Intelligent Soil Compaction Systems (2010)

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.

0 Quality assurance of Pavement Earthwork Using Roller-integrated Continuous Compaction Control (Recommended specification options) 4 into viable performance-based specifications that can be ef- ficiently implemented in practice. Each option can be adopted as the sole method for QA. Alternatively, two or more options can be combined to increase reliability. Uniformity criteria are discussed in Section 7.9 and can be added to any option. The remainder of Chapter 7 is presented in the form of a specification and can be treated as a stand-alone document. The proposed specification provides a discussion on appro- priate applications for the recommended options; defines im- portant CCC-related terms; provides recommendations for roller measurement system requirements, including proce- dures to assess the validity of roller MVs; and concisely pres- ents several important issues related to roller-integrated mea- surement. Chapter 8 presents several case studies illustrating how the various options are implemented and highlights the level and extent of analysis required. c h a p t e r 7 This chapter presents recommended specification options for QA of subgrade, subbase, and aggregate base course com- paction using roller-integrated CCC. Six viable QA options are proposed to accommodate the diverse site conditions and agency needs observed across the United States. Many of the recommended options were inspired by current European specifications (summarized in Chapter 2). Additional recom- mended options stem from the research findings presented in this report. The proposed specifications are technically end product based with required methodological aspects. Tradi- tional method-based approaches (e.g., recording the num- ber of roller passes) do not utilize the roller measurement values (MVs) per se and are therefore not addressed by this specification. None of the recommended options constitute performance-based specifications. Further research is re- quired to implement, for example, the findings from Chapter

0  COnTEnTS 106 7.1 Scope 106 7.2 Definitions 108 7.3 Notation 108 7.4 Important Considerations 110 7.5 Instrumented Roller Requirements 112 7.6 QA Option 1: Spot Testing of Roller-Informed Weakest Area(s) 113 7.7 QA Option 2: Limiting Percentage Change In Roller MV 114 7.8 QA Option 3: Comparison of Roller MV Data to Target MV 117 7.9 Uniformity Criteria

0 7.1 Scope This specification covers the QA of subgrade, subbase, and aggregate base course compaction using roller-integrated CCC. The six recommended QA specification options are summarized in Table 7.1 and Figure 7.1. Options are num- bered 1, 2a, 2b, 3a, 3b, and 3c and are distinguished by three principal categories. In Option 1, CCC is used to assist in QA and acceptance is based on spot-test measurements. Option 2a and 2b acceptance is based on roller MVs, though initial calibration of roller MV to spot-test measurements is not re- quired. Option 3a, 3b, and 3c acceptance is based on roller MVs and requires initial calibration to determine an MV-TV. Each option can be adopted as the sole method for QA; alter- natively, two or more options can be combined to increase reliability. Uniformity criteria discussed in Section 7.9 can be added to any option. Method-based approaches (e.g., using GPS positioning and documentation to record the number of passes) do not utilize roller-based MVs and are therefore not addressed by this specification. However, the implementation of such an approach is straightforward. 7.2 Definitions Automatic Feedback Control: automatic adjustment of roller Operating Parameters such as vibration frequency and am- plitude based on real-time feedback from measurement system. Calibration Area: an area representative of an Evaluation Section but typically smaller and used to establish an MV-TV. Compaction Pass: a static or vibratory roller pass performed during earthwork compaction, not necessarily employing an Instrumented Roller. Continuous Compaction Control (CCC): continuous monitor- ing and documentation of earthwork compaction using an Instrumented Roller. Evaluation Section: an area of earthwork with consistent properties where acceptance is evaluated. Instrumented Roller: a roller compactor outfitted with drum vibration instrumentation or other means to compute a Roller Measurement Value, onboard computer, and posi- tion monitoring equipment. table 7.1. Summary of specification options. Roller-Integrated CCC QA Option Target Measurement Value (MV-TV) Acceptance Criteria Option1: Spot testing of roller-informed weakest area(s) Not required Spot-test measurements in roller-identified weakest area(s) satisfy contract spot-test measurement requirements (QA-TV) Option 2a: Monitoring percentage change in mean MV Not required Achieving ≤ 5% change in mean MV between consecutive roller passes Option 2b: Monitoring spatial percentage change (%∆) in MVs Not required Achieving the %∆-TV between consecutive passes over a defined percentage of an evaluation section Option 3a: Empirically relating MVs to spot-test measurements Based on correlation of MV to spot test measurement: MV-TV = MV corresponding to contract QA-TVa Achieving MV-TV over a set percentage of an evaluation section Option 3b: Compaction curve based on MVs MV-TV = mean MV when the increase in pass-to-pass mean MV in a calibration area ≤ 5% Option 3c: Empirically relating MVs to lab-determined properties (e.g., M r ) Based on correlation of MV to lab soil property: MV-TV = MV corresponding to contract QA-TVb a Assumption is that QA-TV is spot-test-based measurement of density, modulus, etc. b For example, a QA-TV based on M r .

