Laboratory Hot-Mix Asphalt Performance Testing - Transportation ...

Laboratory Hot-Mix Asphalt Performance Testing Asphalt Mixture Performance Tester Versus Universal Testing Machine Lubinda F. Walubita, Jun Zhang, Abu N. M. Faruk, Allex E. Alvarez, and Tom Scullion tion Institute (TTI), College Station. To accomplish this comparative assessment, a laboratory testing plan was undertaken with the following primary objectives:

Reliable and repeatable laboratory testing of hot-mix asphalt (HMA) mixes is of paramount importance to ensure proper HMA mix-design characterization and ultimately to ensure satisfactory field performance. The study reported in this paper was undertaken to evaluate the comparability, repeatability, and reliability of two testing systems, namely, the asphalt mixture performance tester (AMPT) and the traditional universal testing machine (UTM), and to assess if the two systems could be used concurrently or in lieu of one another for dynamic modulus, flow number, and repeated load permanent deformation (RLPD) testing. For the HMA mixes evaluated to accomplish these objectives, the test results from the AMPT and UTM—on the basis of dynamic modulus, flow number, and RLPD testing—were statistically comparable and acceptable at the 95% confidence level. The test repeatability and variability in the AMPT and UTM systems also were statistically acceptable with low coefficient of variation values in the test results. Thus either system could be used confidently in lieu of the other to generate results of similar quality and reliability and with a comparable statistical degree of accuracy with acceptable variability for dynamic modulus, flow number, and RLPD testing. The choice or preference is up to the user, given that each system has its own merits and demerits. However, operator and technician proficiency and equipment calibration are some of the most critical factors never to ignore in laboratory testing work of this nature.

1. To evaluate the statistical comparability of results from the AMPT and the traditional UTM, for the same material (HMA) and test conditions, on the basis of dynamic modulus, flow number, and repeated load permanent deformation (RLPD) testing; 2. To comparatively evaluate the accuracy, operational efficiency, and practicality of the AMPT system relative to the traditional UTM system; and 3. To make recommendations as to which system to use for future HMA performance testing (i.e., RLPD, flow number, dynamic modulus) or to determine that both systems could be used concurrently or in lieu of one another. In the subsequent sections, the AMPT and UTM systems are described, followed by the research methodology and laboratory experimentation plan. Laboratory test results are then presented and comparatively analyzed followed by an evaluation of the systems’ general characteristic features. The paper then concludes with a synthesis and summary of the key findings and recommendations. An extensive literature review conducted at the time that this paper was written found no substantively similar work. Thus the reference citations here to previous AMPT–UTM comparative studies are limited.

The performance of reliable and repeatable laboratory testing of hot-mix asphalt (HMA) mixes constitutes an indispensable element in proper HMA mix-design characterization to ensure satisfactory field performance. Different systems are available for HMA performance testing and material property characterization. This study was conducted to assess the comparability, repeatability, and reliability of two HMA testing systems, namely, the asphalt mixture performance tester (AMPT) and the traditional universal testing machine (UTM), which are available at the Texas A&M Transporta-

AMPT and UTM Systems Figure 1 shows pictures of TTI’s AMPT and UTM systems. One outstanding difference between the two units is the size of their temperature chambers. The UTM’s chamber is more than 10 times the size of the AMPT’s in volume and therefore can permit the concurrent conditioning of multiple specimens without the need for an external chamber. However, a longer time is needed to condition the specimens (i.e., to reach the target temperature). By contrast, the smaller chamber of the AMPT enables better temperature control and consistency during testing, although it would need an external chamber to condition multiple specimens. The specifications details of the two units are comparatively listed in Table 1. As evident in Figure 1, the AMPT unit is more compact with a higher flexibility for mobility than the UTM. The automated

L. F. Walubita, J. Zhang, A. N. M. Faruk, and T. Scullion, Texas A&M Transportation Institute, Texas A&M University System, 601G CE/TTI Building, 3135 TAMU, College Station, TX 77843. A. E. Alvarez, University of Magdalena, Department of Civil Engineering, Carrera 32, Number 22-08, Santa Marta, Magdalena, Colombia 470004. Corresponding author: J. Zhang, [email protected]. Transportation Research Record: Journal of the Transportation Research Board, No. 2447, Transportation Research Board of the National Academies, Washington, D.C., 2014, pp. 61–73. DOI: 10.3141/2447-07 61

62

Transportation Research Record 2447

(a)

(b)

FIGURE 1 TTI’s (a) AMPT and (b) UTM units.

AMPT–linear variable differential transformer (LVDT) gluing and specimen setup jigs (Table 1) translate into simplicity and better accuracy than the manually operated UTM jigs. However, some evident limitations of the AMPT as shown in Table 1 include the lower load cell capacity, shorter LVDT span (i.e., 10 times shorter than that used in the UTM), and shorter temperature range (i.e., the AMPT cannot be used for testing below +4°C or above +60°C). As Table 1 shows, the shorter LVDT span of the AMPT means better resolution and higher accuracy. TABLE 1 Specification Features of AMPT and UTM Systems at TTI Characteristic Feature

UTM

AMPT

Load cell (static) Load cell (dynamic)

25 kN (5.620 kips) 20 kN (4.496 kips)

15 kN (3.372 kips) 13.5 kN (3.035 kips)

Frequency up to (Hz) Loading mechanism LVDT span LVDT accuracy

60 Servohydraulic Varies (±5 mm) NA

Approximate chamber dimension (internal)

H ≅ 1,045 mm; W ≅ 750 mm; B ≅ 475 mm Can handle most specimen dimensions and configurations −40°C to +100°C (−40°F to +212°F) Manual

70 Servohydraulic ±0.5 mm Meets NCHRP 9-29 specifications, resolution > 0.0002 mm (0.04%) ϕ ≅ 285 mm; H ≅ 290 mm

Temperature range LVDT gluing jigs and setup

Designed for 150-mmtall × 100-mmdiameter specimens +4°C to +60°C (+39.2°F to 140°F) Automatic

Note: NA = not available; H = height; W = width; B = breadth; ϕ = diameter.

