Improved Test Methods for Specific Gravity and Absorption of Coarse and Fine Aggregate (2015)

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.

5 C H A P T E R 3 As shown in Table 2-4, 10 candidate test methods were selected for laboratory evaluation. However, since a produc- tion unit of the SG-5 device was not available for evaluation at the time of this study, the list of candidate test methods was revised. Table 3-1 shows the revised list. The laboratory program was originally planned in two experiments, but was later expanded to five experiments. Experiment 1 was a preliminary evaluation to compare results and variability of the test methods listed in Table 3-1. Experi- ment 2 was to further evaluate the test methods selected at the conclusion of Experiment 1 with a broader range of aggregate materials. After reviewing the results of Experiments 1 and 2, Experi- ments 3, 4, and 5 were added to the laboratory testing program to answer specific questions related to the test procedures previously evaluated. Experiment 3 was added to evaluate modifications relative to the drying and soaking methods in AASHTO T 85 and T 84 to reduce the testing time. Experi- ment 4 was added to determine the effect of P200 material on AASHTO T 84 test results. Experiment 5 was added to investigate time-zero reading for Phunque methods. Results of Experiments 1 through 4 are presented in this chapter, and results of Experiment 5 are included in Appendix H, which is available on the project web page. Experiment 1 The objective of Experiment 1 was to compare results and variability of the 10 test methods. Based on the results of this evaluation, promising test methods would be selected for a more detailed evaluation later in the study. Laboratory Testing The 10 test methods listed in Table 3-1 were used to deter- mine the specific gravity and water absorption of the coarse, fine, and combined aggregates shown in Table 3-2. The aggre- gates were selected to include a variety of absorption capacities, shapes, and textures. One operator conducted all testing required in this experiment. The operator was trained to conduct all of the tests, and the equipment was calibrated before testing. Test samples were carefully prepared to ensure they were consistent and homogeneous. Extra samples were prepared for the techni- cian to practice on before running the tests. After samples were prepared, sample numbers were randomized before they were tested. The test methods also were conducted in a randomized order. Triplicate samples were tested using each material and method. To aid in the analysis, the gradation and angularity tests were conducted for all the aggregates, petrographic analy- ses were conducted on natural sand, granite, gravel, and lime- stone aggregates, and the test for insoluble residue in carbonate aggregates was performed on natural sand and limestone aggre- gates. Detailed test results of Experiment 1 are included in Appendix D, which is available on the project web page. Analysis of Test Results for Coarse Aggregates Figure 3-1 compares the specific gravity and water absorp- tion results measured by the three test methods for coarse aggregate—AASHTO T 85, Rapid AASHTO T 85 with the CoreLok vacuum saturation, and Phunque Flask. The results measured by the three test methods appeared more comparable for two aggregates with lower water absorption capacities— Columbus granite (AASHTO T 85 water absorption = 0.43 per- cent) and Elmore gravel (AASHTO T 85 water absorption = 1.13 percent). For Florida limestone (AASHTO T 85 water absorption = 5.53 percent), the Rapid AASHTO T 85 method yielded results that were significantly different from the other two methods, especially the absorption capacity shown in Figure 3-1(d). Figure 3-2 compares the standard deviations of specific grav- ity and water absorption results for the three test methods for coarse aggregate. The single-operator precision (i.e., 1S) limits Evaluation of Candidate Test Methods (text continued on page 8)

6Table 3-1. Revised list of test methods selected for laboratory evaluation. ID Selected Test Method Material Used for Evaluation I. Test Methods for Determining Specific Gravity and Absorption of Coarse Aggregate 1 AASHTO T 85 and ASTM C127 Coarse aggregate retained on No. 4 sieve 2 Rapid AASHTO T 85 with the CoreLok Coarse aggregate retained on No. 4 sieve 3 Volumetric Immersion using Phunque Flasks Coarse aggregate retained on No. 4 sieve II. Test Methods for Determining Specific Gravity and Absorption of Fine Aggregate 1 AASHTO T 84 and ASTM C128 Fine aggregate passing No. 4 sieve 2 Modification to Materials Tested in AASHTO T 84/ASTM C128 Fine aggregate passing No. 4 sieve and retained on No. 200 sieve 3 SSDetect System Fine aggregate passing No. 4 sieve 4 Modification to Materials Tested in SSDetect System Fine aggregate passing No. 4 sieve and retained on No. 200 sieve 5 Volumetric Immersion using Phunque Flasks Fine aggregate passing No. 4 sieve * AASHTO T 133 (with Ethyl Alcohol) Fine material passing No. 200 sieve III. Test Methods for Determining Specific Gravity and Absorption of Combined Aggregate 1 Volumetric Immersion using Phunque Flasks Combined (coarse and fine) aggregate *This test method is used to determine the apparent specific gravity of fine material passing the No. 200 sieve. Table 3-2. Aggregate materials used in Experiment 1. ID Materials (Abbreviation) Source Factors Considered I. Coarse Aggregates 1 Columbus granite (CG) Barin Quarry, Columbus, GA Very low absorption 2 Elmore gravel (EG) Elmore Sand and Gravel, AL Medium absorption 3 Florida limestone (FL) CEMEX, Brooksville, FL Very high absorption II. Fine Aggregates 1 Ottawa sand (OS) Ottawa, IL Low angularity, low P200, low absorption 2 Natural sand (NS) Foley Materials, Phenix City, AL High angularity, low P200, low absorption 3 Limestone screenings (LS) CEMEX, Brooksville, FL High angularity, high P200, high absorption 4 Granite screenings (GS) Barin Quarry, Columbus, GA High angularity, high P200, low absorption III. Combined Aggregates 1 CG + OS 50% Columbus granite and 50% Ottawa sand 2 EG + NS 60% Elmore gravel and 40% natural sand 3 FL + GS 60% Florida limestone and 40% granite screenings Figure 3-1. Comparison of results determined by three test methods for coarse aggregate. (continued on next page) (a) Gsa (b) Gsb 1.90 2.00 2.10 2.20 2.30 2.40 2.50 2.60 2.70 2.80 0 Test Method G sa Columbus Granite Elmore Gravel Florida Limestone Rapid T 85AASHTO T 85 Phunque Flask 1.90 2.00 2.10 2.20 2.30 2.40 2.50 2.60 2.70 2.80 Test Method G sb Columbus Granite Elmore Gravel Florida Limestone Rapid T 85AASHTO T 85 Phunque Flask

7 Figure 3-2. Variability of test results determined by three test methods for coarse aggregate. (a) Variability of Gsa (b) Variability of Gsb (c) Variability of Gssd (d) Variability of Absorption 0.000 0.001 0.002 0.003 0.004 0.005 0.006 0.007 0.008 0.009 0.010 Test Method G sa S ta nd ar d De vi at io n Columbus Granite Elmore Gravel Florida Limestone Rapid T 85AASHTO T 85 Phunque Flask AASHTO T 85 Precision 0.000 0.001 0.002 0.003 0.004 0.005 0.006 0.007 0.008 0.009 0.010 Test Method G sb S ta nd ar d De vi at io n Columbus Granite Elmore Gravel Florida Limestone Rapid T 85AASHTO T 85 Phunque Flask AASHTO T 85 Precision 0.000 0.001 0.002 0.003 0.004 0.005 0.006 0.007 0.008 0.009 0.010 Test Method G ss d St an da rd D ev ia tio n Columbus Granite Elmore Gravel Florida Limestone Rapid T 85AASHTO T 85 Phunque Flask AASHTO T 85 Precision 0.000 0.050 0.100 0.150 0.200 0.250 0 A bs or pt io n St an da rd D ev ia tio n (% ) Test Method Columbus Granite Elmore Gravel Florida Limestone Rapid T 85AASHTO T Phunque Flask AASHTO T 85 Precision Figure 3-1. (Continued). (c) Gssd (d) Absorption 1.90 2.00 2.10 2.20 2.30 2.40 2.50 2.60 2.70 2.80 Test Method G ss d Columbus Granite Elmore Gravel Florida Limestone Rapid T 85AASHTO T 85 Phunque Flask 0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 0 Test Method W at er A bs or pt io n (% ) Columbus Granite Elmore Gravel Florida Limestone Rapid T 85AASHTO T 85 Phunque Flask

8in AASHTO T 85 also are shown in Figure 3-2 for comparison. The AASHTO T 85 precision limits were determined based on aggregates with absorptions of less than 2 percent. In Experi- ment 1, the three test methods met the single-operator preci- sion (1S) limits specified in AASHTO T 85 for specific gravity measurements. However, only the test results determined in accordance with AASHTO T 85 in Experiment 1 met the single- operator precision 1S limit for water absorption for all aggre- gates tested, as shown in Figure 3-2(d). The repeatability of water absorption results for the Phunque flask is not as good as those of the other two methods. However, the Phunque method yielded the most repeatable results for Gsb. Further statistical analyses were conducted for the test results to answer two questions: • Is there a statistically significant difference in the test results (Gsa, Gsb, Gssd, and absorption) determined using the three test methods for each material? • Is there a statistically significant difference in the variance of the results determined using the three test methods (or difference in the repeatability of the three tests)? To answer the first question, analyses of variance (ANOVA) (significance level = 0.05) were conducted. Table 3-3 shows a summary of ANOVA results. In Table 3-3, the three test meth- ods produced statistically different mean results, except for Gsa of Columbus granite (CG). The higher the F-statistic, the more significant the difference would be. The three methods yielded the most significantly different results for the Florida limestone material. This observation was previously shown in Figure 3-1. Bartlett-Box’s F-test (significance level = 0.05) was con- ducted to answer the second question. Table 3-4 summarizes Bartlett-Box’s test results for the variances. The variances of the Gsb and water absorption results determined using the three test methods were statistically different, and the differ- ence in the variances of water absorption was more significant. In summary, the three test methods for coarse aggregate evaluated in this experiment yielded comparable test results for the low absorption aggregate (i.e., the Columbus granite). However, the differences in the test results were greater for higher absorption aggregate materials, especially the Florida limestone. The reasons are briefly discussed as follows: • The AASHTO T 85 method uses dry/damp towels to remove water from the surface of aggregate particles to reach the SSD condition after they have been pre-soaked for 15 to 19 hours. For a high absorption aggregate material (e.g., the Florida limestone), this drying process may remove some water in the surface (permeable) voids of each aggregate particle, resulting in a lower volume of permeable voids, which would yield a higher Gsb value and a lower absorp- tion capacity than the true values. • For the Rapid AASHTO T 85 with the CoreLok vacuum sat- uration, a dry aggregate sample is vacuum saturated; thus, water may penetrate into deeper voids because the pressure inside the surface voids is much lower than the atmosphere pressure. This produces a much higher volume of perme- able voids, resulting in a lower Gsb and a higher absorption capacity. • For the Phunque method, the volume of aggregate plus vol- ume of impermeable and permeable voids is determined at 30 seconds after the first particle has entered the water and Material Response P-Value Significant? Grouping Using Tukey's Test T 85 Rapid T 85 Phunque CG Gsa 0.151 No A A A CG Gsb 0.001 Yes B B A CG Gssd 0.004 Yes B B A CG Abs 0.024 Yes B A A, B EG Gsa 0.001 Yes B A B EG Gsb <0.001 Yes B B A EG Gssd <0.001 Yes C B A EG Abs 0.001 Yes B A B FL Gsa <0.001 Yes C A B FL Gsb <0.001 Yes B C A FL Gssd <0.001 Yes B C A FL Abs <0.001 Yes B A C Note: For each measured property and material, methods that do not share a letter are significantly different. A and C represent the highest and lowest values, respectively. Table 3-3. Summary of ANOVA for coarse aggregate test results. Table 3-4. Summary of Bartlett’s Test for variance of test results. Response Bartlett’s Test Test Statistic P-Value Significant? Gsa 5.26 0.072 No Gsb 7.71 0.021 Yes Gssd 4.40 0.111 No Abs 14.48 0.001 Yes

9 the flask has not been disturbed. There may be two short- comings for this determination: – Air may be trapped between aggregate particles, result- ing in a greater measured volume than the true volume of aggregate and voids. This would result in a lower Gsb and a higher absorption value. – For high absorption aggregates (e.g., Florida limestone), water may penetrate into the surface voids within the first 30 seconds, resulting in a lower measured volume of aggregate and voids and yielding a higher Gsb value and a lower absorption capacity. • Since the aggregate is slowly introduced into the flask, the second shortcoming seems to be more probable; thus, the Phunque method would likely yield a higher Gsb value and a lower absorption capacity than the true values. Compared to the AASHTO T 85 method, the Phunque method does not appear to provide improved accuracy, repeat- ability, or time required for testing. Future implementation of the Phunque method would also require an investment in testing equipment and training. The Rapid T 85 method has the potential to signifi- cantly reduce testing time and yield results comparable to AASHTO T 85 for low absorption aggregates (i.e., less than 2 percent). However, the cost of a CoreLok device should be considered for future implementation. For absorptive aggre- gate, the Rapid T 85 method is believed to yield inaccurate results due to the penetration of water into aggregate voids that would not be permeable under hydrostatic conditions. Analysis of Test Results for Fine Aggregates Figure 3-3 shows comparisons of the specific gravity and water absorption results measured by the five test methods— AASHTO T 84, Modified AASHTO T 84, SSDetect, Modi- fied SSDetect, and Phunque Flask—for fine aggregate and AASHTO T 133 for minus No. 200 materials (P200). Since the Ottawa sand and the natural sand did not contain an appre- ciable amount of P200 (less than 1.0 percent), they were not tested with the Modified T 84 or Modified SSDetect meth- ods. Also, there was not sufficient P200 material from these sources to test using AASHTO T 133 (with ethyl alcohol). Figure 3-3. Comparison of results determined by test methods for fine aggregate. (a) Gsa (b) Gsb 2.10 2.20 2.30 2.40 2.50 2.60 2.70 2.80 Test Method G sa Ottawa Sand Natural Sand Limestone Screening Granite Screening SSDetectT 84 Phunque Flask Modified T 84 Modified SSDetect T 133 2.10 2.20 2.30 2.40 2.50 2.60 2.70 2.80 Test Method G sb Ottawa Sand Natural Sand Limestone Screening Granite Screening SSDetectT 84 Phunque FlaskModified T 84 Modified SSDetect (c) Gssd 2.10 2.20 2.30 2.40 2.50 2.60 2.70 2.80 Test Method G ss d Ottawa Sand Natural Sand Limestone Screening Granite Screening SSDetectT 84 Phunque FlaskModified T 84 Modified SSDetect (d) Absorption 0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 Test Method W at er A bs or pt io n (% ) Ottawa Sand Natural Sand Limestone Screening Granite Screening SSDetectT 84 Phunque FlaskModified T 84 Modified SSDetect

