Supporting Materials for NCHRP Report 626 (2009)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report         PART III—DATA INTERPRETATION AND APPLICATION    Part III of the research report describes the physical characteristics and process for using each NDT technology and device on construction projects to define construction quality. Each system was evaluated in two parts: 1) the systems’ potential to be integrated into the flexible pavement construction process (level of process impact), and 2) the reliability and accuracy of the system (system accuracy and reliability). Chapter 6 focuses on the level of process impact on construction. In other words, what impact will the device have on the contractor’s progress, and will agencies need substantially more manpower to use the technology? Chapters 7 and 8 focus on the system accuracy and reliability of the different technologies and devices included in the field evaluation study. 235

Page intentionally left blank.

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report CHAPTER 6 APPLICABILITY OF NDT TECHNOLOGIES ON CONSTRUCTION PROJECTS Some NDT devices initially were operated by a representative of the manufacturer and then used by field technicians or engineers. Those devices that were found to have a reasonable success rate in identifying anomalies were used by the contractor and agency staff in their daily QA operations, in accordance with manufacturers’ guidelines. Clustered tests were performed using each NDT device to determine the repeatability and accuracy of each system in evaluating its effectiveness in defining construction quality. The time and personnel requirements to perform each test were recorded. This information was considered in rating the level of impact that each device may have on construction. Since the technology was of primary interest (not a particular system or manufacturer), reports on each system are presented under the heading of the technology utilized by the system. 6.1 Ultrasonic—PSPA and DSPA This system is applicable to HMA, unbound aggregate base, and embankment soils. The PSPA is used to test HMA, while the DSPA is used for unbound materials and soils. Both devices consist of a stand linearly connected by a stiff arm to a source and two receivers and by wire to a computer, as shown in Figure 16 in chapter 3. The source contains a hammer which is dropped several times at regular intervals. The receivers, containing quartz-crystal accelerometers, measure the acceleration of the Rayleigh waves induced by the dropping of the hammer and report the resulting electrical charge to the data acquisition system. A FFT transforms the electrical charge or data into the frequency domain. There is also a temperature sensor in the system. Roughness of the laptop is an important feature. The PSPA test can be and was performed on cold material one or multiple days after placement, as well as on surfaces at elevated temperatures, immediately after compaction. The system’s temperature gauge is used to incorporate the temperature into the calculation of the material’s modulus. The rubber pads beneath the receivers deteriorate more rapidly when used on surfaces at elevated temperatures. In fact, they have been known to melt when used on HMA surfaces shortly after placement. The operator needs to check these periodically to ensure adequate coupling between the receivers and surface. These pads are easily replaced. Both devices work properly as long as all points are in firm contact (coupled) with the surface being tested. Adequate coupling is the system’s primary limitation. The speed of data collection makes this technology a good candidate for QC applications, assuming that the temperature of the material is properly considered by the modulus calculation process. None of the PSPA and DSPA devices (including the laptops) used exhibited any problems. The main operational issue was inspecting and replacing the rubber pads of the receivers to ensure good contact with the surface being tested. The data interpretation program that comes with the PSPA and DSPA devices uses this information to provide the output in the form of the mean Young’s modulus to a particular 237

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report depth. The spacing of the receivers determines the depth of measurement. The operator needs to be trained to visually inspect the load pulse and response data on the output screen for judging the suitability of an individual test (see Figure 78). This training is considered more sophisticated than what is required for a nuclear density gauge. The operator also needs to ensure that the spring-loaded receivers are in contact with the surface between each test. If one of the receivers gets stuck, the result will be a data anomaly or “false” reading. False readings are easily identified by the operator viewing the shape of the load pulse and receiver response with proper training. The shapes of the load pulse and receiver response are visually displayed on the laptop screen for each reading. The PSPA is used to test HMA mixtures, while the DSPA is used to test crushed aggregate base layers, embankments, and prepared subgrades. The DSPA was used on the shoulders of the US-280 reconstruction project instead of on the main roadway because the roadway base layer had been chip-sealed. This type of surface reduces the repeatability of the ultrasonic device, as well as other NDT devices, because the points of the receivers and source are not always in good contact with the surface tested. Ensuring good contact with the surface being evaluated is important for both the PSPA and DSPA. The system initially converts the readings of the load pulse and response to a seismic modulus of the material. The seismic modulus is internally adjusted to a modulus at a specific condition (temperature and load frequency for HMA). Each test location requires three to five tests for this system. Each test took 10 to 20 seconds to complete. Therefore, the entire process (3 to 5 readings at a point) takes only slightly longer than the system currently used for QC, the nuclear density gauge, which is generally set for one 60-second reading. This system can also be used to estimate the elastic properties parallel and perpendicular to the direction of the rollers (refer to chapters 7 and 8). Measuring the seismic properties in different directions actually increases the perceived variability of the device. The variability can be reduced slightly by always taking the readings in one direction. All other NDT devices result in an average or equivalent value at a test point. The spacing of the receivers can also be changed easily for testing thin and thick layers. Layer thickness variation that occurs along a construction project can have less of an impact on the resulting seismic modulus values than on the resulting values from other NDT technologies. 238

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report Figure 78. DSPA and PSPA Being Used to Test Different Materials Another advantage of this technology is that the system can be calibrated easily to the specific materials being tested during the mixture design stage for HMA materials or in developing M-D relationships for unbound materials. This calibration procedure allows the PSPA and DSPA to be used to detect volumetric, as well as physical, changes in the materials during construction. The DSPA can be used to develop modulus growth with compaction relationships during the first day of construction for the unbound layers and periodically during the project. Use of the PSPA to develop HMA modulus growth relationships can be problematic because of the elevated temperature. It is more applicable to warm-mix projects. The equipment (including the laptop) was found to be durable, and it does not require more personnel than are now being used for control or acceptance of flexible pavement construction. In fact, the same technician using the nuclear density gauges or taking cores from the HMA layer could also operate the PSPA and DSPA at the same time. Its main disadvantage is training the operators for determining a “false” reading. 239

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report In summary, the ultrasonic technology can be used in day-to-day QA operations to assist contractor and agency personnel in judging construction and materials quality by itself or in tandem with other geophysical and/or ground truth sampling programs. 6.2 Steady-State Vibratory—GeoGauge This system is applicable to HMA and unbound materials and soils, and is similar to the roller-mounted-devices that are discussed at the end of this section of chapter 6. The GeoGauge, however, is only used for testing unbound materials and soils. The GeoGauge provides elastic modulus values that are displayed on the gauge or the readings can be stored in the device and downloaded to a computer at a later date. The resulting values were found to be similar to the resilient modulus values measured in the laboratory or calculated from the resilient modulus regression equations developed through the FHWA-LTPP program (Yau and Von Quintus, 2002). The elastic modulus values from the GeoGauge were found to be a function of the materials moisture content and density. Stiffness readings are also reported by the test equipment and are a function of the structure. The process followed by the GeoGauge operator is almost identical to that followed by an operator of the current state-of-the-art nuclear density gauge, except that the GeoGauge operator spreads a thin layer of sand on the pavement surface to set the instrument on before taking the reading (see Figure 79). The operator clears the surface to be tested with a small broom or other device to remove loose surface particles (see Figure 79). A thin layer of moist sand is used on rough surfaces to fill in surface voids to ensure that the ring under the gauge is in contact with at least 75 percent with the test surface. Moist sand should be used because the gauge vibrations will cause dry sand particles to relocate under the gauge and disturb the reading. The layer of moist sand should only be thick enough to fill the surface voids of the material being tested. A light pressure and rotation of the GeoGauge was also used to ensure good contact with the test surface. Each test takes 75 seconds, as compared to the nuclear density gauge’s 60 seconds. Thus, this test takes about twice as long as the nuclear density gauge, including the time for spreading the sand. The test procedure is still quick enough not to be a hindrance to the contractor’s progress and does not require more personnel now being used for control and acceptance. As for the DSPA, the same technician using the nuclear density gauge or running sand-cone tests could also operate the GeoGauge at the same time. The training and technical capability of the operator is no more than what would be required for operating a nuclear density gauge. The GeoGauge can be easily used to develop modulus growth with compaction effort relationships of unbound layers at the start of the project with periodically throughout the project, similar to the DSPA. This feature becomes advantageous when the water content is significantly varying from the optimum value measured in the laboratory. 240

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report Figure 79. Humboldt GeoGauge The GeoGauge should be calibrated to the project materials and conditions to improve on its accuracy, because of the potential influence from the supporting materials (refer to chapters 3 and 4). This calibration issue requires that laboratory repeated load resilient modulus tests be performed on each unbound layer for judging the quality of construction. Most agencies do not routinely perform resilient modulus tests for design or for forensic evaluations, even though the 1993 AASHTO Design Guide suggests that they be performed (AASHTO, 1993). Eliminating the laboratory resilient modulus tests from the calibration procedure will reduce its accuracy for confirming the design values, but not for identifying construction defects. As a replacement to repeated load resilient modulus test, the regression equations developed from repeated load resilient modulus tests included in the LTPP program (Von Quintus and Yau, 2001) or use of the DCP is permissible. The disadvantage of the GeoGauge is that it will result in high variability when testing non- cohesive, well-graded sands or similar soils. In addition, the elastic modulus readings from the gauge represent an equivalent modulus for the upper 10 to 12 inches of the layer. Thus, 241

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report the gauge in its current form should not be used to test thin (less than 4 inches) or thick (greater than 12 inches) layers without proper material calibration adjustments or changing the diameter of the ring under the gauge. In summary, the GeoGauge has potential use in day-to-day QA programs by both the contractor and agency personnel. 6.3 Deflection-Based Methods 6.3.1 Falling Weight Deflectometer The FWD is a large, expensive apparatus that is mounted on a trailer and pulled behind a tow vehicle from where the operator works a computer and locates the apparatus for testing (see Figure 6 in chapter 3). This system is capable of applying dynamic loads to the pavement surface, similar in magnitude and duration to that of a single heavy moving wheel load. It is being used within the LTPP program, and most state agencies have access to at least one FWD. Thus, it is already being used in most agencies day-to-day practice. The response of the pavement system is measured in terms of vertical deformation, or deflection, over a given area using seismometers or geophones. The use of a FWD enables the user to determine a deflection basin caused by a controlled load. These results make it possible to determine the stiffness of existing pavement structures for use in M-E based rehabilitation design methods. The falling weight strikes a set of rubber buffers mounted to a 300 mm circular foot plate, which transmits the force to the pavement (refer to Figure 6 in chapter 3). A thin-ribbed rubber pad is always mounted under the footplate. By varying the mass or the drop height or both, the impulse load can be varied. This load may be varied between 10 kN and 140 kN. Sensors measure the surface deflections caused by the impulse load. Most agencies use seven sensors at the spacing recommended by LTPP. However, fewer or more sensors can be used, and those can be spaced uniformly or at some other spacing selected by the user. Peak deflections are recorded, stored, and displayed. In some cases, one of the geophones or sensors can be incorrectly placed on the test surface by the sensor bar, especially on rough surfaces. The data acquisition software will identify this anomaly, notifying the operator that the test should be rejected and redone. The test takes about 2 minutes to complete, including the use of seating drops. Seating drops are important and should be used at each test point. This does not include time to configure the trailer and set up the data acquisition system, which should only have to be done once per day for each project. It takes about 30 minutes to configure the trailer and 2-3 minutes to set up the data acquisition program. Similar to the PSPA, the operator requires more technical and sophisticated training in setting up the equipment and visually interpreting the deflection basin data. A separate data interpretation system or software is required for producing elastic modulus values from the measured deflection basins—Young’s modulus for each layer. The 242

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report calculated elastic modulus values are structure dependent. Most data interpretation or analysis programs used backcalculation techniques for calculating layered elastic modulus values. Backcalculation programs do not determine unique modulus values for each layer and are sensitive to layer thickness variations. Forward-calculation procedures have been developed that result in unique layer modulus values for a particular deflection basin, but these values are thickness dependent. Any errors in the layer thickness will increase the error and variability of the processed data. Its use for acceptance of individual layers by the agency should be limited to the use of the forward-calculation procedure. Because the backcalculation procedures do not result in unique layer modulus values, it would be difficult to defend in contractor disputes where material has been rejected or payment penalties issued to the contractor. The device can be used to check or confirm the final flexible pavement for new construction or HMA overlays of existing pavements, but would probably create many disputes with contractor when the entire pavement structure is rejected at the end of the project. In addition, the resulting values for the upper layer are dependent on the stiffness and variability of the supporting layers. More importantly, calculating the elastic modulus of layers is generally restricted to those that are thicker than 3 inches. The FWD may also require one addition field technician and tow vehicle. The expense, size of the system, time needed to perform each test, and data interpretation software make this system less practical for QC and acceptance. Thus, the FWD is believed to be less practical and effective for the QA uses upon which this study is focused. 6.3.2 Lightweight Deflectometer The LWDs use the same theory as the FWD described above, but offer an advantage of being much more portable. In addition, the training and technical requirements for the LWD operators are no different than for nuclear density gauges, with one exception—the operator needs to understand and be aware of the factors and physical features that affect layer modulus calculated from the measured deflections. Results from the LWDs were significantly influenced by the supporting materials on some of the projects. As noted in chapter 5, all three LWD devices used on selected projects have similar features, so only the Dynatest and Carl Bro devices are discussed in the following paragraphs. Dynatest Prima 100 LWD Device The Prima 100 is manufactured by Dynatest and consists of the weight (hammer) on a pole and the sensors (geophones) in a plate on the ground, all encompassed in one, connected, portable structure (see Figure 80). The sensors were connected to a hand-held computer by wireless remote technology. 243

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report Figure 80. The Dynatest Prima 100 LWD The unit tested was somewhat flexible and the frame came apart on multiple occasions. Besides slowing down the process, this resulted in questionable data because the wireless remote would sense the jolt resulting from the frame coming apart as a separate test— resulted in a deflection and modulus value for that anomaly. The wireless remote was troublesome and kept losing contact with the apparatus. This happened anytime the technician carrying the apparatus became closer than a few feet from the technician holding the computer. This was an additional source of aggravation that also slowed down the operation because the computer had to be re-started each time it occurred. When using the system on particularly stiff base material, the hammer can bounce high enough, such that it can strike the apparatus again—resulting in an appreciable rebound load. The rebound load can cause the remote to mistake that rebound as a second or separate test. The software, as written, causes the actual test results to be deleted and replaced by a reading of the rebound. The system, however, is fast. One test takes about 10 seconds, so the five tests conducted (and averaged) at each location take approximately the same amount of time that a nuclear density reading takes at one location. Conversely, the apparatus is bulky to handle, so the time that most non-nuclear systems gain by not having to deal with the steps of transporting the nuclear device are lost. 244

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report Carl Bro LWD Device The Carl Bro system looks exactly like the Dynatest system, except that it has additional sensors that are not attached to the frame. These extended geophones do not change the theory and applications. Although, the algorithms are slightly different to include input from the additional sensors, the theory and application appear to be the same. The geophones are arranged linearly at set distances from the plate. Since the sensors are connected to each other by a bar, but separate from the loading plate, connecting and placing them at a specific distance from the plate for each test becomes problematical. It is expected, however, that these perceived disadvantages can be resolved in future modifications to the equipment, similar to the Prima 100 device. From beginning the process of setting the plate through the last of the five drops, takes on average about 5.5 minutes. The procedure followed for using the system is listed below. 1. Locate test point (surface must be even (flat) and must be cleared of anything that could cause part of the plate to lose contact with the surface). 2. Set the loading plate on the surface to be tested (plate must be flat on the surface). 3. Measure for geophone location. 4. Set the geophone arm and line up the sensors. 5. Set data acquisition key for collecting the deflection data. 6. Drop hammer (first drop “seats” the plate and is not read). 7. Repeat last two steps for five drops at each location (including the one to “seat” the plate). This system had a wired connection to a laptop computer and is more cumbersome to set up due to the additional geophones. More importantly, the seating drop of the plate sometimes moved the plate. This increased the variability in the data gathered from the geophones and increased the number of anomalies. The system is comparable in cost to the Prima 100. Summary This technology was tested on crushed aggregate base material, embankments, and prepared subgrades. However, there should be no difference between the procedures and the device’s reaction to a hard base material and those of HMA mixtures. A key advantage of this technology is that it gives the operator a reading of the elastic modulus in about the same time required to obtain a nuclear density gauge reading. The disadvantages are that the devices have limited reliability because of the range and reliability of the wireless remote and its software logic. In addition, the resulting values for the upper layer are dependent on the stiffness and variability of the supporting layer. It is expected that these disadvantages of the equipment can be easily resolved with future modifications. These devices will likely make the technology and device more expensive. It does, however, provide the agency with elastic modulus values that can be used to confirm design assumptions with proper calibration. In summary, the LWDs are believed to be less practical and effective for the uses upon which this study was focused, similar to the FWD. 245