Research Methodology For the AMPT and UTM systems, the following methodological approach and procedural steps were employed to ensure similar conditions and consistency in the comparative assessment without bias: • Ensure that the AMPT and UTM are calibrated; • Use the same test methods, conditions, and loading parameters; • Use the same HMA mixes; • Use the same number of sample replicates; • Mold and fabricate the samples to the same density (air void content) and dimensions; • Ensure operation by the same trained operator or technician; • Use the same data analysis methods; and • Use different personnel to analyze and verify the data. Experimental Design Plan The experimental design plan was to select commonly used HMA performance test methods and then to devise an appropriate laboratory work plan to complete the comparative assessment. These test methods along with the data analysis models and HMA mix details are discussed in this section. Laboratory Test Methods The work plan involved parallel testing in the AMPT and UTM systems with the use of the same HMA mix, similar test conditions and test loading parameters, and the same operator. Thereafter, the HMA test results were compared for the following three commonly used laboratory test methods for HMA performance testing: (a) RLPD, (b) flow number, and (c) dynamic modulus. Details of these test methods including the loading configuration and test parameters are listed in Table 2 (1, 2). The test description and associated analysis models are discussed in the subsequent text.

Walubita, Zhang, Faruk, Alvarez, and Scullion

63

TABLE 2 Laboratory Test Methods

Load

Load

Specimen dimension Target test temperature Target temperature tolerance Sample temperature conditioning time Loading mode Loading frequency Stress level (vertical) Confining pressure Test termination criterion Test time Measurable parameters and output data

Reference

Stress (psi)

Load

Schematic

Flow Number

Number of Load Cycles

stress

Specimen Schematic Load

Load

Load

Load

Time ∆L PD

100 mm ϕ × 152 mm H (4 in. ϕ × 6 in. H) 50°C (122°F)

±2°C

±2°C

≥2 h

2∼3 h

Compressive repeated Haversine (stress-controlled mode) 1 Hz (0.1-s loading, 0.9-s rest) 138 kPa (20 psi) at 40°C and 69 kPa (10 psi) at 50°C 0 kPa (0 psi) 10,000 load repetitions or 25,000 microstrains ≤3 h Axial permanent deformation, permanent and resilient strains (εp, εr), stress, number of load passes, time, temperature, frequency, viscoelastic properties (α, µ), and resilient modulus (Mr) 1, 2, 3, 4

Compressive repeated Haversine (stress-controlled mode) 1 Hz (0.1-s loading, 0.9-s rest) 207 kPa (30 psi)

The RLPD test is used to characterize the permanent deformation properties of HMA, namely, α and µ, under repeated compressive Haversine loading (3, 4). For the purpose of this study, the viscoelastic properties α and µ were determined as a function of a log-log plot of the accumulated plastic strain (εp) versus the number of load cycles (N) as follows: ε p = aN b

(1)

α =1− b

(2)

ε r (r 200)

≥3 h (4.4°C), 2 h (21.1°C), 2 h (37.8°C), 2 h (54.4°C) Compressive repeated Haversine (stress-controlled mode) 0.1–25 Hz 3.4–1,724 kPa (0.5–250 psi)

0 kPa (0 psi) 10,000 load repetitions or 30,000 microstrains ≤3 h Flow number (cycles), time to tertiary flow (min), temperature, frequency, accumulated microstrain at tertiary flow (microns), and microstrain– flow number ratio

0 kPa (0 psi) Variable preset number of cycles per stress level per loading frequency ≥3 days Load (stress), deformation, phase angle, and dynamic modulus

2, 5, 6

2, 5, 6, 7

Flow Number Test Method As noted in Table 2, the HMA characteristic parameters generated from the flow number test include the following: flow number, time to tertiary flow (tF), accumulated microstrain at tertiary flow (εp), and the microstrain–flow number ratio (FN index). For the purpose of this comparative study, models in the publication by Archilla et al. (5), as shown in Table 3, were used to analyze the UTM and AMPT flow number data (2, 6). Dynamic Modulus Test Method

(3)

Regression parameters a and b are the intercept and slope of the linear portion of the strain–load cycles curve on a log-log scale. The HMA rutting parameters are α and µ, with µ computed at the 200th load cycle for this study. εr(r200) is the resilient microstrain obtained at the 200th RLPD load cycle; see the example in Figure 2 (2, 3). • a, b = 94.0380, 0.3233; • εr(r200) = 57.76;

100 mm ϕ × 152 mm H (4 in. ϕ × 6 in. H) 4.4°C, 21.1°C, 37.8°C, 54.4°C (40°F, 70°F, 100°F, 130°F) ±2°C

• α = 0.6767; and • µ = 0.5264.