10 The results determined by these test methods for fine aggre- gate were more comparable for the three aggregates with lower water absorption capacities—Ottawa sand (AASHTO T 84 water absorption = 0.03 percent), natural sand (AASHTO T 84 water absorption = 0.5 percent), and granite screenings (AASHTO T 84 water absorption = 1.5 percent). However, these methods yielded significantly different results for the limestone screenings (AASHTO T 84 water absorption = 5.0 percent). When comparing the test results conducted on the materials with and without the P200, the differences between the SSDetect and Modified SSDetect results were less than those between the AASHTO T 84 and Modified AASHTO T 84, as shown in Figure 3-3. Figure 3-4 compares the variability of test results for the selected test methods for fine aggregate. The dash lines in these figures represented the single-operator precision 1S limits in AASHTO T 84. As noted in the standard, these specified preci- sion limits were estimated based on manufactured fine aggre- gates with absorptions of less than 1 percent. It appears that all the test methods can meet the specified precision in AASHTO T 84 for the Ottawa and natural sands but not for the granite and limestone screenings. This indicates that the precision informa- tion for AASHTO T 84 should be updated to include data for fine aggregates having water absorptions greater than 1 percent. In addition to the graphical comparisons shown in Fig- ures 3-3 and 3-4, further statistical analysis was conducted for the test results of fine aggregates. The analysis was conducted to evaluate • If there was any statistically significant difference in the test results (Gsa, Gsb, Gssd, and absorption) determined using the five test methods for each material; and • If there was any statistically significant difference in the variance of the results (Gsa, Gsb, Gssd, and absorption) determined using the five test methods. Table 3-5 shows a summary of ANOVA (significance level = 0.05) of the test results. This analysis shows that the five methods yield statistically different results. The largest differences were for the Florida limestone. (a) Variability of Gsa (b) Variability of Gsb (c) Variability of Gssd (d) Variability of Absorption 0.000 0.002 0.004 0.006 0.008 0.010 0.012 0.014 0.016 0.018 0.020 Test Method G sa S ta nd ar d De vi at io n Ottawa Sand Natural Sand Limestone Screening Granite Screening SSDetectT 84 Phunque Flask Modified T 84 Modified SSDetect T 133 0.000 0.002 0.004 0.006 0.008 0.010 0.012 0.014 0.016 0.018 0.020 Test Method G sb S ta nd ar d De vi at io n Ottawa Sand Natural Sand Limestone Screening Granite Screening SSDetectT 84 Phunque FlaskModified T 84 Modified SSDetect 0.000 0.002 0.004 0.006 0.008 0.010 0.012 0.014 0.016 0.018 0.020 Test Method G ss d St an da rd D ev ia tio n Ottawa Sand Natural Sand Limestone Screening Granite Screening SSDetectT 84 Phunque Flask Modified T 84 Modified SSDetect 0.0000 0.0500 0.1000 0.1500 0.2000 0.2500 0.3000 0.3500 Test Method A bs or pt io n St an da rd D ev ia tio n (% ) Ottawa Sand Natural Sand Limestone Screening Granite Screening SSDetectT 84 Phunque Flask Modified T 84 Modified SSDetect Figure 3-4. Variability of test results determined by five test methods for fine aggregate.

11 Table 3-6 summarizes Bartlett’s Test for evaluating the vari- ance of test results determined using the five test methods. The variances of the test results were not statistically different. In Experiment 1, AASHTO T 84 and SSDetect were con- ducted on the limestone and granite screenings containing the P200 fractions, and Modified T 84 and Modified SSDetect were conducted on these screenings without the P200 frac- tions. Thus, these test results may not be comparable. A fur- ther analysis was conducted by mathematically combining the Gsa, Gsb, and Gssd measured using Modified T 84 and Modi- fied SSDetect with the Gsa of the P200 fractions determined in accordance with AASHTO T 133 (with ethyl alcohol). These values are compared in Figure 3-5. These graphs indicated that both T 84 and SSDetect were influenced by removing the P200 from the samples and testing them separately. In summary, the following observations can be offered based on the analysis of the test results for fine aggregates: • The logic of the cone and tamping method for determin- ing the SSD condition for fine aggregates is that the cone of aggregate (~71.6°) is greater than the angle of repose for dry material but less than the angle of repose for aggregate with capillary water between particles. However, the angle of repose of fine aggregate is influenced by several material properties, including the absorption, angularity, and amount and nature (e.g., plasticity) of the P200 fraction. A compari- son of the difference between AASHTO T 84 and Modi- fied AASHTO T 84 results with that between SSDetect and Modified SSDetect results shows that the presence of P200 in a fine aggregate has more effect on the AASHTO T 84 results, and it has less effect on the SSDetect test results. • The SSDetect method for determining the SSD condition of a fine aggregate material does not appear to depend on angularity or presence of P200. • The Phunque flask yields results that are comparable to those of the other test methods. • All of the test methods have problems determining repeat- able absorption results for Florida limestone, which is con- sidered a problematic material in this study. No test method consistently produced more repeatable test results than AASHTO T 84. • Since the SSDetect and Modified SSDetect methods produce comparable results, the SSDetect method may be used to test fine aggregate materials that include the P200 fraction. • In consideration of implementation costs, the most expensive option is the SSDetect method, then the Phunque method and the Modified AASHTO T 84. However, implementation of Modified AASHTO T 84 would require another test for the P200 fraction. Analysis of Test Results for Combined Aggregate Testing also was conducted to evaluate if the Phunque flask used for testing coarse aggregate could be used for testing Material Response P-Value Significant? Grouping Using Tukey's Test T 84 Mod. T 84 SSDetect Mod. SSDetect Phunque GS Gsa 0.079 No A A A A A GS Gsb <0.001 Yes C A B A, B A, B GS Gssd <0.001 Yes C A B A, B A, B GS Abs <0.001 Yes A C C C B LS Gsa <0.001 Yes C A B, C B A, B LS Gsb <0.001 Yes D E C B A LS Gssd <0.001 Yes C D B B A LS Abs <0.001 Yes B A C C B NS Gsa 0.525 No A A A NS Gsb 0.015 Yes B B A NS Gssd 0.072 No A A A NS Abs 0.016 Yes A, B B A OS Gsa 0.001 Yes A B B OS Gsb <0.001 Yes A C B OS Gssd 0.001 Yes A B B OS Abs <0.001 Yes B B A Note: Methods that do not share a letter are significantly different. A and E represent the highest and lowest values, respectively. The Modified AASHTO T 84 and Modified SSDetect were not conducted on NS and OS. Table 3-5. Summary of ANOVA for test results of fine aggregate. Response Bartlett’s Test Test Statistic P-Value Significant? Gsa 2.67 0.614 No Gsb 8.17 0.086 No Gssd 8.13 0.087 No Abs 4.50 0.343 No Table 3-6. Summary of Bartlett’s Test for variance of test results.

12 combined aggregate (i.e., a complete gradation). Figure 3-6 shows comparisons of combined specific gravity and water absorption values measured using the Phunque flask and cal- culated using the AASHTO T 85 and T 84 results and Phunque method results for individual coarse and fine aggregates. In Figure 3-6(a), two data points that were separated from the line of equality were for combined materials that included Florida limestone. The Phunque method has the potential to determine the specific gravity and absorption for a complete gradation. The drawbacks for implementation of this method include the time required for testing, additional cost for equipment, ruggedness of the current flask, and yielding significantly different results determined by AASHTO T 85 and T 84 for a complete grada- tion that includes a high absorption aggregate material. Summary Based on the key findings of Experiment 1, a comparison of the candidate test methods is shown in Table 3-7. After reviewing the results of Experiment 1, the panel selected six test methods for a detailed evaluation in Experiment 2. The other three test methods were not further evaluated in this study because of additional costs of the CoreLok equipment and bags, the repeatability of SSDetect not being improved by removing the P200 fraction, and the poor precision of the Phunque method for testing a complete gradation. The fol- lowing six test methods were chosen for Experiment 2: • Test Methods for Determining Specific Gravity and Absorp- tion of Coarse Aggregate 1. AASHTO T 85-08, Test for Specific Gravity and Absorp- tion of Coarse Aggregate 2. AASHTO TP 77-09, Test for Specific Gravity and Absorp- tion of Aggregate by Volumetric Immersion Method (Phunque Flask for Coarse Aggregate) • Test Methods for Determining Specific Gravity and Absorp- tion of Fine Aggregate 3. AASHTO T 84-09, Test for Specific Gravity and Absorp- tion of Fine Aggregate (a) Gsa (b) Gsb (c) Gssd 2.10 2.20 2.30 2.40 2.50 2.60 2.70 2.80 Test Method G sa Limestone Screening Granite Screening SSDetectT 84 Modified T 84 + T 133 Modified SSDetect Modified T 84 Modified SSDetect + T 133 2.10 2.20 2.30 2.40 2.50 2.60 2.70 2.80 Test Method G sb Limestone Screening Granite Screening SSDetectT 84 Modified T 84 + T 133 Modified SSDetect Modified T 84 Modified SSDetect + T 133 2.10 2.20 2.30 2.40 2.50 2.60 2.70 2.80 Test Method G ss d Limestone Screening Granite Screening SSDetectT 84 Modified T 84 + T 133 Modified SSDetect Modified T 84 Modified SSDetect + T 133 Figure 3-5. Comparison of test results for fine aggregate with and without P200 fraction.

13 Figure 3-6. Comparison of test results for combined aggregates. (a) Specific Gravities (b) Absorption 2.40 2.45 2.50 2.55 2.60 2.65 2.70 2.75 2.40 2.45 2.50 2.55 2.60 2.65 2.70 2.75 Measured Properties (Phunque Method for Combined Agg) Ca lc ul at ed P ro pe rti es Gsa_Meas vs Calc AASHTO Gsa_Meas vs Calc Phunque Gsb_Meas vs Calc AASHTO Gsb_Meas vs Calc Phunque Gssd_Meas vs Calc AASHTO Gssd_Meas vs Calc Phunque Line of Equality 0.00 0.50 1.00 1.50 2.00 2.50 3.00 0.00 0.50 1.00 1.50 2.00 2.50 3.00 Measured Absorption using Phunque Method (%) Ca lc ul at ed A bs or pt io n (% ) Abs_Meas vs Calc AASHTO Abs_Meas vs Calc Phunque Line of Equality Include Florida limestone ID Test Method Comments on Accuracy Gsb Pooled Std. Dev. Equipment Ruggedness Ease of Use Time Eqmt. Cost Overall Comments Total Operator I. Test Methods for Coarse Aggregate 1 AASHTO T 85 and ASTM C127 Yields higher Gsb and lower absorption for absorptive aggregate 0.0053 Good Manual 3 days 30 min. $100 ~ $600 Standard 2 Rapid AASHTO T 85 with the CoreLok Yields lower Gsb and higher absorption for absorptive aggregate 0.0065 Good Manual 2 days 30 min. $6,860 Costs for equipment and bags; not accurate for absorptive aggregate 3 Volumetric Immersion using Phunque Flask Yields higher Gsb and lower absorption for absorptive aggregate 0.0029 Fragile in current design Manual 2 days 2 hrs $500 Small equipment cost; questionable accuracy for absorptive aggregate, may be more repeatable than the T 85 II. Test Methods for Fine Aggregate 1 AASHTO T 84 and ASTM C128 Depends on P200 content, angularity, absorption of aggregate 0.0056 Good Manual 3 days 1.5 hrs $100 ~ $300 Standard 2 Modification to Materials Tested in AASHTO T 84 Depends on angularity, absorption of aggregate 0.0081 Good Manual 3 days 1.5 hrs $100 ~ $300 Does not need new equipment, but requires another test for minus #200 material 3 SSDetect System Depends on absorption of aggregate 0.0050 Good Auto 1 day 1 hr $7,056 Equipment cost, faster results 4 Modification to Materials Tested in SSDetect Depends on absorption of aggregate 0.0099 Good Auto 1 day 1 hr $7,056 No apparent advantage over regular SSDetect method 5 Volumetric Immersion using Phunque Flask Depends on absorption; measure SSD volume within 30 sec 0.0077 Fragile in current design Manual 2 days 2 hrs $500 Small equipment cost, somewhat fragile flasks, less repeatable than T 84 III. Test Method for Complete Gradation 1 Volumetric Immersion using Phunque Flasks Depends on absorption; measure SSD volume within 30 sec 0.0207 Fragile in current design Manual 2 days 2 hrs $500 Small additional equipment cost; lower precision Table 3-7. Comparison of test methods for determining aggregate specific gravity and absorption.

14 4. Modified AASHTO T 84 (Removal of P200), the same as AASHTO T 84, except that the P200 fraction is removed before testing 5. ASTM D7172-06, Test for Determining the Relative Den- sity (Specific Gravity) and Absorption of Fine Aggregates Using Infrared (SSDetect System) 6. AASHTO TP 77-09, Test for Specific Gravity and Absorp- tion of Aggregate by Volumetric Immersion Method (Phunque Flask for Fine Aggregate) Experiment 2 The objective of Experiment 2 was to further evaluate the test methods selected at the conclusion of Experiment 1 with a broader range of aggregate materials. Based on the results of this experiment, precision information, including repeat- ability and reproducibility, would be developed for each test method. The most promising test methods would be selected for further evaluation later in the project. Laboratory Testing This experiment was designed to allow for a statistical analysis addressing the observed variability in the data and preparation of precision information for two test methods for coarse aggregate and four test methods for fine aggregate. This experiment included three operators reviewing the test methods independently and conducting all testing using three different test devices to simulate multi-laboratory operations for evaluating the precision of the selected test methods. As shown in Table 3-8, the materials used in Experiment 2 covered a broader range of aggregate materials used for PCC and HMA mixtures, including crushed, natural, recycled, and manufactured coarse and fine aggregates. The coarse aggregate materials covered a wide range of water absorption capacity that was found to have a significant effect on the variability of test results. The fine aggregate materials had a wide range of amount of materials passing the No. 200 sieve (P200) and water absorption. Three replicates were tested for each factor. For each test method, with three operators and three replicates, nine observations (3 operators and devices × 3 replicates) were determined for each material level. A total of 270 tests were conducted in the laboratory to complete this experimental design, providing data to determine preliminary repeatability and reproducibility of each test method. Test samples were carefully prepared to ensure they were consistent and homogeneous, especially for materials with high P200 content. Extra samples were prepared for the three technicians to practice on before testing. After the samples had been prepared, they were randomly numbered. Three operators and three test devices were used to simu- late a multi-laboratory study. In this experiment, the three operators were all NCAT employees. The order in which each of the test methods was conducted also was randomized for each operator. Before starting Experiment 2, all testing devices were calibrated, and all technicians involved in this program were carefully trained to operate the test devices. After the laboratory study had started, each operator conducted the tests according to the standardized procedures. No further instructions for performing the tests were given, except that forms and instructions for recording the data were developed and distributed to each operator. To aid in the analysis of test results determined in Experi- ment 2, the research team also performed additional tests, as shown in Table 3-9, to characterize gradation, fine aggregate angularity, and other geological and mineralogical proper- ties of the aggregate materials used. Detailed test results of Experiment 2 are included in Appendix E, which is available on the project web page. Results and Analysis Factors Affecting Test Results Analysis of variance (ANOVA) was conducted to determine the effects of material, test method, and operator on the test results, including Gsa, Gsb, Gssd, and water absorption. The Table 3-8. Sources of aggregate materials used in Experiment 2. ID Materials Source I Coarse Aggregates 1 Elmore gravel (EGC) Elmore Sand and Gravel, Elmore, AL 2 Preston sandstone (PSC) Arkhola Sand and Gravel, Fort Smith, AR 3 Blast furnace slag (FSC) Mountain Enterprises, Hager Hill, KY 4 RC limestone (RCC) SHRP Ref. Material from McAdams LS Products, KS 5 Recycled concrete (REC) Recycled Materials, Denver, CO II Fine Aggregates 1 Rounded natural sand (NSF) Dredged from Arkansas River at Van Buren, Crawford County, AR 2 Blast furnace slag (FSF) Mountain Enterprises, Hager Hill, KY 3 Preston sandstone (PSF) Arkhola Sand and Gravel, Fort Smith, AR 4 Texas limestone sand (TLF) 1604 Operations (Vulcan Materials), San Antonio, TX 5 RC limestone (RCF) SHRP Ref. Material from McAdams LS Products, KS