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report 6.4 Dynamic Cone Penetrometer The DCP is used to estimate the strength and modulus of unbound materials and soils. The DCP is much like the LWD in appearance (see Figure 81) but uses a 15-lb (6.8-kg) steel mass falling 20 in (50.8 cm) that strikes the anvil to cause penetration of a 1.5-in (3.8-cm) diameter cone (45° vertex angle) that has been seated at the surface or in the bottom of a hand augured hole (see Figure 3 in chapter 3). The blows required to drive the embedded cone a depth of 1-3/4 in. have been correlated by others to N values derived from the Standard Penetration Test (SPT). Experience has shown that the DCP can be used effectively in augered holes to depths of 15 to 20 ft (4.6 to 6.1 m). The system has been used in the past for the testing of soils more than anything else. Figure 81. The DCP Before Assembly for Use in Measuring the In-Place Strength of Unbound Materials and Layers The technical skills and training requirements for the DCP operator are no different than for a nuclear density gauge. Advantages of the DCP include its simplicity, low maintenance (using disposable tips, making sure that the allen screws are kept tight, etc.), mobility, and low cost. It can also be used to test thick embankment layers, unlike some of the other NDT technologies and devices. Conversely, the manual apparatus is slow (tests took 5 to 10 minutes at each location), its use is physically demanding, and the test is actually destructive to bases and pavements in that a test results in a hole in the material. Use of the device can also be dangerous, if the operator’s hand gets caught between the hammer and base for the hammer. Furthermore, soils or materials with boulders or large aggregate particles (refer to Figure 38 in chapter 5) can cause refusal of the device. When this occurs, the test point should be moved and the test redone. An automated trailer mounted DCP is available, but is more expensive (see Figure 4 in chapter 3). Only the manual DCP was used within the field evaluation of NCHRP Project 10-65. 246

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report The manual DCP is considered to have potential for QC use on a day-to-day basis, but an additional contractor and agency staff person would probably need to be assigned to using the DCP under normal practices. However, the training and maintenance of this device is considered minimal. 6.5 Ground Penetrating Radar GPR is a pulse echo method for measuring pavement layer thickness' and properties. GPR uses radio waves to penetrate the pavement by transmitting the wave energy into the pavement from a moving antenna. These waves travel through the pavement structure and echoes are created at boundaries of dissimilar materials. An air-coupled horn antenna attached to the back of a small SUV (see Figure 9 in chapter 3) was used in the field evaluation of NCHRP Project 1065 to evaluate HMA, unbound aggregate base, and embankment soils. The speed of data collections is one of the biggest advantages of GPR technology. There should be no impact to the contractor’s operation, because this system collects the same information regardless of material temperature and is capable of taking measurements at speeds of up to 40 miles per hour. Higher speeds have been used on more recent projects through enhancements made to the equipment and data acquisition systems. The disadvantages of the technology are the interpretation of the dielectric values that are measured and personnel requirements for calibrating and maintaining the equipment and data interpretation software. The system is simple to operate and provides results immediately, at least in terms of dielectric values. The results are in the form of a “picture” of the pavement system, much like an X-ray. Although the transducer is located above the surface, aimed downward, the picture can be viewed from “plan” or “elevation” (“profile”) perspective. Another huge advantage of this technology is that a continuous profile of the dielectric values is available. In fact, layer thickness profiles or complete contours of the layer can be developed in a short time period. Currently, the technology requires operators with special technical skills to interpret the data that have physical meaning to the quality of construction. Software programs are available that provide color coded charts and contours of the material. This system has been used to determine layer thickness at a reasonable accuracy—when layers with different dielectric values are tested. The accuracy of the analysis programs require cores to accurately measure the in place thickness and other volumetric properties. Most of the data reduction-presentation programs, however, still require some volumetric properties to be assumed in estimating density, air voids, and other volumetric properties. These assumptions result in error of the properties that are calculated from the dielectric values. The assumptions are believed to be a reason why the GPR’s analysis and interpretation from the Part A projects did not coincide with some of the other NDT devices (refer to chapter 5 in Part II). There are programs available that do not require many assumptions, but all of the known programs are proprietary. These proprietary programs were 247

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report no used in the Part A field evaluations, but were included in the Part B summary at a few facilities. Data from some one of these proprietary programs is presented and discussed in chapter 7. Calibration is another issue that is important to the success of GPR antennas in estimating volumetric properties of materials. Cores have to be recovered and the physical properties of those cores determined and correlated to the dielectric values measured by the GPR prior to and during construction. This requires that control strips be used at the beginning of a project and the correlations periodically confirmed during construction. Many agencies are eliminating or not requiring the contractor to use control strips, especially for small projects. Thus, this technology has limited use in QC applications, but has greater potential for use in acceptable programs—especially those for which thickness is included in the price adjustments or pay factors. 6.6 Electric Current/Electronic This family of systems includes those that rely on technology such as electrical sensing fields, impedance, electric current, and radio waves to determine the quality of HMA pavement, base, or embankment (see Figures 23 and 24 in chapter 3). The training and technical skills required to operate this technology are no different than required for nuclear density gauges. In addition, the calibration requirements to improve on the accuracy of testing specific materials with the non-nuclear gauges are similar in detail and extent for nuclear density gauges. 6.6.1 Electrical Density Gauge An electrical density gauge was used in the Part A field evaluation projects, because of the equipment’s perceived ease of use and application to a diverse set of unbound materials and soils. The specific gauge used was the one manufactured by EDG, which is confined for use on aggregate base layers, embankments and subgrades, or any unbound layer (see Figure 82). The system uses 6-inch darts that are driven into the soil within a 1.8 square foot area. This allows the system to measure a 1.0 cubic foot volume of material. The system uses a 3-MHz radio signal, producing a current of a certain voltage and phase, which allows measurements of the capacitance, resistance, and impedance. The connected data acquisition program uses algorithms and ratios of the measured parameters to determine the density and water content of an unbound layer (refer to Figure 82). 248

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report Figure 82. Electric Density Gauge This test takes several minutes to perform, but it appears to have huge potential for use in replacing the nuclear density gauges and other traditional QA tests, such as the sand-cone tests. This technology does not require more personnel than are now being used for QC/QA of unbound layers. The system and devices should be easier to maintain and the operators of the equipment can be easily trained in its use—similar to a nuclear density gauge. The most time-consuming but critical part of the system is developing a proper soil-model for density and moisture content measurements. To date, other more traditional tests (such as sand cones) are performed in specific locations that cover the range in density and water contents. A regression model is then developed based on correlations between the EDG values and the density and water contents measured from other tests. It is expected that this test will be improved with time, but at present, its use as a practical device for controlling construction of unbound layers is limited. 6.6.2 Pavement Quality Indicator—PQI The PQI (see Figure 83.a) uses a constant voltage, radio frequency, electrical impedance approach, in which a toroidal electrical sensing field is established in the material being tested. This allows the PQI to make quick, in-situ measurements of pavement density. The sensor consists of a set of flat plates that are interconnected to form the electrodes of a planar capacitor. Variations in density are determined through changes in the dielectric constant of the medium between the capacitor plates. 249

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report (a) PQI Non-Nuclear Density Gauge (b) PaveTracker Non-Nuclear Density Gauge Figure 83. Non-Nuclear, Non-Roller-Mounted Devices Used to Measure the Density of HMA Layers Using this technology, the PQI can be used like the nuclear density gauge, with the exception that it has the capability to adjust for moisture variations and mix type. The device also has an on-board, real-time system that takes the readings and keeps a record of them, allowing it to be integrated seamlessly into the paving process. 6.6.3 PaveTracker The PaveTracker (see Figure 83.b) is a lightweight non-nuclear device for measuring the uniformity of HMA mixtures. The measurements are practically instantaneous when the device is placed on an HMA surface. Areas of segregation, lower density levels along longitudinal joints or other non-uniformity areas can be detected by the PaveTracker Plus, which allows the operator to correct the problem before construction is complete. The advanced software, built-in reference plate and enlarged display screen are some of the features offered by the PaveTracker. The large display screen is an advantage, because the device is compact and close to the ground. Like the PQI, the PaveTracker can be used exactly like the nuclear density gauge, without the use of any nuclear device. The PaveTracker also has an on-board, real-time system that takes the density readings and keeps a record of them for future use, allowing the device to be easily integrated into the paving process. 250

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report 6.7 Intelligent Compactors/Rollers with Mounted Response Measuring Devices These systems offer real-time pavement quality measurement with no negative impact to the contractor’s progress. They use accelerometers to measure parameters of the compactor’s vibratory signature. Other sensors are also used to gain information about the pavement. Information from the sensors is then used to make decisions about pavement quality. Although these roller-mounted systems have been shown to be beneficial to a contractor from a control standpoint, they have not been used for acceptance and confirmation of the design-modulus values. Two of these systems were used in the demonstrations sponsored by FHWA at the NCAT and MnROAD facilities and included in the NCHRP Project 10-65 field evaluations. Thus, they are discussed in the following paragraphs. 6.7.1 Asphalt Manager and Vari-Control System This system, developed by Bomag, contains an onboard pavement analysis system based on the electrical charge generated by strategically-mounted quartz-crystal accelerometers that measure the acceleration of the vibratory drums of the compactor. An onboard computer transforms the data from the sensors using a FFT into the frequency domain. This transformation allows the computer to calculate the material’s modulus. There is also a temperature sensor in the system, which feeds data into the computer for use in modulus calculations. In addition, the system takes this reading and alters the compaction effort of the roller to avoid the damaging effects of over-compaction. Stiffness readings are taken continuously and presented as a modulus value developed by Bomag and called Evib, in the form of MN/m2. The Evib value should be related to the dynamic modulus of the material being compacted. However, this computed value is expected to be affected by the underlying support conditions. To-date, the Evib value has not been evaluated or checked against dynamic modulus values measured in the laboratory or estimated through other NDT devices. The system is fully integrated into a vibratory roller that is part of an operational paving train (see Figure 19 in chapter 3). The true test of this “intelligent compaction” system is whether it actually saves time (fewer passes), improves uniformity of the mat, and renders accurate, consistent readings. As for this part of the analysis (impact on the contractor’s progress), assuming that the system does what it claims, it can only help the contractor’s progress. 6.7.2 Ammann Compaction Expert Ammann-America, the U.S. branch of the Swiss manufacturer Ammann Compaction, Ltd., has introduced the Ammann Compaction Expert (ACE) to the U.S. market. The goal of the ACE is the same as for the Asphalt Manage. The major difference is that the ACE seems to take the paving environment into account more than does the Asphalt Manager does in an automated fashion. The computer in the ACE system is capable of receiving information such as lift thickness, number of passes, mix or soil type, etc., which is used in the 251

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report calculation of the material’s stiffness or modulus. Just as with the Asphalt Manager, the system is fully integrated into a vibratory roller that is part of an operational paving train. 6.7.3 Summary The true test of this “intelligent compaction” system is whether it actually saves time (fewer passes), improves uniformity of the mat, and results in accurate, consistent readings. As for this part of the analysis (impact on the contractor’s progress), aAssuming that the roller- mounted devices do what they claim, they can only help the contractor’s progress and in making better decisions in real-time regarding compaction of pavement layers. 6.8 Summary of Process Impact Table 64 provides a summary of the level of process impact on flexible pavement construction for different NDT technologies and devices regarding their use in QA programs. 252

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report 253 Table 64. Summary of Process Impact of Different NDT Technologies and Devices on QA Programs NDT Technologies Deflection-Based DCP Non-Nuclear Devices Impact Topics or Issues Ultrasonic Gauges Steady- State Vibratory Trailer Portable Manual Automated GPR Non-Roller- Mounted Roller- Mounted Easily used to develop density or modulus growth curves? HMA- No Unbound- Yes Yes No Yes No No No Yes Yes Resulting Value Seismic Modulus Elastic Modulus Deflection Deflection & Elastic Modulus Penetration Rate or Index Penetration Rate or Index Dielectric Values Density & Water Content Stiffness or Density Conversion required to adjust readings? Yes No No No No No Yes No No Requires calibration to specific materials or soils? Yes Yes Yes Yes Yes Yes Yes Yes No Can readily test thin layers (<3 inches) Yes No No No Yes Yes Yes Yes Yes Can readily test thick layers (>12 inches) Yes No Yes Yes Yes Yes Yes No No Readily applicable to control? Yes Yes No Yes Yes Yes No Yes Yes Readily applicable to acceptance? Yes Yes No; Only final structure Yes Yes Yes Yes Yes No Additional auxiliary equipment needed? No No Yes, tow vehicle No No Yes Yes, vehicle No No Additional man power needed? No No Yes, operator No No Yes Yes No No Equipment readily available on commercial basis? Yes Yes Yes Yes Yes Yes Yes Yes Yes Software readily available on commercial basis? Yes NA Yes Yes NA NA No; for Proprietary NA NA

Page intentionally left blank.

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report CHAPTER 7 MATERIALS TESTING FOR CONSTRUCTION QUALITY DETERMINATION This chapter focuses on the effectiveness of the NDT technology and device for measuring or judging the quality of construction of unbound materials and HMA mixtures. As described in chapter 1, “effectiveness” is defined as the ability or capability of the NDT technology or device to detect changes in unbound materials or HMA mixtures. The research problem statement noted that, with the development of the MEPDG, layer modulus will become a more important property and should be considered a quality characteristic. Thus, the emphasis of the interpretation of data presented in chapter 5 was on identifying those NDT devices that can consistently and accurately determine when changes occur within the construction process, as well as confirm the assumptions used in pavement structural design. 7.1 Identification of Material Anomalies and Differences The testing under the Part A field evaluation was to confirm that the NDT technologies can identify differences in construction quality of unbound pavement layers and HMA mixtures. The specific hypothesis used for this part of the field evaluation was that the NDT technology and device can detect changes in the physical condition of pavement materials and soils that affect flexible pavement performance. Tables 31 and 47 in chapter 5 summarize the anomalies and different conditions placed along each project. A standard t-test and the Student-Newman-Keuls (SNK) mean separation procedure using a 95 percent confidence level were used to determine whether the areas with anomalies were significantly different from the other areas tested. The following subsections summarize the results from the statistical analyses of the data collected within Part A of the field evaluation. 7.1.1 Unbound Layers Table 66 tabulates the results for checking the hypothesis for the unbound material layers. The shaded cells in Table 66 designate those where the hypothesis was incorrectly rejected or accepted. The DSPA accurately identified most of the areas with anomalies or material differences. The GeoGauge did a reasonable identification of the areas, followed by the DCP and LWD. The EDG and GPR devices did a poor job in identifying the different areas. Table 65 demonstrates the success rate by each device in identifying the physical differences of the unbound material within a project. Table 65. Success Rate Demonstrated by each Device in Identifying the Physical Differences of the Unbound Material NDT Device DSPA GeoGauge DCP LWD GPR EDG Success Rate, % 86 79 64 64 33 25 255