RLPD Test Method

ab

Time (s)

Time

100 mm ϕ × 152 mm H (4 in. ϕ × 6 in. H) 40°C, 50°C (104°F, 122°F)

Data Analysis Methods

µ=

Dynamic Modulus

Stress

RLPD

Vertical Deformation

Feature or Test

The typical output parameter that is generated from the dynamic modulus test is the dynamic complex modulus of the HMA, denoted as |E*|, and is expressed as shown in Equation 4 (AASHTO TP 62-03): Ep =

σ0 ε0

(4)

where σ0, is the axial (compressive) stress and ε0 is the axial (compressive) strain in Equation 4. For graphical analysis and easy interpretation

64


120

100

1,600

80

57.76

1,200

57.76

60

800 40 εr (r 200) = 57.76

400

200

0 0

2,000

4,000 6,000 8,000 RLPD Load Cycles (N) (a)

10,000

Resilient Microstrain (εr)

Accumulated Permanent Microstrain (εp)

2,000

20

0 12,000

Accumulated Permanent Microstrain (εp)

1.00E+04

1.00E+03

1.00E+02 y = 94.03807x 0.32326 R 2 = .97413

1.00E+01 1.00E+00

1.00E+01

a = 94.038; b = 0.3233

1.00E+02 RLPD Load Cycles (N) (b)

FIGURE 2 Plot of RLPD strain and load cycles.

1.00E+03

1.00E+04


65

TABLE 3 Flow Number Data Analysis Models Number

Item or Parameter

Model

Description

1

General relationship between the accumulated permanent strain and the number of load cycles Probabilistic distribution (Weibull) model for the relationship between εp and N

εp = aN b

εp is the accumulated permanent strain due to dynamic vertical loading, N is the number of load cycles to produce εp, and a and b are regression constants that depend on the material and stress state conditions. β, α, and γ are the probability distribution and shape parameters. The parameter γ has the simple interpretation of being the maximum number of load cycles that the specimen would last if the testing machine could apply an arbitrary deformation to the sample (i.e., the number of load cycles at which the rate dεp /dN → ∞). εp (predicted) is the predicted accumulated permanent strain as a function of N; where N, β, α, and γ are as previously defined. FN is flow number or number of load cycles at the onset of tertiary zone; at which d 2 εp /d 2 N = 0.

2

3

Predicted permanent strains [εp (predicted)]

4

Flow number (cycle)

5

Accumulated permanent strain at tertiary flow [εp (F); microns] Time to tertiary flow [t(F); minutes]

6 7

FN index (microstrains/cycle)

8

References

−β N = γ 1 − e ( ε p )    α

ε p ( predicted ) =

N  1   × − ln  1 −    β  γ 

( ) ( )

1 −1 FN = γ 1 − exp   α ε pflow =

1 1 1− α β

t ( FN ) =

1α

FN 60

FN index =

1α

εp(F) is accumulated permanent strain at the onset of tertiary flow, that is, at d 2 εp /d 2 N = 0. t(F) is time at the onset of tertiary flow (1 Hz loading frequency) or time count in minutes at d 2 εp /d 2 N = 0.

ε p (F ) FN

Derived composite parametric ratio that simultaneously incorporates the strain at tertiary flow, εp(F), and flow number at tertiary flow. na

1, 2, 6, 8

Note: na = not applicable.

of the dynamic modulus data, |E*| mastercurves also were generated as a function of the loading frequency with the Pellinen et al. time–temperature superposition sigmoidal model (7 ) shown in Equations 5 and 6. log E p = δ +

α 1 + eβ−γ log(ξ)

log ( ξ) = log ( f ) + log ( aT )

(5) (6)

where ξ = reduced frequency (Hz), δ = minimum dynamic modulus value (ksi or MPa), α = span of modulus values, and β and γ = shape parameters. Parameters f and aT are the loading frequency and temperature shift factor to a reference temperature (Tref), respectively. For this study, the selected Tref was 21.1°C (70°F). Statistical Analysis Tools Statistical tools to analyze the test data included an analysis of variance program, Tukey’s honest significant difference (HSD) test, and t-tests along with average, standard deviation, and coefficient of variation (CV) computations. Two confidence levels, 90% and 95%, were used to define the statistical reliability with a CV of 30% tentatively selected as the statistical threshold for repeatability and acceptable variability (1, 2).

HMA Mix-Design Details A dense-graded Type C mix was used, common in Texas, with 19-mm (3⁄4-in.) nominal maximum aggregate size. The HMA mix design comprised 4.8% PG 64-22 (Jebro) asphalt binder, limestone or dolomite aggregate, 1% lime, 17% reclaimed asphalt pavement, and 3% reclaimed asphalt shingle. In accordance with the standards of the Texas Department of Transportation, Austin, all of the plantmixed HMA samples were molded, compacted, and fabricated to a final target density of 93 ± 1%—air voids content of 7 ± 1% (8). To ensure consistency and lack of bias, all of the HMA samples were fabricated and tested by the same technician. For each test type and test condition, three sample replicates were used. Laboratory Test Results and Analysis This section presents the results and the corresponding analyses of the AMPT and UTM on the basis of RLPD, flow number, and dynamic modulus testing. However, these laboratory test results pertained only to the HMA mix and the laboratory test conditions defined in this study. The overall findings and conclusions may not be exhaustive. HMA Sample Dimensions and Air Voids Content Measurements To comparatively study two testing systems (i.e., AMPT and UTM), it is imperative that both the HMA sample dimensions and air voids content are consistently similar and within a set tolerance limit so as to avoid bias in the final results. As shown in Table 4, the dimensions and

66


TABLE 4 HMA Specimen Dimensions and Air Void Content Measurements AMPT

UTM

Sample Test

AMPTa

UTMa

RLPD 40°C

Statistic

H

ϕ

Air Voids (%)

H

ϕ

Air Voids (%)

Average CV (%)

6.06 in. 0.23

3.97 in. 0.26

7.17 4.40

6.06 in. 0.23

3.97 in. 0.26

7.13 5.60

Range

6.04–6.08 in.