0  Figure 7.1. Summary of CCC specification options.

0 Intelligent Compaction: the combined use of an Instrumented Roller and Automatic Feedback Control in an attempt to improve earthwork compaction. Layer: a component of the pavement earthwork with distinct soil properties (e.g., subgrade, subbase, or base course). Lift: a unit of material within a Layer that is deposited at one time for compaction. A Layer may be comprised of a single lift or multiple lifts. Measurement Depth: the soil depth to which Roller Measure- ment Values or Spot-Test Measurements are representative. Measurement Pass: a pass performed by an Instrumented Roller during which all required information, including Roller Measurement Values and machine position, are recorded. Roller Operating Parameters must be held constant, and thus no Automatic Feedback Control is permitted during a Measurement Pass. MV Reporting Rate: the time-dependent rate at which new Roller Measurement Values are reported. MV-TV: a target Roller Measurement Value (e.g., the measure- ment value corresponding to a QA-TV). Operating Parameters: roller machine parameters used dur- ing operation, including forward speed, driving direction, vibration frequency, and eccentric force amplitude. Pass Sequence: a record of the roller pass history (pass num- ber, Operating Parameters) over a specified area. Quality Assurance (QA): evaluation methods and procedures administered by the owner or owner’s representative to ensure that the constructed earthwork meets contract obligations. QA-TV: the spot-test measurement–based QA target value specified in the project contract. Quality Control (QC): testing performed by the contractor or contractor’s representative to ensure that the constructed earthwork meets contract obligations. Roller Measurement Value (MV): the roller-based parameter used for assessment of soil stiffness during compaction and based on roller vibration measurements. Rolling Pattern: the path traversed by the roller during a Mea- surement Pass. Spot-Test Measurement: a field test used during earthwork QC and QA that provides a measurement at a discrete location; common examples include the nuclear gauge for density and moisture and the lightweight deflectometer. 7.3 notation The following symbols are used throughout this specification: A theoretical vertical drum vibration amplitude f excitation frequency of eccentric mass within drum M r resilient modulus (e.g., per AASHTO T-307) MV i spatial Roller MV data from pass i. This is an array of data. µ MVi mean of spatial Roller MV data from pass i σ MVi standard deviation of spatial Roller MV data from pass i %∆ percent difference %∆µ MVi percent difference of the mean of spatial Roller MV data from pass i – 1 to pass i (for Option 2a) %∆MV i spatial percent difference in Roller MV data from pass i – 1 to pass i. This is an array of values. µ %∆MVi mean of spatial percent difference in Roller MV data from pass i – 1 to pass i σ %∆MVi standard deviation of spatial percent difference in Roller MV data from pass i – 1 to pass i v forward travel velocity of roller w opt optimum moisture content (e.g., from standard or modified Proctor density testing) 7.4 Important Considerations 7.4.1 applicable Soil Types This specification is applicable to cohesive and cohesionless soils and aggregate base materials. Research has shown that current Roller MVs are less reliable on cohesive soils and that particular attention must be given to soil moisture content. 7.4.2 personnel requirements The implementation of roller-integrated CCC for earth- work QA requires knowledgeable field personnel. A certifica- tion process is highly recommended. QA personnel must be familiar with the aspects of CCC and IC described in the pre- vious chapters of this report and with the CCC and IC equip- ment on the job site. QA personnel must be able to verify the proper working condition of an Instrumented Roller, iden- tify appropriate Evaluation Sections and Calibration Areas, and conduct calibration if Option 3 is pursued. QA person- nel must understand how to direct Measurement Passes and evaluate Measurement Pass data in a timely and reliable man- ner. To this end, QA personnel must direct the roller operator regarding required Operational Parameters and Measurement Pass procedures. 7.4.3 roller mv dependence on operating parameters Current Roller MVs depend on machine Operating Pa- rameters such as theoretical vertical drum vibration ampli- tude A, excitation frequency f, forward velocity v, and travel

