Cover Image

Not for Sale



View/Hide Left Panel
Click for next page ( 27


The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 26
26 Morphology Analysis from Profile Images and 3-D Images: 2.4.4 Relationships Between Fine Aggregate Similar techniques have been applied by Wang and Moham- Shape, Angularity, and Texture and HMA mad (57) and Kecham and Shashidhar (58) to evaluate parti- Performance cle size, shape, angularity, and texture of aggregate. 2.4.4.1 Introduction The following section describes 12 studies relating FAA 2.4.3.2 Indirect Tests to HMA performance. Because of the controversy over the fine aggregate uncompacted voids test, the studies are dis- Standard Test Method for Index of Aggregate Particle cussed individually and in some detail. Shape and Texture (ASTM D3398): In this test method, the sample is first broken down into individual sieve frac- tions. Thus, the gradation of the sample is determined. Each 2.4.4.2 NCAT National Rutting Study size of material is then separately compacted in a cylindrical by Cross and Brown mold using a tamping rod at 10 and 50 drops from a height of 2 in. The mold is filled completely by adding extra mate- Cross and Brown (17) reported relationships between rial so that it levels off with the top of the mold. The weight aggregate properties and field rut depth obtained from a of the material in the mold at each compactive effort is deter- national rutting study. The study indicated the aggregate prop- mined, and the percent voids is computed. A particle index erties had little relationship with rutting when the in-place air for each size fraction is then computed, and, using the gra- voids of the pavement section were less than 2.5%; however, dation of the sample, a weighted average particle index for relationships between aggregate properties and field rut depths were observed for pavement sections with in-place air void the entire sample is also calculated (16). contents in excess of 2.5%. A relationship with an R2 = 0.67 was determined between the National Aggregate Association Direct Shear Test (ASTM D 3080): The direct shear test (NAA) Flow Test Method A, which is the basis of AASHTO (DST) method is used to measure the angle of internal frac- T304, and the pavement rut depth divided by the square root tion of a fine aggregate under different normal stress condi- of the applied ESAL. The relationship was developed from tions. A prepared sample of the aggregate under considera- the analysis of data from 13 pavements. The pavement rut tion is consolidated in a shear mold. The sample is then placed depth divided by the square root of ESALs was used to in a direct shear device and sheared by a horizontal force while account for the fact that greater truck traffic was likely to known normal stress is applied (16). DST is probably the most produce greater pavement rut depths. straightforward way to determine the stress-dependent shear A rutting model with an R2 = 0.77 was developed between strength of fine aggregate. Research conducted by Fernandes rate of rutting and aggregate properties with data from pave- et al. (59) found that direct shear strength may provide a more ments with in-place air voids in excess of 2.5%. The aggre- relevant parameter to evaluate fine aggregates. The researchers gate properties considered included coarse aggregate crushed also stated that the DST is significantly more complex and faces, uncompacted voids in fine aggregate, gradation pa- less repeatable than the FAA test, and its relation to the per- rameters, and both nominal and maximum aggregate size formance of fine aggregates needs to be further verified and divided by lift thickness (17). Only two factors--percent of developed. coarse aggregate with two or more crushed faces and uncom- pacted voids in fine aggregate--were included in the model CAR Test: The CAR test method was developed to evaluate (Equation 1). shear resistance of compacted fine aggregate (60, 61). It is similar to the Florida bearing ratio test (61). In this method, P = 0.080038 - 0.00008(CF ) - 0.00151( NAA) (1) fine aggregates are compacted in a 100-mm mold following the Marshall hammer method using 50 blows applied to only one face of the specimen. The compacted sample height was where maintained as 63.5 mm. The CAR stability was measured by P = predicted rate of rutting, rut depth (mm)/square applying a compressive load using the Marshall test machine. root ESAL; The compacted sample, while still in the mold, is placed in CF = two or more crushed faces in coarse aggregate the Marshall test machine in the upright position. A load of (%); and 50 mm/min is transmitted through a 37.5-mm-diameter steel NAA = NAA uncompacted voids, (%). cylinder on the plane surface of the compacted sample. The highest load that one specimen can carry was reported as the In 1992, Cross and Brown (10) reported that a rutting rate of CAR stability value. This test is believed to be a performance- 0.005842 mm per square root ESALs delineated good per- related method of measuring FAA (61). forming pavements from rutted pavements. Using this crite-

OCR for page 26