3.95–3.98 in.

6.90–7.57

6.04–6.07 in.

3.95–3.98 in.

6.59–7.72

Average CV (%) Range Average CV (%) Range Target

6.06 in. 0.34 6.04–6.08 in. 6.07 in. 0.25 6.05–6.08 in. 6.00 ± 0.10 in.

3.97 in. 0.25 3.96–3.98 in. 3.96 in. 0.39 3.95–3.98 in. 4.08 ± 0.08 in.

7.21 5.49 6.80–7.59 7.45 2.58 7.28–7.66 7±1

6.06 in. 0.17 6.05–6.07 in. 6.06 in. 0.25 6.06–6.07 in. 6.00 ± 0.10 in.

3.97 in. 0.39 3.95–3.98 in. 3.96 in. 0.25 3.95–3.97 in. 4.08 ± 0.08 in.

7.36 3.65 7.15–7.66 7.46 4.00 7.26–7.80 7±1

50°C

Flow number 50°C

Dynamic modulus

4.4°C–54.4°C

3(4 in. ϕ × 4 in. μ).

a

air void contents of the HMA specimens used in the AMPT and UTM systems were comparably consistent and within the tolerance limits. With respect to field cores for in-service pavement structures with a layer thickness less than 6 in., prismatic specimens should be used and cognizance maintained of the geometrical and anisotropic requirements, where applicable (9). RLPD Test Results and Statistical Analysis Table 5 shows that the overall RLPD test results in terms of the computed α and µ were statistically equivalent in the two systems TABLE 5 RLPD Test Results: a and m

Sample Replicate Sample 1 Sample 2 Sample 3 Average CV (%)

Parameter α µ α µ α µ α µ α µ

RLPD at 40°C and 20,138 kPa (20 psi)

RLPD at 50°C and 69 kPa (10 psi)

UTM

AMPT

UTM

AMPT

0.6922 0.6276 0.7462 0.8218 0.6354 0.2086 0.6913 0.5527 8.02 10.03

0.7198 0.5997 0.7185 0.6182 0.7508 0.5691 0.7297 0.5957 2.51 2.76

0.7873 0.9800 0.7922 0.9403 0.7540 0.7580 0.7779 0.8928 2.67 2.16

0.7258 0.6382 0.7262 0.4750 0.6671 0.4406 0.7064 0.5179 4.82 5.81

as subsequently discussed. The magnitudes of these HMA rutting parameters were comparable. Therefore both systems could be used reliably and accurately for RLPD testing to characterize the HMA permanent deformation properties at the given test conditions. Statistical variability, as measured in terms of the CV, for the computed α and µ parameters in both systems, also was reasonably acceptable and comparable. All of the CV values computed on the basis of three replicate RLPD tests in Table 5 were below 15%, which suggested that the AMPT and UTM systems were repeatable and comparable for RLPD testing at 40°C and 50°C, respectively. Likewise, the analysis of variance program and Tukey’s HSD analysis at the 95% confidence level reaffirmed that the α and µ results from both systems were statistically similar (i.e., there were no statistically significant differences). Corresponding results are presented in Table 6. The AMPT-40°C and AMPT-50°C results were statistically similar to the UTM-40°C and UTM-50°C results, respectively. The consistency and repeatability in the RLPD test data may have been attributable to the consistency in the HMA sample dimensions and air voids shown in Table 4. Therefore, it is imperative to ensure consistent air voids in the HMA samples when performance studies of this nature are conducted. Flow Number Test Results and Statistical Analysis The computed flow number tests results are shown in Tables 7–10 on the basis of the actuator (RAM) and LVDT displacement mea-

TABLE 6 Analysis of Variance Program Analysis at 95% Confidence Level with RLPD Test Data α Group UTM at 40°C AMPT at 40°C UTM at 50°C AMPT at 50°C