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report Table 66. Summary on the Effectiveness of NDT Devices to Identify Areas of Unbound Layers With Anomalies or Different Physical Conditions NDT Device Project Hypothesis GPR EDG, pcf Geo., ksi DSPA, ksi DCP, ksi Defl., ksi Lane A 14.65 107.6 12.6 25.2 5.20 --- Pre-IC Rolling Lanes B,C,D 15.99 108.1 16.3 34.0 5.62 --- Lane A is weaker No Yes Yes Yes No --- Area 1 21.61 108.3 17.1 39.4 6.93 9.99 Post-IC Area 2 23.00 107.7 19.0 40.4 6.21 11.78 No Planned Difference Yes No No Yes Yes No Pre-IC 15.65 108.0 15.4 31.8 5.51 --- All areas Post-IC 22.31 108.0 17.7 39.9 6.57 --- I-85 Low Plasticity Soil Embankment Post-IC area is stronger Yes No Yes Yes Yes --- Area 2 No IC --- --- 19.6 23.6 11.9 --- Area 1 With IC --- --- 22.9 27.1 9.1 --- Area 1 is stronger --- --- Yes Yes No --- Lane C --- --- 20.1 30.4 9.9 12.9 With IC Rolling Lanes A,B --- --- 24.4 25.4 8.7 8.00 SH-21 High Plasticity Clay Lane C is stronger --- --- No Yes No Yes So. Area Laanes A,B 18.24 122.7 10.5 43.6 15.16 5.65 No. Area Lanes B,C 29.16 124.1 10.1 35.7 19.01 4.77 No Planned Difference No No Yes No No No Lane C 19.33 122.9 7.5 31.1 11.47 5.58 So. Area Lanes A,B 18.24 122.7 10.5 43.6 15.16 5.65 Lane C is weaker No No Yes Yes Yes No Lane A 20.32 123.9 12.6 51.7 18.52 4.69 No. Area Lanes B,C 29.16 124.1 10.1 35.7 19.01 4.77 TH-23 Silt- Sand-Gravel Mix Embankment Lane A is stronger No No Yes Yes No No Lane A 10.29 123.2 25.4 33.9 21.60 24.2 Lane B 9.30 123.0 25.5 34.7 20.95 27.8 All lanes Lane C 9.78 123.8 24.77 33.3 20.74 21.2 No Planned Difference Yes Yes Yes Yes Yes No Area 1,2 9.74 123.5 26.3 36.5 20.64 24.6 All areas Area 3 9.88 123.1 22.3 28.9 22.01 24.1 SH-130 Granular Improved Embankment No Planned Difference Yes Yes No No Yes Yes Lanes A,B 9.37 129.8 14.4 100.4 42.05 16.75 South & Middle Sections Lane C 10.62 129.8 10.8 50.7 21.33 8.31 Lane C is weaker No No Yes Yes Yes Yes So. Area Lanes A,B 9.79 129.9 15.0 110.7 46.45 19.38 Middle Section Lane C 10.38 129.8 9.8 28.0 18.55 7.95 All other areas 9.54 129.8 12.8 75.0 33.14 12.31 Lane C, middle section, is weaker No No Yes Yes Yes Yes TH-23 Crushed Aggregate Base Lanes A & B, south section, are stronger No No Yes Yes Yes Yes Lane 4 11.57 148.2 35.1 117.4 34.31 18.53 All areas Lanes 1,2,3 11.95 147.4 47.9 198.6 50.29 46.46 US-280 Crushed Stone Base Lane 4 is weaker No No Yes Yes Yes Yes NOTE: The shaded or black cells are those areas were the hypothesis was rejected based on a 95 percent confidence interval, and are inconsistent with the construction records and experimental plan. 256

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report The DSPA and GeoGauge have acceptable success rates, while the EDG and GPR have unacceptable rates. More importantly, the modulus measuring devices (DSPA, GeoGauge, DCP, and LWD) found all of the hypotheses to be true for the crushed aggregate materials (TH-23 and US-280 projects), while the volumetric devices (GPR and EDG) rejected all of the hypotheses. This observation suggests systematic differences between the technologies. Some of the important differences observed between the technologies and devices and the reason for the higher success rates for the DSPA and GeoGauge are listed below. • The DSPA and GeoGauge induce small dynamic stress waves into the material being tested. These small responses emphasize the effect of changes in the density and moisture content of the material being tested. More importantly, both devices measure the responses in a relatively limited area and depth. In fact, the sensors for the DSPA (refer to Figures 16 and 78) were spaced so the measured responses would be confined to the layer being tested. The GeoGauge measurements have a deeper influence, so its results can be influenced by the supporting layer. The depth of influence depends on the thickness and stiffness of the material being tested. • The DCP is a point-based test and estimates the modulus of the material from the average penetration rate through the material. The penetration rate is dependent on the dry density of the material. However, there are other physical properties that have a greater effect on the penetration rate. The amount and size of the aggregate particles can have a larger effect on the estimated modulus than for the DSPA or GeoGauge, especially for fine-grained soils with some aggregates. For example, the DCP found all of the hypotheses to be true for the coarse-grained materials and rejected many of the hypotheses for the fine-grained embankment materials with varying amounts of coarse aggregate. • The LWD induces larger strains into the underlying materials. The measured deflections or responses are affected by a much larger area and depth than for the DSPA, GeoGauge, and DCP. The modulus calculated from the deflections is dependent on the thickness and stiffness of the material being tested, as well as the thickness and stiffness of the supporting layers. In fact, some resulting modulus values were lower than expected for the type of material being tested (TH-23 embankment and areas of the US-280 crushed stone). More importantly, the LWD found all of the hypotheses to be true were the layer thicknesses were well defined, but rejected many of the hypotheses for the materials where the layer thickness was less defined—the embankments. • Both the GPR and EDG devices are dependent on the density and water content measurements made with other traditional test methods. Any errors within those traditional methods are included in the GPR and EDG results. More importantly, average water contents were assumed for each area in calculating the wet densities from the dielectric values measured with the GPR. Obviously, water contents are not constant within a specific area. Errors in the water content will be reflected in the wet density for a specific test. More importantly, varying plasticity of the fines and in the 257

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report gradation of the material is difficult to identify with the GPR and EDG by themselves. • Variability of the measurements is another reason for the outcome. The GeoGague had lower variability, followed by the DSPA and DCP. The deflection-based methods had the greatest variability. The lower the variability, the higher the probability to identify a difference, if a difference exists, given the same number of tests (refer to section 7.3). In summary, the DSPA and GeoGauge are considered acceptable in identifying localized differences in the physical condition of unbound materials. 7.1.2 HMA Layers Table 68 tabulates the results for checking the hypotheses for the HMA layers. The shaded cells in Table 68 designate those areas where the hypothesis was incorrectly rejected. Another difference that was found but not planned (so it was excluded from Table 68) was the difference between the initial and supplemental sections of the US-280 project (see chapter 5). All NDT devices found a significant difference between these two areas—the supplemental section had the higher dynamic modulus, which was confirmed with laboratory dynamic modulus tests. Both the PSPA and FWD resulted in higher modulus values and the GPR estimated lower air voids, but the PQI resulted in much lower densities. The PSPA did identify all but one of the areas with anomalies or differences. The non- nuclear density gauge did a reasonable job, while the GPR and FWD only identified slightly more than 50 percent of the areas with differences. The GPR, however, did measure the HMA lift thickness placed that was confirmed through field cores. Table 67 summarizes the success rates for identifying the physical differences of the HMA mixtures within a project. Table 67. Summary of the Success Rates for Identifying the Physical Differences of the HMA Mixtures Within a Project NDT Device PSPA PQI GPR FWD Success Rate, % 93 71 54 56 The PSPA had an excellent success rate, while the PQI had an acceptable rate. The GPR and FWD had lower rates that are considered unacceptable. Some of the important differences observed between the technologies and devices and the reasons for the lower success rates of the GPR and FWD are listed below. • The FWD is believed to have been influenced by the supporting layers creating noise and additional variability making it more difficult to identify the localized areas. In addition, its loading plate probably bridged some of the localized anomalies making it difficult to detect differences near the surface of the layer evaluated (e.g., segregation). 258

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report • The dielectric values measured by the GPR are minimally affected by some of the properties that can change within a project, and its success is heavily dependent on the number of cores taken for calibration purposes—similar to that for unbound materials. In summary, the PSPA and non-nuclear density gauges (PQI) are considered acceptable in identifying localized differences in the physical condition of HMA mixtures. Table 68. Summary of the Effectiveness of NDT Devices to Identify Areas of HMA Layers With Anomalies or Different Physical Conditions NDT Device Project Hypothesis PSPA FWD GPR PQI Section 2 Lanes A,B 285.0 568.9 6.18 149.9 Sections 1,3 Lanes A,B 262.0 405.4 10.14 146.6 Section 2 is Stronger or Stiffer Yes Yes Yes Yes Lane C Section 2,3 288.5 NA 8.51 141.6 Lane C Sections 1 215.4 NA 8.62 140.3 I-85 SMA Overlay Section 1 is Weaker/Less Dense Yes NA No Yes Section 2 All Lanes 454.4 NA 7.04 145.2 Sections 1,3 All Lanes 489.8 NA 6.64 146.6 Section 2 is Weaker Yes NA Yes Yes Section 4 All Lanes 499.5 NA NA 143.9 TH-23 HMA Base No Planned Difference; Sections 1,3,4 Yes NA NA No Initial Sections Section 1 499.9 203.3 7.03 148.0 Supplemental Sections Sections 1,2 555.0 877.2 5.50 140.4 US-280 HMA Base Supplemental Area is Stronger/Denser Yes Yes Yes No Section 1 All Lanes 499.9 203.3 7.03 148.0 Section 2 All Lanes 423.9 125.9 6.81 154.5 Section 1 is Stronger/Denser Yes Yes No No Longitudinal Joints Confined Joint 305.8 125.5 7.70 145.7 Joints are Less Dense/Weaker Yes No Yes Yes Segregated Areas All Lanes 329.9 144.5 7.28 147.1 US-280 HMA Base, Initial Sections Segregated Areas are Less Dense/Stiff Yes No No Yes Section 1 All Lanes 559.8 569.0 5.55 140.4 Section 2 All Lanes 550.2 1185.3 5.45 140.5 No Planned Difference Yes No Yes Yes Longitudinal Joints All Lanes 596.0 379.0 5.78 135.8 Joints are Les Dense/Weaker No Yes No Yes Segregated Areas All Lanes 391.3 707.0 5.64 136.6 US-280 HMA Base, Supplemental Sections Segregated Areas are Less Dense/Stiff Yes No No Yes Section 1 All Lanes 384.9 NA 5.95 126.5 Section 2 All Lanes 292.6 NA 5.61 124.0 Section 3 All Lanes 461.7 NA NA 125.1 Section 2 is Weaker/Less Dense Yes NA Yes Yes Joints All Lanes 297.5 NA 5.08 118.8 I-35/SH-130 HMA Base Joints are Less Dense/Stiff Yes NA No Yes 259

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report 7.2 Estimating Target Modulus Values Laboratory measured modulus of a material is an input parameter for all layers in the MEPDG. Resilient modulus is the input for unbound layers and soils, while the dynamic modulus is used for all HMA layers. As discussed in chapter 5, none of the NDT devices accurately predicted the modulus values that were measured in the laboratory for the unbound materials and HMA mixtures (see Figures 76 and 77 in chapter 5). All of the modulus estimating NDT devices, however, did show a trend of increasing moduli with increasing laboratory measured moduli. The following subsections discuss the use of adjustment factors for confirming the assumptions used for structural design. 7.2.1 Unbound Layers It has been previously reported that layer moduli calculated from deflection basins must be adjusted (multiplied) by a factor for pavement structural design procedures that are based on laboratory derived values at the same stress state (AASHTO, 1993; Von Quintus and Killingsworth, 1998). In the 1993 AASHTO Pavement Design Manual, the adjustment factor is referred to as the “C-factor,” and the value recommended for use is 0.33. Thus, there are differences between the field and laboratory conditions that can cause significant bias when using NDT modulus values. Von Quintus and Killingsworth (1998) found that this adjustment factor was structure or layer dependent but not material type dependent. Adjustment factors were determined for different types of structures. The C-factor found for embankment or subgrade soils ranged from 0.35 to 0.75 and averaged 0.62 for aggregate base materials. However, none of the deflection basins measured in this study was measured on the surface of the unbound layers themselves. Conversely, all testing under this study was directly on the surface of the layer being evaluated. To compensate for differences between the laboratory and field conditions, an adjustment procedure was used to estimate the laboratory resilient modulus from the different NDT technologies for making relative comparisons. The adjustment procedure assumes that the NDT response and modulus of laboratory prepared test specimens are directly related and proportional to changes in density and water content of the material. Figures 84 to 86 compare the seismic (PSPA) modulus measured on the samples used in preparing an M-D relationship. The PSPA modulus-water content relationship follows the M-D relationship. Thus, the assumption is believed to be valid. For simplicity, the adjustment factors were derived using the same methodology within the FHWA-LTPP study, with the exception that a constant, low stress state was used to determine the adjustment factor. In other words, the average laboratory measured modulus (triplicate repeated load resilient modulus tests were performed) was divided by the average moduli estimated with each NDT device. Table 69 summaries the adjustment factors equating the NDT moduli to the resilient modulus measured in the laboratory (see Tables 60 and 62 in chapter 5) for the Part A field evaluation 260

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report projects. The adjustment factors do not appear to be related to the percent compaction, percent of optimum water content, or material type. The adjustment factors for the deflection-based devices are approximately the inverse of the values reported from the FHWA-LTPP study. Thus, the adjustment factors derived from testing on bound pavement surfaces should not be used when testing directly on the unbound layer being evaluated. 100 102 104 106 108 110 112 114 116 8.5 10 12 14.5 16 18 20 Moisture Content, I-85 Embankment, percent Dr y De ns ity , p cf 0 10 20 30 40 50 60 70 80 90 Se si m ic M od ul us , k si Sesimic Modulus, ksi Dry Density, pcfi Poly. (Sesimic Modulus, ksi) Poly. (Dry Density, pcfi) Figure 84. Comparison of the PSPA Moduli to the M-D Relationship for the I-85 Low Plasticity Soil Embankment 114 116 118 120 122 124 126 128 7.5 8.5 9.6 10.5 12 13 14 15 16 Moisture Content, SH-130 Embankment, percent Dr y De nb si ty , p cf 0 10 20 30 40 50 60 70 80 90 100 Se is m ic M od ul us , k is Dry Density Seismic Modulus Poly. (Seismic Modulus) Poly. (Dry Density) Figure 85. Comparison of the PSPA Moduli to the M-D Relationship for the SH-130 Improved Granular Embankment 261

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report 127 127.5 128 128.5 129 129.5 130 130.5 131 3.8 4.8 5.8 7 8 9 10 Moisture Content, US-280 Crushed Stone, percent Dr y De ns ity , p cf 0 20 40 60 80 100 120 140 S ei sm ic M od ul us , k si Dry Density Seismic Modulus Poly. (Seismic Modulus) Poly. (Dry Density) Figure 86. Comparison of the PSPA Modulus to the M-D Relationship for the US-280 Crushed Stone Base Table 69. Adjustment Factors or Ratios Applied to the NDT Modulus Values to Represent Laboratory Conditions or Values at Low Stress States; Part A Projects Ratio or Adjustment Factor Project Material Percent Compaction Percent of Optimum Moisture Geo. DSPA DCP LWD I-85 Embankment Low Plasticity Clay 91 165 0.19 0.087 0.53 0.39 TH-23 Embankment Silt-Sand-Gravel Mix 100 132 0.90 0.41 0.95 3.13 SH-21 Subgrade High Plastic Clay 99 84 1.16 0.99 2.94 2.78 TH-23 Base Crushed Aggregate 104 55 0.71 0.30 0.68 1.69 SH-130 Embankment Improved Granular Mix 105 101 1.39 1.04 1.67 1.43 US-280 Base Crushed Stone 101 52 1.01 0.24 0.96 1.04 The adjustment ratio or factor was determined by dividing the average resilient modulus measured in the laboratory (for a specific stress state, see table 60) by the average modulus from the NDT device. Another important observation from the Part A projects is that the adjustment factors for all NDT devices for the I-85 low plasticity clay embankment prior to IC rolling are significantly lower than for any of the other materials. This observation suggests that the resilient moduli measured in the laboratory are much lower than for any of the other soils and materials. The reason for the low values is unknown. This embankment soil had the lowest dry density and highest water content relative to its maximum dry density and optimum water content also see Tables 60, in chapter 5). However, these data were excluded from developing the 262

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report adjustment factors and selection of an NDT device that can be used to confirm the structural design parameters because they were consistent across all NDT devices. Table 70 summarizes the adjustment factors for all projects included in the field evaluation (Parts A and B). The LWD is not included in Table 70 because it was excluded from the Part B projects. On the average, the GeoGauge and DCP provided a reasonable estimate to the laboratory measured values, with the exception of the fine-grained, clay soils. The GeoGauge deviated significantly from the laboratory values for the fine-grained soils. The results also show that both the GeoGauge and DCP over- or under-predicted the laboratory measured values for the same material, with a few exceptions. Table 70. Adjustment Factors Applied to the NDT Modulus Values to Represent Laboratory Conditions or Values at Low Stress States, All Projects Resilient Modulus, ksi Adjustment Factors Relating Laboratory Values to NDT Values Project Identification Laboratory Measured Value Predicted with LTPP Equations Geo Gauge DSPA DCP Fine-Grained Clay Soils Before IC Rolling 2.5 10.5 0.154 .0751 0.446 I-85 Low- Plastic Soil After IC Rolling 4.0 13.1 0.223 0.113 0.606 NCAT; OK High Plastic Clay 6.9 19.7 0.266 0.166 0.802 SH-21, TX High Plastic Clay 26.8 19.6 1.170 0.989 3.045 Average Ratios for Fine-Grained Soil 0.454 0.336 1.225 Embankment Materials; Soil-Aggregate Mixture South Embankment 16.0 15.7 0.696 0.367 1.053 TH-23, MN North Embankment 16.4 16.3 0.735 0.459 0.863 US-2, ND Embankment 19.0 19.5 1.450 0.574 0.856 SH-130, TX Improved Soil 35.3 21.9 1.337 1.029 1.657 Average Ratios for Soil-Aggregate Mixtures; Embankments 1.055 0.607 1.107 Aggregate Base Materials Co. 103, TX Caliche Base --- 32.3 1.214 --- 1.436 NCAT, SC Crushed Granite 14.3 36.1 0.947 0.156 --- NCAT, MO Crushed Limestone 19.2 40.9 0.747 0.198 --- Crushed Stone, Middle 24.0 29.9 0.851 0.303 0.725 TH-23, MN Crushed Stone, South 26.0 35.6 0.788 0.235 0.560 US-53, OH Crushed Stone 27.5 38.3 1.170 0.449 0.862 NCAT, FL Limerock 28.6 28.1 0.574 0.324 0.619 US-2, ND Crushed Aggregate 32.4 39.8 1.884 0.623 1.129 US-280, AL Crushed Stone 48.4 49.3 1.010 0.244 0.962 Average Ratios for Aggregate Base Materials 1.021 0.316 0.899 Overall Average Values 0.942 0.422 1.084 NOTES: 1. The adjustment ratio is determined by dividing the resilient modulus measured in the laboratory at a specific stress state by the NDT estimated modulus. 2. The average ratios listed above exclude the data from the I-85 low plasticity clay prior to IC rolling. The resilient modulus regression equations are provided in equations 34 to 48. 263

