Non-Nuclear Methods for HMA Density MeasurementsMBTC 2075Final ReportbyStacy G. Williams, Ph.D., P.E.Research Assistant ProfessorDepartment of Civil EngineeringUniversity of Arkansas700 Research Center Blvd.Fayetteville, AR 72701479-575-2220479-575-7639 (FAX)[email protected] 2008

ABSTRACTNon-nuclear methods for the measurement of hot-mix asphalt (HMA) density offer theability to take numerous density readings in a very short period of time, without the needfor intensive licensing, training, and maintenance efforts common to nuclear gauges.The Pavement Quality Indicator (PQI) and the PaveTracker use electricalimpedance to estimate density. Early models of these gauges were deemed inadequatefor quality control and quality assurance testing, but improvements have been made toeach.In this project, a number of field sites were used to evaluate the non-nuclear gauges interms of ruggedness, accuracy, and precision. A thorough investigation of calibrationmethods was also performed.In the ruggedness study, three pavement sites were used to determine potentialprocedural factors that significantly affected the non-nuclear density results. Moisture,the presence of sand or debris, gauge orientation, gauge type, and presence of paintmarkings were determined to significantly impact the accuracy of non-nuclear gaugereadings.Four calibration methods were investigated, including screed offset, core offset, twopoint, and data pair techniques. None were found to possess all of the necessarycomponents for generating significant correlations with field core densities. A screedcore method was developed as a method to more comprehensively adjust themagnitude of the offset as well as the sensitivity of the device over a large range of truedensities.Overall, neither non-nuclear gauge was able to predict core densities as accurately orprecisely as the nuclear gauge. Of the non-nuclear devices, the PQI generated moreconsistent results but was less sensitive to actual changes in density. The PaveTrackerwas more sensitive to actual changes in density, but exhibited a higher level ofvariability. Existing specifications for use of non-nuclear devices should be edited toinclude guidance on gauge orientation during testing, as well as calibration proceduresfor a screed-slope type of technique.

ACKNOWLEDGMENTSThe Mack-Blackwell Transportation Center (MBTC) and the Arkansas State Highwayand Transportation Department (AHTD) are gratefully acknowledged for their supportand sponsorship of this research project. The author also wishes to express sincereappreciation for the assistance provided by APAC-Arkansas, Inc., Kiewit Southern, Inc.,and Rogers Group, Inc. The efforts of Artem Popov, Roselie Conley, Mary Fleck, andthe personnel of AHTD RE#43 are also recognized.The contents of this paper reflect the views of the author who is responsible for the factsand accuracy of the data presented herein. The contents do not necessarily reflect theofficial views or policies of the Federal Highway Administration or the Arkansas StateHighway and Transportation Department. This paper does not constitute a standard,specification, or regulation, and should not be considered an endorsement of anycommercial product or service.

TABLE OF CONTENTSIntroduction . 1Background . 2Core Method . 2Nuclear Method . 3Non-Nuclear Methods . 3Calibration . 6Cost . 11Literature Review . 12Analysis and Discussion . 21Ruggedness Study . 21Phase 1 . 21Phase 1 Analysis . 24Phase 1 Conclusions. 28Phase 2 . 30Phase 2 Conclusions. 36Influence of Surface Moisture . 37Effect of Paint . 39Method Comparisons . 40Calibration . 48Evaluation of Methods . 48Results and Discussion . 49Variability . 63Conclusions . 63Screed-Core Calibration . 65Variability and Sample Size. 68Practicality . 68Conclusions . 68Conclusions and Recommendations. 69Ruggedness . 69Method Comparisons . 69Calibration . 70Recommendations . 70References . 71