Count 3 3 3 3

Sum 2.0738 2.1891 2.3336 2.1191

µ Average

Variance

0.6913 0.7297 0.7779 0.7064

0.0031 0.0003 0.0004 0.0012

Sum

Average

Variance

1.6580 1.7870 2.6783 1.5537

0.5527 0.5957 0.8928 0.5179

0.0982 0.0006 0.0140 0.0111


surements. Discussion and analysis of these results are provided in the subsequent pages. As evident in Tables 7–10, HSD and t-tests were performed to evaluate any statistical differences between the AMPT– and UTM– FN test results at the 90% and 95% confidence levels. On the basis of these results, the following could be inferred: • In terms of statistical variability, all the flow number results were statistically acceptable and comparable. Thus, both systems (UTM and AMPT) exhibited acceptable in-laboratory repeatability for the flow number test and could be used with a fairly similar level of reliability. However, although the CV results between the systems were comparable, the CV values for the flow number (cycles) parameter were higher than the AASHTO TP 79-12 specification. • With the exception of the FN index parameter, the AMPT in general exhibited lower variability on the basis of its lower CV values, which indicated superior repeatability. This result was not u nexpected, among others, given the better temperature consistency of the r elatively smaller temperature chamber of the AMPT (Table 2). • The HSD and t-test statistical analyses showed statistically insignificant differences at the 90% and 95% confidence levels with the exception of the FN index parameter at 95% for data analysis on the basis of the LVDT displacement measurements with the UTM system and actuator (RAM) in the AMPT system. Therefore, either system (UTM or AMPT) could be used confidently and reliably in lieu of the other. • In general, the LVDT and axial RAM (actuator) deformation measurements could be used satisfactorily to characterize the HMA mix permanent deformation properties and compute the flow number parameters with the UTM system. Further, use of either the AMPT or the UTM did not have a significant impact on, or cause a change in, the flow number test results. Nonetheless, caution should be exercised with the FN index computation, particularly at higher confidence levels such as 95%. It would be best if the data analyses were based on actuator (RAM) displacement measurements in both systems. • For UTM–AMPT comparison purposes, flow number data analyses should have their basis in axial RAM (actuator) deformation measurements. Unlike the UTM with longer-span LVDTs, the current AMPT setup at TTI uses only the actuator (RAM) deformation measurements (without LVDTs) when the destructive flow number test is conducted, which is associated with relatively larger HMA vertical deformations. It is recommended that consideration be given to the provision of longer-span LVDTs (>1.0 mm) in the AMPT system. Dynamic Modulus Test Results and Statistical Analysis As shown in Figure 3, the |E*| mastercurves exhibited a reasonably comparable moduli overlap among the HMA mix replicate specimens from the UTM and AMPT systems, particularly at the high moduli values that corresponded to the low temperature domain. As expected theoretically, the overlap was not pronounced at the high temperature domain, in part because of the HMA’s viscoelastic nature. Therefore, caution should be exercised when results at the high tempera ture domain are analyzed and interpreted. To accurately calibrate the equipment and to use trained operators also are imperatives to generate quality laboratory dynamic modulus test results.

67

All of the results were statistically acceptable and comparable as evident in Figure 3 (i.e., CV < 30%). Thus either system could be used to yield comparable and statistically acceptable results. However, operator proficiency and equipment calibration are factors that should not be ignored. As theoretically expected given HMA’s viscoelastic nature, the AMPT CV values showed an increasing trend in the level of variability with increasing temperature. As shown in Figure 3, the lower CV values (i.e., overall average of 3.70% versus 13.06% for the UTM), of the AMPT system indicated superiority over the UTM in terms of repeatability and lower variability in the moduli values. These results were not unexpected, in part because of the better accuracy in the automated LVDT stud setup and better temperature consistency in the smaller chamber of the new AMPT unit. Thus the newer AMPT unit would be preferred to the traditional UTM, because it provided more confidence and reliability in the test results. The AMPT CV results also were consistent with the AASHTO TP 79-12 specification for DM testing with the AMPT.

General Characteristic Features General characteristic features (e.g., LVDT setup, LVDT accuracy, temperature consistency) were compared and evaluated. These aspects are discussed in the subsequent text.

HMA Mix Sample and LVDT Setup The HMA mix sample and LVDT setup (e.g., gluing the studs, cleaning) was much simpler and faster with the automated AMPT jigs than with the UTM’s manually operated jigs (Table 1). For instance, it took approximately 10 min to glue the studs and set up the LVDTs with the AMPT system for one HMA specimen. The same processes took nearly 80 min with the UTM system. Thus the AMPT system was considered more efficient and cost-effective in this respect.

Temperature Consistency and Tolerances Because of the smaller chamber (less than one-tenth that of the UTM chamber in volume), it was much quicker to obtain and maintain temperature consistency with the AMPT than with the UTM system. It took more than twice the time to heat from room temperature (approximately 25°C) to 40°C and about 1.5 times more to heat from 40°C to 50°C with the UTM system than with the AMPT system. For all the tests performed, the AMPT system exhibited better temperature consistency than the UTM system, which was attributable mainly to its smaller chamber size in volume. In the RLPD test, both systems were within the 50 ± 2°C temperature tolerance range. The example in Figure 4, however, shows less temperature fluctuation with the AMPT system (CV of 0.01% with a temperature range from 49.98°C to 50.00°C) than with the UTM system (CV of 0.46% with a temperature range of 49.90°C to 50.60°C). The need for an external chamber for multiple sample conditioning may, however, negate these AMPT characteristics in terms of cost effectiveness, which was not the case with the UTM. Given all of the HMA tests conducted in this study, the following temperature operational tolerances were noted for each system: AMPT ≤ ±0.25°C and UTM ≤ ±1.0°C. These results suggested that

TABLE 7 FN Test Results and Tukey’s HSD Statistical Analysis, UTM (RAM) Versus AMPT (RAM) HSD Pairwise Comparison UTM (RAM)

AMPT (RAM)

Average

CV

HMA Characteristic Parameter

Sample 1

Sample 2

Sample 3

Sample 1

Sample 2

Sample 3

UTM

AMPT

Flow number (cycles) εp (flow) (microstrains) Time (flow) (min) FN index (microstrains/cycle)

4,538 13,757 76 3.03

5,115 14,463 85 2.83

2,966 12,345 49 4.16

4,125 13,127 69 3.18

4,373 11,335 73 2.59

3,067 13,390 51 4.37

4,206 13,522 70 3.34

3,855 12,618 64 3.38

UTM (%)

AMPT (%)