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report These ratios were compared to the percent compaction, percent of optimum water content, and material type, but no relationship could be found. The GeoGauge and DSPA adjustment ratios appear to be related to the amount of fines in the material (percent passing number 200 sieve), as shown in Figure 87. 0 0.5 1 1.5 2 0 20 40 60 80 10 Percent Passing Number 200 Sieve, % A dj us tm en t R at io fo r G eo G au ge 0 Fine-Grained Soil Aggregate-Soil Mixture Crushed Aggregate Base (a) GeoGauge 0 0.2 0.4 0.6 0.8 1 1.2 0 20 40 60 80 10 Percent Passing Number 200 Sieve, % A dj us tm en t R at io fo r D SP A 0 Fine-Grained Soil Soil-Aggregate Mixture Crushed Aggregate Base (b) DSPA Figure 87. Effect of the Amount of Fines of the Adjustment Ratio for the GeoGauge and DSPA Devices 264

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report In summary, the GeoGauge can be used to estimate the resilient modulus measured in the laboratory for aggregate base materials and coarse-graded soil-aggregate embankments, while the DCP provided a closer estimate for the fine-grained soils. However, the ratios for both of these devices were variable—even within the same soil or material group. The DSPA resulted in a positive bias (over-predicted the laboratory resilient modulus) with variable ratios. It is suggested that repeated load resilient modulus tests be performed to determine the target or design value and those results be used to calibrate the NDT devices for a specific soil or aggregate base, because of the variability of these ratios. The resilient modulus test should be performed on bulk material sampled from the stockpiles or the roadway during construction (control strips). Most state agencies do not have a resilient modulus testing capability, so other procedures will need to be used to establish the design or target value during construction (Darter et al., 1997). The resilient modulus was calculated at the same stress state shown in Table 60 using the regression equations that were developed from an FHWA-LTPP study (Von Quintus and Yau, 2001). The regression equations used are provided in equations 34 to 48. ( ) 32 11 k a oct k a aR pp pkM ⎟⎟⎠ ⎞ ⎜⎜⎝ ⎛ +⎟⎟⎠ ⎞ ⎜⎜⎝ ⎛= τθ (34) Where: θ = Bulk Stress, psi 321 σσσθ ++= (35) τ = Octahedral shear stress, psi ( ) ( ) ( )( ) 3 5.02 13 2 32 2 21 σσσσσστ −+−+−= (36) = Atmospheric pressure, 14.7 psi. ap 3,2,1σ = Principal stress, psi. = Regression constants from laboratory resilient modulus test results. 3,2,1k The k regression constants are material specific. The following defines the regression constants for the different materials that were tested within the field evaluation projects. These relationships for these regression constants were developed from the FHWA-LTPP study (Von Quintus and Killingsworth, 1998) Crushed Stone Base Materials: ( )( ) ( ) ( ) dryswLLPk 0001.0037.00088.0008.07632.0 8/31 + − − γ+= (37) ( ) ( ) ( ) ( ) ⎟⎟⎠ ⎞ ⎜⎜⎝ ⎛+− −+−= 40# 2 8/32 00000024.00006.0 038.00008.00016.02159.2 P wLLPk dry dry s γγ (38) ( ) ( ) dryswLLk γ0005.00014.00082.01720.13 +−−−= (39) 265

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report Embankments, Soil-Aggregate Mixture, Coarse-Grained ( ) ( ) ( ) ( ) ( ) ( ) − + + ⎟⎟⎠ ⎞ ⎜⎜⎝ ⎛− ⎟⎟⎠ ⎞ ⎜⎜⎝ ⎛+⎟⎟⎠ ⎞ ⎜⎜⎝ ⎛+−+ += 40# 2 200#4#8/31 00000082.0 3932.06456.10426.00000016.0 0149.00027.00174.00130.05856.0 P w w w PIPPPk Max Max s Max dry SMax γ γ γγ (40) ( ) ( ) ( ) ⎟⎟⎠ ⎞ ⎜⎜⎝ ⎛+ ⎟⎟⎠ ⎞ ⎜⎜⎝ ⎛−+−−= 40# 2 200#2 00000027.0 1483.00001.00081.00060.07833.0 P w wPIPk dry opt s Max γ γ (41) ( ) ⎟⎟⎠ ⎞ ⎜⎜⎝ ⎛+−−= 40# 2 200#3 00000081.00026.01906.0 P Pk opt γ (42) Embankments, Soil-Aggregate Mixture, Fine-Grained ( ) ( ) ( ) ( ) + + ⎟⎟⎠ ⎞ ⎜⎜⎝ ⎛+− += Max dry optw LLPPk γ γ 179.1051.0 0030.00128.00051.07668.0 200#4#1 (43) ( ) ( ) ⎟⎟⎠ ⎞ ⎜⎜⎝ ⎛+−−= Max dryPPk γ γ 3941.10061.00141.04951.0 200#4#2 (44) ( ) ( ) ⎟⎟⎠ ⎞ ⎜⎜⎝ ⎛−++= Max dryLLPk γ γ 8903.30036.00293.09303.0 8/33 (45) Fine-Grained Clay Soil ( ) ( )swClayk 0437.00106.03577.11 −+= (46) ( ) ( ) ( ) ( ) ( )swLL PPPk 0049.00030.0 0027.00095.00073.05193.0 200#40#4#2 −− −+−= (47) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ⎟⎟⎠ ⎞ ⎜⎜⎝ ⎛−+−−+ +−+−= opt s dryopt w wwLL SiltPPPk 6055.00025.00026.00672.00535.0 025.00521.00303.00288.04258.1 max 200#40#4#3 γγ (48) Figure 88 compares the laboratory measured resilient modulus values and those calculated from the regression equations (see Table 70). Use of the regression equations, on the average, resulted in a reasonable prediction of the laboratory measured values. Von Quintus and Yau (2001), however, reported that the regression equations can result in significant error and recommended that repeated load resilient modulus tests be performed. 266

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report 0 10 20 30 40 50 0 10 20 30 40 50 Resilient Modulus Measured in Laboratory, ksi Re si lie nt M od ul us C al cu la te d fr om L TP P E qu at io ns , k si Line of Equality Fine-Grained Soils Embankment Soils; Coarse-Grained Granular Base Figure 88. Comparison of the Resilient Modulus Values Measured in the Laboratory to the Resilient Modulus Values Predicted with the LTPP Regression Equations 7.2.2 HMA Layers Table 63 in chapter 5 listed the laboratory dynamic moduli measured at a loading frequency of 5.0 Hz for the in-place average mixture temperature measured during NDT. As for the unbound materials, it is expected that the modulus values determined from the deflection- based methods are affected by the supporting materials. To compensate for differences between the laboratory and field conditions, an adjustment procedure was used to estimate the modulus values from the PSPA and FWD for making relative comparisons. This field adjustment procedure is the same as used for the unbound materials. The adjustment ratios were determined for the areas without any anomalies or physical differences from the target properties and are given in Table 71. The PSPA adjustment ratios were found to be relatively close to unity, with the exception of the I-35/SH-130 HMA base mixture. This HMA base mixture is a very stiff mixture in the laboratory but was estimated to be similar to the US-2 HMA base with the PSPA (see Table 63 in chapter 5). The reason for the large difference between the laboratory and field or deviation from unity for this one mixture is unknown. Conversely, the FWD adjustment factors are significantly different from unity. The FWD over estimated the SMA modulus for the overlay project and under estimated the HMA base modulus for the reconstruction projects suggesting that the calculated values from the deflection basins are being influenced by the supporting materials. On the average, the PSPA can be used to estimate the dynamic modulus measured in the laboratory HMA mixtures, while the FWD was found to be extremely variable. The PSPA ratios are variable, but that variability is less than the ratios for the unbound materials. These ratios were compared to the binder type, gradation, and other volumetric properties but no 267

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report relationship was found. It is suggested that dynamic modulus tests be performed to determine the target or design value and those results be used to calibrate the PSPA for a specific mixture. The dynamic modulus test can be performed on bulk mixture compacted to the expected in-place density during the mixture verification process or during construction of a control strip. Table 71. Summary of Dynamic Modulus Values Measured in the Laboratory and Adjustment Factors for the Modulus Estimating NDT Devices Ratio or Adjustment Factor Project/Mixture Dynamic Modulus, ksi PSPA FWD I-85 AL, SMA Overlay 250 1.055 0.556 TH-23 MN, HMA Base 810 1.688 NA US-280 AL, HMA Base; Initial Area 650 1.407 3.939 US-280 AL, HMA Base; Supplemental Area 780 1.398 2.516 I-35/SH-130 TX, HMA Base 1,750 5.117 3.253 I-75 MI, Dense-Graded Type 3-C 400 0.919 NA I-75 MI, Dense-Graded Type E-10 590 0.756 NA US-47 MO, Fine-Graded Surface 530 1.158 NA US-47 MO, Coarse-Graded Base Mix 420 0.694 NA I-20 TX, HMA Base, CMHB 340 0.799 NA US-53 OH, Coarse-Graded Base 850 1.275 NA US-2 ND, Coarse-Graded Base, PG58-28 510 1.482 NA NCAT SC, PG67 Base Mix 410 0.828 NA NCAT FL, PG67 Base Mix 390 0.872 NA NCAT FL, PG76 Base Mix 590 1.240 NA NCAT AL, PG76 with RAP and Sasobit 610 1.3760 NA NCAT AL, PG76 with RAP and SBS 640 1.352 NA NCAT AL, PG67 with RAP 450 0.881 NA Overall Average Ratio or Adjustment Factor 1.128 2.566 NOTES: 1. The adjustment factor or ratio was determined by dividing the dynamic modulus measured in the laboratory for the in place temperature and at a loading frequency of 5 Hz by the modulus estimated with the NDT device. 2. The laboratory dynamic modulus values listed above are for a test temperature of a loading frequency of 5.0 Hz at the temperature of the mixture when the NDT was performed (see table 63). 3. The overall average adjust factor excludes the SH-130 mixture because it was found to be significantly different than any other mixture tested in the laboratory; which has been shaded. 268

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report 7.3 Accuracy and Precision Important parameters in QA are the accuracy and precision of a test method. The higher precision of a test method, the fewer tests need to be completed at some confidence level for estimating properties of the population or lot and making the “right” decision regarding the quality of the lot. This section evaluates and compares the variability measured within the field evaluation projects with different NDT devices. The more precise the result, however, does not automatically imply that the test method can identify physical differences or information about the population related to performance. 7.3.1 NDT Devices for Unbound Layers Variability of Response Measurements Figures 89 through 92 compare the COV to the average modulus measured by each device. All COV point comparisons were for the same test area. Thus, the material variance should be the same between the different NDT devices. The GeoGauge consistently has the lower COV, and that value decreases with increasing material stiffness (Figure 92). The variations of the GeoGauge measurements were found to be less dependent of type and size of aggregate, as well as less dependent on the underlying materials for the thicker layers tested. The reason for the higher COV values for the other devices is that the DCP penetration rate is dependent on the amount and size of coarse aggregate particles, while the LWD modulus values are more dependent on the underlying materials. The DSPA is dependent on the water content variations nearer the surface (water content-density gradients), and the amount of fines in coarse-gained materials. 0 10 20 30 40 50 60 70 0 10 20 30 40 50 6 Mean Elastic Modulus, DCP, ksi C oe ffi ci en t o f V ar ia tio n, % 0 Fine-Grained Coarse-Grained Figure 89. Coefficient of Variation versus the Mean Modulus Calculated from the DCP Penetration Rates 269

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report 0 10 20 30 40 50 60 70 80 90 0 10 20 30 40 50 6 Mean Elastic Modulus, LWD, ksi C oe ffi ci en t o f V ar ia tio n, % 0 Fine-Grained Coarse-Grained Log. (Coarse-Grained) Figure 90. Coefficient of Variation versus the Mean Modulus Calculated from the LWD Deflections 0 10 20 30 40 50 60 70 80 90 0 50 100 150 200 250 Mean Elastic Modulus, DSPA, ksi C oe ffi ci en t o f V ar ia tio n, % Fine-Grained Coarse-Grained Figure 91. Coefficient of Variation versus the Mean Modulus Determined from the DSPA Responses 270

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report 0 5 10 15 20 25 30 0 10 20 30 40 50 6 Mean Elastic Modulus, GeoGauge, ksi Co ef fic ie nt o f V ar ia tio n, % 0 Fine-Grained Coarse-Grained Log. (Coarse-Grained) Figure 92. Coefficient of Variation versus the Mean Modulus Determined from the GeoGauge Responses The DSPA had higher variability when testing stiff materials that had water contents significantly below the optimum value or where the surface had been primed. Some layers tested had a significant modulus gradient near the surface, which has a much larger effect on the DSPA responses. Some sites had a positive gradient (modulus increases with depth), while other sites had a negative gradient. Those sites with positive modulus gradients generally had higher adjustment ratios, while those with negative gradients had lower ratios. These modulus gradients were confirmed with the DCP—the only device that could readily measure these gradients in real-time. Figure 92 shows some examples of the change in modulus with depth, as calculated from the penetration rate (see equation 33 in chapter 5). The DSPA was also placed in different directions relative to the roller direction for measuring modulus, while the other NDT devices do not have this capability—only an equivalent or average modulus value is reported for all directions. Figure 93 compares the difference between the modulus values parallel and perpendicular to the roller’s direction to the modulus measured parallel to roller direction. For less stiff materials (especially fine- grained materials) there is no difference between the two readings. For stiffer, coarse-grained materials, however, there is a slight bias. The moduli measured parallel to roller direction were slightly higher, on the average. This difference and bias result in a higher COV for the clustered measurements. The LWD had the higher variability in test results and lower success rates. The higher COV value is related to the variability in the underlying layers and their influence on the measured response with the deflection measuring devices, as well as thickness variations of the layer being evaluated—a constant layer thickness and subsurface condition were used. 271

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report 10 20 30 40 50 60 0 2 4 6 8 10 12 14 Depth Below Surface, inches R es ili en t M od ul us C al cu la te d fro m D C P P en et ra tio n Ra te , k si Ohio Crushed Stone North Dakota Crushed Stone Florida Limerock Base (a) Aggregate Base Materials/Layers 0 5 10 15 20 25 0 2 4 6 8 10 12 14 Depth Below Surface, inches Re si lie nt M od ul us Ca lc ul at ed fr om D CP Pe ne tr at io n Ra te , k si Oklahoma High PI Clay North Dakota Embankment (b) Subgrade and Embankment Materials/Layers Figure 93. Modulus Gradients in Unbound Layers, as Determined with the DCP The variability of the GPR and EDG for measuring the volumetric properties (density and fluids content) were found to be significantly different from each other, as well as from the agencies’ QA data, when available. Both of these devices had very poor success rates in identifying physical differences between different sections. The EDG resulted in very low variability in its estimates of dry density and water content within a specific area or test section. Most of the COV values for both properties were less than 2 percent (see Tables 43 and 44 in chapter 5). Thus, the average values determined at a test point and within a test section did not deviate significantly from the project average that was determined from nuclear density gauges and/or sand-cone tests. 272