15 ANOVA results for coarse and fine aggregate materials are summarized in Tables 3-10 and 3-11, respectively. Detailed ANOVA results are included in Appendix E, which is available on the project web page. Based on the ANOVA results, the effect of material, test method, and simulated lab was statistically significant on all the test results at a significance level of 0.05, except for the effect of operator on water absorption for coarse aggregate. The F-statistics for the first two sources of effect (material and test method) were much higher than those of the third effect (simulated lab) on all measured responses. Therefore, to simplify analysis and presentation, the test results were later evaluated only at each combined level of material and test method. Table 3-9. Additional tests for characterizing properties of materials for Experiment 2. Material Test Methods Grad. Angularity Insoluble Residue Float Test Elemental Analysis Sand Equivalent Petrograph Exam I Coarse Aggregate 1 Elmore gravel (EGC) x x 2 Preston sandstone (PSC) x x 3 Blast furnace slag (FSC) x x 4 RC limestone (RCC) x x x 5 Recycled concrete (REC) x x1 II Fine Aggregate 1 Rounded natural sand (NSF) x x x3 x x2 2 Blast furnace slag (FSF) x x x x 3 Preston sandstone (PSF) x x x x 4 Texas limestone sand (TLF) x x x x x 5 RC limestone (RCF) x x x x x Notes: 1 Estimate percent cement paste; 2 Determine clay and shale content; 3 Determine light-weight particles. Table 3-10. Summary of ANOVA results for coarse aggregate. Response Source of Effect F-Statistic P-Value Gsa Material (BF Slag Coarse, Elmore Gravel, Sandstone Coarse, RC LS Coarse, Recycled Concrete) 904.20 <0.0001 Test Method (Phunque Coarse, T 85) 188.59 <0.0001 Simulated Lab (3 Operators, 3 Devices) 7.01 0.002 Gsb Material 3322.81 <0.0001 Test Method 3125.38 <0.0001 Simulated Lab 6.48 0.003 Gssd Material 1993.85 <0.0001 Test Method 2108.39 <0.0001 Simulated Lab 8.11 0.001 Absorption Material 1820.27 <0.0001 Test Method 3752.23 <0.0001 Simulated Lab 0.71 0.497 Table 3-11. Summary of ANOVA results for fine aggregate. Response Source of Effect F-Statistic P-Value Gsa Material (BF Slag Fine, Natural Sand, Sandstone Fine, RC LS Fine, TX LS Sand) 6916.48 <0.0001 Test Method (Modified T 84, Phunque Fine, SSDetect, T 84) 46.94 <0.0001 Simulated Lab (3 Operators, 3 Devices) 16.04 <0.0001 Gsb Material 879.07 <0.0001 Test Method 163.04 <0.0001 Simulated Lab 6.36 0.002 Gssd Material 1869.02 <0.0001 Test Method 164.95 <0.0001 Simulated Lab 4.96 0.008 Absorption Material 465.33 <0.0001 Test Method 135.21 <0.0001 Simulated Lab 10.24 <0.0001

16 Comparing Test Results for Coarse Aggregate Pairwise comparisons among different levels of material and test method were conducted using Tukey’s tests. A summary of Tukey’s test results for coarse aggregate materials is presented in Table 3-12. The “difference of means” values shown in Table 3-12 indicate that the Phunque method yielded higher Gsa, Gsb, and Gssd results and lower water absorption values than the AASHTO T 85 method. Based on a significance level of 0.05, the differences in the test results between the Phunque flask and AASHTO T 85 methods were statistically significant for each of the coarse aggregate materials tested in this study, as shown in the “significant” column of Table 3-12. These differences can be seen in Figure 3-7, which presents graphical comparisons of the average Gsa, Gsb, Gssd, and water absorption results for the five coarse aggregate materials tested in Experiment 2. The differences in Gsb and Gssd values for the two methods were considered significant from a practical point of view. Part of the differences can be attributed to how the two methods account for absorption. In essence, the Phunque method does not take into account the water absorption that takes place from the time when the first aggregate particles enter the water in the flask until the initial reading is made of the water level in the neck of the flask. This may be as much as 30 seconds. Obviously, water is absorbed into the aggregate particles during this time. From Figure 3-7(d), it can be seen that the materials with the greatest differences in water absorption were the recycled concrete and the blast furnace slag. The aggregate particles for these materials have the larg- est pore sizes and therefore are capable of absorbing water more quickly than the other materials, resulting in lower water absorption values. To correct the problem, timing for the initial water level reading should be varied based on the absorption of each material. However, it is difficult to shorten the 30-second period for the initial reading because it would take that much time to introduce all of the sample into the flask, clean the neck, and take the reading. Further analysis of the test results using a correlation between the water level and time at which the water level reading is taken is conducted in Experiment 5 to estimate the water level prior to 30 seconds. Comparing Test Results for Fine Aggregate Tukey’s tests were also conducted for test results of fine aggregate materials. A summary of Tukey’s test results for Gsa of fine aggregates is presented in Table 3-13. Based on a Table 3-12. Tukey’s pairwise comparisons for coarse aggregate test results. Response Variable Gsa Difference Adjusted Materials Method of Means T-Value P-Value Significant? BF Slag Coarse T 85 - Phunque -0.04912 -15.83 <0.0001 Yes Elmore Gravel T 85 - Phunque -0.01189 -3.83 0.0107 Yes PS Coarse T 85 - Phunque -0.01021 -3.29 0.0492 Yes RC LMS Coarse T 85 - Phunque -0.01218 -3.92 0.0080 Yes Recycled Concrete T 85 - Phunque -0.01188 -3.830 0.0107 Yes Response Variable Gsb Difference Adjusted Materials Method of Means T-Value P-Value Significant? BF Slag Coarse T 85 - Phunque -0.1189 -31.45 <0.0001 Yes Elmore Gravel T 85 - Phunque -0.0609 -16.12 <0.0001 Yes Sandstone Coarse T 85 - Phunque -0.0765 -20.24 <0.0001 Yes RC LS Coarse T 85 - Phunque -0.0866 -22.92 <0.0001 Yes Recycled Concrete T 85 - Phunque -0.1296 -34.28 <0.0001 Yes Response Variable Gssd Difference Adjusted Materials Method of Means T-Value P-Value Significant? BF Slag Coarse T 85 - Phunque -0.09014 -28.33 <0.0001 Yes Elmore Gravel T 85 - Phunque -0.0423 -13.30 <0.0001 Yes Sandstone Coarse T 85 - Phunque -0.0513 -16.13 <0.0001 Yes RC LS Coarse T 85 - Phunque -0.0589 -18.51 <0.0001 Yes Recycled Concrete T 85 - Phunque -0.08400 -26.40 <0.0001 Yes Response Variable Abs (%) Difference Adjusted Materials Method of Means T-Value P-Value Significant? BF Slag Coarse T 85 - Phunque 28.87 <0.0001 Yes Elmore Gravel T 85 - Phunque 16.00 <0.0001 Yes Sandstone Coarse T 85 - Phunque 22.798 <0.0001 Yes RC LS Coarse T 85 - Phunque 24.84 <0.0001 Yes Recycled Concrete T 85 - Phunque 44.46 <0.0001 Yes 1.335 0.7399 1.0540 1.1482 2.055

Figure 3-7. Comparison of results determined by two test methods for coarse aggregate. T85 Phunque (a) Gsa (b) Gsb (c) Gssd (d) Absorption 2.30 2.35 2.40 2.45 2.50 2.55 2.60 2.65 2.70 2.75 Elmore Gravel PS Coarse RC LMS Coarse BF Slag Coarse RE Concrete Test Material G sa 0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00 4.50 Elmore Gravel PS Coarse RC LMS Coarse BF Slag Coarse RE Concrete Test Material A bs or pt io n (% ) 2.30 2.35 2.40 2.45 2.50 2.55 2.60 2.65 2.70 2.75 Elmore Gravel PS Coarse RC LMS Coarse BF Slag Coarse RE Concrete Test Material G sb 2.30 2.35 2.40 2.45 2.50 2.55 2.60 2.65 2.70 2.75 Elmore Gravel PS Coarse RC LMS Coarse BF Slag Coarse RE Concrete Test Material G ss d Response Variable Gsa Difference Adjusted Materials Method of Means T-Value P-Value Significant? BF Slag Fine Phunque - MT 84 0.0169 5.24 0.0002 Yes BF Slag Fine SSDetect - MT 84 0.0301 9.32 0.0001 Yes BF Slag Fine T 84 - MT 84 -0.0110 -3.41 0.0914 No BF Slag Fine SSDetect-Phunque 0.0131 4.08 0.0117 Yes BF Slag Fine T 84 - Phunque -0.0279 -8.65 0.0001 Yes BF Slag Fine T 84 - SSDetect -0.0411 -12.73 0.0001 Yes Natural Sand Phunque - MT 84 -0.001103 -0.342 1.0000 No Natural Sand SSDetect - MT 84 -0.006213 -1.926 0.9234 No Natural Sand T 84 - MT 84 -0.006201 -1.923 0.9246 No Natural Sand SSDetect-Phunque -0.005110 -1.584 0.9886 No Natural Sand T 84 - Phunque -0.005098 -1.581 0.9889 No Natural Sand T 84 - SSDetect 0.000012 0.0036 1.0000 No Sandstone Fine Phunque - MT 84 0.00493 1.530 0.9923 No Sandstone Fine SSDetect - MT 84 -0.01150 -3.565 0.0596 No Sandstone Fine T 84 - MT 84 -0.00064 -0.198 1.0000 No Sandstone Fine SSDetect-Phunque -0.01643 -5.095 0.0003 Yes Sandstone Fine T 84 - Phunque -0.00557 -1.728 0.9717 No Sandstone Fine T 84 - SSDetect 0.01086 3.367 0.1034 No RC LS Fine Phunque - MT 84 0.00928 2.88 0.3204 No RC LS Fine SSDetect - MT 84 -0.01909 -5.92 0.0001 Yes RC LS Fine T 84 - MT 84 -0.01235 -3.83 0.0266 Yes RC LS Fine SSDetect-Phunque -0.02837 -8.80 0.0001 Yes RC LS Fine T 84 - Phunque -0.02163 -6.71 0.0001 Yes RC LS Fine T 84 - SSDetect 0.00674 2.09 0.8550 No TX LS Sand Phunque - MT 84 0.022998 7.1310 0.0001 Yes TX LS Sand SSDetect - MT 84 0.019817 6.1446 0.0001 Yes TX LS Sand T 84 - MT 84 -0.001152 -0.3572 1.0000 No TX LS Sand SSDetect-Phunque -0.00318 -0.986 1.0000 No TX LS Sand T 84 - Phunque -0.02415 -7.488 0.0001 Yes TX LS Sand T 84 - SSDetect -0.02097 -6.502 0.0001 Yes Table 3-13. Tukey’s pairwise comparisons for Gsa of fine aggregate.

18 significance level of 0.05, the differences in Gsa of the natu- ral sand measured by the four test methods (Modified T 84, Phunque fine, SSDetect, AASHTO T 84) were not statistically significant. Since the natural sand was rounded and had a low P200 content with no clay fines, it was not expected to cause discrepancies among the four test methods. The differences in Gsa of the Preston sandstone fine aggregate, except for Phunque vs. SSDetect, also were not statistically significant. For other materials, the four test methods generally produced statistically different Gsa results. Table 3-14 shows a statistical analysis of Gsb results mea- sured using the four test methods for fine aggregate. Based on a significance level of 0.05, the differences in Gsb results measured using the AASHTO T 84 and Modified AASHTO T 84 procedures were not statistically significant, except for the RC limestone fine aggregate, which has a high percent- age (26.5 percent) of P200 material. For the natural sand, the differences in Gsb measured by the four test methods, except for Phunque vs. SSDetect, were not statistically significant. However, for the RC limestone fine aggregate, the four test methods produced statistically different Gsb results. Table 3-15 shows a statistical analysis for Gssd results measured by the four test methods for fine aggregate. The comparison results of Gssd values were very similar to those for the Gsb results discussed previously. This observation was expected because Gsb and Gssd are both dependent on how the SSD condition and water absorption capacity are determined. Tukey’s pairwise comparisons of water absorption values measured by the four methods for fine aggregate are presented in Table 3-16. The SSDetect and Phunque flask methods had the tendency to yield lower and higher absorption values than AASHTO T 84, respectively. In addition, the Modified AASHTO T 84 procedure produced higher water absorption results than the AASHTO T 84 method; however, the differ- ence was only statistically significant for the RC limestone fine aggregate. Figure 3-8 shows graphical comparisons of the means of Gsa, Gsb, Gssd, and water absorption measured by the four test methods for fine aggregate. As shown in Figure 3-8(a), the AASHTO T 84 procedure yielded Gsa results that were closest to those of the Modified AASHTO T 84 procedure, then the Phunque method, and finally the SSDetect. Of the four test methods, the SSDetect is the only one that vacuum- saturates fine aggregate samples for measuring the Gsa, which may contribute to the difference of Gsa. Response Variable Gsb Difference Adjusted Materials Method of Means T-Value P-Value Significant? BF Slag Fine Phunque - MT 84 0.0524 6.72 0.0001 Yes BF Slag Fine SSDetect - MT 84 0.0439 5.63 0.0001 Yes BF Slag Fine T 84 - MT 84 0.0048 0.61 1.0000 No BF Slag Fine SSDetect-Phunque -0.0085 -1.09 0.9999 No BF Slag Fine T 84 - Phunque -0.0477 -6.11 0.0001 Yes BF Slag Fine T 84 - SSDetect -0.0392 -5.02 0.0004 Yes Natural Sand Phunque - MT 84 0.0125 1.61 0.9868 No Natural Sand SSDetect - MT 84 -0.0277 -3.54 0.0634 No Natural Sand T 84 - MT 84 0.0000 0.00 1.0000 No Natural Sand SSDetect-Phunque -0.0402 -5.15 0.0002 Yes Natural Sand T 84 - Phunque -0.0125 -1.61 0.9869 No Natural Sand T 84 - SSDetect 0.0277 3.54 0.0633 No Sandstone Fine Phunque - MT 84 0.0278 3.56 0.0598 No Sandstone Fine SSDetect - MT 84 -0.0522 -6.69 0.0001 Yes Sandstone Fine T 84 - MT 84 0.0085 1.08 0.9999 No Sandstone Fine SSDetect-Phunque -0.0800 -10.25 0.0001 Yes Sandstone Fine T 84 - Phunque -0.0194 -2.48 0.6013 No Sandstone Fine T 84 - SSDetect 0.06065 7.770 0.0001 Yes RC LS Fine Phunque - MT 84 0.14003 17.938 0.0001 Yes RC LS Fine SSDetect - MT 84 -0.03610 -4.625 0.0016 Yes RC LS Fine T 84 - MT 84 0.09214 11.804 0.0001 Yes RC LS Fine SSDetect-Phunque -0.1761 -22.56 0.0001 Yes RC LS Fine T 84 - Phunque -0.0479 -6.13 0.0001 Yes RC LS Fine T 84 - SSDetect 0.1282 16.43 0.0001 Yes TX LS Sand Phunque - MT 84 0.04457 5.710 0.0001 Yes TX LS Sand SSDetect - MT 84 -0.01525 -1.953 0.9142 No TX LS Sand T 84 - MT 84 0.00961 1.231 0.9995 No TX LS Sand SSDetect-Phunque -0.05982 -7.663 0.0001 Yes TX LS Sand T 84 - Phunque -0.03496 -4.479 0.0028 Yes TX LS Sand T 84 - SSDetect 0.02486 3.184 0.1643 No Table 3-14. Tukey’s pairwise comparisons for Gsb of fine aggregate.