26.45 7.97 26.45 21.53

17.99 8.86 17.99 26.72

Are the UTM–AMPT Results Significantly Different at 90% and 95% Confidence Levels? No No No No

TABLE 8 FN Test Results and Tukey’s HSD Statistical Analysis, UTM (LVDT) Versus AMPT (LVDT) HSD Pairwise Comparison UTM (LDVT)

AMPT (RAM)

Average

CV


Sample 1

Sample 2

Sample 3

Sample 1

Sample 2

Sample 3

UTM

AMPT


5,074 11,870 85 2.34

5,069 10,528 84 2.08

3,004 8,217 50 2.74

4,125 13,127 69 3.18

4,373 11,335 73 2.59

3,067 13,390 51 4.37

4,382 10,205 73 2.38

3,855 12,618 64 3.38

UTM (%)

AMPT (%)

27.23 18.11 27.23 13.90

17.99 8.86 17.99 26.7

Are the UTM–AMPT Results Significantly Different at 90% and 95% Confidence Levels? No No No No


69

TABLE 9 FN Test Results and T-Test Statistical Analysis, UTM (RAM) Versus AMPT (RAM) Are the UTM–AMPT Results Significantly Different at UTM (RAM)

AMPT (RAM)


Average (x1bar)

SD (S1)

Average (x2bar)

SD (S2)

SP


4,206 13,522 70 3.34

1,112 1,078 19 0.72

3,855 12,618 64 3.38

1,460 1,634 24 0.47

1,297.82 1,384.10 21.63 0.61

SE (X1bar − X2bar) 1,059.66 1,130.11 17.66 0.49

t=

x1 − x 2 SE

0.331 0.800 0.331 0.080

90% Confidence Level?


No No No No

No No No No

Note: SD = standard deviation; SE = standard error.

TABLE 10 FN Test Results and T-Test Statistical Analysis, UTM (LVDT) Versus AMPT (LVDT) Are the UTM–AMPT Results Significantly Different at UTM (RAM)

AMPT (RAM)


Average (x1bar)

SD (S1)

Average (x2bar)

SD (S2)

SP


4,382 10,205 73 2.38

1,193 1,848 20 0.33

3,855 12,618 64 3.38

1,460 1,634 24 0.47

1,333.32 1,744.11 22.22 0.40

the newer AMPT system was superior and more cost effective than the UTM in terms of temperature operational efficiency, which might also have contributed to the more consistent LVDT readings for the AMPT system that are discussed subsequently. LVDT Accuracy and Repeatability As shown in Figure 5 for the RLPD test as a demonstration example, the LVDT measurements from the AMPT system exhibited more consistency and repeatability than the ones for the UTM system. The CV values computed on the basis of the average LVDT measurements from three individual LVDTs were comparatively higher for the UTM than those computed from the AMPT system at both test temperatures (e.g., 34.64% versus 14.67% at 40°C and 39.50% versus 21.99% at 50°C) (Figure 5). Therefore, although the overall α and µ results were comparable and acceptable, the LVDT readings suggested more statistical confidence and reliability in the use of the AMPT system than of the UTM system (see Table 1 for the LVDT accuracy associated with the AMPT). As observed in other studies (1, 2, 12), variability in the LVDT measurements at the higher test temperature (i.e., 50°C) in general was greater than at lower test temperatures (see the CV values in Figure 5). This variability was in part attributed to the HMA viscoelastic behavior, particularly at elevated temperatures. Nonetheless, the AMPT system exhibited statistical superiority, with LVDT variability that had CV values less than 30% (i.e., 14.67% and 21.99% at 40°C and 50°C, respectively). By contrast, however, Figure 5 shows that the CV values from the UTM LVDT measurements were higher than 30% (i.e., 34.64% and 39.50% at 40°C and 50°C, respectively). However, the magnitude of the LVDT measurements indicated relatively less permanent deformation in the AMPT than in the UTM

SE (X1bar − X2bar) 1,088.65 1,424.06 18.14 0.33

t=

x1 − x 2 SE

0.484 1.694 0.484 3.019



No No No No

No No No Yes

system [e.g., 1,531 (AMPT) versus 1,806 (UTM) microstrains at 40°C (Figure 5) and 1,745 (AMPT) versus 2,613 (UTM) microstrains at 50°C]. These results may in part be attributed to the smaller AMPT chamber, which may act as confinement to the HMA mix specimen during testing. For the RLPD test, however, the final results were not affected significantly, because computation of the α and µ parameters depended predominantly on the shape characteristics of the strain–cycle response curve rather than on the strain magnitude. Synthesis and Discussion of Results After the fundamental questions and study objectives of AMPT–UMT comparability, repeatability, and reliability were addressed, a synthesis of the results presented here led to the following conclusions: 1. The results of all tests (i.e., RLPD, flow number, dynamic modulus) were statistically comparable and acceptable at the 95% confidence level in both systems. 2. Test repeatability and variability in both systems were statistically acceptable with low CV values, less than 30%. 3. Either system could be used confidently in lieu of the other to generate results of similar quality and reliability and with a comparable statistical degree of accuracy at the 95% or 90% confidence level with acceptable variability (i.e., CV < 30%). The choice and preference for one system or the other would be up to the user. Table 11 provides a subjective comparison of the AMPT and UTM systems solely on the basis of the HMA mix evaluated in this study and on the laboratory experience with these systems in this study.