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report Conversely, the GPR resulted in high variability of the dielectric values (see Table 42 in chapter 5), as well as for the dry densities. More importantly, the dry densities determined in some areas exceeded 160 pcf (see Figure 51 in chapter 5)—an unlikely value. The reasons for the improbable high as well as low values within a project is the assumption used to convert the dielectric values to dry densities—a constant water content for all areas within a lot was assumed. As a result, the GPR data interpretation technique needs to be improved and determine the dry density and water content along the project prior to day-to-day use in QA programs. -40 -30 -20 -10 0 10 20 30 40 0 10 20 30 40 50 60 70 80 Adjusted Seismic Modulus Parallel to Roller Direction, DSPA, ksi Re si du al (E , P ar re lle l - E , P er de nd ic ul ar ), ks i Coarse-Grained Fine-Grained Zero Residual Line Figure 94. DSPA Modulus Values Measured Parallel to Roller Direction versus the Difference Between Modulus Values Parallel and Perpendicular to Roller Direction Standard Error Another reason for using the adjustment ratios in evaluating each NDT device is to eliminate or reduce bias in assuming that the target value is the laboratory resilient modulus measured at a specific stress state. Figure 95 compares the laboratory measured resilient modulus values to those estimated with different NDT devices but adjusted to laboratory conditions, while Figure 96 presents the residuals (laboratory resilient modulus minus the NDT modulus), assuming that the laboratory value is the target value. On the average, the adjusted elastic modulus from all devices compare reasonably well with the laboratory measured resilient modulus. Table 72 tabulates the mean of the residuals and standard error for the NDT devices that provide a direct measure of material stiffness. Table 72. Tabulation of Mean of the Residuals and Standard Error for NDT Devices NDT Device GeoGauge DSPA DCP LWD Mean Residual, ksi -0.117 0.149 -0.078 0.614 Standard Error, ksi 2.419 4.486 3.768 5.884 273

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report In summary, the GeoGauge, DSPA, and DCP all provide good estimates with negligible bias (effect of adjustment ratios) of the laboratory measured resilient modulus values. The GeoGauge has the lower standard error. The LWD has a higher bias and over two times the standard error, in comparison to the GeoGauge. 0 10 20 30 40 50 60 0 10 20 30 40 50 6 Laboratory Resilient Modulus, ksi Ad ju st ed E la st ic M od ul us fr om N DT D ev ic es , k si 0 Geo., Fine-Grained Geo., Coarse-Grained Line of Equality DSPA, Fine-Grained DSPA, Coarse-Grained (a) DSPA and the GeoGauge. 0 10 20 30 40 50 60 0 10 20 30 40 50 6 Laboratory Resilient Modulus, ksi A dj us te d El as tic M od ul us fr om N D T D ev ic es , k si 0 DCP, Fine-Grained DCP, Coarse-Grained Line of Equality LWD, Fine-Grained LWD, Coarse-Grained (b) Deflection-Based and DCP methods. Figure 95. Laboratory Resilient Modulus versus Adjusted NDT Modulus 274

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report -15 -10 -5 0 5 10 15 0 10 20 30 40 50 60 Laboratory Resilient Modulus,Target Value, ksi Re si du al fr om T ar ge t V al ue , ks i GeoGauge DSPA Zero Residual (a) GeoGauge and DSPA. -15 -10 -5 0 5 10 15 20 25 0 10 20 30 40 50 6 Laboratory Resilient Modulus, Target Value, ksi Re si du al fr om T ar ge t V al ue , ks i 0 DCP LWD Zero Residual (b) DCP and LWD. Figure 96. Residuals (Laboratory Minus NDT Modulus Values) versus Adjusted NDT Modulus 275

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report 7.3.2 NDT Devices for HMA Mixtures Variability of Response Measurements Figure 97 compares the COV between different technologies and devices (PSPA, FWD, PQI, and GPR). The PQI consistently had a low COV relative to the other devices, while the FWD has the largest value. It should be noted that a low COV does not necessarily mean that the device is providing an accurate measure of the HMA mixture property and variability. One reason for the lower COV values for the PQI relative to the other devices is that five tests were performed at each test point. In other words, the testing and sampling error or differences get averaged out through the testing sequence. Two versions of the GPR air-coupled antennas were used. The first version was a single- antenna method and only used in Part A of the field evaluation. The second version included the use of multiple antennas and the EPIC Hyper OpticsTM proprietary data interpretation system. The EPIC GPR system was supposed to be used along the NCAT, Missouri (US-47), and Texas (I-20) sections; however, weather delays and equipment/plant problems resulted in changes to the testing schedule. These schedule changes resulted in conflicts with other projects, so ultimately, this system was used only on the NCAT test sections, at a later date. Data were made available for use from other projects built in Florida, which were not included in the original field evaluation (Greene, 2007; Greene and Hammons, 2006). The EPIC system is reported to have much more accurate and repeatable estimates of the HMA volumetric properties. One reason for this increased accuracy and precision is that it does not rely on the assumptions that were included in the single antenna method used along the Part A projects. The precision and bias for both devices and systems is provided in the next section. 0 5 10 15 20 25 30 35 0 5 10 15 20 25 30 35 PSPA Coefficient of Variation, percent Co ef fic ie nt o f V ar ia tio n, O th er D ev ic es , p er ce nt FWD PQI GPR Line of Equality Figure 97. Comparison of Coefficient of Variations of Different NDT Devices 276

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report Standard Error As for the unbound materials, the adjustment ratios were used in evaluating the PSPA and FWD to reduce bias in assuming that the target value is the laboratory dynamic modulus measured at a specific load frequency and average in place mix temperature. Figure 98 compares the PSPA and FWD modulus values that have been adjusted to laboratory conditions using the factors or ratios listed in Table 71. On the average, the adjusted modulus values compare reasonably well to one another. Table 73 tabulates the mean of the residuals (laboratory dynamic modulus minus the NDT modulus) and standard error from the expected laboratory value—excluding all measurements taken in areas with anomalies, segregation, and along longitudinal joints. Table 73. Tabulation of Mean of the Residuals and Standard Error for NDT Devices From the Expected Laboratory Value NDT Device PSPA FWD Mean Residual, ksi 13.5 39.0 Standard Error, ksi 76 87 While the difference between the two NDT devices is small, the PSPA had the lower residual and standard error. 100 200 300 400 500 600 700 100 200 300 400 500 600 700 Adjusted Seismic Modulus, PSPA, ksi A dj us te d E la st ic M od ul us , FW D , k si Adjusted to Laboratory Conditions Line of Equality Figure 98. Comparison of the PSPA and FWD Modulus Values Adjusted to Laboratory Conditions 277

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report 7.3.3. Summary Table 74 summarizes the statistical analyses of the NDT devices included in the field evaluation projects. This information is grouped into two areas—those NDT devices with an acceptable to excellent success rate and those with poor success rates in identifying material/layer differences. Table 74. NDT Device and Technology Variability Analysis Summary; Standard Error Material/Layer Property Structural Volumetric Material NDT Devices Thickness, in. Modulus, ksi Density, pcf Air Voids, % Fluids Content NDT Devices with Good Success Rates Based on Modulus or Volumetric Properties; see section 7.1.1 GeoGauge NA 2.5 NA NA NA Fine-Grained Soils DSPA NA 4.5 NA NA NA GeoGauge NA 2.5 NA NA NA Coarse-Grained Soils & Aggregate Base DSPA NA 4.5 NA NA NA PSPA NA 76 NA NA NA HMA Mixtures PQI & PT NA NA 1.7 NA NA NDT Devices with Poor Success Rates Based on Modulus or Volumetric Properties; see section 7.1.2 DCP NA 3.8 NA NA NA LWD NA 5.9 NA NA NA GPR NA NA NA NA NA Fine-Grained Soils EDG NA NA 0.8 NA 0.2 DCP NA 3.8 NA NA NA LWD NA 5.9 NA NA NA GPR 0.8 NA 3.4 NA NA Coarse-Grained Soils & Aggregate Base EDG NA NA 1.0 NA 0.2 FWD NA 87 NA NA NA GPR; Single 0.25 NA NA 0.40 NA HMA GPR; Multiple 0.27 NA 1.6 0.22 0.18 NOTES: 1. The standard error for the modulus estimating devices is based on the adjusted modulus values that have been adjusted to laboratory conditions. 2. The US-280 project with the PATB was removed for the GPR (single antenna) thickness data – it was the only site that resulted in a significant bias of layer thickness and the only one with a PATB layer directly beneath the layer tested. 278

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report Table 75. NDT Device and Technology Variability Analysis Summary; 95 Percent Precision Tolerance Material/Layer Property Structural Volumetric Material NDT Devices Thickness, in. Modulus, ksi Density, pcf Air Voids, % Fluids Content NDT Devices with Good Success Rates Based on Modulus or Volumetric Properties; see section 7.1.1 GeoGauge NA 4.9 NA NA NA Fine-Grained Soils DSPA NA 8.8 NA NA NA GeoGauge NA 4.9 NA NA NA Coarse-Grained Soils & Aggregate Base DSPA NA 8.8 NA NA NA PSPA NA 150 NA NA NA HMA Mixtures PQI & PT NA NA 3.4 NA NA NDT Devices with Poor Success Rates Based on Modulus or Volumetric Properties; see section 7.1.2 DCP NA 7.4 NA NA NA LWD NA 11.6 NA NA NA GPR NA NA NA NA NA Fine-Grained Soils EDG NA NA 1.6 NA 0.4 DCP NA 7.4 NA NA NA LWD NA 11.6 NA NA NA GPR 1.5 NA 6.7 NA NA Coarse-Grained Soils & Aggregate Base EDG NA NA 2.0 NA 0.4 FWD NA 170.5 NA NA NA GPR; Single 0.49 NA NA 0.8 NA HMA GPR; Multiple 0.55 NA 3.1 0.4 0.36 NOTES: 1. The precision tolerance for the modulus estimating devices is based on the adjusted modulus values that have been adjusted to laboratory conditions. 2. The US-280 project with the PATB was removed for the GPR (single antenna) thickness data – it was the only site that resulted in a significant bias of layer thickness and the only one with a PATB layer directly beneath the layer tested. 7.4 Comparison of Results—Between NDT Technologies This section provides a brief evaluation and comparison of the test results between different technologies to determine the reasons for the low success rates of the DCP, LWD, GPR, and EDG. 7.4.1 NDT Modulus Comparisons Figure 99 compares the NDT modulus values used to identify areas with physical differences in the unbound layers, except that the NDT values have been adjusted to laboratory conditions with the adjustment ratios listed in Table 70. Figure 99.a includes a comparison of the individual test points, while Figure 99.b compares the data on a project basis. Figure 98 compared the adjusted PSPA and FWD modulus for the HMA layers using the adjustment ratios listed in Table 71. 279

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report Table 76. NDT Device and Technology Variability Analysis Summary; Combined or Pooled Standard Deviation Material/Layer Property Structural Volumetric Material NDT Devices Thickness, in. Modulus, ksi Density, pcf Air Voids, % Fluids Content NDT Devices with Good Success Rates Based on Modulus or Volumetric Properties; see section 7.1.1 GeoGauge NA 1.1 NA NA NA Fine-Grained Soils DSPA NA 1.2 NA NA NA GeoGauge NA 1.8 NA NA NA Coarse-Grained Soils & Aggregate Base DSPA NA 1.5 NA NA NA PSPA NA 56 NA NA NA HMA Mixtures PQI & PT NA NA 2.5 NA NA NDT Devices with Poor Success Rates Based on Modulus or Volumetric Properties; see section 7.1.2 DCP NA 1.9 NA NA NA LWD NA 2.0 NA NA NA GPR NA NA 4.2 NA NA Fine-Grained Soils EDG NA NA 0.7 NA 0.5 DCP NA 5.3 NA NA NA LWD NA 2.0 NA NA NA GPR 0.6 NA 3.0 NA NA Coarse-Grained Soils & Aggregate Base EDG NA NA 0.8 NA 0.6 FWD NA 55 NA NA NA GPR; Single 0.3 NA NA 2.1 NA HMA GPR; Multiple NA NA NA NA NA NOTES: 1. The pooled standard deviations for the modulus estimating devices are based on the adjusted modulus values that have been adjusted to laboratory conditions. 2. The US-280 project with the PATB was removed for the GPR (single antenna) thickness data – it was the only site that resulted in a significant bias of layer thickness and the only one with a PATB layer directly beneath the layer tested. As noted in the previous section, the adjustment procedure reduced the bias between the different devices, but not the dispersion. Thus, any of these NDT modulus estimating devices can be used to estimate the resilient modulus of the material with proper calibration at the beginning of the project, with some exceptions. • Deflection-Based Devices: The calculated modulus values from the deflection-based devices can be affected greatly by the underlying materials and soils. For example, the crushed stone base material placed in area 4 along US-280 near Opelika, Alabama, is a stiff and dense material, even though the deflection-based devices found it to be weaker than the other areas tested with a value less than 20 ksi. All other NDT devices estimated the modulus for area 4 to be about 35 ksi or higher. An in place modulus of 20 ksi for this material is too low. Thus, variations in the subsurface layers or materials/soils can incorrectly result in significant bias in the resilient modulus. • DSPA: The DSPA can significantly over-estimate the laboratory measured resilient modulus values. The US-280 crushed stone base was dry or significantly below the 280

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report optimum water content during testing in some areas. It is believed that the surface of this dense-dry crushed stone is responding like a bound layer—resulting in a much higher modulus of the entire layer. In fact, the surface of this material actually exhibited radial cracks during the seating drop of the DCP. Figure 100 shows the estimated modulus with depth from the DCP. 0 10 20 30 40 50 60 0 10 20 30 40 50 6 Elastic Modulus, GeoGauge, ksi El as tic M od ul us fr om O th er N D T D ev ic es , k si 0 DSPA DCP LWD Line of Equality (b) Comparison of adjusted modulus values on a project basis 0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 6 Elastic Modulus, GeoGauge, ksi El as tic M od ul us fr om O th er N D T D ev ic es , k si 0 DCP DSPA LWD Line of Equality (a) Comparison of adjusted modulus values on a point-by-point basis Figure 99. Comparison of Adjusted Modulus Values Determined from Different NDT Devices 281

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report 0 10 20 30 40 50 60 0 2 4 6 8 10 1 Depth Below the Surface, inches Re si lie nt M od ul us E st im at ed fro m th e D CP P en et ra tio n R at e, k si 2 US-280 Base; Area 1-C US-280; Area 1-A Figure 100. Modulus Gradient Measured with the DCP for the US-280 Crushed Stone Base Material 7.4.2 NDT Volumetric Property Comparisons Unbound Layers The EDG and GPR were used to estimate the volumetric properties of the unbound materials. The following provides a summary of the response measurements to the dry densities obtained from construction records and traditional volumetric tests. • Figure 101 compares the dielectric values to the dry densities measured with the EDG. No good correlation was found between the different materials tested. In addition, no defined relationship was found between the two response measurements for the same material. This observation suggests that there are different properties affecting the EDG and GPR results—none of which could identify the physical differences at a reasonable success rate. • Figure 102 compares the GPR dielectric values to the dry density measured with different devices—the EDG, nuclear density gauges, and sand-cone tests. No good correlation was found; only a trend was identified between the GPR results and the densities obtained from construction records. As the dry density increased, the GPR dielectric values decreased, but across significantly different materials. Changes in material density along the same project were poorly correlated to changes in the dielectric value. • Figure 103 compares the dry densities measured with the EDG to those measured with a traditional nuclear density gauge. As shown, there are two definite groups of data—one for fine-grained soils and the other for crushed aggregate base materials. As the dry density increased between different materials, the density from the EDG also increased. Within each group, however, no reasonable relationship was found. 282

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report 0 10 20 30 40 50 100 110 120 130 140 150 Dry Density, EDG, pcf Di el ec tri c Va lu es , G PR I-85 Embankment TH-23 Embankment SH-130 Subgrade TH-23 Base US-280 Base Figure 101. GPR Dielectric Values versus the EDG Dry Densities Measured along Different Projects 90 100 110 120 130 140 150 0 10 20 30 40 GPR Dielectirc Values Dr y De ns ity , p cf 50 Electrical Density Gauge Nuclear Density Gauge Power (Nuclear Density Gauge) Figure 102. GPR Dielectric Values versus Dry Densities Measured with Nuclear and Non-Nuclear Density Gauges 283

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report 90 95 100 105 110 115 120 125 130 135 140 90 95 100 105 110 115 120 125 130 135 140 Dry Density, EDG, pcf Dr y De ns ity , N uc le ar D en si ty G au ge Series1 Line of Equality Figure 103. Dry Densities Measured with the EDG and Nuclear Density Gauges HMA Layers Figure 104 compares the air voids measured with the GPR to the results from other devices and methods. Figure 104.a compares the densities measured directly with the nuclear density gauge and PQI. There is a general trend between the air void measurements and densities— as air voids increase, the density decreases, but any correlation is poor. There are significant differences between the volumetric properties measured with these different devices. Figure 104.b compares the air voids calculated from the maximum theoretical density provided for each mixture to the air voids estimated from the GPR dielectric values. As shown, no correlation exists between the devices from the field evaluation projects included in this study. Figure 105 compares the densities measured with the nuclear density gauge and the PQI along the longitudinal joints and in areas with localized segregation. These densities are compared with the values measured away from the joints and outside any noticeable segregation. There is a greater variation in density measured with the nuclear device than with the PQI. As noted previously, however, the wet surface may have affected the PQI readings when the measurements were recorded. 7.4.3 Volumetric—Modulus Comparisons Unbound Layers The in-place modulus of the unbound materials is dependent on its density. The FHWA- LTPP study reported that the laboratory resilient modulus was dependent on dry density for all unbound materials (Von Quintus and Yau, 2001). In fact, density and water content are two volumetric properties that have a significant affect on the modulus of the material. Thus, it follows that the NDT devices resulting in a material modulus should be related to the 284