INTRODUCTIONIn-place density is a key indicator used to judge the quality of hot-mix asphalt (HMA)pavements. Traditionally, this property has been measured by determining the densityof cores cut from the compacted pavement, or by the use of a nuclear gauge. Coredensities are typically believed to provide the most accurate results, but this process isdestructive to the newly compacted pavement. Nuclear technology offers a nondestructive method for density measurements, but is burdened with intense regulationsassociated with the handling, storage, and transportation of radioactive materials.Within the last decade, non-nuclear technology has been developed for the purpose ofmeasuring the density of in-place HMA materials. These devices operate based on theprinciples of electrical impedance for a current passing through the HMA material.These devices have many advantages in that they are capable of providing densitymeasurements very quickly and in a completely non-destructive manner, are easy tohandle, and are not subject to complicated regulations.In order to adopt a new test method, certain advantages must be realized. The accuracyof the new method should be equivalent to, or better than existing technology. Otherjustifications relate to efficiency, in terms of time, effort, and money. There are a numberof practical reasons to move toward the non-nuclear technology, but the method mustfirst be proven to perform adequately.In this project, the use of two non-nuclear gauges, the Pavement QualityIndicator (PQI) and the PaveTracker were investigated in order to determine whethernon-nuclear technology was appropriate for use in quality control and quality assurance(QC/QA) applications. A ruggedness study was performed in order to determine theeffects of a number of factors on the results obtained by the non-nuclear devicesincluding temperature, moisture, gauge type, gauge orientation, and presence of debris.The accuracy and precision of the gauges were assessed by comparisons withtraditional methods of density measurement, including field cores and the nuclear gauge.In addition, a thorough consideration of calibration procedures was conducted, andsuggestions were made for incorporation into existing specifications.MBTC 20751

BACKGROUNDIn-place density of asphalt pavement is a vital property which can indicate the long-termperformance of a flexible pavement. It is also a primary characteristic used to measurequality during construction. Traditionally, the in-place density of HMA pavements wasmeasured from core samples cut from the pavement after compaction. While thismethod offered a measure of density that was believed to be accurate, the process wastime-consuming, labor-intensive, and destructive to the pavement. Nuclear technologywas later developed as a non-destructive alternative for density determinations. Thisadvancement was significant because a nuclear density measurement could becompleted in less than five minutes, which provided the contractor with reasonablyaccurate information for “real-time” quality control. The greatest disadvantage of thenuclear device was that it contained radioactive materials, which required significantefforts relating to training, licensing, calibration, maintenance, handling, storage, andtransportation. During the last decade, non-nuclear technology has been developed,which uses the impedance of an electrical current to measure dielectric constant andestimate pavement density. These devices do not require intensive safety procedures,are lightweight and easy to handle, and provide density measurements within a fewseconds. Early models of these devices demonstrated poor correlations with traditionaldensity measurements, and were significantly affected by factors such as temperatureand moisture.Core MethodDensity from field cores is determined by cutting the core from the compacted pavement,then measuring bulk specific gravity the saturated-surface-dry method specified inAASHTO T166 or similar, as shown in Figure 1. (1) The bulk specific gravity is thendivided by the maximum theoretical density (MTD) of the mix and density is expressedas a percentage of MTD. This measure of density has traditionally been accepted as thebest available estimate of “true” density. Concerns have been expressed regarding theaccuracy of this method, especially for coarse-graded and large NMAS mixes; however,this test is still the most commonly specified method for QA measures of field density.Figure 1. AASHTO T 166MBTC 20752

Nuclear MethodNuclear density gauges measure density by emitting gamma rays from a Cesiumsource. These rays pass through the compacted material to detectors. For a denselycompacted material, the gamma rays do not easily pass through to the detector,resulting in a low number of counts. Lower density materials allow the gamma rays topass through to the detectors more readily, resulting in a higher number of counts. Thedensity of the mat is inversely proportional to the counts, and can be expressed as a unitweight or as a percentage of MTD. Standard procedures for tests performed on HMApavements are outlined in ASTM D 2950. (2) A Troxler Model 3430 nuclear gauge isshown in Figure 2. Although the nuclear method is a relatively quick and non-destructivemethod for obtaining field densities, poor correlations between nuclear and coredensities have been documented. (3, 4, 5) As a result, most states require field coresfor QA purposes.Figure 2. Troxler Model 3430 Nuclear GaugeNon-Nuclear MethodsNon-nuclear gauges estimate density by measuring the change in electromagnetic fieldwhen an electrical current is transmitted through an asphalt pavement. Specifically, anelectrical current passes from the transmitter, is forced around an isolation ring, throughthe pavement, and is detected by the receiver. The impedance, or resistance toelectrical flow, is measured and used to determine the dielectric constant. (6) Aschematic of this process is shown in Figure 3.MBTC 20753

