflection changes due to temperature variations.
Figure 13. The TxDOT FWD used for testing.
When comparing RWD results to the FWD it is important to keep in mind that each device produces a slightly different deflection basin, and therefore, it is not expected that the deflection magnitudes agree exactly. The FWD applies an impact load by dropping a weight on a stationary load plate, whereas the RWD is an actual moving wheel load, which overcomes the inertia of the at-rest pavement system through the deflection basin that is traveling ahead of the truck tires. The effective loading frequencies of the FWD and RWD vary as well, which leads to slightly different stiffness responses from the viscoelastic AC layer. Finally, while the FWD applies its impact load through a single load plate, the RWD distributes its 9-kip load over two tires spaced approximately 13 in. apart.
Nevertheless, to be effective pavement evaluation tools, both devices should be capable of identifying changes in deflection with variations in pavement stiffness, even though the relationships between the two are not expected to be identical. Also, given that the FWD is the current standard for pavement structural evaluation, comparison between the RWD and FWD is a logical process in the evaluation of the RWD’s effectiveness.
Figure 14 presents representative maximum deflection profiles collected on the six pavements included in the general study. Profiles collected on all 38 test sections are included in appendix F. The deflections have been normalized to a standard load of 9 kips; however, they have not been standardized to a common temperature.
Figure 14. Example FWD deflection profiles for all roads.
Figure 14 shows that four of the pavements exhibit low to medium deflections, ranging from approximately 3 to 18 mils. Of these four, the IH 45 CRCP pavement gave the lowest deflections (mean of about 5 mils), while the SH 47 AC section produced a mean deflection of about 13 mils. The SH 47 cement-treated base and SH 21 bumpy sections produced nearly identical deflections, averaging about 7 mils, while the SH 21 texture pavement had slightly higher deflections, averaging approximately 10 mils. The two surface treatment pavements produced the highest deflections, as is expected. Deflections for FM 50 and the rutted sections on FM 2154 averaged approximately 38 and 45 mils, respectively.
Figure 14 also illustrates FWD deflection variability within a given section. In general, as deflection magnitudes increase, the in-section variability also increases. For example, the IH 45 CRCP and FM 2154 rutted pavements produced the least and most variability, respectively.
The results of repeat passes made on the SH 47 instrumented sections (R5 and R6) are presented in figure 15. The FWD shows mean deflections of approximately 5 and 12 mils for the AC pavements with cement-treated and granular bases, respectively. Deflections in the afternoon increased by 0.41 and 0.75 mils (8.0 and 6.3 percent) for sections R5 and R6, respectively. For the range of temperatures experienced throughout the day, this is a very small change in deflection with respect to temperature. This is due to mainly to the thin AC layer making the total pavement deflection only slightly temperature
Figure 15. FWD deflection profiles for the instrumented sections in the morning and afternoon.