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report density and/or water content of the material. Dry densities and water contents were extracted from the QA reports for the different projects included in the field evaluation. 130 140 150 160 170 4 6 8 10 12 14 Air Voids, GPR, percent De ns ity , p cf 16 PQI - TH-23 Base PQI - SMA Overlay PQI - US-280 Base PQI - US-280 Base Nuclear - SMA Overlay Nuclear - US-280 Base (a) Density measured with the different devices. 0 2 4 6 8 10 12 14 0 2 4 6 8 10 12 14 Air Voids, GPR, percent Ai r V oi ds , O th er D ev ic es , pe rc en t PQI, TH-23 PQI, SMA Nuclear, SMA PQI, US-280 Nuclear, US-280 PQI, US-280 Line of Equality (b) Air voids calculated from the maximum theoretical density for the mixture. Figure 104. Air Voids Measured with the GPR versus Densities Measured with the PQI and Nuclear Density Gauges for Different HMA Mixtures 285

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report 120 130 140 150 160 170 120 130 140 150 160 170 Density, nuclear, pcf D en si ty , P Q I, pc f Joint Readings Segregated Areas HMA Mixture Line of Equality Figure 105. Nuclear Density Gauge Measurements Compared to the PQI Values Along Longitudinal Joints and in Areas with Segregation Figure 106 compares the average modulus values estimated from the different NDT devices and dry densities reported by the individual agencies during construction. The important observation from this comparison is that there is a good relationship between dry density and the DCP estimated modulus, prior to adjusting the modulus values to laboratory conditions (Figure 106.a). The resilient modulus from the GeoGauge is also related to the dry density of the material, but appears to be become insensitive to dry density for less dense, fine-grained soils with high water contents. The resilient modulus from the LWD is related to dry density but has the greatest variation because of the influence of the underlying materials. Figure 106.b graphically presents the same comparison included in Figure 106.a, but using the adjusted modulus values. The GeoGauge and DSPA have similar relationships to dry density for both conditions. Conversely, the relationship for the DCP becomes less defined while it is improved for the LWD. Overall, the modulus values resulting from each NDT device are related to the dry density across a wide range material. The GeoGauge has the better correlation to dry density using the adjusted values, followed by the DSPA and DCP. Thus, the GeoGauge was the primary device used in comparing the elastic modulus to the EDG and GPR results. The dry density and water contents from the QA records were fairly dispersed and were not taken at each NDT test location or individual area. As such, the QA data can only be used to evaluate the results for different types of materials, rather than actual density variations within a project or lot. The EDG was used to measure the density and water contents at specific test locations for the other NDT devices. 286

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report 0 10 20 30 40 50 60 100 110 120 130 140 150 160 Dry Density, QA Records, pcf M od ul us fr om N D T Te st s, U na dj us te d, k si GeoGauge DCP LWD Poly. (DCP) Poly. (GeoGauge) (a) Unadjusted modulus values. 0 10 20 30 40 50 60 100 110 120 130 140 150 160 Dry Density, QA Records, pcf A dj us te d M od ul us fr om N D T Te st s, k si GeoGauge DCP LWD DSPA Poly. (GeoGauge) Poly. (DCP) (b) Modulus values adjusted to laboratory conditions. Figure 106. Dry Density versus NDT Adjusted Modulus Values for Different Materials Figure 107 compares the dry densities measured with the EDG and modulus values estimated from the GeoGauge and DCP. The NDT modulus increases with increasing dry density over a wide range of material types, which is consistent with previous experience. However, there are clusters of data for the EDG that correspond to similar unbound materials that were tested. Within each data cluster, the correspondence between dry density and NDT modulus is poor for both devices. 287

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report This observation suggests that there are other factors that impact the modulus within a specific area; for example, water content and amount of coarse aggregate varying within each data cluster. The EDG did not measure large variations in water content within each area. In summary, the within-project area variation of the modulus values appears to be more dependent on properties other than dry density (e.g., water content, gradation)—assuming that the EDG is providing an accurate estimate of the in-place dry density. That assumption is questionable based on the data accumulated to-date. Figure 108 compares the GeoGauge modulus to the GPR dielectric values. No clear correspondence was found between the dielectric values and modulus values. Specifically, a wide range of dielectric values and moduli were measured, but no consistent relationship was found between the two properties. Thus, material/layer properties that affect modulus within an area have little effect on the dielectric values. 0 10 20 30 40 50 60 70 80 100 110 120 130 140 150 160 Dry Density, EDG, pcf A dj us te d El as tic M od ul us , k si GeoGauge DCP Poly. (DCP) Figure 107. NDT Modulus Values versus Dry Density Measured by the EDG HMA Layers Figure 109 compares the PSPA modulus and the GPR air voids. There is a general trend within this data set—decreasing air voids and increasing PSPA modulus, but no good correlation. All NDT devices did correctly identify the difference between the US-280 initial and supplemental sections, with the exception of the PQI. This difference was not planned but was confirmed through the use of laboratory dynamic modulus tests. The state agency’s and contractor’s QA data did not identify any difference between these two areas or time periods. Figure 110 compares the PSPA modulus and the PQI density. A general trend exists for a specific mixture, but no correlation exists between these devices that can be used in day-to- 288

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report day construction operations for control or acceptance. A more important observation is that the volumetric measuring devices are not being influenced by those properties that affect the modulus measuring NDT devices. As an example, changes in the asphalt content and gradation in relation to density, air voids, and stiffness changes within an area. This finding or conclusion is applicable to all of the NDT devices used to test the HMA mixtures. 0 10 20 30 40 50 60 0 10 20 30 40 GPR Dielectric Values Re si lie nt M od ul us , G eo G au ge , ks i 50 Figure 108. GPR Dielectric Values versus the GeoGauge Modulus 0 2 4 6 8 10 12 14 16 100 200 300 400 500 600 700 PSPA Seismic Modulus, ksi G P R A ir Vo id s, p er ce nt TH-23 Base SMA Overlay US-280 Base, Initial US-280 Base, Supp. Figure 109. PSPA Modulus versus GPR Air Voids 289

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report 130 135 140 145 150 155 160 165 100 200 300 400 500 600 700 PSPA Seismic Modulus, ksi PQ I D en si ty , p cf TH-23 Base SMA Overlay US-280 Base, Initial US-280 Base, Supp. Figure 110. PSPA Modulus versus PQI Density of HMA Mixtures 7.5 Supplemental Comparisons This section provides an overview of three areas of supplementary information and data that were collected during the Part B field evaluation projects: use of multiple gauges, development of modulus and density-growth relationships, and contractor-agency personnel use of selected NDT devices. 7.5.1 Modulus and Density-Growth Relationships for Monitoring the Rolling Operation Instrumented rollers were used on projects to monitor the increase in density and stiffness of the unbound and HMA layers, where the rollers could be scheduled for use. In a couple of cases, the Asphalt Manager was on the project site but exhibited hardware or software problems. In other cases, the unbound base layer had already been compacted by the contractor, and the instrumented roller was only used to test the surface. The contractor did not want to take the risk of potentially disturbing the aggregate base, requiring it to be re- compacted and tested. Figures 73 to 75 in chapter 5 presented some of the IC roller data, as related to HMA densities measured with other devices. Overall, the densities and stiffness measured with other devices compared with the output from the instrumented rollers in the areas without localized anomalies. The instrumented rollers did not identify differences caused by localized anomalies (i.e., anomalies significantly less than the width of the roller). Different NDT devices were also used to monitor the compaction operation of HMA and unbound layers to demonstrate the value of these devices in real-time. The PSPA, DSPA, GeoGauge, and PaveTracker devices were used on some of the Part A and most of the Part B field evaluation projects. The following summarizes important observations from the use of selected NDT devices for controlling the placement and compaction of both unbound and HMA layers in real-time. 290

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report Unbound Materials and Layers Overall, the GeoGauge, DCP, and DSPA were successful in monitoring the build-up of modulus with the number of roller passes for the unbound materials placed within the field evaluation, and they were beneficial in assisting the contractor in making decisions on the compaction operation used along the project. Some examples are noted below. • Figure 111 presents data collected on a caliche base material placed along an entrance roadway from County Road 103 near Pecos, Texas. Both the GeoGauge and DCP were used to determine the increase in material modulus with compaction. The DCP was used along this project because it was on a private facility and delaying the compaction of this base material was not an issue. Both devices found an increase in modulus with increasing number of roller passes. 5 10 15 20 25 30 0 2 4 6 8 10 1 Number of Roller Passes R es ili en t M od ul us fr om N D T D ev ic es , k si 2 DCP GeoGauge Log. (GeoGauge) Log. (DCP) Figure 111. Modulus-Growth Relationships for a Caliche Base Along an Entrance Roadway to a Facility from County Road 103 near Pecos, Texas • Figure 112 presents data collected during the compaction of a Missouri crushed limestone base material. The first roller pass within this figure is after the material had been preliminary compacted from other construction equipment and roller passes. The maximum modulus for this material was achieved at about eight passes of the roller over a specific area. The number of passes obviously is dependent on the water content of the in-place material; for the Missouri crushed limestone, the in-place water content was just below the optimum value. • Figure 113 presents data collected during the compaction of a South Carolina crushed granite base material. This crushed granite base material was difficult to compact with the roller on the project site when compaction was initiated. In addition, the water content of this base material was well below the optimum value. As shown, both the DSPA and GeoGauge modulus values did not increase with the number of 291

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report roller passes. A nuclear density gauge was also used along the project, and it also showed no increase in density with the number of roller passes. Thus, rather than waste additional compaction effort, the contractor had to use a heavier roller, but more importantly, increase the water content of the material to obtain the specified density. This example shows the benefit and advantage of using the GeoGauge or DSPA to make decisions in real-time. These examples show the benefit of developing modulus-growth curves using the DSPA or GeoGauge during construction for monitoring and optimizing the rolling pattern. 10 12 14 16 18 20 22 24 26 28 0 1 2 3 4 5 6 7 8 9 10 11 12 Number of Roller Passes G eo G au ge M od ul us , k si Missouri Crushed Limestone Base 0 10 20 30 40 50 60 70 80 90 0 2 4 6 8 10 12 Number of Roller Passes D SP A M od ul us , k si 0 5 10 15 20 25 30 G eo G au ge M od ul us , ks i DSPA Modulus Geogauge Modulus Figure 112. Modulus-Growth Relationships for a Missouri Crushed Limestone Base Material for Two Different Areas 292

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report 40 45 50 55 60 65 70 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 Number of Roller Passes DS P A M od ul us , k si 6 7 8 9 10 11 G eo G au ge M od ul us , ks i DSPA Modulus GeoGauge Modulus 5 6 7 8 9 10 0 2 4 6 8 10 12 14 Number of Roller Passes G eo G au ge M od ul us , k si South Carolina Crushed Granite Figure 113. Modulus-Growth Relationships for a South Carolina Crushed Granite Base Material for Two Different Areas HMA Mixtures and Layers Overall, the PSPA and PaveTracker were successful in monitoring the build up of modulus and density with the number of roller passes for the HMA layers placed within the field evaluation projects. Some examples are noted below. • Figure 114 presents data collected along the Missouri widening project (US-47) for two different areas. Figure 114.a compares the densities measured with the contractor’s nuclear density gauge being used on site for QC to those values measured with the PaveTracker. The densities from the nuclear gauge were related to the non-nuclear density gauge values with mixture specific calibration values. The 293

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report contractor was using one-test point readings with the nuclear gauge, while four readings at a test point were made with the PaveTracker within the same time. More importantly, the contractor was using the cold-side pinch method for compacting the longitudinal joint adjacent to the old pavement. This HMA was tender based on visual observations of its behavior under the roller—shoving of the mat was observed in front of, as well as across, the roller’s direction. Rollers marks were also present after the last pass of the finish roller. The HMA was being pushed away from the confined longitudinal joint, rather than being pushed down into the joint. Joint densities were made with both the nuclear and non-nuclear density gauges along the joint, and the densities were found to be very low—about 5 to 10 pcf below the densities measured within the center of mat. The contractor was asked to change the rolling pattern for the confined longitudinal joint using the hot-side method. Using the hot-side method, the first pass of the roller is along the confined longitudinal joint, with about a 6-inch overhang off the hot mat. Densities were measured with both devices after changing the rolling pattern. Figure 114.b shows the densities along the longitudinal joint, as compared to those in the center of the mat. As shown, the densities significantly increased by eliminating the roller pass on the cold side of the joint. Thus, the contractor was able to use the non-nuclear density gauge in real-time to significantly increase the joint density by slightly revising the rolling pattern of the joint. The PSPA was also used along this project, but the results were erratic during or immediately after compaction of the mat—the wave form was not consistent with HMA mixtures. The mixture was found to be too tender to obtain reliable readings, until the mix cooled below about 150 °F. This HMA mixture was being used as the base for the shoulder or in a non-critical area. It was initially believed that the PSPA had been damaged in transport, but that was found to be incorrect from latter testing of the HMA after it had cooled down. At lower temperatures, the PSPA provided reasonable results. Thus, its use would have been a benefit in identifying a tender mix, if this mix had been used in a critical area under heavy traffic. The PSPA was attempted to be used on a couple of other projects, but the temperature of those mixtures was too high to obtain reliable results. Mix temperature is a limitation on testing HMA mixtures during rolling. • Figure 115 presents density data collected on a Missouri HMA base mixture that was not tender, but was rolled within the temperature sensitive zone. As shown, the first pass of the rubber-tired roller increased the density, but additional passes of that roller significantly decreased the density of the mat. The nuclear density gauge being used on site for QC gave the same results. The nuclear gauge, however, was not being used after each roller pass. This mixture did not exhibit the traditional mix “checking” or tearing under the rollers, but the non-nuclear density gauge did identify the detrimental effect of rolling within the temperature sensitive zone. More roller passes were required to regain the density that was lost by rolling within the temperature sensitive zone. Many of the other HMA mixtures that were included within the field evaluation projects also exhibited this temperature sensitivity under the rollers. 294

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report Selecting HMA mixtures that checked and tore for the field evaluation was not planned. (a). PaveTrack versus nuclear gauge density measurements (b) PaveTracker density measurements made along a confined joint and within the center of the mat. 145 147 149 151 153 155 157 159 161 163 165 0 1 2 3 4 5 Number of Roller Passes De ns ity M ea su re d w ith P av eT ra ck er , p cf 170 180 190 200 210 220 230 Te m pe ra tu re o f M ix tu re , F Mat Density Joint Density Mat Temperature Joint Temperature 136 138 140 142 144 146 148 150 152 154 0 1 2 3 4 5 Number of Roller Passes H M A M at D en si ty , p cf 0 50 100 150 200 250 M at T em pe ra tu re , F PaveTracker Density Nuclear Gauge Density Temperature Figure 114. Typical Density-Growth Curve Measured with PaveTracker and Nuclear Density Gauge for the Missouri US-47 Project 295

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report 140 141 142 143 144 145 146 147 148 149 150 0 2 4 6 8 10 12 14 16 Number of Roller Passes De ns ity M ea su re d w ith Pa ve Tr ac ke r, pc f Missouri HMA Mixture Compaction Operation: Pass 1-2; Vibratory Roller Pass 3-4; Static Steel Wheel Pass 5-6; Vibratory Roller Pass 7-11; Rubber Tired Pass 11-14; Finish Roller Figure 115. Density-Growth Relationship for an HMA Base Mixture from Missouri • The I-75 Michigan overlay project was another project where a HMA mixture was rolled within its temperature sensitive zone. With three passes of a SAKAI vibratory roller in the primary roller position, the HMA mixture density was greater than the specified value (see Figure 116). However, an intermediate roller continued to roll the mix, and was followed by two additional rollers. The use of the PaveTracker determined that the contractor was rolling in the temperature sensitive zone—the density began to decrease. By monitoring the density of the mat during rolling, the result was that the contractor could eliminate two of the rollers and use fewer passes to obtain the required density, as long as the rollers stayed out of the temperature sensitive zone. • Figure 117 shows another example, but for polymer modified asphalt (PMA) and conventional neat asphalt mixtures. These mixtures were placed during the same time period. The conventional neat asphalt mixture exhibited the traditional checking and tearing of the mat when it is rolled within the temperature sensitive zone (photographs are included in Appendix B), while the PMA mixture did not exhibit tearing or checking. As shown, after pass 3 for the neat asphalt mix and after pass 5 for the PMA mix, the densities decreased. The mix tearing and checking was observed under the roller, to confirm that the mix was rolled within the temperature zone. Thus, the mat had to be rolled much more to increase density to the specified value for both mixtures. Similar to the benefit for unbound layers, the non-nuclear density gauges provide significant benefit to a contractor to optimize the rolling pattern within the center of the mat, as well as along longitudinal joints. The non-nuclear gauges can be also used to determine when the rollers are being operated within the temperature sensitive zone, so a contractor does not 296