0  orientation (e.g., forward, reverse). The nature of the Roller MV dependency on each Operating Parameter is soil, stratig- raphy, and roller dependent and can vary considerably. As a result, machine Operating Parameters, including A, f, and v, must remain constant during a Measurement Pass. To this end, Automatic Feedback Control should not be used during a Measurement Pass. Variations in roller Operating Parameters should remain within the following tolerances: ±2 Hz for f, ±0.2 mm (0.0008 in) for A, and ±0.5 km/h (0.3 mph) for v. Manufacturer-recommended amplitudes should be used if provided. Otherwise, theoretical amplitudes A between 0.7 and 1.1 mm are recommended for Measurement Passes. Typi- cal vibration frequencies range from 28 to 32 Hz. Typical roller speeds range from 3.0 to 5.5 km/h (1.9 to 3.4 mph). Once de- cided on, these Operating Parameters should remain constant for all Measurement Passes within an Evaluation Section. Roller MVs collected during startup, stopping, and turning typically violate these tolerances and should not be used for QA. 7.4.4 roller mv dependence on driving direction Current roller-integrated vibration-based measurement systems include sensors on one side of the drum, resulting in MVs that are more representative of soil on the instrumented side of the drum. Accordingly, MVs can exhibit a strong de- pendence on driving direction in areas where material prop- erties are heterogeneous. In extreme cases, MVs can vary by as much as 100% when the roller is driven over the same track in opposite directions. Therefore, care must be taken to en- sure that the same Rolling Pattern is used for all Measurement Passes as well as any Compaction Passes used for QA. 7.4.5 measurement depth Current roller-integrated vibration-based measure- ment systems for 11- to 15-ton vibratory roller compactors provide a composite measure of soil stiffness to depths of 0.8 to 1.2 m (2.6 to 3.9 ft) and three to four times deeper than typical 0.2 to 0.3 m (8 to 12 in) Lifts of subgrade, sub- base, or base material. An increase in theoretical amplitude slightly increases the Measurement Depth. Typical spot-test measurements such as the nuclear density gauge and light weight deflectometer (LWD) typically reflect the properties of the 0.2- to 0.3-m (8- to 12-in)-placed Lift of soil, whereas Roller MVs reflect the properties of multiple Lifts. Conse- quently, sublift soil properties are reflected in Roller MVs much more significantly than in spot-test measurements. Therefore, CCC options that rely on obtaining correlations between Roller MVs and spot-test results are increasingly difficult to implement if the sublift conditions are variable. Further, as the sublift properties change, so too will the Roller MV and spot-test measurement correlations, even if the Lift material is the same. 7.4.6 evaluation Section Acceptance testing for all specification options is per- formed on Evaluation Sections. An Evaluation Section is an area of production earthwork that exhibits homogeneity, or consistently distributed heterogeneity, both in the longitudi- nal and transverse directions, as evidenced by the Roller MV map. An Evaluation Section is commonly the full width of the earthwork section by a length that varies depending on the pace of construction, longitudinal heterogeneity, and other factors (e.g., schedule). Typical lengths can vary from 100 to 500 m (330 to 1,640 ft). A number of factors can contribute to heterogeneity and thus the sizing of Evaluation Sections. A change in borrow material or a transition from a cut to a fill section may induce a distinct change in Roller MV, particu- larly in the longitudinal direction. In the transverse direction, edge material or shoulders can sometimes exhibit markedly different stiffness behavior. Figure 7.2 provides examples of appropriate and inappropriate Evaluation Sections. Figure 7.2. Acceptable and unacceptable evaluation sections based on roller MV data map.