Figure 3. Schematic of Non-nuclear Gauge FunctionThe overall dielectric constant of an asphalt pavement is directly proportional to itsdensity. (6) The dielectric constant of air is approximately 1.0, and that for aggregateand asphalt cement is in the range of 5 to 6. Because air has a smaller dielectricconstant than the other HMA components, higher air void levels (i.e., lower densities)are indicated by a lower overall dielectric constant. Water has a very high dielectricconstant (approximately 80), and small amounts of water can significantly impact themeasured overall dielectric constant, such that densities estimated by dielectric constantare higher when water is present. Because the dielectric constant of air is less than thatof HMA, and that of water is greater than HMA, the presence of a small amount of watercan have the same numerical effect as a large decrease in air void spaces. (7)The first impedance-type density gauge, the Pavement Quality Indicator (PQI), wasdeveloped by TransTech, Inc. in 1998 as part of the NCHRP-IDEA Program. (6, 8, 9, 10)The earliest model of the device, the PQI 100, did not have a moisture sensor, whichcreated obvious problems with accuracy. This gauge did not correlate well with othermeasures of density and was not recommended for use. The next model (PQI 200) wascapable of measuring moisture and temperature, and improved accuracy was reported.Subsequent improvements have been made to the PQI since that time, resulting in thePQI 300, PQI 301, PQI 302, and PQI 303. The PQI 301 is shown in Figure 4. Additionsto the device have included settings for layer type, depth settings for layer thickness, andupdated algorithms. (11)MBTC 20754

Figure 4. Pavement Quality Indicator (PQI) Model 301The PaveTracker is another non-nuclear density gauge, which was developed in 2000and is currently marketed by Troxler Electronic Laboratories, Inc. (8, 12) This device isbased on the same principles as the PQI, and is shown in Figure 5.Figure 5. Troxler PaveTracker Model 2701-BNon-nuclear gauges have primarily been used for determining mat density duringconstruction, and results have been varied. (6, 8, 13, 14) Additional uses for thesegauges have also been investigated, including density profiling, longitudinal joint densitytesting, and the quantification of segregation in compacted pavements. (15, 16, 17)MBTC 20755

CalibrationA critical step in using impedance gauges effectively is to calibrate them in a mannerthat will increase the accuracy of results. Density measurements are relative measuresof compaction, and can thus be adjusted mathematically in order to more accuratelyrepresent the “true” density of the pavement. Although an “absolutely true” measure ofpavement density cannot be reasonably achieved, the most accurate measures aretypically believed to result from the bulk specific gravity measurement of a pavementcore cut from the compacted mat. Therefore, an alternative measure of density isbelieved to be accurate if it can produce results similar to those generated by coredensities.A number of methods are available for use in calibrating impedance gauges. TheAASHTO TP 68 method, “Density of In-Place Hot-Mix Asphalt (HMA) Pavement byElectronic Surface Contact Devices”, outlines three such methods. (1) The first is arelative method, which is recommended primarily for establishing rolling patterns duringfield compaction. In this method, the density of the mat is recorded after each of aseries of roller passes. This process continues until the density no longer increases, andthe number of roller passes is recorded. The second method is a screed calibration, inwhich the density behind the screed is estimated and used to generate an offset for themix. The density behind the screed is typically in the range of 75 to 85 percent, and theactual value is usually based on operator experience. The third method involves a corecalibration, and is the method recommended by AASHTO. In this procedure, one to fivelocations are chosen and impedance gauge readings are taken at each location. Theoffset is calculated based on the average differences in the gauge and core densities.A method for non-nuclear density measurements has also been published as ASTM D7113, “Standard Test Method for Density of Bituminous Paving Mixtures in Place by theElectromagnetic Surface Contact Methods”. (18) This specification acknowledges that anumber of calibration methods are available, but recommends a core calibration methodin which three to ten locations are selected. A minimum of four non-nuclear gaugereadings are taken at each location and compared to corresponding core densities. Theaverage of the differences serves as the calibration offset.Manufacturer’s instructions for each gauge also provide insight to appropriate calibrationprocedures. TransTech recommends a core calibration procedure using five corelocations and five gauge readings at each location. (19) A one point method is alsodescribed, which utilizes an offset based on the estimated density of the matimmediately behind the screed. An offset is determined based on the difference in theestimated screed density and the average of five gauge readings. In addition, a twopoint method is described, which utilizes an estimate of density behind the screed aswell as an estimate of density after the mat has been ‘peaked’ (i.e., the density no longerincreases with additional roller passes). TransTech suggests that the density of an HMApavement behind the screed is approximately 82 percent, and that the typical density ofa peaked mat is approximately 95 percent. Based on linear modeling, the estimated andmeasured values are used to generate slope and intercept calibration constants.Troxler recommends either a density offset method or a mix calibration method. (12) Forthe offset method, a series of readings can be taken with the non-nuclear gauge andcompared to values generated by some other method, such as core densities, nucleargauge densities, or gyratory-compacted core densities. The average difference is takento be the offset. The mix calibration method involves taking pairs of density readings atMBTC 20756