27 rion and the relationship between the NAA flow test and rut- uated cylinder, which eliminates the need for a bulk specific ting rate (17), Kandhal et al. (20) determined a minimum gravity; however, the results are affected by the aggregate uncompacted voids content of 43.3%. Cross and Brown (17) absorption. Based on Michigan DOT's survey responses, this developed several additional models relating uncompacted test method has been replaced by AASHTO T304. The flow voids content and air void contents of recompacted speci- rate test uses the NAA apparatus. The flow rate is determined mens using various compaction methods. by dividing the volume of a 500-g sample of fine aggregate by the time it takes to flow through the NAA orifice. A shape- texture index is calculated from the flow time by dividing the 2.4.4.3 Evaluation of Particle Shape and Texture flow time from a standard set of steel balls by the flow time of Mineral Aggregates Used in for the fine aggregate. Standardized gradations were used for Pennsylvania by Kandhal et al. the study. Previous studies had evaluated the as-received gra- dations and recommended a standardized gradation for the Kandhal et al. (20) evaluated 18 sources (8 natural and 10 NAA uncompacted voids test, Michigan Test Method 118-90, manufactured) of fine aggregate from Pennsylvania using and flow rate test (63). ASTM D3398 and both Methods A and B of the NAA uncom- The twelve sands were ranked by each of the test methods pacted voids test. They observed an overlap between the nat- based on the average test value. The best method of differen- ural and manufactured sands in that one manufactured sand, tiation was the flow time test. This was also the easiest para- a limestone, produced both a particle index (12.8) and an meter to obtain. ASTM D3398 correctly differentiated all of the NAA uncompacted void contents (Method A = 43.1) that poor-quality sands from the good-quality sands. The weighted were lower than those of several natural sands. The authors particle index that divided good- and poor-performing materi- concluded that a minimum particle index of 14 and NAA als was between 11.7 and 13.9. NAA uncompacted voids uncompacted voids content Method A of 44.5 separated Method A ranked one of the poor-quality materials the same between natural and manufactured sands with confidence lev- as one of the good-quality materials. Both had an uncom- els of 86% and 82%, respectively. pacted voids content of 44.7%. However, the test procedure During the development of the Superpave method, an for the poor sand was violated because the sand did not have expert panel using a modified Delphi process determined the size fractions retained above the 0.600-mm sieve. Thus three consensus aggregate properties (1). During the fifth round of size fractions were excluded from the standard. Mogawer and questionnaires used as part of the Delphi process, the expert Stuart (63) concluded that 44.7% uncompacted voids would panel recommended minimum uncompacted voids of 42.8% divide good- and poor-performing sands for high traffic lev- for pavements with design traffic levels less than 300,000 els. The remaining methods, Michigan Test Method 118-90 ESALs and 44.2 for pavements with design traffic levels less and the DST, did not differentiate the sands as well. The than 10 million ESALs. These values represented the expert authors noted that the DST was time consuming. panel's average recommendations for pavement layers in the Attempts were made to differentiate between the rutting top 50 mm of the pavement structure. The recommended performance of HMA produced with four of the sands, two uncompacted void levels were reduced to 41.4% and 42.8%, of good quality and two of poor quality. Twelve aggregate respectively, for layers at a depth of 127 mm. blends with levels of 10%, 20%, and 30% of each of the sands were tested with the GLWT, the French Laboratorie Central des Ponts et Chausses (LCPC) Pavement Rutting Tester, and 2.4.4.4 Evaluation of Natural Sands Used in the USACE's Gyratory Testing Machine. The remainder of Asphalt Mixtures by Stuart and Mogawer the mix was made up of a good-quality traprock coarse aggre- gate and traprock crushed sand. Unfortunately, none of the Stuart and Mogawer (62) conducted a study to evaluate rutting tests differentiated between the performance of the different methods of measuring fine aggregate shape and tex- sands. This was most likely due to the high quality of the other ture. Twelve materials were evaluated in the study: five nat- aggregates (crushed traprock) used in the blend (62). ural sands with a poor performance history, four natural sands Stuart and Mogawer (62) presented three additional impor- with a good performance history, and three manufactured tant conclusions: (crushed) sands with a good performance history. Five meth- ods were used to characterize the sands: NAA uncompacted voids Method A, DST, ASTM D3398, Michigan Test Method 1. Methods for measuring shape and texture can only be 118-90, and a flow rate method. Michigan test method 118-90 expected to group sands into performance categories, is similar to the NAA uncompacted voids test in that the vol- such as high or low potential for rutting. The perfor- ume of voids in a loosely compacted sample is used to deter- mance of a sand depends on its quality, the quantity mine the air voidstosolids ratio and, in turn, an angularity used, the qualities of the other aggregates, and the traf- index. The volume of voids is determined in water in a grad- fic level.