0 7.4.7 Calibration area Option 3 requires the establishment of an MV-TV prior to acceptance testing in an Evaluation Section. If possible, a por- tion of the Evaluation Section should be used as the Calibra- tion Area. A Calibration Area can range in size from a single roller width by 30 m (100 ft) long to the full width of the pave- ment earthwork section [10 to 30 m (33 to 100 ft)] by 100 m (330 ft) long. The underlying or sublift conditions of the Calibration Area must be representative of the Evaluation Sec- tion for which the MV-TV will be used. For this reason, larger Calibration Areas are preferred [e.g., full width of Evaluation Section by 100 m (330 ft) long]. Smaller Calibration Areas may be suitable for homogeneous site conditions. A Roller MV map of the sublift material or of the first roller pass of the Evalu- ation Section is helpful in selecting a Calibration Area. As a general guideline, the Calibration Area should capture 50% to 75% of the variation observed in the Evaluation Section. The Calibration Area material Lift should be constructed in the same manner as the Evaluation Section. Material type, ma- terial placement procedures, moisture conditioning, and Lift thickness all influence Roller MVs and should be consistent between the Calibration Area and the Evaluation Section. It is important to consider the effects of edge lanes and other special features (e.g., shallow pipe crossings) that may exist in varying concentrations within the Calibration Area and the Evaluation Section. Roller MVs are influenced by the presence or absence of lateral confinement (e.g., an edge lane often only has lateral confinement on one side). Data from edge lanes or from other areas with special features should be used judiciously when developing correlations. If these features are predominant in some Evaluation Sections, QA personnel should consider developing separate correlations for these sections or using another option. Recalibration is required if Evaluation Section conditions are sufficiently different from the Calibration Area. The ap- propriateness of a correlation should be periodically verified by comparing roller MV and spot-test measurements from the Evaluation Section to the correlation developed in the Calibration Area. 7.5 Instrumented Roller Requirements An Instrumented Roller must meet minimum requirements regarding Roller MV reliability, documentation, and Roller MV and position reporting. In addition, QA personnel should ver- ify that the selected roller is capable of producing the required level(s) of compaction in a reasonable amount of time. 7.5.1 roller mv and position reporting Roller MVs should be reported with a constant spatial reso- lution within the acceptable range of 0.2 to 1.0 m (8 to 40 in). All reported Roller MVs should be unique measurements (i.e., repeat Roller MVs should not be reported). To ensure full coverage of earth material properties, each Roller MV should reflect a spatial average over a distance not less than the Roller MV reporting resolution. Each reported Roller MV should be accompanied by a three-dimensional position (e.g., northing, easting, elevation, determined via roller-mounted GPS). RTK differential GPS is highly recommended. Each reported posi- tion should reflect the geometric center of the earthwork over which the corresponding Roller MV was determined. 7.5.2 documentation The Instrumented Roller must clearly document the follow- ing parameters: • Roller MV • Three-dimensional position (including time stamp and GPS quality) • Vibration amplitude (theoretical A or actual z d ) • Vibration frequency, f • Travel speed, v • Forward/reverse driving direction • Automatic feedback control on/off • Indication of jumping • Vibration on/off • Pass Sequence These recorded data must be easily accessible via the In- strumented Roller’s onboard computer, and units should be clearly documented (standard SI or English units are ac- ceptable). In addition, these data should be easily export- able for postprocessing and record keeping. Simple text files are preferred, and files in a proprietary format (e.g., only compatible with the roller manufacturer’s software) are not acceptable. Basic statistics about the Roller MV data and the Operat- ing Parameters should be readily available. Statistics of inter- est for the Roller MV data include the minimum, maximum, mean, standard deviation, and histogram. Statistics of in- terest for the Operating Parameters include the minimum, maximum, mean and standard deviation of the amplitude, frequency, and speed. This information allows QA personnel to efficiently determine if a data map meets the requirements of a Measurement Pass.