19 Table 3-15. Tukey’s pairwise comparisons for Gssd of fine aggregate. Response Variable Gssd Difference Adjusted Materials Method of Means T-Value P-Value Significant? BF Slag Fine Phunque - MT 84 0.0405 7.76 0.0001 Yes BF Slag Fine SSDetect - MT 84 0.0385 7.40 0.0001 Yes BF Slag Fine T 84 - MT 84 -0.0006 -0.11 1.0000 No BF Slag Fine SSDetect-Phunque -0.0019 -0.37 1.0000 No BF Slag Fine T 84 - Phunque -0.0410 -7.87 0.0001 Yes BF Slag Fine T 84 - SSDetect -0.0391 -7.51 0.0001 Yes Natural Sand Phunque - MT 84 0.00740 1.42 0.9969 No Natural Sand SSDetect - MT 84 -0.01952 -3.75 0.0346 Yes Natural Sand T 84 - MT 84 -0.00233 -0.45 1.0000 No Natural Sand SSDetect-Phunque -0.0269 -5.17 0.0002 Yes Natural Sand T 84 - Phunque -0.0097 -1.87 0.9418 No Natural Sand T 84 - SSDetect 0.01720 3.30 0.1232 No Sandstone Fine Phunque - MT 84 0.01927 3.70 0.0401 Yes Sandstone Fine SSDetect - MT 84 -0.03661 -7.02 0.0001 Yes Sandstone Fine T 84 - MT 84 0.00506 0.97 1.0000 No Sandstone Fine SSDetect-Phunque -0.05588 -10.72 0.0001 Yes Sandstone Fine T 84 - Phunque -0.01422 -2.73 0.4186 No Sandstone Fine T 84 - SSDetect 0.04166 7.995 0.0001 Yes RC LS Fine Phunque - MT 84 0.09180 17.617 0.0001 Yes RC LS Fine SSDetect - MT 84 -0.02920 -5.603 0.0001 Yes RC LS Fine T 84 - MT 84 0.05397 10.356 0.0001 Yes RC LS Fine SSDetect-Phunque -0.1210 -23.22 0.0001 Yes RC LS Fine T 84 - Phunque -0.0378 -7.26 0.0001 Yes RC LS Fine T 84 - SSDetect 0.08316 15.96 0.0001 Yes TX LS Sand Phunque - MT 84 0.036435 6.9920 0.0001 Yes TX LS Sand SSDetect - MT 84 -0.002430 -0.4663 1.0000 No TX LS Sand T 84 - MT 84 0.005578 1.0705 0.9999 No TX LS Sand SSDetect-Phunque -0.03887 -7.458 0.0001 Yes TX LS Sand T 84 - Phunque -0.03086 -5.922 0.0001 Yes TX LS Sand T 84 - SSDetect 0.008008 1.537 0.9919 No Table 3-16. Tukey’s pairwise comparisons for water absorption of fine aggregate. Response Variable Abs (%) Difference Adjusted Materials Method of Means T-Value P-Value Significant? BF Slag Fine Phunque - MT 84 -0.473 -3.87 0.0235 Yes BF Slag Fine SSDetect - MT 84 -0.227 -1.86 0.9434 No BF Slag Fine T 84 - MT 84 -0.201 -1.64 0.9831 No BF Slag Fine SSDetect-Phunque 0.245 2.01 0.8931 No BF Slag Fine T 84 - Phunque 0.272 2.22 0.7800 No BF Slag Fine T 84 - SSDetect 0.026 0.22 1.0000 No Natural Sand Phunque - MT 84 -0.1981 -1.622 0.9854 No Natural Sand SSDetect - MT 84 0.3174 2.598 0.5129 No Natural Sand T 84 - MT 84 -0.0889 -0.727 1.0000 No Natural Sand SSDetect-Phunque 0.5156 4.2198 0.0071 Yes Natural Sand T 84 - Phunque 0.1093 0.8944 1.0000 No Natural Sand T 84 - SSDetect -0.4063 -3.325 0.1154 No Sandstone Fine Phunque - MT 84 -0.3574 -2.925 0.2911 No Sandstone Fine SSDetect - MT 84 0.6541 5.354 0.0001 Yes Sandstone Fine T 84 - MT 84 -0.1346 -1.102 0.9999 No Sandstone Fine SSDetect-Phunque 1.0115 8.279 0.0001 Yes Sandstone Fine T 84 - Phunque 0.2227 1.823 0.9529 No Sandstone Fine T 84 - SSDetect -0.789 -6.456 0.0001 Yes RC LS Fine Phunque - MT 84 -2.063 -16.89 0.0001 Yes RC LS Fine SSDetect - MT 84 0.336 2.75 0.4024 No RC LS Fine T 84 - MT 84 -1.613 -13.20 0.0001 Yes RC LS Fine SSDetect-Phunque 2.3994 19.639 0.0001 Yes RC LS Fine T 84 - Phunque 0.4505 3.687 0.0414 Yes RC LS Fine T 84 - SSDetect -1.949 -15.95 0.0001 Yes TX LS Sand Phunque - MT 84 -0.3505 -2.869 0.3249 No TX LS Sand SSDetect - MT 84 0.5157 4.221 0.0071 Yes TX LS Sand T 84 - MT 84 -0.1607 -1.315 0.9988 No TX LS Sand SSDetect-Phunque 0.8662 7.090 0.0001 Yes TX LS Sand T 84 - Phunque 0.1898 1.554 0.9909 No TX LS Sand T 84 - SSDetect -0.6764 -5.536 0.0001 Yes

20 Since the Phunque flask consistently produced lower water absorption results for all the fine aggregates, as shown in Figure 3-8(d), it was suggested that some water penetrates into the surface voids of the aggregate particles within the first 30 seconds of the test before the initial water level read- ing is taken. Differences between the results of AASHTO T 84 and Modified AASHTO T 84 suggest that the presence of P200 material only significantly affected the RC limestone, especially the water absorption. To gain a better understanding of the difference in the results of the RC limestone when tested with the AASHTO T 84 (with P200) and Modified AASHTO T 84 (without P200) meth- ods, a small experiment was performed to quantify the impact of removing the P200 material on the results of the cone test. As previously discussed, the RC limestone material contained 26.5 percent P200 and had a sand equivalent value of 39.6. For this small experiment, two samples of RC limestone fine aggregate were prepared. One sample was dry-sieved over the No. 200 sieve to remove the P200 material while the other sample was not. The samples were placed in a plastic bag with approximately 12 percent moisture and allowed to soak overnight (as per the AASHTO T 84 procedure). The samples were then air-dried using a hair drier, and the cone test was performed at various times during the drying pro- cess. Each time the cone test was performed on the material, a photograph was taken of the aggregate cone. The material from the cone was then removed from the overall sample and weighed. This material was later dried out in order to deter- mine the moisture content of the aggregate sample that cor- responded to the composition of that particular cone test. A total of five cone tests were performed on each sample. Figure 3-9 shows photographs of the cone tests performed on the aggregate samples at various moisture conditions. Fig- ure 3-9 (A1) and (B1) show photographs of cone tests performed prior to the samples reaching an SSD condition, and the color of materials in the two figures was very similar. Figure 3-9 (A2), (A3), (B2), and (B3) show cones that were very close to the typi- cal definition of an SSD condition in AASHTO T 84. Figure 3-9 (A4), (A5), (B4), and (B5) show the aggregate materials after being dried beyond the typical SSD condition (with progres- sively less of the material retaining the cone shape). The lighter color of aggregate samples with P200 material (on the right) at and after SSD conditions indicated that these samples were drier than those with the P200 material removed. (a) Gsa (b) Gsb (c) Gssd (d) Absorption 2.200 2.300 2.400 2.500 2.600 2.700 2.800 2.900 Ark NS BF Slag Fine PS Fine RC LMS Fine TX Sand Test Material G sa 2.200 2.300 2.400 2.500 2.600 2.700 2.800 2.900 Ark NS BF Slag Fine PS Fine RC LMS Fine TX Sand Test Material G sb MT84 Phunque SSDetect T84 2.200 2.300 2.400 2.500 2.600 2.700 2.800 2.900 Ark NS BF Slag Fine PS Fine RC LMS Fine TX Sand Test Method G ss d 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 Ark NS BF Slag Fine PS Fine RC LMS Fine TX Sand Test Method A bs or pt io n (% ) Figure 3-8. Comparison of results determined by four test methods for fine aggregate.

21 Without P200 Material (A1) Cone #1: Wet Condition (A2) Cone #3: Early SSD Condition (A3) Cone #5: SSD Condition (A4) Cone #7: Past SSD Condition (A5) Cone #8: Past SSD Condition (B5) Cone #10: Past SSD Condition With P200 Material (B1) Cone #2: Wet Condition (B2) Cone #4: SSD Condition (B3) Cone #6: SSD Condition (B4) Cone #9: Past SSD Condition Figure 3-9. Cone tests at various moisture conditions for RC limestone. Figure 3-10 shows a comparison of the measured mois- ture contents taken from the cone test material for each of the photographs shown in Figures 3-9. The data clearly showed a significant difference between the moisture contents of the material with and without P200 material after the start of SSD condition. The moisture contents varied by approximately 2 percent for the samples tested after the material had reached SSD by the definition set forth in AASHTO T 84. This pro- vided the evidence that the P200 material of the RC limestone caused a difference in the moisture contents determined by AASHTO T 84 for the two types of material. This was likely due to the cohesive nature of these fines that allowed the material to retain the cone shape even after it had reached the SSD condition. Variance Estimates for Coarse Aggregate Test Methods The next analysis step was to determine estimates of within- and between-laboratory variances for each test

22 method. Detailed results of the within- and between-laboratory variance components for each material evaluated in Experi- ment 2 are presented in Appendix E, which is available on the project web page. Figures 3-11 and 3-12 compare the within-lab and between- lab standard deviations, respectively, for the AASHTO T 85 and Phunque methods for Gsa, Gsb, Gssd, and water absorption. The dash lines in these figures represent the single-operator standard deviations from AASHTO T 85 for comparison. It should be noted that the AASHTO T 85 precision informa- tion was estimated based on aggregates with absorptions of less than 2 percent. In these figures, the AASHTO T 85 procedure was more repeatable and reproducible than the Phunque method in Figure 3-10. Moisture contents corresponding to various cone tests. 0 1 2 3 4 5 6 7 M oi st ur e Co nt en t ( %) Cone Test Photograph Number RC Limestone -#200 x #4 RC Limestone -#4 A1, B1 A2, B2 A3, B3 A4, B4 A5, B5 (a) Gsa (b) Gsb (c) Gssd (d) Absorption 0.000 0.002 0.004 0.006 0.008 0.010 0.012 0.014 0.016 0.018 Elmore Gravel PS Coarse RC LMS Coarse BF Slag Coarse Recycled Concrete Test Material W /L S ta nd ar d De vi at io n fo r G sa 0.000 0.002 0.004 0.006 0.008 0.010 0.012 0.014 0.016 0.018 Elmore Gravel PS Coarse RC LMS Coarse BF Slag Coarse Recycled Concrete Test Material W /L S ta nd ar d De vi at io n fo r G sb T85 Phunque 0.000 0.002 0.004 0.006 0.008 0.010 0.012 0.014 0.016 0.018 Elmore Gravel PS Coarse RC LMS Coarse BF Slag Coarse Recycled Concrete Test Material W /L S ta nd ar d De vi at io n fo r G ss d 0.000 0.050 0.100 0.150 0.200 0.250 Elmore Gravel PS Coarse RC LMS Coarse BF Slag Coarse Recycled Concrete Test Material W /L S ta nd ar d De vi at io n fo r A bs or pt io n Figure 3-11. Within-lab standard deviations for coarse aggregate test methods.

23 almost all of the comparisons. The comparisons (also con- firmed by Bartlett’s test) show that there should be two lev- els of precision: one for materials with absorptions of less than 2.5 percent (Elmore gravel, Preston sandstone, and RC limestone) and another level for more absorptive materials (blast furnace slag and recycled concrete). Thus, the precision information determined in this study for the two methods for coarse aggregate are shown in Table 3-17. The precision information was determined based on limited data generated in this project, so a round robin is needed to verify this preci- sion information prior to future implementation. Table 3-17 also includes the precision information (based on aggregates (a) Gsa (b) Gsb (c) Gssd (d) Absorption 0.000 0.002 0.004 0.006 0.008 0.010 0.012 0.014 0.016 0.018 Elmore Gravel PS Coarse RC LMS Coarse BF Slag Coarse Recycled Concrete Test Material B /L S ta nd ar d De vi at io n fo r G sa 0.000 0.002 0.004 0.006 0.008 0.010 0.012 0.014 0.016 0.018 Elmore Gravel PS Coarse RC LMS Coarse BF Slag Coarse Recycled Concrete Test Material B /L S ta nd ar d De vi at io n fo r G sb T85 Phunque 0.000 0.002 0.004 0.006 0.008 0.010 0.012 0.014 0.016 0.018 Elmore Gravel PS Coarse RC LMS Coarse BF Slag Coarse Recycled Concrete Test Material B /L S ta nd ar d De vi at io n fo r G ss d 0.000 0.050 0.100 0.150 0.200 0.250 Elmore Gravel PS Coarse RC LMS Coarse BF Slag Coarse Recycled Concrete Test Material B /L S ta nd ar d De vi at io n fo r A bs or pt io n Figure 3-12. Between-lab standard deviations for coarse aggregate test methods. Standard Deviation (1s) Standard Deviation (1s) Determined in NCHRP 4-35 Given in AASHTO T 85 Response Method Absorption W/L B/L W/L B/L Gsa T 85 < 2.5% 0.003 0.004 0.007 0.011 2.5% 0.007 0.011 Phunque < 2.0% 0.006 0.012 2.0% 0.001 0.011 Gsb T 85 < 2.5% 0.004 0.004 0.009 0.013 2.5% 0.008 0.013 Phunque < 2.0% 0.008 0.011 2.0% 0.010 0.012 Gssd T 85 < 2.5% 0.003 0.004 0.007 0.011 2.5% 0.007 0.011 Phunque < 2.0% 0.006 0.011 2.0% 0.010 0.011 Abs (%) T 85 < 2.5% 0.06 0.07 0.088 0.145 2.5% 0.08 0.20 Phunque < 2.0% 0.01 0.10 2.0% 0.15 0.20 Table 3-17. Precision information for two test methods for coarse aggregate.