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report waste compaction effort or time, but more importantly, does not tear or damage the HMA mix by operating the rollers within the temperature sensitive zone. 146.0 148.0 150.0 152.0 154.0 156.0 158.0 0 2 4 6 8 10 12 Number of Passes of the Rollers D en si ty o f M at M ea su re d w ith Pa ve Tr ac ke r, pc f 200 210 220 230 240 250 260 Te m pe ra tu re o f M ix tu re , F Density Temperature 142.0 144.0 146.0 148.0 150.0 152.0 154.0 0 2 4 6 8 10 12 Number of Roller Passes D en si ty o f M at M ea su re d w ith Pa ve Tr ac ke r, p cf 210 220 230 240 250 260 270 Te m pe tu ru e of M ix tu re , F SAKAI Vibratory SAKAI Pneumatic & Other Rollers Temperature Vibratory Breakdown Roller Intermediate & Finish Rollers Figure 116. Density-Growth Curves for the Michigan Mixture Measured with PaveTracker and Effects of Rolling within the Temperature Sensitive Zone; Two Different Areas of Project 297

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report Figure 117. Density-Growth Curves for Two Florida Mixtures Measured with PaveTracker and Effects of Rolling within the Temperature Sensitive Zone 7.5.2 Multiple Operators and NDT Gauges For most of the Part B projects, multiple GeoGauges and PaveTrackers were used by different operators to determine the effects of multiple operators on the variability of the devices. Figure 118 compares the measured responses from the two GeoGauges that were used for testing unbound materials, while Figure 119 compares the measured densities from the two PaveTracker devices used to monitor HMA mixtures. At the end of the field evaluation testing for each project, one of each device was left with the agency and contractor personnel. The following summarizes observations from this comparative testing. • Use of different GeoGauges and operators resulted in some bias that was modulus dependent for some materials; more bias was exhibited for the higher modulus values or stiffer material. It is recommended that material specific calibration or adjustment factors be determined and used for each material tested (see Table 70). This material specific calibration with a sufficient number of replicate tests should minimize the 132 134 136 138 140 142 144 0 2 4 6 8 10 Number of Roller Passes De ns ity M ea su re d w ith Pa ve Tr ac ke r, pc f Compaction Operation: Pass 1-4; Vibratory Roller Pass 5; Static Steel Drum Pass 6-8; Vibratory Roller Pass 9- ; Finish Roller Florida PMA Base Mix (b) PMA Mixture 132 134 136 138 140 142 144 0 2 4 6 8 10 1 Number of Roller Passes D en si ty M ea su re d w ith P av eT ra ck er , p cf 2 Florida Neat Base Mix Compaction Operation: Pass 1-4; Static Steel Drum Pass 5-7; Rubber Tired Roller Pass 8-10; Finish Roller (a) Conventional Neat HMA Mixture 298

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report bias between the different gauges. The variability between different gauges, however, will still exist. • Use of different PaveTrackers and operators resulted in almost no bias between the two gauges, with the exception of dense or high specific gravity mixtures. It is also recommended for these devices that material specific adjustments be determined for each mixture tested. The mixture specific factors should minimize bias, but the variability between different gauges will still exist. Comparison between Geogauge B 24C and B 25C (All Part B test data) y = 0.9025x + 1.6607 R2 = 0.7615 0 5 10 15 20 25 30 35 40 45 0 5 10 15 20 25 30 35 40 45 GeoGauge, B 24C; Modulus, ksi G eo G au ge , B 2 5C ; M od ul us , k si Comparison of Geogauge B 24C and B 25C (By Section) 15 16 17 18 19 20 21 22 23 24 25 26 27 28 15 16 17 18 19 20 21 22 23 24 25 26 27 28 GeoGauge, B 24C; Modulus, ksi G eo G au ge , B 2 5C ; M od ul us , k si Comparison between Geogauge B 24C and B 25C (Modulus, ksi) 5 10 15 20 25 30 35 40 45 5 10 15 20 25 30 35 40 4 GeoGauge, B 24C; Modulus, ksi G eo G au ge , B 25 C ; M od ul us , k si 5 ND US-2 Base NCAT, Missouri N10, Base NCAT, Oklahoma, N8&N9, Subgrade OH, SR-53, Base Linear (ND US-2 Base) Linear (NCAT, Missouri N10, Base) Linear (NCAT, Oklahoma, N8&N9, Subgrade) Linear (OH, SR-53, Base) (a) Comparison on a point-by-point basis. (b) Comparison on a project basis. Figure 118. Comparison of Modulus Measurement with Two Independent GeoGauges (c) Comparison on a material basis 299

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report Comparison between Pavetracker 10232 and 10233 (All Density Data) y = 1.1315x - 17.635 R2 = 0.2007 110 120 130 140 150 160 170 110 120 130 140 150 160 170 Pavetracker 10232; Density, pcf Pa ve tra ck er 1 02 33 ; D en si ty , p cf Comparison of Pavetracker 10232 and 10233 (Average Density, pcf) y = 0.9506x + 7.1232 R2 = 0.9179 120 125 130 135 140 145 150 155 160 165 170 120 130 140 150 160 170 Pavetracker 10232; Density, pcf Pa ve tra ck er 1 02 33 ; D en si ty , p cf (a) Comparison on a point-by-point basis (b) Comparison on a project basis Comparison between Pavetracker 10232 and 10233 (Density, pcf) 110 120 130 140 150 160 170 110 120 130 140 150 160 170 PaveTracker 10232; Density, pcf P av eT ra ck er 1 02 33 ; D en si ty , p cf MI I-75 Surface MO US-47 Surface ND US-2 Surface OH SR-53 Base OH SR-53 Surface (c) Comparison on a material or mixture basis Figure 119. Comparison of the Density Measurements with Two Non-Nuclear PaveTracker Devices Used Within the Part B Field Evaluation 300

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report 7.5.3 Agency and Contractor Use of NDT Devices During Part B of the field evaluation, one of the multiple gauges being used on a project was left with agency and contractor construction personnel for continued use on a day-to-day QA bases. Those NDT devices left with the construction personnel included the GeoGauge, PSPA, and PaveTracker. Data from this additional use were included in the comparison of multiple operators and devices at specific project sites. This information was used in the evaluation described in chapter 8, in determining the parameters needed to set up control and acceptance plans when using these NDT devices. The projects where construction personnel continued to use the devices included Missouri, North Dakota, and Texas. The NDT devices were going to be left at the Michigan I-75 project, but issues with the HMA mixture resulted in the project being stopped for a short term, so the construction personnel did not actually use the devices. For the Missouri project, weather delays resulted in the contractor moving to a different project so the devices were not used on the same project, as included in the Part B field evaluation. The devices were used for more than 2 weeks on the North Dakota and Texas projects. In actuality, the contractor had already been using the PaveTracker and PSPA on the Texas I-20 project. The PaveTracker was a part of the contractor’s standard or day-to-day QC plan, while the PSPA had been used on a research basis. 7.6 Summary of Evaluations In summary, the steady-state vibratory (GeoGauge) and seismic (DSPA) technologies are recommended for use in judging the quality of unbound layers, while the seismic (PSPA) and non-nuclear density gauges (the PaveTracker was used in Part B) are recommended for use of HMA layers. The GPR is recommended for layer thickness acceptance, while the IC rollers are recommended for use on a control basis for compacting unbound and HMA layers. The following identifies and lists some of the reasons for these recommendations. 7.6.1 NDT Devices for Unbound Layers and Materials • The DSPA and GeoGauge devices had the highest success rates for identifying an area with anomalies with rates of 86 and 79 percent, respectively. The DCP and LWD identified about two-thirds of the anomalies, while the GPR and EDG had unacceptable rates below 50 percent. • Three to five repeat measurements were made at each test point with the NDT devices, with the exception of the DCP. o The LWD exhibited low standard deviations that were less dependent on material stiffness with a pooled standard deviation less than 0.5 ksi. One reason for the low values is that the moduli were less than for the other devices. The COV, however, was higher. It is expected that the supporting layers had an effect on the results by reducing the modulus. o The GeoGauge had a standard deviation for repeatability measurements varying from 0.3 to3.5 ksi and are material dependent. 301

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report o The DSPA had the lowest repeatability with a standard deviation varying from 1.5 to 21.5 ksi. The reason for this higher variation in repeat readings is that the DSPA sensor bar was rotated relative to the direction of the roller, while the other devices were kept stationary or do not have the capability to detect anisotropic conditions. No significant difference was found relative to the direction of testing for fine-grained soils, but there was a slight bias for the stiffer coarse-grained materials. o The EDG was highly repeatable with a standard deviation in density measurements less than 1 pcf, while the GPR had poor repeatability—based on point measurements. Triplicate runs of the GPR were made over the same area or sublot. For comparison to the other NDT devices, the values measured at a specific point, as close as possible, were used. Use of point specific values from successive runs could be a reason for the lower repeatability, which are probably driver specific. One driver was used for all testing with the GPR. • The COV was used to compare the normalized dispersion measured with different NDT devices. The EDG consistently had the lowest COV with values less than 1 percent. The GeoGauge had a value of 15 percent, followed by the DSPA, LWD, DCP, and GPR. The GPR and EDG are dependent on the accuracy of other tests in estimating volumetric properties (density and moisture contents). Any error in the calibration of these devices for the specific material is directly reflected in the resulting values. A probable reason why the GPR and EDG devices did not consistently identify the areas with anomalies or physical differences. • Repeated load resilient modulus tests were performed in the laboratory for characterizing and determining the target resilient modulus for each material. Adjustment ratios were determined based on uniform conditions. The overall average ratio for the GeoGauge for the stiffer coarse-grained materials was near unity (1.05). For the fine-grained, less stiff soils, the ratio was about 0.5. After adjusting for laboratory conditions, all NDT devices that estimate resilient modulus resulted in low residuals (laboratory resilient modulus minus the NDT elastic modulus). However, the GeoGauge and DCP resulted in the lowest standard error. The LWD had the highest residuals and standard error. • The DSPA and DCP measured responses represent the specific material being tested. The DCP, however, can be significantly affected by the varying amounts of aggregate particles in fine-grained soils and the size of the aggregate in coarse-grained soils. The GeoGauge measured responses are minimally affected by the supporting materials, while the LWD can be significantly affected by the supporting materials and thickness of the layer being tested. Thickness deviations and variable supporting layers are reasons that the LWD had a low success rate in identifying areas with anomalies or physical differences. • No good or reasonable correlation was found between the NDT devices that estimate modulus and those devices that estimate volumetric properties. • The instrumented rollers were used on too few projects for a detailed comparison to the other NDT devices. The rollers were used to monitor the increase in density and stiffness 302

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report with increasing number of roller passes. One potential disadvantage with these rollers is that they may bridge localized soft areas. These rollers are believed to be worth future investment in monitoring the compaction of unbound materials. • The GPR resulted in reasonably accurate estimates to the thickness of aggregate base layers. None of the other NDT devices have the capability or same accuracy to determine the thickness of the unbound layer. 7.6.2 NDT Devices for HMA Mixtures and Layers • The PSPA had the highest success rate for identifying an area with anomalies with a rate of 93 percent. The PQI identified about three-fourths of the anomalies, while the FWD and GPR identified about half of those areas. The seismic and non-nuclear gauges were the only technologies that consistently identified differences between the areas with and without segregation. These two technologies also consistently found differences between the longitudinal joint and interior of the mat. • The non-nuclear density gauges (PaveTracker) was able to identify and measure the detrimental effect of rolling the HMA mat within the temperature sensitive zone. This technology was beneficial on some of the Part B projects to optimize the rolling pattern initially used by the contractor. • Three to four repeat measurements were made at each test point with the NDT devices. o The PSPA had a repeatability value, a median or pooled standard deviation, of about 30 ksi for most mixtures, with the exception of the US-280 supplemental mixture that was much higher. o The FWD resulted in comparable value for the SMA mixture (55 ksi), but a higher value for the US-280 mixture (275 ksi). o The non-nuclear density gauges had repeatability values similar to nuclear density gauges with a value less than 1.5 pcf. o The repeatability for the GPR device was found to be good and repeatable, with a value of 0.5 percent for air voids and 0.05 inches for thickness. • The PSPA moduli were comparable to the dynamic moduli measured in the laboratory on test specimens compacted to the in place density at a loading frequency of 5 Hz and the in place mixture temperature, with the exception of one mixture—the US-280 supplemental mixture. In fact, the overall average ratio or adjustment factor for the PSPA was close to unity (1.1). This was not the case for the FWD. More importantly, without making any corrections for volumetric differences to the laboratory dynamic modulus values, the standard error for the PSPA was 76 ksi (laboratory values assumed to be the target values). The PSPA was used on HMA surfaces after compaction and the day following placement. The PSPA modulus values measured immediately following compaction were found to be similar to the values one or two days after placement— making proper temperature corrections in accordance with the master curves measured in the laboratory. 303

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report • A measure of the mixture density or air voids is required in judging the acceptability of the modulus value from a durability stand point. The non-nuclear gauges were found to be acceptable, assuming that the gauges have been properly calibrated to the specific mixture—as for the PSPA. • Use of the GPR single antenna method, even with mixture calibration, requires assumptions on specific volumetric properties that do vary along a project. As the mixture properties change, the dielectric values may or may not be affected. Use of the proprietary GPR analysis method on other projects was found to be acceptable for the air void or relative compaction method. This proprietary and multiple antenna system, however, was not used within Part A of the field evaluation to determine its success rate in identifying localized anomalies and physical differences between different areas. Both GPR systems were found to be very good for measuring layer thickness along the roadway. • Water can have a definite affect on the HMA density measured with the non-nuclear density gauges (PQI). The manufacturer’s recommendation is to measure the density immediately after compaction, prior to allowing any traffic on the HMA surface. Within this project, the effect of water was observed on the PQI readings, as compared to dry surfaces. The measured density of wet surfaces did increase, as compared to dry surfaces. From the limited testing completed with wet and dry surfaces, the PaveTracker was less affected by surface conditions. However, wet versus dry surfaces were not included in the field evaluation plan for different devices—only the technology. Based on the data collected within the field evaluation, wet surfaces did result in a bias of the density measurements with this technology. • Another important condition is the effect of time and varying water content on the properties of the HMA mixture during construction. There have been various studies completed on using the PSPA to detect stripping and moisture damage in HMA mixtures. For example, Hammons et al. used the PSPA (in combination with GPR) to locate areas with stripping along selected interstate highways in Georgia (Hammons et al., 2005). The testing completed within this study also supports the use of the seismic-based technology to identify such anomalies. • The instrumented rollers used to establish the increase in stiffness with number of passes was correlated to the increases in density, as measured by different devices. These rollers were used on too few projects to develop or confirm any correlation between the NDT response and the instrumented roller’s response. One issue that will need to be addressed is the effect of decreasing temperature on the stiffness of the mixture and how the IC roller perceives that increase in stiffness related to increases in density of the mat. A potential disadvantage with these rollers is that they will bridge segregated areas and may not accurately identify cold spots in the HMA mat. These rollers are believed to be worth future investments in monitoring the compaction of HMA mixtures. 304