  7.5.3 verification of roller mv repeatability and gpS position reporting Roller MVs must be repeatable (i.e., repeated Measurement Passes over the same, fully compacted material must exhibit similar magnitudes and trends). It is also important to verify the accuracy of roller-mounted GPS position and that a posi- tion offset between the receiver and the center of the drum (i.e., where MVs are computed) and/or errors due to data averaging are properly considered. Accurate position report- ing is particularly important for specification options that involve Roller MV correlation to spot-test measurements or spatial comparison of Roller MVs from consecutive passes. 7.5.3.1  Procedure to Verify Roller MV Repeatability 1. Perform two Measurement Passes on a fully compacted test strip at least 100 m (330 ft) long. Measurement Passes should be performed in the same direction, with static passes performed in the reverse direction between Mea- surement Passes. 2. From the two Measurement Passes, compute the spatial percent difference array %∆MV i per Equation 7.1: % – – – ∆MV MV MV MVi i i i = ×1 1 100 (7.1) where MV i and MV i-1 represent the MV data array from pass i and pass i – 1, respectively. If necessary, linear in- terpolation may be used to transform the data onto a grid for precise spatial comparison. If the mean of the %∆MV array (µ %∆MVi ) is greater than 5%, the test strip may not have been fully compacted and the procedure should be repeated. 3. Compute the standard deviation of the %∆MV i array (σ%∆MVi). The σ%∆MVi quantifies the Roller MV repeatability. The recommended acceptable maximum σ%∆MVi is 10%, although visual inspection and engineering judgment should be employed when investigating repeatability and when deciding what level of repeatability is required and acceptable on a given project. Figure 7.3 illustrates accept- able and unacceptable levels of σ%∆MVi as well as an invalid test (due to increase in compaction of the test strip). 7.5.3.2   Procedure to Verify Roller Position  Reporting The accuracy of Roller MV position reporting should be verified when the Instrumented Roller is stationary as well as moving. When the roller is stationary, the roller-mounted GPS position can be compared with the position from a handheld RTK GPS unit (i.e., rover) placed at the drum cen- Figure 7.3. Roller MV repeatability testing: (left) acceptable repeatability, (middle) invalid test, and (right) unac- ceptable repeatability.

 ter. This can also be accomplished by establishing a marker of known position on the ground and approaching the marker with the roller from different directions. In addition, the fol- lowing procedure is recommended to ensure Roller MV posi- tion reporting accuracy while the roller is moving (see Figure 7.4). This procedure may be combined with the evaluation of Roller MV repeatability (see Section 7.5.3.1). 1. Create two obstructions in the earthwork spaced at least 10 m apart that will induce noticeable spikes in the Roller MV data record. Example obstructions include a wooden beam and narrow trenches perpendicular to the direction of roller travel. Note that the obstructions should span the drum width. 2. Perform two Measurement Passes over the obstructions, one in each direction. 3. Superimpose the Measurement Pass records to observe if the spikes in the Roller MV data occur simultaneously. Any difference in location of a spike in the Roller MV data re- cords is a reflection of position error. If this position error is greater than one-half of the Roller MV reporting resolu- tion (see Section 7.5.1) or the accuracy of the GPS, this position error must be corrected. 7.6 QA Option 1: Spot Testing of Roller-Informed Weakest Area(s) QA Option 1 utilizes roller-integrated CCC to identify the weakest area(s) of the Evaluation Section. The weakest area is defined by the lowest Roller MVs recorded during a Mea- surement Pass. More than one weakest area may be selected. Acceptance is based on spot-test measurements from the weakest area(s) (see Figure 7.5). If spot-test measurements from the weakest area(s) meet specified criteria, the Evalua- tion Section meets acceptance. The frequency of spot testing in the weakest area(s) and acceptance criteria for spot-test measurements should be based on existing spot-test-based QA practice. The selection of Evaluation Sections must be performed in accordance with Section 7.4.6. Figure 7.4. MV data records with position error (requiring correction) and without position error. Figure 7.5. Acceptance testing via Option 1.