24 with absorptions of less than 2 percent) given in the current AASHTO T 85 test method for comparison. Variance Estimates for Fine Aggregate Test Methods Figure 3-13 compares the within-lab standard deviations, and Figure 3-14 compares the between-lab standard devia- tions for the four test methods (Modified AASHTO T 84, Phunque, SSDetect, and AASHTO T 84) for Gsa, Gsb, Gssd, and water absorption. The dash lines represent the AASHTO T 84 single-operator standard deviations in Figure 3-13 and multi-laboratory standard deviations in Figure 3-14 for com- parison. The AASHTO T 84 precision information was based on aggregates with absorptions of less than 1 percent. Based on the comparisons shown in Figures 3-13 and 3-14, the following observations are offered: • Compared to the within-lab standard deviation for Gsa set forth in AASHTO T 84, the Phunque, SSDetect, and AASHTO T 84 methods had higher standard deviations for the blast furnace slag but lower standard deviations for the other materials, Figure 3-13(a), and the Modified AASHTO T 84 method showed lower within-lab stan- dard deviations for Gsa for all the materials. In addition, each of the test methods had lower between-lab standard deviations for Gsa than the standard deviation specified in AASHTO T 84 for all of the materials except the blast furnace slag, Figure 3-14(a). • Test results were generally most repeatable and repro- ducible for the natural sand that had low angularity, low P200 content, and low absorption capacity. High contents of P200 and/or clay materials in the fine aggregates (e.g., RC limestone and Preston sandstone) can significantly increase the variability of the test results. • The Phunque method appeared to yield the most repeat- able within-lab results for most materials. The Phunque method also had the most reproducible (between-lab com- parison) results for absorption. The within- and between- lab standard deviations for the Phunque method seemed to be more consistent across the range of materials in this experiment. In other words, the repeatability and repro- ducibility of the Phunque method did not appear to be as sensitive to material properties. (a) Gsa (b) Gsb (c) Gssd (d) Absorption 0.000 0.010 0.020 0.030 0.040 0.050 0.060 AR Natural Sand Blast furnace slag Preston sandstone RC limestone TX limestone sand Test Material W /L S ta nd ar d De vi at io n fo r G sa 0.000 0.010 0.020 0.030 0.040 0.050 0.060 AR Natural Sand Blast furnace slag Preston sandstone RC limestone TX limestone sand Test Material W /L S ta nd ar d De vi at io n fo r G sb MT84 Phunque SSDetect T84 0.000 0.010 0.020 0.030 0.040 0.050 0.060 AR Natural Sand Blast furnace slag Preston sandstone RC limestone TX limestone sand Test Material W /L S ta nd ar d De vi at io n fo r G ss d 0.00 0.20 0.40 0.60 0.80 1.00 1.20 1.40 AR Natural Sand Blast furnace slag Preston sandstone RC limestone TX limestone sand Test Material W /L S ta nd ar d De vi at io n fo r A bs or pt io n Figure 3-13. Within-lab standard deviations for fine aggregate test methods.

25 • The AASHTO T 84 method appeared to be the least repeat- able and reproducible in the comparisons. • The variability of the Gsb and Gssd test results appeared to have a trend similar to that of the water absorption capacity. To develop precision information for the four test methods based on the results of this study, a statistical analysis (Bartlett’s test) was conducted to evaluate the effects of water absorp- tion, sand equivalent, and P200 content on the variability of each measured parameter (Gsa, Gsb, Gssd, and water absorp- tion). Table 3-18 summarizes the analysis results, and the shaded cells indicate the material properties that significantly affect the variability of the corresponding test results. The variability of the Modified AASHTO T 84 test results was significantly influenced by the absorption capacity of the material tested. The variability of the Phunque test was not affected by the three material properties. The variability of the SSDetect method was influenced by the water absorption capacity and the P200 content with the absorption capacity being more significant. The variability of the AASHTO T 84 results was affected by the absorption capacity, clay content (sand equivalent), and P200 content of the material tested with the P200 content showing the most influence. Based on the comparisons shown in Figures 3-13 and 3-14, as well as the analysis results presented in Table 3-18, the fol- lowing observations were made: • There is one level of precision for Gsa for each of the four test methods and one level of precision for Gsb, Gssd, and water absorption for the Phunque method. • There are three levels of precision for Gsb, Gssd, and water absorption measured by the Modified AASHTO T 84 and SSDetect methods. The three levels are applied for fine aggre- gate based on the measured water capacity: less than 1 per- cent, between 1 and 2.5 percent, and greater than 2.5 percent. • There are also three levels of precision for Gsb, Gssd, and water absorption for AASHTO T 84 corresponding to the amount of P200 material: less than 1 percent, between 1 and 10 percent, and greater than 10 percent. Based on the above observations, the precision informa- tion for the four test methods for fine aggregate are proposed in Tables 3-19 through 3-22. The precision information was (a) Gsa (b) Gsb (c) Gssd (d) Absorption 0.000 0.010 0.020 0.030 0.040 0.050 0.060 AR Natural Sand Blast furnace slag Preston sandstone RC limestone TX limestone sand Test Material B /L S ta nd ar d De vi at io n fo r G sa 0.000 0.010 0.020 0.030 0.040 0.050 0.060 AR Natural Sand Blast furnace slag Preston sandstone RC limestone TX limestone sand Test Material B /L S ta nd ar d De vi at io n fo r G sb MT84 Phunque SSDetect T84 0.000 0.010 0.020 0.030 0.040 0.050 0.060 AR Natural Sand Blast furnace slag Preston sandstone RC limestone TX limestone sand Test Material B /L S ta nd ar d De vi at io n fo r G ss d 0.00 0.20 0.40 0.60 0.80 1.00 1.20 1.40 AR Natural Sand Blast furnace slag Preston sandstone RC limestone TX limestone sand Test Material B /L S ta nd ar d De vi at io n fo r A bs or pt io n Figure 3-14. Between-lab standard deviations for fine aggregate test methods.

26 Method Parameter Is Effect of Material Property on W/L Variability Significant? Is Effect of Material Property on B/L Variability Significant? % Abs Sand Equiv. P200 % Abs Sand Equiv. P200 MT 84 Gsa No N/A N/A No N/A N/A Gsb Yes N/A N/A Yes N/A N/A Gssd Yes N/A N/A Yes N/A N/A % Abs Yes N/A N/A Yes N/A N/A Phunque Gsa No No No No No Yes Gsb Yes No No No No No Gssd No No No No No No % Abs No No No Yes No No SSDetect Gsa No No No No No No Gsb Yes No Yes Yes No Yes Gssd Yes No Yes Yes No Yes % Abs Yes No Yes Yes No Yes T 84 Gsa No No No Yes No No Gsb Yes No Yes Yes Yes Yes Gssd Yes No Yes Yes Yes Yes % Abs Yes No Yes Yes Yes Yes Note: N/A = Not applicable because materials passing the #200 sieve were removed from the fine aggregate tested. Table 3-18. Correlation between standard deviation of measured parameters and absorption, sand equivalent, and P200. Parameter Absorption Standard Deviation Standard Deviation Determined in NCHRP 4-35 Specified in AASHTO T 84 W/L B/L W/L B/L Gsa 0.006 0.011 0.0095 0.020 Gsb < 1% 0.004 0.009 0.011 0.023 1 - 2.5% 0.015 0.028 2.5% 0.030 0.038 Gssd < 1% 0.003 0.010 0.0095 0.020 1 - 2.5% 0.010 0.019 2.5% 0.019 0.024 Abs (%) < 1% 0.04 0.11 0.11 0.23 1 - 2.5% 0.22 0.45 2.5% 0.53 0.67 Table 3-21. Precision information for SSDetect method for fine aggregate. Parameter Standard Deviation Standard Deviation Determined in NCHRP 4-35 Specified in AASHTO T 84 W/L B/L W/L B/L Gsa 0.006 0.012 0.010 0.020 Gsb 0.007 0.015 0.011 0.023 Gssd 0.006 0.013 0.010 0.020 Abs (%) 0.10 0.12 0.11 0.23 Table 3-20. Precision information for Phunque method for fine aggregate. Parameter Absorption Standard Deviation Standard Deviation Determined in NCHRP 4-35 Specified in AASHTO T 84 W/L B/L W/L B/L Gsa 0.005 0.011 0.0095 0.020 Gsb < 1% 0.009 0.012 0.011 0.023 1 - 2.5% 0.015 0.024 2.5% 0.027 0.046 Gssd < 1% 0.006 0.011 0.0095 0.020 1 - 2.5% 0.012 0.018 2.5% 0.018 0.029 Abs (%) < 1% 0.11 0.19 0.11 0.23 1 - 2.5% 0.22 0.41 2.5% 0.40 0.74 Table 3-19. Precision information for Modified AASHTO T 84 for fine aggregate.

27 Table 3-22. Precision information for AASHTO T 84 method for fine aggregate. Parameter Content of Minus #200 Standard Deviation Standard Deviation Determined in NCHRP 4-35 Specified in AASHTO T 84 W/L B/L W/L B/L Gsa 0.008 0.01 0.0095 0.020 Gsb < 1% 0.008 0.010 0.011 0.023 1 - 10% 0.017 0.030 10% 0.025 0.053 Gssd < 1% 0.007 0.008 0.0095 0.020 1 - 10% 0.013 0.021 10% 0.016 0.041 Abs (%) < 1% 0.08 0.10 0.11 0.23 1 - 10% 0.22 0.48 10% 0.42 0.93 determined based on limited data in this project. A round robin is needed to verify this precision information prior to future implementation. These tables also include the preci- sion information (based on aggregates with absorptions of less than 1 percent and results from more than 10 participat- ing laboratories) set forth in the current AASHTO T 84 test method for comparison. Summary Based on the results of Experiment 2, the following conclu- sions are offered for the six test methods: • Of the two test methods for coarse aggregate, the AASHTO T 85 method was more repeatable and reproducible than the Phunque method. The disadvantage of AASHTO T 85 is that it takes a day longer than the Phunque method to complete a test. • The Phunque method can be used to test both coarse and fine aggregate materials; however, different types of flasks are required. Use of the initial water level reading at 30 seconds is likely to yield inaccurate absorption, Gsb, and Gssd results for absorptive aggregate materials. This method requires two or more flasks of the same type if two or more replicates are tested simultaneously. The current design makes the flasks very fragile; if the method continues to be used, the use of a more durable material for the flasks is suggested. • The AASHTO T 84 method can yield precise results for aggregate materials with less than 1 percent of P200. However, the precision of this method was significantly decreased when used for materials with higher P200 con- tent, especially for fine aggregates with more than 10 per- cent of P200 material. The accuracy of the test was also questionable when used for materials with more than 10 percent of P200. • The difference between the AASHTO T 84 and Modified AASHTO T 84 procedures was that the material used for the Modified AASHTO T 84 did not include the P200 material. This modification improved the precision of the AASHTO T 84 procedure. For the materials with high P200/clay content and high water absorption (i.e., Preston sandstone and RC limestone), the Modified AASHTO T 84 procedure yielded results that were as comparable, repeat- able, and reproducible as those of the SSDetect method. The disadvantage of this test method is that it does not measure the properties of the P200 material, which would require a separate test. • The SSDetect utilizes a rational and objective method of determining the SSD condition of fine aggregate. This test method can be used to measure the full gradation of fine aggregate. It produced repeatable and reproducible test results in this study. It offers substantial time savings (1 day), but at a substantially higher equipment cost. After meeting with the technical panel, the following pro- posals were made for further research in this project: • The AASHTO T 85 procedure was proposed for determin- ing specific gravity and water absorption of coarse aggregate. Further experiments should focus on utilizing a vacuum saturation method to reduce the soak time period, which currently requires 15 to 19 hours, so that this test can be done within two working days. • If the properties of the P200 material are not required for design and construction of PCC and HMA mixtures, the Modified AASHTO T 84 method can be utilized. Further experiments with the Modified AASHTO T 84 method should focus on utilizing a vacuum saturation method to reduce the soak time period. • If the design and construction of PCC and/or HMA mixtures require the properties of the P200 material, the AASHTO T 84 procedure can be used for testing fine aggre- gate materials that have less than 10 percent of P200 material. As for the Modified AASHTO T 84 procedure, further experi- ments should focus on utilizing a vacuum saturation method to reduce the soak time period.

28 • The SSDetect method is probably a better method for testing fine aggregate that contains 10 percent or more of P200 material. Ruggedness and round robin studies for this test method have been completed under another study published in NCAT Report 05-07 (25); thus, no further research is needed. • If the Phunque method (for coarse and/or fine aggregate) is selected, it is proposed that further research focus on determining optimum time for taking the initial water level reading and a vacuum saturation method to reduce the time required for taking the final water level reading. In addition, more durable flasks are needed. Additional Experiments After reviewing the results of Experiment 2, the research team was tasked with conducting additional experiments to answer specific questions related to the test procedures previ- ously evaluated. The objectives of these experiments were as follows: • Experiment 3 was conducted to evaluate modifications relative to the drying and soaking methods in AASHTO T 85 and T 84 to reduce the testing time. • Experiment 4 was conducted to determine the effect of P200 on AASHTO T 84 test results. • Experiment 5 was conducted to investigate the determina- tion of a time-zero reading for Phunque methods. Results of Experiments 3 and 4 are presented in this sec- tion, and results of Experiment 5 are included in Appendix H, which is available on the project web page. Experiment 3: Evaluation of Drying and Soaking Methods in AASHTO T 85 and T 84 Experiment 3 was planned to investigate the viability of modifications in the drying and soaking methods to reduce the testing time for AASHTO T 85 and T 84. These modifica- tions are underlined in Tables 3-23 and 3-24. Experiment 3 was divided into three parts, as shown in Figure 3-15. The materials included in the experimental design were previously used in Experiment 2 (see Table 3-8). Sufficient quantities of these materials were available in the NCAT laboratory. The first part of Experiment 3 was to evaluate two alterna- tives to the initial oven drying of test samples. One alternative was to use a vacuum drying method (i.e., CoreDry according Current AASHTO T 85 Modifications to AASHTO T 85 1. Wash the aggregate to remove P200 and deleterious materials. 2. Dry the sample using a conventional oven at a temperature range of 105-115°C until there is no mass change in the aggregate. The aggregate is often dried overnight. Let the aggregate cool until it is comfortable to handle. 3. Submerge the aggregate in water at room temperature for 15 to 19 hours. 4. Remove the sample from the water, and roll it in an absorbing cloth. This is to take excess surface moisture off the aggregates. The aggregate should no longer be shiny. 5. Weigh the sample. This is saturated surface dry mass. The measurement should be taken to the nearest 0.5 g. 6. The aggregate then must be weighed in water. Shake the container to remove entrapped air. Take this weight. 7. Dry the sample to a constant mass using a conventional oven at a temperature range of 105- 115° C. Remove and cool the test sample for 1 to 3 hours in air. Determine the mass. 1. Wash the aggregate to remove P200 and deleterious materials. 2. Dry the sample using a vacuum drying method (such as the CoreDryTM device, ASTM D 7227) at room temperature until there is no mass change in the aggregate; or test the aggregate sample without oven-drying. 3. Vacuum saturate the aggregate using a vacuum saturation method (such as the procedure described in AASHTO T 209).* 4. Remove the sample from the water, and roll it in an absorbing cloth. This is to remove excess surface moisture from the aggregates. The aggregate should no longer be shiny. 5. Weigh the sample. This is saturated surface dry mass. The measurement should be taken to the nearest 0.5 g. 6. The aggregate then must be weighed in water. Shake the container to remove entrapped air. Take this weight. 7. Dry the test sample using a vacuum drier at room temperature until there is no mass change in the aggregate. Determine the mass. *Note 1: The vacuum procedure described in AASHTO T 209-09 is used to saturate the aggregate sample. “Remove air trapped in the sample by applying gradually increased vacuum until the residual pressure manometer reads 3.7 ± 0.3 kPa (27.5 ± 2.5 mm Hg). Maintain this residual pressure for a specific time period. Agitate the container and contents using the mechanical device during the vacuum period. At the end of the vacuum period, release the vacuum by increasing the pressure at a rate not to exceed 8 kPa (60 mm Hg) per second.” The estimated price for a complete set of equipment for conducting AASHTO T 209-09 is approximately $1,536. The complete set includes a 2- stage vacuum pump ($625), a digital absolute pressure gauge ($450), an in-line dewatering filter ($79), a pycnometer ($357), and a 10-foot vacuum hose ($25). Table 3-23. Modifications to AASHTO T 85 drying and soaking methods.