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report 7.6.3 Limitations and Boundary Conditions The following lists the limitations and boundary conditions observed during the field evaluation for the NDT devices recommended for QA application on an immediate, effective, and practical basis. • All NDT devices recommended for QA application, with the exception of the GPR and IC rollers, are point-specific tests. Point-specific tests are considered a limitation because of the number of samples that would be required to identify localized anomalies that deviate from the population distribution. o Ultrasonic scanners are currently under development so that relatively continuous measurements can be made with this technology. These scanners are still considered in the research and development stage and are not ready for immediate and practical use in a QA program. o GPR technology to estimate the volumetric properties of HMA mixtures is available for use on a commercial basis, but the proprietary system has only had limited verification of its potential use in QA applications and validation of all volumetric properties determined with the system. o Similarly, the IC rollers take continuous measurements of density or stiffness of the material being compacted. During the field evaluation, some of these rollers had both hardware and software problems. Thus, these devices were not considered immediately ready for use in a day-to-day QA program. The equipment, however, has been improved and its reliability has increased. The technology is recommended for use on a control basis but not for acceptance. • Ultrasonic technology (PSPA) for HMA layers and materials are recommended for use in control and acceptance plans. o Test temperature is the main boundary condition for the use of the PSPA. Elevated temperatures during mix placement can result in erratic response measurements. Thus, the gauge may not provide reliable responses to monitor the compaction of HMA layers and define when the rollers are operating within the temperature sensitive zone for the specific mixture. o These gauges need to be calibrated to the specific mixture being tested. However, this technology can be used in the laboratory to measure the seismic modulus on test specimens during mixture design or verification prior to measuring the dynamic modulus in the laboratory. o A limitation of this technology is that the results (material moduli) do not provide an indication on the durability of the HMA mixture. Density or air void measurements are needed to define durability estimates. o The DSPA for testing unbound layers is influenced by the condition of the surface. High modulus values near the surface of the layer will increase the modulus estimated with the DSPA. Thus, the DSPA also needs to be calibrated to the specific material being evaluated. • Steady-state vibratory technology (GeoGauge) for unbound layers and materials; recommended for use in control and acceptance plans. 305

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report o This technology or device should be used with caution when testing fine-grained soils at high water contents. In addition, it should not be used to test well-graded, non-cohesive sands that are dry. o The condition of the surface of the layer is important and should be free of loose particles. A layer of moist sand should also be placed to fill the surface voids and ensure that the gauge’s ring is in contact with about 75 percent of the material’s surface. Placing this thin layer of moist sand takes time and does increase the time needed for testing. o These gauges need to be calibrated to the specific material being evaluated, and are influenced by the underlying layer when testing layers that are less than 8 inches thick. o These gauges are not applicable for use in the laboratory during the preparation of M-D relationships that are used for monitoring compaction. The DSPA technology is applicable for laboratory use to test the samples used to determine the M-D relationship. o A relative calibration process is available for use on a day-to-day basis. However, if the gauge does go out calibration, it must be returned to the manufacturer for internal adjustments and calibration. o These gauges do not determine the density and water content of the material. The water content and density of the unbound layer should be measured with other devices. • Non-nuclear density gauges (electric technology) for HMA layers and materials; recommended for use in control and acceptance plans. o The results from these non-nuclear density gauges can be dependent on the condition of the layer’s surface—wet versus dry conditions. It is recommended that the gauges be used on relatively dry surfaces until additional data become available relative to this limitation. Free water should be removed from the surface to minimize any affect on the density readings. However, water penetrating the surface voids in a segregated areas will probably affect the readings—incorrect or high density reading, in comparison to the actual density from a core. The PSPA was able to identify areas with segregation. o These gauges need to be calibrated to the specific material under evaluation. • GPR technology for thickness determination of HMA and unbound layers are recommended for use in acceptance plans. o The data analysis or interpretation is a limitation of this technology. The GPR data requires some time to estimate the material property—the time for layer thickness estimates is much less than for other layer properties. o This technology requires the use of cores for calibration purposes. Cores need to be taken periodically to confirm the calibration factors used to estimate the properties. o Use of this technology, even to estimate layer thickness, should be used with caution when measuring the thickness of the first lift placed above PATB layers. o GPR can be used to estimate the volumetric properties of HMA mats, but that technology has yet to be verified on a global basis. o The technology and devices are not applicable to the use of laboratory data for calibration purposes. 306

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report 307 • IC rollers are recommended for use in a control plan, but not within an acceptance plan. o The instrumented rollers may not identify localized anomalies in the layer being evaluated. These rollers can bridge some defects—insufficient sensitivity to identify defects that are confined to local areas. o Temperature is an issue with the use of IC rollers for compacting HMA layers. Although most of these rollers have a capability to measure the surface temperature of the mat, the effect of temperature on the mat stiffness is an issue— as temperature decreases the mat stiffness will increase, not necessarily because of an increase in density of the mat. Delaying the compaction would increase the stiffness of the mat measured under the rollers because of the decrease in temperature. o The instrumented rollers also did not properly identify when checking and tearing of the mat occurred during rolling. The non-nuclear density gauges (PaveTracker) did identify this detrimental condition. o The technology and devices are not applicable to the use of laboratory data for calibration purposes.

This page intentionally left blank.

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report CHAPTER 8 CONSTRUCTION QUALITY DETERMINATION The approach taken for this project is to use fundamental properties needed for mixture and structural design for the control and acceptance of flexible pavements and HMA overlays. The NDT technologies included in the field evaluation were evaluated for their ability to determine these properties accurately on a practical and effective QA program. Basically, the NDT technology or QA tests are used to confirm the design assumptions for the materials placed. Chapter 5 presented the test results measured using each technology, and chapter 7 identified those devices that were able to identify or discriminate areas with different material properties or conditions. This chapter presents the evaluation of the NDT devices recommended for further use (chapters 5 and 7) to determine the quality of the unbound and HMA mixtures placed on some of the projects discussed in Appendix B. These devices include the GeoGauge for unbound materials and the PSPA for HMA mixtures. Other devices can also be used, such as the DCP, but these were not as successful in identifying anomalies. In addition, the intent of this chapter is to show the use of NDT devices that estimate modulus for defining construction quality. 8.1 Quality Control and Acceptance Application As stated earlier in the report, of the many process control procedures that can be used in highway construction, process control charts, particularly statistical control charts, are most commonly used by contractors and material producers for verifying that their process is under control. Although there are different approaches that can be taken in implementing NDT technologies to verify that the process is in control, statistical control charts are being used within this project. As a result, the NDT test methods must produce results that can be adapted to existing AASHTO procedures in pavement construction. The ASTM Manual on Presentation of Data and Control Chart Analysis was used for preparing practical procedures that contractors can use in deciding whether their process is in control (ASTM, 1992). Similarly, there are different acceptance procedures that are used in judging whether the pavement material meets the required specifications. Two of the more common methods that have been used and adopted by most agencies are PWL and AAD. PWL is the procedure used by over 75 percent of the agencies that have adopted statistical-based acceptance specifications. As a result, AASHTO R9 entitled Acceptance Sampling Plans for Highway Construction was used for preparing practical but effective procedures that agencies can use in deciding whether the product meets their specifications (AASHTO, 2003). In summary, statistical control charts are the primary method for determining whether the construction is in-control or out-of-control, and PWL is the primary method for judging the acceptability of construction. To demonstrate the use of the NDT technology for use in QA program, specific projects were selected to cover the range of conditions encountered during 309

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report construction. The following lists the steps recommended for use in setting up a QA program for using NDT devices to judge the quality of construction of unbound materials and HMA mixtures using the material modulus. Unbound Materials HMA Mixtures 1. Develop M-D relationships in the laboratory prior to construction for the unbound material to determine the maximum dry unit weight. Select the target density and water content for compacting the unbound layer. 1. Conduct an HMA mixture design to determine the target gradation and asphalt content. Select the target density and job mix formula for the project mixture or lift being tested. The target job mix formula will likely be revised based on plant produced and placed material. 2. Prepare and compact test specimens at the average water content and dry density expected during construction; based on the project specifications. 2. Prepare and compact test specimens at the target asphalt content and the average density expected during construction; based on the project specifications. 3. Measure the repeated load resilient modulus in accordance with the agency’s procedure (AASHTO T307 or NCHRP 1-28A, as required by the MEPDG). Determine the resilient modulus at a selected stress state. The resilient modulus should equal or exceed the value used during design. If the agency does not have a resilient modulus testing capability, the FHWA-LTPP regression equations can be used to estimate the target value, until the laboratory resilient modulus test has been completed (see equations 34 to 48). 3. Measure the dynamic modulus in accordance with the agency’s procedure or the test protocol in accordance with the MEPDG. Determine the dynamic modulus for the test temperature expected during acceptance testing. Two values should be extracted from the test results or master curve; one for the day of paving (an elevated temperature expected after compaction) and the other for one or multiple days following placement. This target value for one or more days following placement will need to be adjusted back to a standard temperature depending on the actual pavement temperature. 4. Define the adjustment factor or ratio for the unbound material to laboratory conditions. Low stress states were used in establishing the ratios for this project. 4. Define the adjustment factor for the HMA mixtures to laboratory conditions. A load frequency of 5 Hz was used in establishing the adjustment ratios for this project. 5. Determine the combined or pooled standard deviation of the modulus for setting up the control limits of the unbound layer for the contractor (see section 8.3). Establish the action, as well as warning, limits for the statistical control charts; upper and lower control limits (see section 8.2. 5. Determine the combined or pooled standard deviation of the seismic modulus for setting up the control limits of the HMA mixture for the contractor (see section 8.3). Establish the action, as well as warning limits for the statistical control charts; upper and lower control limits (see section 8.2). 6. Determine the upper and lower specification limits (see section 8.3) for the resilient modulus of the unbound material. This includes the upper and lower specification limits for the resilient modulus of the unbound layer. 6. Determine the upper and lower specification limits (see section 8.3) for the dynamic modulus of the HMA mixture. This includes the upper and lower specification limits for the dynamic modulus of the HMA mixture. 7. Prepare the statistical control charts 7. Prepare the statistical control charts. 8. Determine the PWL criteria for different conditions. 8. Determine the PWL criteria for different conditions. 8.2 Control Limits for Statistical Control Charts The upper and lower control or action limits are calculated from the NDT modulus tests in accordance with the following equations. ( )( )sAXUCLX 3+= ...................................................................................................... (49) ( )( )sAXLCLX 3−= ...................................................................................................... (50) 310

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report Where: XUCL = Upper control limit for the sample means. XLCL = Lower control limit for the sample means. X = Target value for a project. s = Pooled standard deviation that represents the process variance. The target value of the control chart for each material is the average value measured in the laboratory in accordance with AASHTO T 307 or the test protocol used by the agency. Both action and warning limits are normally included on the statistical control charts. The upper and lower action limits are set at three standard deviations from the target value, while the warning limits are set at two standard deviations from the target. 8.2.1 Target Modulus or Critical Value As noted above, the target value of the control chart for each material and project is the modulus measured in the laboratory. This average laboratory value should be the same as the input to the MEPDG for structural design. Tables 77 and 78 list the target values for the unbound and HMA layers included in the field evaluation projects, respectively. The repeated load resilient modulus (low stress state) and dynamic modulus (5 Hz load frequency) test results are provided in Appendix B and summarized in chapter 5. Table 77. Parameters Used to Prepare Statistical Control Charts for the Unbound Layers Included in the Field Evaluation Projects Action Warning Limits, ksi Project Identification Material Target Modulus, ksi Pooled Standard Deviation, ksi UCL LCL I-85, AL Low Plasticity Clay 4.0 0.8 5.6 2.4 NCAT, OK High Plasticity Clay 6.9 2.0 10.8 3.00 SH-21, TX High Plasticity Clay 26.8 2.5 30.4 23.2 TH-23, MN Soil-Aggregate Embankment 16.4 1.0 17.8 15.0 US-2, ND Soil-Aggregate Embankment 19.0 2.6 22.7 15.3 SH-130, TX Improved Soil Embankment 35.3 2.8 39.3 31.3 NCAT, SC Crushed Granite Base 36.1 2.7 41.4 30.8 NCAT, MO Crushed Limestone Base 19.2 2.7 24.5 13.9 TH-23, MN Crushed Stone Base 24.0 2.6 27.7 20.3 US-53, OH Crushed Stone Base 27.5 1.6 30.6 24.4 NCAT, FL Limerock Base 28.6 3.5 35.4 25.5 US-2, ND Crushed Aggregate Base 32.4 4.5 38.8 26.0 US-280, AL Crushed Limestone Base 48.4 10.0 62.7 33.7 NOTE: The target modulus for the South Carolina crushed granite base was determined using the FHWA-LTPP regression equation, because the densities were significantly below the maximum dry unit weight of the material during NDT testing. The pooled standard deviation for this project was assumed to be equal to the Missouri limestone base because the same contractor placed both materials. 311

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report Table 78. Parameters Used to Prepare Statistical Control Charts for the HMA Layers Included in the Field Evaluation Projects Action Warning Limits, ksi Project Identification Material Target Modulus, ksi Pooled Standard Deviation, ksi UCL LCL I-85, AL SMA 250 14 270 230 TH-23, MN HMA Base 810 35 860 760 US-280, AL HMA Base 650 45 715 585 I-35, TX HMA Base 800 57 910 690 I-75, MI Type 3-C 400 86 520 280 I-75, MI Type E-10 590 86 715 465 US-47, MO Surface Mix 530 60 615 445 US-47, MO Base Mix 420 36 470 370 I-20, TX CMHB Base 340 40 420 260 US-53, OH HMA Base 850 44 915 785 US-2, ND HMA Base 510 33 555 465 NCAT, SC HMA Base 410 58 525 295 NCAT, FL HMA Base 390 40 470 310 NCAT, FL PMA Base 590 45 675 505 NCAT, AL PG76-Sasobit 610 40 690 530 NCAT, AL PG76-SBS 640 45 725 555 NCAT, AL HMA Base 450 50 550 350 NOTE: The Texas SH-130 target modulus was determined from Witczak’s regression equation because changes were made to mixture just prior to NDT testing. 8.2.2 Combined or Pooled Standard Deviation The pooled standard deviation was calculated in accordance with the AASHTO R 9-03, Standard Recommended Practice for Acceptance Sampling Plans for Highway Construction. The pooled standard deviation was determined for each project and unbound material using the NDT results for the areas without anomalies or physical differences included at the end chapter 5. The pooled standard deviations for each project and material are listed in Tables 77 and 78 for the unbound and HMA layers, respectively. These values were used to determine whether the projects were in-control or out-of-control, using the action limits or upper control limit (UCL) and lower control limit (LCL) provided in Tables 77 and 78. 8.3 Parameters for Determining PWL 8.3.1 Determining Quality Indices The upper and lower quality indices are calculated in accordance with equations 51 and 52, respectively. The upper and lower specification limits were determined using data from all projects with similar materials. s LSLXQL −= .............................................................................................................. (51) 312

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report s XUSLQL −= .............................................................................................................. (52) Where: = Lower quality index. LQ = Upper quality index. UQ USL = Upper specification limit. LSL = Lower specification limit. s = Sample standard deviation of the lot. X = Sample mean of a lot. The upper and lower quality indices are used to determine the total PWL for each lot of material using equation 53. The upper and lower PWL values are then determined from the Q-tables provided in the AASHTO QC/QA Guide Specification. ....................................................................................... (53) 100−+= UL PWLPWLPWL Where: PWL = Percent Within Limits. PWLL = Percent Within Limits from the lower specification limit. PWLU = Percent Within Limits from the upper specification limit. 8.3.2 Determining Specification Limits Tables 79 and 80 list the target values for the unbound and HMA layers included in the field evaluation projects, respectively. These values were used to determine the PWL for the different materials used in the field evaluation projects and compared to the control limits determined for each project. 313

NCHRP Project 10-65—Volume 2: Research Report June 2008 Part III—Data Interpretation and Application Final Report 314 Table 79. Upper and Lower Specification Limits for the Unbound Layers and Materials Included in the Field Evaluation Projects Project Identification Material Median Standard Deviation, ksi Specification Tolerance, (-) ksi I-85, AL Low Plasticity Clay NCAT, OK High Plasticity Clay SH-21, TX High Plasticity Clay 2.0 3.3 TH-23, MN Soil-Aggregate Embankment US-2, ND Soil-Aggregate Embankment SH-130, TX Improved Soil Embankment 2.1 3.5 NCAT, SC Crushed Granite Base NCAT, MO Crushed Limestone Base TH-23, MN Crushed Stone Base US-53, OH Crushed Stone Base NCAT, FL Limerock Base US-2, ND Crushed Aggregate Base US-280, AL Crushed Limestone Base 3.0 5.0 Table 80. Upper and Lower Specification Limits for the HMA Layers and Mixtures Included in the Field Evaluation Projects Project Identification Material Median Standard Deviation, ksi Specification Tolerance, + ksi I-85, AL SMA 15 30 TH-23, MN HMA Base US-280, AL HMA Base I-35, TX HMA Base I-75, MI Type 3-C 50 100 I-75, MI Type E-10 US-47, MO Surface Mix 70 140 US-47, MO Base Mix I-20, TX CMHB Base US-53, OH HMA Base US-2, ND HMA Base NCAT, SC HMA Base NCAT, FL HMA Base NCAT, FL PMA Base 50 100 NCAT, AL PG76-Sasobit NCAT, AL PG76-SBS 45 90 NCAT, AL HMA Base 50 100