  An important premise of Option 1 is that there is a positive correlation between Roller MV and soil compaction (i.e., the lowest Roller MVs correspond to lowest compaction). This positive correlation may not exist if the Evaluation Section ex- hibits localized soil variability (e.g., pockets of high clay con- tent in an otherwise granular material) or significant variabil- ity in the sublift material. In cases where the lift and/or sublift material exhibits variability, the correlation between Roller MVs and spot-test measurements should be verified. This can be accomplished by comparing spot-test measurements with low, medium, and high Roller MVs across the Evaluation Section. In addition, low Roller MVs can result from fluctua- tions in machine operating parameters and surface uneven- ness. For this reason the weakest area(s) selected for testing should not be based on Roller MVs from areas less than 3 m (10 ft) long in the driving direction. Option 1 requires mini- mal changes to existing spot-test-based QA specifications and should be relatively easy to implement. To improve reliability, Option 1 may be implemented in combination with another recommended option. 7.7 QA Option 2: Limiting Percentage Change in Roller MV QA Option 2 utilizes the pass-to-pass percentage change in Roller MVs to determine acceptance of an Evaluation Sec- tion. Acceptance is based on achieving a threshold or target %∆MV i (i.e., %∆-TV) between two consecutive Measurement Passes over the same Evaluation Section. There are two ways in which Option 2 can be implemented. Acceptance may be based on the pass-to-pass percentage change in mean MV (%∆µ MVi ) of the Evaluation Section. Alternatively, acceptance may be based on a spatial analysis of the %∆MV i data array. The dependence of Roller MVs on driving direction and the influence of soil heterogeneity within the drum length re- quire that Measurement Passes must be performed with an identical pass-to-pass Rolling Pattern, particularly for Option 2b. The selection of Evaluation Sections must be performed in accordance with Section 7.4.6. 7.7.1 option 2a: monitoring percentage Change in mean mv QA Option 2a involves computing %∆µ MVi from two con- secutive Measurement Passes over the Evaluation Section ac- cording to Equation 7.2: % – – – ∆µ µ µ µMV MV MV MV i i i i =       × 1 1 100 (7.2) The recommended target value for %∆µ MVi (i.e., %∆-TV) is 5%. If %∆µ MVi between consecutive passes is less than the %∆-TV, acceptance is met (see Figure 7.6). 7.7.2 option 2b: monitoring Spatial percentage Change in roller mvs QA Option 2b involves evaluating the spatial percent change in %∆MV i from two consecutive Measurement Passes over the Evaluation Section according to Equation 7.3: Figure 7.6. Acceptance testing via Option 2a.

 % – – – ∆MV MV MV MVi i i i = ×1 1 100 (7.3) This spatial analysis involves transforming the Roller MV data onto a fixed grid for direct comparison. The percentage change in Roller MV is then computed for each grid location. The recommended %∆-TV for Option 2b is two times the uncertainty σ%∆MV determined from repeatability testing (as described in 7.1.5.1). Acceptance is based on achieving the %∆-TV over a specified proportion or percent area threshold of the Evaluation Section, as shown in Figure 7.7. The recom- mended range for a specified proportion/percent area thresh- old is 80% to 95% (e.g., acceptance is met when the Roller MV increases by ≤%∆-TV for 90% or more of the Earthwork Section). The process of transforming spatial Roller MV data onto a fixed grid is not trivial and has not been proven fully reliable. The simplest methods (e.g., nearest neighbor grid- ding, linear interpolation) may be more favorable than the more complex methods (e.g., nonlinear interpolation, krig- ing) until the geostatistical nature of Roller MV data is better understood. QA Option 2 requires moderate changes to existing con- struction practice in that two Measurement Passes must be performed with similar Rolling Patterns. If implemented cor- rectly, Option 2 ensures that compaction has been achieved to the capability of the CCC roller and the selected Operat- ing Parameters but does not necessarily ensure that adequate compaction has been achieved. The ability of the roller to produce adequate compaction should be verified before earthwork compaction begins. To improve the reliability of compaction QA, Option 2 may be used in combination with another QA option. 7.8 QA Option 3: Comparison of Roller MV Data to Target MV Options 3a, 3b, and 3c each require that a specified percent- age of Roller MVs in an Evaluation Section exceed a Roller MV target value (MV-TV). The MV-TV must be determined dur- ing field calibration prior to acceptance testing. In the future, MV-TVs may stem from a database of project information, documented case histories, and literature; however, extreme care should be exercised given the wide range of variables that influence Roller MVs. The approach to establish an MV-TV differs for each option and is described in the following sec- tions. Acceptance for each option is similar and is based on Figure 7.7. Acceptance testing via Option 2b.