29 to ASTM D7227), which would eliminate the cooling period for the sample to return to room temperature. The second alter- native was to skip the initial drying of the aggregate and test the sample in its in-situ moisture condition as permitted for concrete mix design. Three replicates were prepared and tested for each factor, resulting in a total of 60 tests conducted for this part of Experiment 3. The second part of Experiment 3 was to evaluate a vacuum saturation method for replacing 15-hour soaking using oven-dried aggregate samples. In Experiment 1, the Rapid AASHTO T 85 procedure using the CoreLok (vacuum sealing) device was used to remove entrapped air within the dry coarse aggregate that had been placed in a plastic bag. The vacuum sealed bag with the aggregate was then opened under water and immersed for 30 min. The aggregate sample was then tested according to AASHTO T 85. This test method allowed water to penetrate into deeper voids because the pressure inside the surface voids was much lower than the atmosphere pressure. This method yielded a much higher volume of per- meable voids, resulting in a lower Gsb and a higher absorp- tion capacity. The vacuum method described in AASHTO T 209 was utilized in a previous study by Mills-Beale et al. (30). Three vacuum saturation periods, including 10, 20, and 30 minutes, appeared to yield test results that were comparable to those using the 15-hour soak time for coarse aggregate. The authors recommended that a vacuum period of 10 min- utes be used. AASHTO T 209 requires asphalt mixtures be vacuum saturated for 15 ± 2 minutes. Hence, for this evalu- ation, three time periods of 5, 10, and 15 minutes were used for vacuum saturating the aggregate samples. Three replicates were prepared and tested for each factor, resulting in a total of 90 tests conducted for this part. Current AASHTO T 84 Modifications to AASHTO T 84 1. This is used to test materials passing the No. 4 sieve. 2. Dry the sample using a conventional oven at a temperature range of 105°-115°C until there is no mass change in the aggregate. Let the aggregate cool until it is comfortable to handle. 3. Submerge the aggregate in water at room temperature for 15 to 19 hours. 4. Place the soaked aggregate sample on the flat baking sheet and spread it out. Begin to dry the sample with a hair dryer. 5. At regular intervals, determine if the sample reaches an SSD condition using a cone and tamping rod. 6. Partially fill the pycnometer with distilled water and place a funnel in the top part of the neck. Tare this entire apparatus on the scale. Add 500 ± 10 grams of SSD aggregate to the pycnometer. Record this mass as the weight of SSD material. 7. Place the exact amount of a separate sample of SSD aggregate used in Step 6 into a separate bowl (labeled properly). 8. Manually or mechanically shake the pycnometer to allow all the air bubbles to escape from the sample. 9. Fill the pycnometer to the calibrated fill line with distilled water. Use a rolled up towel to remove any deleterious material and air bubbles from the surface of the water. Use the towel to dry the neck of the flask as well. When the flask is filled to the calibration mark with distilled water, record the mass of the pycnometer filled with aggregate. 10. Dry the sample (Step 7) in the 110°C conventional oven to a constant mass. Remove the sample, cool in air for 1.0 ± 0.5 hours, and determine the mass. 1. This is used to test materials passing the No. 4 sieve and retained on the No. 200 sieve. 2. Dry the sample using a vacuum drying method (such as the CoreDryTM device, ASTM D 7227) at room temperature until there is no mass change in the aggregate; or test the aggregate sample without oven-drying. 3. Vacuum saturate the aggregate using a vacuum saturation method (such as the procedure described in AASHTO T 209).* 4. Place the soaked aggregate sample on the flat baking sheet and spread it out. Begin to dry the sample with a hair dryer. 5. At regular intervals, determine if the sample reaches an SSD condition using a cone and tamping rod. 6. Partially fill the pycnometer with distilled water and place a funnel in the top part of the neck. Tare this entire apparatus on the scale. Add 500 ± 10 grams of SSD aggregate to the pycnometer. Record this mass as the weight of SSD material. 7. Place the exact amount of a separate sample of SSD aggregate used in Step 6 into a separate bowl (labeled properly). 8. Mechanically shake the pycnometer to allow all the air bubbles to escape from the sample. 9. Fill the pycnometer to the calibrated fill line with distilled water. Use a rolled up towel to remove any deleterious material and air bubbles from the surface of the water. Use the towel to dry the neck of the flask as well. When the flask is filled to the calibration mark with distilled water, record the mass of the pycnometer filled with aggregate. 10. Dry the test sample (Step 7) using a vacuum drier at room temperature to a constant mass. Remove the sample and determine the mass. *Note 1: The vacuum procedure described in AASHTO T 209-09 is used to saturate the aggregate sample. “Remove air trapped in the sample by applying gradually increased vacuum until the residual pressure manometer reads 3.7 ± 0.3 kPa (27.5 ± 2.5 mm Hg). Maintain this residual pressure for a specific time period. Agitate the container and contents using the mechanical device during the vacuum period. At the end of the vacuum period, release the vacuum by increasing the pressure at a rate not to exceed 8 kPa (60 mm Hg) per second.” The estimated price for a complete set of equipment for conducting AASHTO T 209-09 is approximately $1,536. The complete set includes a 2- stage vacuum pump ($625), a digital absolute pressure gauge ($450), an in-line dewatering filter ($79), a pycnometer ($357), and a 10-foot vacuum hose ($25). Table 3-24. Modifications to AASHTO T 84 drying and soaking methods.

30 The third part of Experiment 3 was to investigate the same vacuum method used in the second part, but the testing was conducted on aggregate samples with in-situ moisture instead of oven-dried samples. For each material, four sets of three replicates were tested, based on four combinations of drying and soaking methods. A detailed summary of specific gravity and water absorp- tion results of coarse and fine aggregates tested for Experi- ment 3 is included in Appendix F, which is available on the project web page. The results of the three parts were analyzed to determine if modifications to drying and soaking methods would be suitable, and revisions to the AASHTO T 85 and T 84 would be proposed. Results of this analysis follow. Part 1: Evaluation of Initial Drying Methods for AASHTO T 85 and T 84 Figures 3-16 and 3-17 show how the initial drying methods were incorporated into the laboratory testing procedures for AASHTO T 85 and 84, respectively. Three drying methods Part 1 -- Evaluate drying methods: Oven-drying vs. alternatives (vacuum drier, natural condition) Part 2 -- Evaluate soaking methods: 15-hour immersion vs. vacuum saturation for 5, 10, or 15 minutes using oven-dried aggregate Analyze test results to determine suitable modifications Recommend possible revisions to AASHTO T 85 and T 84 Prepare test samples and randomize sample numbers Part 3 -- Evaluate soaking methods: 15-hour immersion vs. vacuum saturation for 5, 10, or 15 minutes using in-situ moisture aggregate Figure 3-15. Plan for evaluating modifications to AASHTO T 85 and T 84. Figure 3-16. Laboratory testing plan for evaluating initial drying methods for AASHTO T 85. (a) Oven Dry - Tested in Experiment 2 Drum Split replicates Sieve over #4 Wash dust Oven Dry Soak 15 hrs Test Oven Dry Back (b) Natural Condition - Tested in Experiment 3 Drum Split replicates Sieve over #4 Wash dust Soak 15 hrs Test Oven Dry Back (c) Vacuum Dry - Tested in Experiment 3 Drum Split replicates Sieve over #4 Wash dust Vacuum Dry Soak 15 hrs Test Vacuum Dry Back

31 were evaluated for each test method: oven drying, natural condition, and vacuum drying. The effect of initial drying methods on test results was analyzed in two steps. First, ANOVA and Tukey’s pairwise comparisons among levels of material and drying method were conducted to determine the effects of three initial dry- ing methods (oven drying, natural moisture condition, and vacuum drying) on the test results (Gsa, Gsb, Gssd, and water absorption). Tables 3-25 and 3-26 summarize the results of the statistical analyses for AASHTO T 85 and 84, respectively. Based on a significance level of 0.05, the effect of the three initial drying methods was determined, as shown in the “significance” column. Grouping of test results that were not significantly different was conducted according to Tukey’s test method. In the last three columns (note that Tukey’s comparisons should be made across the rows of the table, not down the columns), initial drying methods that did not share a letter yielded statistically different aggregate proper- ties. For all materials except the gravel coarse aggregate and natural sand fine aggregate, the results using the initial drying methods were significantly different for at least one of the measured aggregate properties. An analysis of the variability of test results showed that the initial drying methods did not significantly affect the precision of AASHTO T 85 and T 84. Second, the research team used graphical comparisons to illustrate the trends of the differences in the measured properties. Figure 3-17. Laboratory testing plan for evaluating initial drying methods for AASHTO T 84. (a) Oven Dry - Tested in Experiment 2 (b) Natural Condition - Tested in Experiment 3 (c) Vacuum Dry - Tested in Experiment 3 Drum Split replicates Sieve over #4 Wash over #200 Oven dry Soak for 15 hrs Test Oven dry back Drum Split replicates Sieve over #4 Wash over #200 Soak 15 hrs Test Oven Dry Back Drum Split replicates Sieve over #4 Wash over #200 Vacuum Dry Soak 15 hrs Test Vacuum Dry Back Results Materials P-Value Significant? Grouping Using Tukey’s Method Oven-Drying Natural CoreDry Gsa BF Slag Coarse 0.000 Yes B A C Elmore Grav 0.093 No A A A PS Coarse 0.002 Yes B A B RC LMS Coarse 0.000 Yes B A B RE Concrete 0.000 Yes B A C Gsb BF Slag Coarse 0.022 Yes B B A Elmore Grav 0.787 No A A A PS Coarse 0.111 No A A A RC LMS Coarse 0.003 Yes B A A RE Concrete 0.000 Yes B B A Gssd BF Slag Coarse 0.015 Yes B A B Elmore Grav 0.586 No A A A PS Coarse 0.332 No A A A RC LMS Coarse 0.001 Yes B A A RE Concrete 0.127 No A A A Abs BF Slag Coarse 0.000 Yes B A C Elmore Grav 0.246 No A A A PS Coarse 0.000 Yes B A C RC LMS Coarse 0.002 Yes A A B RE Concrete 0.000 Yes B A C Note: For each measured property and material, methods that do not share a letter are significantly different. A and C represent the highest and lowest values, respectively. Table 3-25. ANOVA results for AASHTO T 85—initial drying methods.

32 Figure 3-18 compares the coarse aggregate test results measured in accordance with AASHTO T 85 using the three initial dry- ing methods. The effects of the initial drying methods were more profound for absorption and Gsa and less significant for Gsb and Gssd. Compared to the oven drying method, the natural condition yielded the same or higher Gsa and absorp- tion results, whereas the vacuum drying (CoreDry) methods yielded the same or lower absorption and Gsa values, which was likely because the CoreDry method was not able to dry back the aggregate completely. Within each measured property (Gsa, Gsb, Gssd, or water absorption), the effects of the initial drying methods were more significant for BF slag and recycled concrete coarse aggregates that had higher water absorption. Figure 3-19 compares the fine aggregate test results con- ducted according to AASHTO T 84 using the three initial dry- ing methods. The effects of the initial drying methods on fine aggregate properties were more significant for absorption and less significant for Gsa. This observation suggests that the effect of the initial drying method is less significant than the method of determining the SSD condition, as the mea- surement of Gsa does not require the determination of SSD condition. Within each measured property, the effects were more significant for BF slag and RC limestone fine aggregates that had higher water absorption. Part 2: Evaluation of Soaking Methods for AASHTO T 85 and T 84 Using Oven-Dried Samples Figures 3-20 and 3-21 show how the soaking methods were incorporated in the laboratory testing procedures for AASHTO T 85 and 84, respectively. For each test method, the aggregate samples were soaked for 15 hours or vacuum saturated for 5, 10, or 15 minutes before being tested. The research team conducted the analysis of test results in two steps. First, ANOVA and Tukey’s pairwise comparisons among levels of material and soaking method were conducted to determine the effects of four soaking procedures (15-hour hydrostatic soak, 5-minute vacuum soak, 10-minute vacuum soak, and 15-minute vacuum soak) on the test results (Gsa, Gsb, Gssd, and water absorption). Tables 3-27 and 3-28 sum- marize the results of the statistical analyses for AASHTO T 85 and 84, respectively. Based on a significance level of 0.05, the effect of the four soaking methods was determined, as shown in the “significance” column. Grouping of test results that were not significantly different was conducted according to Tukey’s test method. In the last four columns for each row, soaking methods that did not share a letter yielded statisti- cally different results. For coarse aggregate, 5- and 10-minute vacuum soaking was not significantly different from the standard 15-hour hydrostatic soak on Gsb and Gssd results. However, each of the coarse aggregate soaking methods yielded statistically dif- ferent results for Gsa and water absorption. For fine aggregate, the 5- and 10-minute vacuum soaking methods did not yield statistically different results from the standard 15-hour hydrostatic soak for all of the measured properties, except for the water absorption of PS fine aggre- gate. An analysis of the variability of test results showed that the soaking methods did not significantly affect the precision of AASHTO T 85 and T 84. Results Materials P-Value Significant? Grouping Using Tukey’s Method Oven-Drying Natural CoreDry Gsa Ark NS 0.118 No A A A BF Slag Fine 0.000 Yes B A B PS Fine 0.030 Yes B B A RC LMS Fine 0.015 Yes B B A TX Sand 0.491 No A A A Gsb Ark NS 0.059 No A A A BF Slag Fine 0.012 Yes A B A PS Fine 0.002 Yes B A A RC LMS Fine 0.228 No A A A TX Sand 0.013 Yes B A A Gssd Ark NS 0.058 No A A A BF Slag Fine 0.047 Yes A A A PS Fine 0.001 Yes B A A RC LMS Fine 0.286 No A A A TX Sand 0.012 Yes B A A Abs Ark NS 0.228 No A A A BF Slag Fine 0.001 Yes B A B PS Fine 0.007 Yes A B B RC LMS Fine 0.132 No A A A TX Sand 0.031 Yes A B A Note: For each measured property and material, methods that do not share a letter are significantly different. A and B represent the highest and lowest values, respectively. Table 3-26. ANOVA results for AASHTO T 84—initial drying methods.

(a) Gsa (b) Gsb (c) Gssd (d) Absorption Figure 3-18. Effect of initial drying methods on AASHTO T 85 test results.

(b) Gsb (c) Gssd (d) Absorption (a) Gsa Figure 3-19. Effect of initial drying methods on AASHTO T 84 test results.