a series of three to ten locations such that the range of density is at least 3 pcf. Thenon-nuclear gauge densities are paired with density readings from another method (inwhich the other method is assumed to provide ‘true’ results) such as core densities ornuclear gauge density readings, and linear modeling is employed as a means togenerate slope and intercept calibration constants.A calibration method seeks to reconcile the differences between two measures of thesame property. In graphical form, if two measures are plotted against each other andagree perfectly, then the resulting relationship will follow the line of equality, which is astraight line having an intercept at the origin and a slope of 1 (see Figure 6). In mostcases, perfect agreement is not present, and coefficients of regression (i.e., slope andintercept values) are used to transform measurements by one method to an equivalentmeasure by the second method.Relationship of Two Variables25Line of Equality20Y1510500510152025XFigure 6. Relationship of Two Variables Having Perfect AgreementMBTC 20757

An offset method of calibration is used to shift data vertically in order to create a datasetthat most closely follows the line of equality. In Figure 7, the relationship of the originaldata is shown, as well as the corrected data after an offset calibration has been applied.In order to correct data using an offset calibration, a constant is added to the originaldata value.Relationship of Two VariablesUsing an Offset Calibration25Line of e 7. Relationship of Two Variables Using an Offset CalibrationMBTC 20758

A slope calibration is used to change the slope of a relationship to create a dataset thatmost closely follows the line of equality. In essence, the slope calibration will “twist” therelationship about the origin. In Figure 8, the relationship of the original data is shown,as well as the data after a slope calibration has been applied. In order to correct datausing a slope calibration, the original data value is multiplied by a constant.Relationship of Two VariablesUsing a Slope Calibration25Line of e 8. Relationship of Two Variables Using a Slope CalibrationMBTC 20759

To account for errors in the vertical position and slope of the relationship concurrently, aslope-intercept calibration should be used. This type of calibration method will “twist” thedata to match the line of equality, and then apply an offset, or intercept, to shift the datavertically toward the line of equality. By including the both the slope factor and verticaloffset, the data appears to be twisted about a point other than the origin. In Figure 9, therelationship of the original data is shown, as well as the data after a slope-interceptcalibration has been applied. In order to correct data using a slope-intercept calibration,the data value is multiplied by a factor (slope) and then a constant (offset) is added tothe result. The slope-intercept is the most difficult method to compute, but is usuallyable to provide a more complete adjustment to a dataset.Relationship of Two VariablesUsing a Slope-Intercept Calibration25Line of e 9. Relationship of Two Variables Using a Slope-Intercept CalibrationMBTC 207510

CostThe costs associated with performing field density tests vary according to method. Theterm “cost” can include money, time, or effort, but these factors are typically assessed bysome equivalent monetary value. In general, the non-nuclear devices have beenadvertised to provide significant savings as compared to the nuclear gauge. The initialcost of the non-nuclear devices is similar or slightly less than a nuclear gauge(depending on the model), but the majority of the savings is generated through theelimination of costs associated with licensing, training, and maintenance. One sourcereported that the annual operating cost of the PQI was 210 per year, as compared to 3075 per year for the nuclear gauge. (7) Another source reported that over a 5 yearperiod, the non-nuclear devices could save as much as 50318. (20)MBTC 207511