Dynamic Modulus |E*| (ksi)

Reduced Frequency (Hz)

4.4

40

21.1

70

37.8

100

54.4

130

25.0 10.0 5.0 1.0 0.5 0.1 25.0 10.0 5.0 1.0 0.5 0.1 25.0 10.0 5.0 1.0 0.5 0.1 25.0 10.0 5.0 1.0 0.5 0.1

2613 2340 2154 1779 1584 1170 1543 1270 1113 802 663 431 807 557 440 235 175 102 246 159 125 73 55 38

(a) (b) FIGURE 3 DM |E*| mastercurves and statistical variability.

2753 2560 2405 2057 1898 1532 1573 1353 1191 854 731 478 625 481 387 211 162 90 211 145 107 51 41 26

17.71 18.15 18.94 20.01 21.77 24.44 8.49 2.71 3.92 4.77 4.64 6.64 22.77 20.34 19.86 14.74 14.68 16.45 9.24 7.94 7.94 7.44 5.95 13.91

1.70 1.31 1.51 1.52 2.46 4.03 0.41 1.14 1.96 2.87 3.28 4.53 3.90 4.98 6.44 8.26 9.02 9.82 2.44 0.91 0.67 2.46 4.31 8.93


71

Test Temperature (°C)

52.00

51.00

50.60

50.40

50.40

50.40

50.30 50.00

50.00

49.90

49.90

50.30

50.30

49.99

50.00

50.00 50.00

50.00

49.99

49.00

50.00

49.98

UTM (Average = 50.28°C, CV = 0.46%) AMPT (Average = 50.00°C, CV = 0.01%)

48.00 0

20

40

60 80 Test Time (min)

100

120

FIGURE 4 Comparison of temperature consistency during RLPD testing at 50 6 28C.

Permanent Microstrain (εp)

3,000

40 C

2,500 2,000

UTM

1,806 1,531

1,500

AMPT Average_LVDT (UTM)

1,000

Average_LVDT (AMPT)

500 0 0

2,000

4,000

6,000

8,000

10,000

12,000

RLPD Load Cycles (N )

(a)

Permanent Microstrain (εp)

3,000 UTM

50 C

2,500

2,613

2,000 1,745 1,500

AMPT

1,000

Average_LVDT (UTM) Average_LVDT (AMPT)

500 0 0

2,000

4,000

6,000

8,000

10,000


(b) FIGURE 5 LVDT variability comparison for RLPD testing at 408C and 508C. (continued on next page)

12,000

140

72


80%

40 C CV (UTM) (Average CV = 34.64%)

60%

CV

CV (AMPT) (Average CV = 14.67%)

40%

UTM

Average = 34.64% 20%

AMPT

Average = 14.67% 0% 0

2,000

4,000

6,000

8,000

10,000

12,000

10,000

12,000


(c) 80%

50 C

CV (UTM) (Average CV = 39.50%)

60%

CV (AMPT) (Average CV = 21.99%)

CV

UTM

40%

Average = 39.50% AMPT

20% Average = 21.99% 0% 0

2,000

4,000

6,000

8,000


(d) FIGURE 5 (continued) LVDT variability comparison for RLPD testing at 408C and 508C.

TABLE 11 Comparison of AMPT and UTM Systems System

Advantage and Application

Limitation and Challenge

AMPT

Compact system for easy mobility Small chamber for better temperature consistency Automatic LVDT setup jigs for improved efficiency and accuracy Robust LVDTs with high resolution and accuracy Shorter-span LVDTs means high accuracy for measuring relatively small displacements; see Table 1 Easy, simple, and more user-friendly software used to conduct the tests, especially for DM testing

UTM

High load cell capacity for high load applications Longer span LVDT for large deformation measurements Wider temperature range that permits testing below zero and over 60°C (from −40 to 100°C) Big temperature chamber permits the conditioning of multiple specimens Bigger temperature chamber means no need for external chamber Both the actuator (RAM) and LVDTs can sufficiently be used to measure deformations under most test methods Can accommodate different specimen dimensions and configurations, which allows for performing different tests

Relatively low load cell capacity Shorter-span LVDTs limit the measurement of larger deformations in destructive tests such as flow number (actuator [RAM] deformation measurements should instead be used to sufficiently characterize the HMA tertiary stage during flow number testing). Requires external chamber for conditioning multiple specimens Narrow temperature range limits testing below +4°C or above +60°C. Designed specifically for specimens 150 mm (6 in.) tall by 100 mm (4 in.) in diameter Manually operated LVDT setup jigs means longer setup time and less accurate and less stable LVDT measurements Bigger chambers means longer time in reaching target temperature Software not very user-friendly, especially for DM testing


Besides the limitations and challenges listed in Table 11, the newer AMPT system in general exhibited superiority in terms of • Operational efficiency, • Temperature consistency (i.e., < ±0.25 versus ±1.0°C tolerance for the UTM), • LVDT measurement consistency [it had about twice the accuracy of the UTM in terms of the CV in the three LVDT readings], • Simplicity of sample setup and practicality (i.e., it took about 1/8th the time to set up the LVDT studs), • Statistical reliability (lower CV values), and • Cost-effectiveness (i.e., setup took at least 40% less time than for the UTM) and temperature heating and cooling time (i.e., it was at least 30% more efficient than the UTM).

73

Last, operator and technician proficiency and equipment calibration are some of the most critical factors in laboratory studies of this nature and should not be ignored. Caution should be exercised when dynamic modulus tests in the high temperature domain are compared, because variability in the test results could occur in part as a result of the HMA’s viscoelastic nature. Acknowledgments The authors thank the Texas Department of Transportation and FHWA for their financial support and all those who helped during the course of this research. In particular, the authors thank Fujie Zhou for assistance with the use of the AMPT at TTI and Jacob Hoeffner, Jason Huddleston, David Contreras, and Tao-I Tang of TTI for their assistance with laboratory work and documentation.