  achieving the MV-TV over a specified proportion or percent area threshold of the Evaluation Section. The range of recom- mended proportions is 80% to 95% and should be defined prior to testing. Figure 7.8 illustrates the acceptance process. The selection of Evaluation Sections must be performed in accordance with Section 7.4.6. 7.8.1 mv-Tv determination for Qa option 3a: relating roller mv to Spot-Test measurements In Option 3a, Roller MVs are related to Spot-Test Measure- ments using statistical regression analysis. The methodology is illustrated in Figure 7.9. A Calibration Area should be se- lected per Section 7.4.7. The Calibration Area is sequentially compacted to low, intermediate, and full (complete) compac- tion states. The full compaction state should meet or exceed the required QA-TV. At each compaction state a Measurement Pass is performed, followed immediately by spot testing at several locations in the Calibration Area. A minimum of five locations is recommended for spot testing at each compaction level. The Measurement Pass data can be used to identify loca- tions for spot testing. To improve reliability, locations where Roller MVs are locally constant are recommended and loca- tions where Roller MVs are highly variable should be avoided (see Figure 7.9). The collection of locations should represent the range of variability observed in the Measurement Pass data. At each selected location, three to five spot-test mea- surements should be collected across the roller drum lane and averaged to generate one regression point. The Roller MVs at each location may be averaged over a 1-m distance in the direction of roller travel to generate one regression point. Regression analysis is performed with the Roller MV as the dependent variable and Spot-Test Measurement as the inde- pendent variable. In general, the coefficient of determination R2 is used to judge correlations, with acceptable correlations being defined by R2 ≥ 0.5 (e.g., the correlation depicted in Figure 7.9 is acceptable). This level of acceptability for R2 is commonly used when correlating soil properties measured from different devices because loading conditions, seating is- sues, and representative volumes vary. In Ping et al. (2002), R2 = 0.3 was found acceptable when correlating modulus back calculated from a FWD with lab data. In Vennapusa & White (2009), R2 = 0.5 to 0.9 was found acceptable when correlating data collected from different LWDs and R2 = 0.40 to 0.66 was found acceptable for correlations between LWD and static PLT moduli. When correlating Roller MVs to stiffness-based spot tests (e.g., PLT, LWD), single-variable regression is usually suffi- cient. When correlating Roller MVs to density, single-variable regression may not be successful. Multivariate regression may be required if the material behavior is moisture dependent. Figure 7.8. Acceptance testing via Option 3.

 Figure 7.9. Determination of MV-TV via Option 3a. It may also be desirable to perform multivariate regression analysis that takes sublift conditions into consideration to account for the significant difference in measurement depth of Roller MVs and Spot-Test Measurements. Prior experience and a sound fundamental understanding of both Roller MVs and Spot-Test Measurements will increase the probability of achieving acceptable correlations. If a suitable correlation is determined (R2 ≥ 0.5), an MV- TV is selected from the regression equation based on the existing Spot-Test Measurement–based target value (QA-TV; e.g., 95% maximum dry density; see Figure 7.9). The QA- TV used to determine the MV-TV may be increased to in- crease confidence. Prediction limits can be employed when performing correlations to select the MV-TV. An MV-TV can be established using the upper prediction limit, as shown in Figure 7.10. The greater the prediction interval specified, the higher the MV-TV. 7.8.2 determination of mv-Tv for Qa option 3b: Compaction Curve In Option 3b the change in Roller MV is monitored to de- termine when compaction is complete. During compaction Figure 7.10. Illustration of use of prediction limits when selecting an MV-TV. of the Calibration Area (selected per Section 7.4.7), the Roller MV and pass number are continuously recorded. The MV-TV is established as the µ MVi when the %∆MV i ≤ 5% for 90% of the Calibration Area, as shown in Figure 7.11. The calibration procedure for Option 3b is similar in principle to acceptance testing of Option 2b.