OCR for page 26
28 2. Each sand should be tested to determine its rutting poten- ments to characterize aggregate particle shape and texture tial. The methods are not sensitive enough to evaluate the instead of percent crushed particles are modified ASTM C1252 for the coarse aggregate fraction and ASTM C1252 blend of materials found in a job-mix formula gradation. for the fine aggregate fraction. 3. The discrepancies provided by the NAA and the Michi- gan DOT methods may be related to gradation. A sin- gle, standard gradation should be used in these methods so that the voids that they provide are only a function of 2.4.4.6 Study of the Contribution of FAA and shape and texture. Particle Shape to Superpave Mixture Performance by Huber et al. 2.4.4.5 Investigation of the Influence of Aggregate Huber et al. (34) conducted a study to assess the contribu- Properties on Performance of Heavy-Duty tion of FAA and particle shape to the rutting performance of HMA Pavements by Ahlrich a Superpave-designed HMA. Four fine aggregates were selected for the study: a Georgia granite; Alabama limestone; Ahlrich (19) reported an investigation of aggregate parti- Indiana crushed sand (geology not identified, most likely cle shape and texture on the permanent deformation proper- limestone); and Indiana natural sand. The uncompacted void ties of HMA meeting the Federal Aviation Administration's contents (AASHTO T304 Method A) of the four aggregates P-401 specification. Eleven blends meeting the P-401 grada- were measured as 48, 46, 42, and 38, respectively. A refer- tion band were produced with varying amounts of crushed ence mixture was prepared with the Georgia granite (coarse coarse aggregate (0%, 30%, 50%, 70%, and 100%) and vary- and fine aggregate) and a PG 67-22 binder. The other three ing amounts of natural sand (0%, 10%, 20%, 30%, and 40%). aggregates were sieved into size fractions and substituted for The blends were produced using crushed limestone, crushed the granite fine aggregate to produce four mixtures, keeping gravel, and uncrushed gravel. The fine aggregate portion of the gradation constant. All four blends were mixed at the the blends was evaluated by visual inspection of the percent optimum asphalt content determined for the granite blend. crushed particles according to CRD-C-171, ASTM D3398 No adjustment was made for variances in asphalt absorption (Particle Index Test), and ASTM C1252 Methods A and C between the fine aggregates. (FAA test). The uncompacted voids contents of the fine The resulting mixtures were tested in the Couch Wheel aggregate portion of the 11 blends as measured by ASTM Tracker (a modified Hamburg Wheel Tracker), the APA, and C1252 Method A ranged from 38.4% to 47.1%. ASTM D3398 the SST using the frequency sweep test. The rutting tests did and ASTM C1252 Method A both produced strong correla- not appear to differentiate between the blends in a consistent tions (R2 = 0.98) with the percent crushed particles (mini- manner or at all in some cases. The authors concluded that the mum two fractured faces). ASTM C1252 Method A produced choice of coarse aggregate may have masked the effect of the the best correlation with the percent of (rounded) natural sand fine aggregate (34). There was not a correlation between any in the blend (R2 = 0.94). ASTM C1252 Method C produced of the tests and the uncompacted void contents. This finding lower R2 values with both the percent crushed faces and per- is not unexpected because there were not significant differ- cent natural sand (R2 = 0.66 and R2 = 0.71, respectively). ences between the rutting results. A volumetric mix design was performed for each of the 11 blends using the USACE's Gyratory Testing Machine. The samples were prepared with AC-20 (approximately PG 64- 22). Samples were tested using a triaxial (confined) repeated 2.4.4.7 NCHRP Project 4-19 by Kandhal load creep test at 60C. Three properties were used to evalu- and Parker ate the rutting propensity of the