Conclusions and Recommendations This study was undertaken to evaluate the comparability of the AMPT and UTM systems in terms of accuracy, repeatability, and reliability, and to assess if the two units could be used concurrently or interchangeably for HMA mix performance testing (i.e., RLPD, flow number, dynamic modulus). For the HMA mix evaluated, the results of the tests (i.e., RLPD, flow number, dynamic modulus) with the UTM and AMPT systems were statistically comparable and acceptable at the 90% and 95% confidence levels. The test repeatability and variability in both systems were statistically acceptable, with low CV values less than 30%. Thus either system could be used confidently in lieu of the other to generate results of similar quality and reliability and of a comparable statistical degree of accuracy with acceptable variability. Thus choice and preference should be left to the user. Each system has its own merits and limitations. As expected, however, the AMPT system, which was newer, exhibited superiority in terms of (a) operational efficiency and practicality, (b) efficiency and consistency of temperature control (i.e., less than ±0.25 versus ±1.0°C tolerance for the UTM during testing), (c) LVDT measurement consistency (about twice the accuracy of the UTM in terms of the CV values calculated on the basis of three LVDT readings), (d) simplicity of sample setup, (e) statistical reliability (i.e., lower CV values in the test parameters), and ( f ) costeffectiveness (i.e., setup time was at least 40% shorter than for the UTM), and temperature heating and cooling time (i.e., it was at least 30% more efficient than the UTM). Compared with the UTM, the limitations associated with the AMPT included the lower load cell capacity, shorter LVDT span, shorter temperature range, and the need for an external chamber to condition m ultiple specimens. Thus, if feasible, provision and installation of longer-span LVDTs (> 1 mm) would be welcome, if without compromise of the resolution and accuracy of the AMPT system to accommodate destructive testing, such as flow number. The other advantages of the UTM included the potential to simultaneously use LVDTs and the actuator (RAM) in destructive testing, such as flow number, and the ability to accommodate different specimen dimensions and configurations, which made it possible to perform different tests. To sufficiently characterize and model the HMA tertiary stage with the current AMPT system, flow number data analysis should instead have its basis in the actuator (RAM) deformation measurements.

References 1. Walubita, L. F., J. Zhang, G. Das, X. Hu, C. Mushota, A. E. Alvarez, and T. Scullion. Hot-Mix Asphalt Permanent Deformation Evaluated by Hamburg Wheel Tracking, Dynamic Modulus, and Repeated Load Tests. In Transportation Research Record: Journal of the Transportation Research Board, No. 2296, Transportation Research Board of the National Academies, Washington, D.C., 2012, pp. 46–56. 2. Walubita, L. F., S. Lee, J. Zhang, A. N. Faruk, S. T. Nguyen, and T. Scullion. HMA Shear Resistance, Permanent Deformation, and Rutting Tests for Texas Mixes: Year-1 Report. Draft Technical Report 0-6744-1. Texas A&M Transportation Institute, Texas A&M University System, College Station, 2013. 3. Zhou, F., and T. Scullion. Laboratory and Field Procedures Used to Characterize Materials. FHWA/TX-09/0-5798-P1. Texas A&M Transportation Institute, Texas A&M University System, College Station, 2009. 4. Zhou, F., E. G. Fernando, and T. Scullion. Development, Calibration, and Validation of Performance Prediction Models for the Texas M-E Flexible Pavement Design System. Technical Report 0-5798-2. Texas A&M Transportation Institute, Texas A&M University System, College Station, 2010. 5. Archilla, A. R., L. G. Diaz, and S. H. Carpenter. Proposed Method to Determine the Flow Number in Bituminous Mixtures from Repeated Axial Load Tests. Journal of Transportation Engineering, Vol. 133, No. 11, 2007, pp. 610–617. 6. Zhang, J., A. E. Alvarez, S. I. Lee, A. Torres, and L. F. Walubita. Comparison of Flow Number, Dynamic Modulus, and Repeated Load Tests for Evaluation of HMA Permanent Deformation. Journal of Construction and Building Materials, Vol. 44, 2013, pp. 391–398. 7. Pellinen, T. K., and M. W. Witczak. Stress Dependent Master Curve Construction for Dynamic (Complex) Modulus. Journal of the Association of Asphalt Paving Technologists, Vol. 71, 2002. 8. Standard Specifications for Construction and Maintenance of Highways, Streets, and Bridge. Texas Department of Transportation, Austin, 2004. 9. Walubita, L. F., and T. Scullion. Perpetual Pavements in Texas: The Fort Worth SH 114 Perpetual Pavement in Wise County. Technical Research Report 0-4822-2. Texas A&M Transportation Institute, College Station, 2007. http://tti.tamu.edu/documents/0-4822-2.pdf. The contents of this paper reflect the views of the authors, who are responsible for the facts and accuracy of the data presented here, and do not necessarily reflect the official views or policies of any agency or institute. This paper does not constitute a standard or specification, nor is it intended for design, construction, bidding, contracting, tendering, or permit purposes. Trade names were used solely for informational purposes and not for product endorsement. The Characteristics of Asphalt Paving Mixtures to Meet Structural Requirements Committee peer-reviewed this paper.