35 Figure 3-21. Laboratory testing plan for evaluating soaking methods for AASHTO T 84. (a) Soak for 15 Hours - Tested in Experiment 2 (b) Vacuum Saturate for 5, 10, 15 minutes Drum Splitreplicates Sieve over #4 Wash over #200 Oven dry Soak 15 hrs Test Oven dry back Drum Splitreplicates Sieve over #4 Wash over #200 Oven dry Vacuum saturate in 5, 10, 15 min. Test Oven dry back Table 3-27. ANOVA results for AASHTO T 85—oven dried. Results Materials P-Value Significant? Grouping Using Tukey’s Method 15-hr Soak 5-min Vacuum 10-min Vacuum 15-min Vacuum Gsa BF Slag Coarse 0.000 Yes A A A B Elmore Grav 0.001 Yes A B B B PS Coarse 0.148 No A A A A RC LMS Coarse 0.002 Yes A B B B RE Concrete 0.000 Yes A B B B Gsb BF Slag Coarse 0.605 No A A A A Elmore Grav 0.517 No A A A A PS Coarse 0.166 No A A A A RC LMS Coarse 0.000 Yes A A A B RE Concrete 0.088 No A A A A Gssd BF Slag Coarse 0.000 Yes A A A B Elmore Grav 0.513 No A A A A PS Coarse 0.663 No A A A A RC LMS Coarse 0.004 Yes A A A B RE Concrete 0.072 No A A A A Abs BF Slag Coarse 0.000 Yes A A A B Elmore Grav 0.209 No A A A A PS Coarse 0.000 Yes A A B B RC LMS Coarse 0.000 Yes A B B B RE Concrete 0.006 Yes A B B B Note: For each measured property and material, methods that do not share a letter are significantly different. A and B represent the highest and lowest values, respectively. (a) Soak for 15 Hours - Tested in Experiment 2 (b) Vacuum Saturate for 5, 10, 15 minutes Drum Splitreplicates Sieve over #4 Wash dust Oven dry Soak 15 hrs Test Oven dry back Drum Splitreplicates Sieve over #4 Wash dust Oven dry Vacuum saturate in 5, 10, 15 min. Test Oven dry back Figure 3-20. Laboratory testing plan for evaluating soaking methods for AASHTO T 85.

36 Results Materials P-Value Significant? Grouping Using Tukey Method 15-hr Soak 5-min Vacuum 10-min Vacuum 15-min Vacuum Gsa Ark NS 0.396 No A A A A BF Slag Fine 0.039 Yes A A A B PS Fine 0.033 Yes A A A B RC LMS Fine 0.200 No A A A A TX Sand 0.498 No A A A A Gsb Ark NS 0.269 No A A A A BF Slag Fine 0.745 No A A A A PS Fine 0.002 Yes A A A B RC LMS Fine 0.480 No A A A A TX Sand 0.056 No A A A A Gssd Ark NS 0.361 No A A A A BF Slag Fine 0.792 No A A A A PS Fine 0.001 Yes A A A B RC LMS Fine 0.517 No A A A A TX Sand 0.053 No A A A A Abs Ark NS 0.227 No A A A A BF Slag Fine 0.483 No A A A A PS Fine 0.003 Yes A B B B RC LMS Fine 0.386 No A A A A TX Sand 0.054 No A A A A Note: For each measured property and material, methods that do not share a letter are significantly different. A and B represent the highest and lowest values, respectively. Table 3-28. ANOVA results for AASHTO T 84—oven dried. Second, graphical comparisons were conducted to illus- trate the statistical analysis results. Figures 3-22 and 3-23 show graphical comparisons of the coarse and fine aggregate properties, respectively, determined using the four soaking methods. The effect of the soaking methods on the measured properties was more profound for Gsa and water absorption results. For each measured property, the effect was more sig- nificant for more absorptive and vesicular aggregates, espe- cially the BF slag material. However, the effect of the soaking methods on the test results for fine aggregates was less signifi- cant compared to that for the coarse aggregates, likely because the fine aggregates have less connected voids than the coarse aggregates. The effect was more significant for water absorp- tion but less profound for the other measured properties. Part 3: Evaluation of Soaking Methods for AASHTO T 85 and T 84 Using In-Situ Moisture Samples Figure 3-24 shows the procedures for testing coarse aggre- gates. As shown in Figure 3-24(a), aggregate samples were oven-dried and then soaked for 15 hours before testing. In Figure 3-24(b), samples were tested in their in-situ mois- ture conditions and vacuum-soaked up to 15 minutes before testing. Testing of fine aggregates was different depending on the P200 content and composition. Based on results of Experi- ment 4 presented later in this report, the P200 should be tested separately according to ASTM C110, Section 21 or ASTM D5550 if the sand equivalent (AASHTO T 176) value of the fine aggregate is less than 75. Thus, two test procedures for fine aggregates are shown in Figures 3-25 and 3-26, as follows: • The Preston sandstone (sand equivalent is 26.1 percent) and RC Limestone (sand equivalent is 39.2 percent) materials were tested according to the procedure shown in Figure 3-25 in which the P200 materials should be tested separately. • The Arkansas natural sand (sand equivalent is 100 per- cent), Blast Furnace Slag (sand equivalent is 86.5 percent), and Texas limestone fine aggregate (92.8 percent) were tested without separating the P200 materials. • Instead of drying and then soaking samples for 15 hours, as shown in Figure 3-25(a) and Figure 3-26(a), the samples can be tested in their natural moisture conditions and vac- uum soaked for much less time, as shown in Figure 3-25(b) and Figure 3-26(b). For Part 3, four combinations of drying and soaking methods were tested. In the first combination (control), the aggregate samples were oven-dried and soaked for 15 hours as specified in AASHTO T 84 and T 85. The other three com- binations were a combination of one natural moisture con- dition and three vacuum soaking periods. For each material, 14 replicates were prepared. Four sets of three replicates were randomly selected for testing based on the four combinations of drying and soaking methods. The random selection was carried out to make sure there was no bias toward any sets of replicates or the order in which the samples were prepared. The remaining two samples were kept in reserve. The in-situ water content and water absorption capacity of the aggregates are presented in Table 3-29. For production of

(a) Gsa (b) Gsb (c) Gssd (d) Absorption Figure 3-22. Effect of soaking methods on AASHTO T 85 test results.

(a) Gsa (b) Gsb (c) Gssd (d) Absorption Figure 3-23. Effect of soaking methods on AASHTO T 84 test results.

39 Figure 3-24. Laboratory testing plan for evaluating soaking methods on in-situ moisture samples for AASHTO T 85 (coarse aggregate). (a) Control--Samples are oven-dried and soaked for 15 hours (b) Alternatives--Samples are in natural moisture condition and vacuum-soaked for 5, 10, 15 minutes Drum Split replicates Sieve over #4 Wash dust Vacuum saturate for 5, 10, 15 min. Test Oven dry back Drum Split replicates Sieve over #4 Wash dust Oven dry Soak 15 hrs Test Oven dry back Figure 3-25. Laboratory testing plan for evaluating soaking methods for AASHTO T 84 (fine aggregate) using in-situ moisture, Preston sandstone, and RC limestone (low sand equivalent and P200 removed). (a) Control--Oven-dried and soaked for 15 hours (b) Alternatives--Natural moisture condition and vacuum-soaked for 5, 10, 15 minutes Drum Split replicates Sieve over #4 Wash over #200 Vacuum saturate for 5, 10, 15 min. Test Oven dry back Drum Split replicates Sieve over #4 Wash over #200 Oven dry Soak 15 hrs Test Oven dry back Figure 3-26. Laboratory testing plan for evaluating soaking methods for AASHTO T 84 (fine aggregate) using in-situ moisture, natural sand, blast furnace slag, and Texas limestone (high sand equivalent and P200 not removed). (a) Control--oven-dried and soaked for 15 hours (b) Alternatives--Natural moisture condition and vacuum-soaked for 5, 10, 15 minutes Drum Split replicates Sieve over #4 Vacuum saturate for 5, 10, 15 min. Test Oven dry back Drum Split replicates Sieve over #4 Oven dry Soak 15 hrs Test Oven dry back Material In-Situ Water Content (%) Measured Water Absorption (%)* Coarse Aggregates BF Slag Coarse 2.7 3.0 Elmore Gravel 0.6 1.1 PS Coarse 0.9 2.1 RC Limestone Coarse 2.2 2.2 RE Concrete 2.2 4.4 Fine Aggregates AR Sand 6.1 0.2 BF Slag Fine 2.0 3.4 PS Fine 7.3 1.0 RC Limestone Fine 0.7 2.3 Texas Limestone Fine 4.6 1.3 *Measured in accordance with AASHTO T 85 and T 84 (oven-dried and soaked for 15 hours). Table 3-29. Water contents of aggregates in barrels.

40 portland cement concrete, aggregate stockpiles are sprinkled to maintain the moisture content that is often higher than the aggregate absorption capacity, so testing of aggregates in their natural moisture conditions is allowed according to AASHTO T 84 and T 85. However, for production of asphalt concrete, the moisture content of aggregate stockpiles varies, but is often lower than the aggregate absorption capacity, so AASHTO T 84 and T 85 require aggregates be oven-dried before testing. The effect of using the vacuum method for soaking coarse aggregate samples in their in-situ moisture conditions instead of soaking oven-dried aggregate samples for 15 hours in the AASHTO T 85 procedure was analyzed in two steps. First, graphical comparisons were conducted to determine if there were any trends in the differences in the measured proper- ties and their variances. Second, statistical analyses were per- formed to check if the differences in the measured results and their variances were statistically significant. Figures 3-27 and 3-28 compare the average values and stan- dard deviations of coarse aggregate test results based on four combinations of drying and soaking methods. The single- operator standard deviation of AASHTO T 85 also is plotted in these figures for comparison. The effect of the alternative initial drying and soaking methods on the average values was more profound for Gsa and absorption and less so for Gsb and Gssd. Within each measured property (Gsa, Gsb, Gssd, or water absorption), the effect of the alternative methods was most significant for the vesicular BF slag coarse aggregate. The effect of the alternative methods on the variability (standard deviation) was not clear. Statistical analyses also were conducted to determine if the differences seen in Figure 3-27 were significant. Table 3-30 summarizes the results of ANOVA and Tukey’s tests for the measured properties. Based on a significance level of 0.05, the effect of the alternative methods was determined, as shown in the “significance” column. Grouping of test results that were not significantly different was conducted according to Tukey’s test method. In the last four columns, groups of initial drying and soaking methods in each row that did not share a letter yielded statistically different aggregate properties, with letter “A” representing the highest value. The alternative initial drying and soaking methods yielded slightly higher results (letter “A” in Table 3-30). The effect of the alternative initial drying and soaking methods on Gsb was not statistically significant for all of the materials. For the other properties (Gsa, Gssd, and water absorption), the effect of the alternative soaking methods was significant for the BF slag and RC limestone coarse aggregates. For these two materials, the 10-minute vacuum soaking method appeared to yield results that were closer to those of the control (15-hour soaking) method. In addition, Bartlett’s and Levene’s statistical tests also were conducted to determine if the variability of test results shown in Figure 3-28 was statistically different. Table 3-31 summarizes the results of the analysis. Although the two sta- tistical analyses are used to test equal variances, the Levene test is less sensitive to departures of test results from normal- ity. Based on a significance level of 0.05, the difference in the variability of test results was determined, as shown in the “significance” column. Based on Bartlett’s analysis, the vari- ability of the test results was statistically different for three groups. Since the Gsb results for Elmore gravel and the Gssd results for Preston sandstone were very consistent, the sig- nificant difference in the variance was not a concern. The variability of the absorption results for the control method was higher than that for the alternative methods. Based on the Levene test, the difference in the test variability was not statistically significant. The effect of using the vacuum method for soaking fine aggregate samples in their in-situ moisture conditions in the AASHTO T 84 procedure also was analyzed in two steps— graphical comparison and statistical analysis. Figures 3-29 and 3-30 compare the averages and standard deviations of fine aggregate test results measured according to AASHTO T 84 using four combinations of drying and soak- ing methods as well as the single-operator standard deviation of AASHTO T 84. The Preston sandstone and RC limestone fine aggregates were tested without the P200 materials, and the other three fine aggregate materials—natural sand, blast furnace slag, and Texas limestone—were tested with the P200 in accordance with AASHTO T 84. The effect of the alternative initial drying and soaking meth- ods on the mean values was more profound for absorption, Gsb, and Gssd (Gsb and Gssd can be determined based on Gsa and absorption), and less so for Gsa. Within each measured property (Gsa, Gsb, Gssd, or water absorption), the effect of the alternative methods was more significant for the vesicular BF slag and highly absorptive RC limestone. The effect of the alternative methods on the variability (standard deviation) was inconsistent. Statistical analyses, including ANOVA, Tukey’s test, Bartlett’s test, and Levene’s test, were conducted to determine if the differences seen in Figures 3-29 and 3-30 were significant. Table 3-32 summarizes the results of ANOVA and Tukey’s tests for the measured properties. Based on a significance level of 0.05, the effect of the alternative methods was determined, as shown in the “significance” column. Grouping of test results that were not significantly different was conducted according to Tukey’s test and is shown in the last four columns. The effect of the alternative initial drying and soaking methods was most significant (very low p-value) on the measured properties of BF slag and RC limestone materials, which are highly absorp- tive. Where the difference was significant, the 10-minute vacuum soaking method appeared to yield results that were closer to those of the control (15-hour soaking) method.

Figure 3-27. Effect of soaking methods and use of in-situ moisture samples on AASHTO T 85 test results. (a) Gsa (b) Gsb (c) Gssd (d) Absorption

(a) Gsa (b) Gsb (c) Gssd (d) Absorption Note: Dashed horizontal line = T 85 Single-Operator Std. Deviation Figure 3-28. Effect of soaking methods and use of in-situ moisture samples on variability of AASHTO T 85 test results.

43 Results Materials ANOVA Grouping Using Tukey’s Method* P-Value Significant? Oven- Drying, 15-hr Soak Natural Moist., 5-min Vacuum Natural Moist., 10-min Vacuum Natural Moist., 15-min Vacuum Gsa BF Slag Coarse 0.000 Yes B A A A Elmore Grav 0.041 Yes B A A, B A, B PS Coarse 0.113 No A A A A RC LMS Coarse 0.001 Yes B A A A RE Concrete 0.388 No A A A A Gsb BF Slag Coarse 0.140 No A A A A Elmore Grav 0.475 No A A A A PS Coarse 0.096 No A A A A RC LMS Coarse 0.437 No A A A A RE Concrete 0.680 No A A A A Gssd BF Slag Coarse 0.000 Yes B A A A Elmore Grav 0.101 No A A A A PS Coarse 0.095 No A A A A RC LMS Coarse** 0.060 Yes/No B A, B A A, B RE Concrete 0.858 No A A A A Abs BF Slag Coarse 0.000 Yes B A A A Elmore Grav 0.726 No A A A A PS Coarse 0.179 No A A A A RC LMS Coarse 0.013 Yes B A A, B A RE Concrete 0.147 No A A A A *For each measured property and material, methods that do not share a letter are significantly different. A and B represent the highest and lowest values, respectively. **Differences in results of the four methods were not significant in ANOVA but significant in Tukey’s method. Table 3-30. Results of ANOVA and Tukey’s analyses for coarse aggregates. Table 3-31. Results of Bartlett’s and Levene’s analyses for equal variances. Results Materials Bartlett’s Test Levene’s Test P-Value Significant? P-Value Significant? Gsa BF Slag Coarse 0.996 No 0.999 No Elmore Grav 0.099 No 0.307 No PS Coarse 0.091 No 0.550 No RC LMS Coarse 0.835 No 0.920 No RE Concrete 0.673 No 0.826 No Gsb BF Slag Coarse 0.941 No 0.941 No Elmore Grav 0.007 Yes 0.185 No PS Coarse 0.404 No 0.648 No RC LMS Coarse 0.637 No 0.812 No RE Concrete 0.375 No 0.668 No Gssd BF Slag Coarse 0.573 No 0.774 No Elmore Grav 0.330 No 0.414 No PS Coarse 0.049 Yes 0.490 No RC LMS Coarse 0.310 No 0.836 No RE Concrete 0.341 No 0.678 No Abs BF Slag Coarse 0.188 No 0.531 No Elmore Grav 0.557 No 0.687 No PS Coarse 0.893 No 0.946 No RC LMS Coarse 0.680 No 0.737 No RE Concrete 0.043 Yes 0.473 No

(c) Gssd (d) Absorption (a) Gsa (b) Gsb Figure 3-29. Effect of soaking methods and use of in-situ moisture samples on AASHTO T 84 test results.