  7.8.3 determination of mv-Tv for Qa option 3c: relating roller mv to lab- determined properties In this approach the MV-TV is established through em- pirical correlation to laboratory-determined engineering properties (e.g., M r ) based on a preselected combination of moisture contents and dry unit weights. This option requires a significant investment of time to establish the MV-TV. It must be noted that performing laboratory resilient modulus tests is time consuming, which can potentially impact the ap- plication of this specification option if heterogeneous materi- als exist on a project. Laboratory testing should be performed on samples pre- pared at preselected moisture-density combinations in ac- cordance with standard protocols used by state agencies. For example, moisture contents may vary from w opt – 4% to w opt + 4% and dry unit weight from 90% to 110% of γ dmax es- tablished from laboratory Proctor compaction (standard or modified) specified by the agency for QA. Using the test re- sults, a multiple regression model is developed as a function of moisture and dry unit weight to predict the laboratory soil property values (Figure 7.12b). For stress-dependent M r val- ues, multiple regression relationships may be developed for a selected stress condition. Compaction of the Calibration Area (selected per Section 7.4.7) is performed using variable moisture content (e.g., w opt –4% to w opt + 4%) and obtaining Roller MVs in paral- lel with in situ moisture and dry unit weight measurements from multiple passes (e.g., 1, 2, 4, 8, and 12). Using the Spot- Test Measurement data and corresponding Roller MV data, a multiple regression relationship is developed to predict Roller MV as a function of moisture and dry unit weight (similar to Option 3a; shown in Figure 7.12c using prediction limits). Using the multiple regression relationships from laboratory and field testing, laboratory soil properties are correlated to Roller MVs as shown in Figure 7.12a. An MV-TV can be determined based on a target laboratory soil property (QA- TV) from the linear regression relationship. Prediction inter- vals are employed as shown in Figure 7.12. The compaction method used for laboratory-prepared specimens can affect the laboratory-determined values, especially for fine-grained cohesive soils. Laboratory vibratory compaction is recom- mended for granular soils compacted using smooth drum rollers. Laboratory impact compaction is recommended for soils compacted using pad foot rollers. 7.9 Uniformity Criteria Uniformity is recognized as an important component of quality compaction (e.g., Davis 1953, Sherman et al. 1966). Results from numerical studies indicate that considering av- erage values in design may not capture actual performance (e.g. White et al. 2004, Griffiths et al. 2006). With the ability of real-time viewing of compaction data, roller-integrated measurement technology offers an opportunity to construct more uniform earthwork layers. Current CCC specifications address uniformity using percentage limits based on an MV- TV. The International Society for Soil Mechanics and Geo- technical Engineering (ISSMGE)/Austrian CCC earthwork specifications, for example, require that roller MVs in the production area should fall within 0.8 to 1.5 MIN-TV with a coefficient of variation < 20% (MIN-TV corresponds to the MV at 0.95 QA-TV established from calibration). Using a slightly different approach, the Minnesota Department of Figure 7.11. Determination of MV-TV via Option 3b.

 Figure 7.12. Determination of MV-TV via Option 3c. Transportation (MnDOT) implemented a predetermined target percentage limits distribution criterion (Mn/DOT 2007) on a full-scale earthwork project in the state (White et al. 2007, 2008a, 2008b). The acceptance requirement was that at least 90% of roller MV data in the production area should fall within 90% to 120% of the MV-TV; none should be below 80% of the MV-TV; and if any are above 120%, a new MV-TV should be established. If uniformity criteria are desired as part of the specifi- cation, the ISSMGE and Mn/DOT approaches described above can be implemented for specification Options 2 and 3. However, it must be realized that these approaches are limited to conditions where Evaluation Sections have simi- lar spatial heterogeneity in compaction layer properties and support conditions to the Calibration Area. If not, achiev- ing these uniformity targets is challenging. For such cases, information of underlying support conditions may help in evaluating compaction layer data and selecting repre- sentative Calibration Areas. Further, these approaches do not address uniformity from a spatial standpoint. More research is needed in relating uniformity to performance for a better understanding of the level of uniformity de- sired and how field operations can be improved to control nonuniformity. An alternate approach to quantify uniformity is to use spa- tial statistics in combination with univariate statistics (mean and standard deviation; Brandl 2001, Vennapusa et al. 2009, Facas et al. 2010). Using spatial statistics requires developing semivariogram models using spatially referenced GPS coor- dinate measurements, which describe the spatial relationship in the measured roller MVs. The three main characteristics by which a semivariogram is often summarized are range, sill, and nugget (Isaaks & Srivastava 1989). Comparatively, a semivar- iogram with a lower sill and longer range represents reduced nonuniformity and improved spatial continuity. Vennapusa et al. (2009) describe an approach for using spatial statistics to target areas for compaction that results in improved spatial continuity and reduced nonuniformity.