LITERATURE REVIEWRogge and JacksonA number of studies have been performed to determine the ability of the non-nucleardevices to accurately and precisely measure in-place density of asphalt pavements.One of the earliest was performed in Oregon in 1999. (21) In this study, a Humboldtnuclear gauge and the original PQI model were compared to field cores in order toassess the compaction and field density for open-graded asphalt pavements having anominal maximum aggregate size (NMAS) of 25mm and a typical air void range of 17 to26 percent. Six projects were tested such that nuclear and non-nuclear densities weremeasured in 45 locations for each. Cores were also cut in each location, resulting in 270cores and corresponding density measurements. Although a large amount of data wascollected, it was reported that neither nuclear nor non-nuclear densities correlated wellwith core densities, and neither method was determined to be adequate for controllingfield compaction.Sully-MillerIn 2000, the Sully-Miller Contracting Company reported on a study in which the PQI wascompared to a Troxler 3440 nuclear gauge. The variability was compared for the twogauges, and standard deviations of 0.95, 0.79, and 0.84 were reported for the PQI, andstandard deviations of 1.51, 2.12, and 0.90 were reported for the nuclear gauge. Theresults of the nuclear gauge were widely varied, and it was concluded that these effectswere due to surface texture. The PQI, however, did not appear to be affected by surfacetexture. Overall, it was determined that the PQI was reliable and accurate for measuringthe in-place density of compacted HMA.HenaultIn another early study, the PQI 300 was evaluated in conjunction with the nuclear gaugein thin-lift mode, and the two were compared to field core densities with the intention ofdetermining whether the PQI could be used for quality assurance (QA) testing. (6) Tensites were tested, and a 5-core offset method was used to calibrate the PQI.Correlations between the PQI and core densities were poor, averaging 0.28, which wassuspected to be due to moisture from the roller. Correlations between the nuclear gaugeand core densities were better, but not good, having an average R2 value of 0.55.Additionally, PQI densities did not correlate well with the nuclear gauge. Overall, the QAtesting with the PQI was not recommended.Pooled Fund StudyBy 2002, the results of several non-nuclear gauge studies were published, including apooled fund study. (8) In this study, laboratory and field evaluations of the PQI 300 wereperformed. The laboratory study sought to determine the effects of several factors,including density, NMAS, aggregate source, temperature, and moisture.Threeaggregate sources and three aggregate sizes were used to compact slabs in the linearkneading compactor at varying densities. The results indicated that PQI readings weresensitive to temperature and moisture, even though improvements to the device hadbeen made in an effort to combat these effects. Small amounts of moisture were notsignificant provided the moisture level remained fairly constant, and NMAS was notsignificant. The PQI 300 was recommended for use to indicate changes in density underconstant temperature and humidity conditions, as long as a mixture specific calibrationwas used. A slope and intercept method of calibration was recommended.MBTC 207512

As part of the pooled fund study, two field evaluations were performed. The first tookplace during the 2000 construction season, in which a nuclear gauge and PQI wereevaluated with respect to field cores. It was found that nuclear gauge density readingsprovided the stronger correlation to field core densities. Although this relationship wasmerely fair, it was suggested that PQI readings were probably as reasonable as nucleargauge readings, which are currently accepted by industry, and many practical reasonsfor using the PQI were cited. Due to poor correlations, the PQI was not yetrecommended for field use; however, updated algorithms were recommended to furtherimprove the device.A second field study was held in the 2001 construction season, with five statesparticipating. The PaveTracker, marketed by Troxler, was available at this time and wasincluded in the study. In Pennsylvania, the PQI 300 (improved) demonstrated thebetter performance, providing density values similar to that of cores for 6 of 9 projects.The PaveTracker demonstrated significant similarities to field cores in just 3 of 6projects. The Pennsylvania project was regarded as highly successful, and part of thesuccess was attributed to the experience of the technicians who made the decision touse a calibration based on an assumed density behind the screed of 87 percent.In contrast, testing performed in Maryland indicated that the PQI was unable tosufficiently correlate with field cores in two of two projects. The PaveTracker was testedon three projects, and was able to successfully correlate with field cores in two of thethree.Minnesota also reported greater success with the PaveTracker, whichdemonstrated high correlation with field cores for 4 of 7 projects and no poorcorrelations. The PQI performed poorly in 2 of 5 projects, and very well in 1 of the 5projects. Similar results were generated for one project in Oregon in which thePaveTracker demonstrated a much stronger correlation to core densities than the PQI.In New York, similar results were achieved for the PQI and PaveTracker devices.Overall, the PaveTracker demonstrated a somewhat better correlation to core densitiesthan did the PQI, but neither performed as well as the nuclear gauge. It was noted thatthe PQI did not appear to be sensitive to actual changes in density in spite of its recentimprovements. The non-nuclear devices were recommended for QC or supplementaltesting, but not for QA testing.RomeroBecause several groups were analyzing the differences in non-nuclear densitymeasurements, Romero offered guidance on the most appropriate ways to statisticallyanalyze a dataset of this type. (22) T-tests were demonstrated to be inappropriate forcomparing non-nuclear density data because the t-test can be misleading for highlyvariable data. The correlation coefficient was cited as the most accurate way tocompare test methods because it provides an idea of the sensitivity of each parameter;however, small sample sizes can be detrimental to the reliability of thi

(2) A Troxler Model 3430 nuclear gauge is shown in Figure 2. Although the nuclear method is a relatively quick and non-destructive method for obtaining field densities, poor correlations between nuclear and core densities have been documented. (3, 4, 5) As a