(a) Gsa (b) Gsb (c) Gssd (d) Absorption Note: Dashed horizontal line = T 84 Single-Operator Std. Deviation Figure 3-30. Effect of soaking methods and use of in-situ moisture samples on variability of AASHTO T 84 test results.

46 As shown in Table 3-32, the control and 10-minute vacuum methods often had the same letter, which indicated that the difference between these two methods was not significant at the 95 percent confidence interval. Table 3-33 shows the results of Bartlett’s and Levene’s sta- tistical tests for equal variances. Based on both the Bartlett and Levene tests, the difference in the test variability was not statistically significant. Summary The key findings of Experiment 3 can be summarized as follows: • The use of an alternative initial drying method can save up to 4 hours, and the use of a vacuum soaking method can save approximately 15 hours. However, they must yield test results comparable to those of the current Table 3-32. Results of ANOVA and Tukey’s analyses for fine aggregates. Results Materials ANOVA Grouping Using Tukey’s Method* P-Value Significant? Oven- Drying, 15-hr Soak Natural Moist., 5-min Vacuum Natural Moist., 10-min Vacuum Natural Moist., 15-min Vacuum Gsa AR Sand 0.680 No A A A A BF Slag Fine 0.003 Yes B A A A PS Fine 0.013 Yes B A B A, B RC LMS Fine 0.056 No A A A A TX LMS Fine 0.094 No A A A A Gsb AR Sand 0.284 No A A A A BF Slag Fine 0.000 Yes C A A, B B PS Fine 0.162 No A A A A RC LMS Fine 0.004 Yes A B A, B B TX LMS Fine 0.038 Yes A, B B A A, B Gssd AR Sand 0.623 No A A A A BF Slag Fine 0.004 Yes B A A A PS Fine 0.826 No A A A A RC LMS Fine 0.007 Yes A B A, B B TX LMS Fine 0.531 No A A A A Abs AR Sand 0.030 Yes B A, B A A, B BF Slag Fine 0.000 Yes A B B A PS Fine 0.001 Yes B A B B RC LMS Fine 0.001 Yes B A A, B A TX LMS Fine 0.028 Yes A, B A A, B B *For each measured property and material, methods that do not share a letter are significantly different. A and C represent the highest and lowest values, respectively. Table 3-33. Results of Bartlett’s and Levene’s analyses for equal variances. Results Materials Bartlett’s Test Levene’s Test P-Value Significant? P-Value Significant? Gsa AR Sand 0.109 No 0.758 No BF Slag Fine 0.907 No 0.973 No PS Fine 0.596 No 0.632 No RC LMS Fine 0.188 No 0.607 No TX LMS Fine 0.382 No 0.577 No Gsb AR Sand 0.432 No 0.560 No BF Slag Fine 0.899 No 0.935 No PS Fine 0.289 No 0.815 No RC LMS Fine 0.978 No 0.994 No TX LMS Fine 0.415 No 0.798 No Gssd AR Sand 0.197 No 0.572 No BF Slag Fine 0.768 No 0.912 No PS Fine 0.440 No 0.794 No RC LMS Fine 0.881 No 0.918 No TX LMS Fine 0.341 No 0.765 No Abs AR Sand 0.730 No 0.881 No BF Slag Fine 0.397 No 0.644 No PS Fine 0.809 No 0.904 No RC LMS Fine 0.570 No 0.747 No TX LMS Fine 0.964 No 0.988 No

47 AASHTO T 85 and 84 methods. Testing the aggregate in the natural moisture condition yielded results closer to those of the oven-dried material, but the differences in the test results were still statistically significant for some materials. The vacuum drying method would be promising if it was able to completely dry highly absorptive materials, such as blast furnace slag and recycled concrete. • For all coarse aggregates, the 10-minute vacuum soaking pro- cedure yielded Gsb and Gssd results that were not statistically different from those measured using the 15-hour hydrostatic soaking method. For all fine aggregates, the 10-minute vac- uum soaking procedure yielded all properties that were not statistically different from those measured using the 15-hour hydrostatic soaking method. Since the asphalt mix design process requires only Gsb, aggregate samples can be vacuum soaked for approximately 10 minutes according to the vacuum saturation method described in AASHTO T 209. This would replace the 15-hour soak required in AASHTO T 85 and 84 and substantially reduce the testing time. Experiment 4: Evaluation of Effects of P200 on AASHTO T 84 Test Results Experiment 4 was performed to (1) determine the amount of P200 in a fine aggregate sample at which the P200 fraction should be tested separately; and (2) conduct a sensitivity ana- lysis of the impact of P200 on the AASHTO T 84 test results. Based on the sand equivalent test results shown in Table G-7 (Appendix G, available on the project web page), both the Preston sandstone and RC limestone fine aggregates had low sand equivalent values, which indicated the P200 frac- tions contained significant amounts of clay-like material. The blast furnace slag and Texas limestone sand had higher sand equivalent values, which indicate that the P200 fines were not clay-like. The natural sand had no clay-like material. The term “clay-like material” is used in AASHTO T 176 for determining the amount of plastic fines in soils and fine aggregate. This term refers to the flocculated material in this test, and this flocculated material may include other materials that are not clay. The following materials were proposed in the testing plan to assess the sensitivity and breakpoint of the P200 fraction: • The fractions retained on the No. 200 sieve from the natu- ral sand and RC limestone fine aggregate. These materials have the lowest and highest water absorption, respectively. • The fine fraction passing the No. 200 sieve from the Texas limestone. This is relatively clean with a minimum amount of clay-like material. • Clay used to determine the effect of clay present in P200 on the test results. Based on the panel’s recommendation, sodium bentonite clay was used this study. The sodium bentonite contains a significant amount of montmoril- lonite that can absorb a large amount of water and swell 15 to 18 times its dry size. This sodium bentonite material is commercially available and often used as a pond sealer. The Atterberg limits (LL, PL, and PI) were conducted for this material before it was used for testing in this study (Table 3-34). The replacement rate was based on mass. Figure 3-31 shows the laboratory testing plan for Experi- ment 4. The aggregate materials used in this testing plan are shown in Table 3-35. The testing plan was conducted in the following steps: • Wash natural sand, RC limestone, and Texas limestone sand over the No. 200 sieve. • Keep the materials retained on the No. 200 sieve from the natural sand and RC limestone and the P200 material from the Texas limestone sand. • Discard the P200 materials from the natural sand and RC limestone and the material retained on the No. 200 sieve from the Texas limestone sand. • Create 12 blends for each of the two materials retained on the No. 200 sieve (Table 3-35). • Conduct AASHTO T 84 on the 24 blends (2 materials × 12 blends/material) using three replicates. • Conduct AASHTO T 84 on the two materials retained on the No. 200 sieve (2 materials × 3 replicates = 6 tests). • Conduct ASTM C110 on the clay and P200 material from the Texas limestone sand (2 materials × 3 replicates = 6 tests). • Determine the sand equivalent for the P200 material from the Texas limestone sand (1 material × 3 replicates = 3 tests). A detailed summary of the testing results for Experiment 4 is included in Appendix G, which is available on the project web page. The following analyses were conducted to deter- mine (1) the impact of P200 on AASHTO T 84 test results and (2) the amount of P200 at which the P200 fraction of the fine aggregate should be tested separately. Atterberg Limits Limestone P200 Clay (Sodium Bentonite, montmorillonite) Liquid Limit (LL) 18 600 Plastic Limit (PL) 16 98 Plastic Index (PI) 2 502 Table 3-34. Atterberg limits of limestone P200 and clay.

48 Step 1: Prepare samples for Arkansas Natural Sand and RC Limestone Step 2: Prepare P200 samples from TX Limestone Fine Aggregate Step 3: Prepare samples for testing Step 4: Test samples as follows: • Conduct AASHTO T 84 on the 24 blends (2 materials × 12 blends/material) using three replicates (= 72 tests) • Conduct AASHTO T 84 on NS and RC Lms materials retained on the No. 200 sieve (2 mats × 3 reps = 6 tests) • Conduct ASTM C110 on the blend of clay and P200 from the Texas limestone (2 mats × 3 reps = 6 tests) • Determine the sand equivalent for P200 material from the Texas limestone (1 mat × 3 reps = 3 tests) Drums Split 36 replicates Sieve over #4 Wash over #200 Keep +#200 Oven dry Drums Sample enough material Oven dry Dry sieve over #200 Keep all P200 Mix NS and RC Lms with P200 + clay (oven dry) to create 12 blends Soak 15 hours Test Figure 3-31. Laboratory testing plan for evaluating effect of P200 on AASHTO T 84 test results. Material Fraction Material Blend No. 1 2 3 4 5 6 7 8 9 10 11 12 Plus #200 from natural sand or RC limestone fine aggregate 100 95 95 95 90 90 90 80 80 80 70 70 Minus #200 material from Texas limestone and clay 0 5 5 5 10 10 10 20 20 20 30 30 P200 from Texas limestone 0 5 3.75 2.5 10 7.5 5 20 15 10 30 22.5 % Clay 0 0 1.25 2.5 0 2.5 5 0 5 10 0 7.5 No. of Replicates 3 No. of Tests T 84: 2 (+#200 mats) × 12 (blends) × 3 (reps) + 6 = 78 tests C110: 6 tests Sand Equivalent: 3 tests Table 3-35. Materials used in Experiment 4. Results and Analysis Figure 3-32 shows graphical comparisons for 12 blends that included (1) the material retained on the No. 200 sieve from the Arkansas natural sand (shown as “+200” in the graphs); (2) the relatively clean fine material passing the No. 200 sieve from the Texas limestone (shown as “-200” in the graphs); and (3) the sodium bentonite (shown as “clay” in the graphs). Tukey’s groupings (significance level = 0.05) of the measured results are included in Tables G-4 through G-7 in Appendix G. For the blends with no clay, the measured results (Gsa, Gsb, Gssd, and absorption) were not statistically different even though these blends included up to 30 percent of the rela- tively clean P200 material. It should be noted that the absorp- tion capacity for all of these blends was less than 1 percent. The measured results started deviating (statistically different) when even a small amount (1.25 percent) of clay was added to the blend. In addition, the variability of the test results also increased for the blends to which the clay was added. Figure 3-33 compares measured results for the other 12 blends in which the material retained on the No. 200 sieve (shown as “+200” in the graphs) was from the RC limestone. Tukey’s groupings of the results are included in Tables G-8 through G-11 in Appendix G. Based on the statistical analysis, the dif- ference in the Gsa results was not significant when the amount of clay added to the blend was less than 7.5 percent. However,

(a) Gsa (b) Gsb (c) Gssd (d) Absorption Figure 3-32. Effect of P200 materials on AASHTO T 84 test results for natural sand blends.

(a) Gsa (b) Gsb (c) Gssd (d) Absorption Figure 3-33. Effect of P200 materials on AASHTO T 84 test results for RC limestone blends.

51 for the other measured results (Gsb, Gssd, and absorption), the differences were statistically significant between two groups of blends—with and without adding the clay. Figures 3-34 and 3-35 show the correlations between the clay content and the Gsb and absorption results for the natural sand and RC limestone, respectively. The correlations suggest that the presence of a clay material (even in a small amount) can significantly affect the measured results, especially Gsb, Gssd, and absorption. It should be noted that the clay material (sodium bentonite) used in this study can absorb a large amount of water and swell 15 to 18 times its dry size. As shown in Table 3-34, the Atterberg limits of the clay were very differ- ent from those of the limestone P200. The amount of the relatively clean P200 material also affected the test results, especially Gsb, Gssd, and absorption. Although this effect was less significant for the natural sand, it was more profound for the RC limestone. As shown in Fig- ures 3-32 and 3-33, when the amount of the relatively clean P200 material was 20 percent or more, it affected the mea- sured Gsb, Gssd, and absorption values; however, the differ- ences were not statistically significant based on a 95-percent confidence interval (significance level of 0.05). Due to the significant effect of clays and P200 on AASHTO T 84 test results, it is desirable to have a test method, such as the sand equivalent test (AASHTO T 176), that can be used to identify fine aggregate materials containing P200 that may have adverse effects on AASHTO T 84 test results. An analysis was conducted to assess the correlation between P200 con- tents and sand equivalent test results of the fine aggregate blends tested in this study. Figures 3-36 and 3-37 show the correlations for the natural sand and RC limestone blends, respectively. The effect of 1.25 percent of clays was similar to that of 10 percent of relatively clean P200 on the sand equiva- lent test results, and the effect of 2.5 percent of clays was simi- lar to that of 20 percent of relatively clean P200. Due to the significant effect of clay on test results, a sand equivalent threshold should be selected so that P200 materials (containing clays) having adverse effects on test results should be tested separately. Based on Figure 3-36, this cut-off sand equivalent value is approximately 75 percent, and it is approxi- mately 65 percent based on Figure 3-37. Thus, it is proposed that a sand equivalent threshold of 75 percent is selected, and it can be used as an option to determine if the “+200” and P200 materials for a fine aggregate should be tested separately. (a) Gsb vs. Clay Content (b) Absorption vs. Clay Content Figure 3-34. Effect of clay content on Gsb and Abs results for natural sand blends. (a) Gsb vs. Clay Content (b) Absorption vs. Clay Content Figure 3-35. Effect of clay content on Gsb and Abs results for RC limestone blends.

52 However, since the sand equivalent value of 75 percent was determined based on limited data, it should be verified in the future. Summary Based on the results of the evaluation of the effect of P200 materials with and without clay, the following observations and proposals are offered: • If the P200 material does not contain clays, inclusion of the P200 material in the fine aggregate sample will have a minimal effect on AASHTO T 84 test results. If more than 10 percent P200 material is present in the fine aggregate sample, the measured absorption of the fine aggregate sample will be less than that of the “+200” portion. This is due to over-drying of the test sample containing the P200 to determine the SSD condition. • If the P200 portion contains clays, the clays can signifi- cantly affect the AASHTO T 84 test results. In this case, the measured absorption of the “+200” portion will be lower than that of the test sample containing the P200 material because the AASHTO T 84 procedure is sensitive to clays in the P200, causing erroneous measurements. • Due to the potential error caused by the presence of clays in P200, it is proposed that “+200” and P200 materials of a fine aggregate be tested separately when the sand equiva- lent of the fine aggregate is lower than 75 percent. The sand equivalent value was determined based on limited data col- lected in this project, and it should be verified in the future. 0 20 40 60 80 100 120 100 95 90 80 95 95 90 90 80 80 0 5 10 20 3.75 2.5 7.5 5 15 10 0 1.25 2.5 5 10 Sa nd E qu iv al en t ( %) +200 -200 Clay Note: Dashed horizontal line = sand equivalent of 75. Figure 3-36. Correlation of P200 contents and sand equivalent results for natural sand blends. 0 20 40 60 80 100 120 100 95 90 80 95 95 90 90 80 80 0 5 10 20 3.75 2.5 7.5 5 15 10 0 1.25 2.5 5 10 Sa nd E qu iv al en t ( %) +200 -200 Clay Note: Dashed horizontal line = sand equivalent of 65. Figure 3-37. Correlation of P200 contents and sand equivalent results for RC limestone blends.