mixtures: permanent strain, creep modulus, and slope of the deformation curve. The com- NCHRP Project 4-19, "Aggregate Tests Related to Asphalt posite (coarse and fine aggregate) particle index measured by Concrete Performance in Pavements," (2) evaluated fine ASTM D3398 produced the best correlation with all three aggregate tests related to rutting performance. Three tests were parameters (R2 = 0.78, 0.69, and 0.71, respectively). ASTM used in the study: ASTM D3398, AASHTO T304 Method A, C1252 Method A produced better correlations with all three and particle shape from image analysis (the University of parameters than the other two fine aggregate tests (ASTM Arkansas Method). Used in this study were nine fine aggre- D3398 and percent crushed particles). The R2 values ranged gate sources with a range in uncompacted void contents of from 0.29 to 0.41. The better correlation with the compos- 40.3% to 47.5%. Three of the materials were natural sands. The ite aggregate index from ASTM D3398 is not unexpected fine aggregates were mixed with an uncrushed gravel coarse because the coarse aggregate fraction was also varied between aggregate. All of the mixes were produced using the same gra- the blends. Ahlrich (19) concluded, dation, above the maximum density line. The coarse aggregate and gradation were chosen to emphasize the response of the On the basis of the strong correlations and simple test proce- fine aggregate. The aggregate was mixed with a PG 64-22 dure, the promising alternatives for specification require- binder. A mix design was conducted for each mixture using

OCR for page 26
29 an Ndesign level of 119 gyrations to determine optimum asphalt height and repeated shear at constant height were performed content. according to AASHTO TP7-94. The resulting mixtures were tested using the GLWT and the Correlation analysis between the three fine aggregate tests SST. Simple shear at constant height and frequency sweep at and rutting performance based on both repeated shear at con- constant height were performed using the SST. Poor correla- stant height and the PurWheel rut depths indicated that the tion coefficients were observed between all three fine aggre- uncompacted voids content was most correlated with rutting gate tests and the SST results. The index of aggregate shape performance (64). A stepwise regression was performed to and particle texture from ASTM D3398 produced the best predict the rutting performance of the mixtures using the Pur- correlation with the GLWT rut depths (R2 = 0.67). The Wheel. The independent variables considered were uncom- uncompacted void contents produced a slightly lower corre- pacted voids content, asphalt content, air voids content (of the lation (R2 = 0.60). The authors noted that the uncompacted PurWheel samples), dust to asphalt ratio, gradation parame- voids were highly correlated with the aggregate index (R2 = ters, the interaction between uncompacted voids and asphalt 0.99) and that the uncompacted voids test was much simpler content, and the number of loading cycles to 2% shear strain to run. They therefore recommended AASHTO T304 to quan- from the repeated shear at constant height test. Six of the tify fine aggregate particle shape, angularity, and surface tex- eight variables were included in the model by the stepwise ture. The Roundness Index from the University of Arkansas regression: uncompacted voids, asphalt content, air voids, digital image analysis produced a fair correlation with the the interaction between uncompacted voids and asphalt con- GLWT rut depth (R2 = 0.56). tent, cycles to 2% strain in the SST, and gradation. The uncompacted voids content was the most significant param- eter (F-value = 41.00). Comparing the aggregate properties individually to the rutting results from the PurWheel device 2.4.4.8 Study of the Effect of FAA on Asphalt and repeated shear at constant height, FAA had the highest Mixture Performance by Lee et al. correlation with the PurWheel results (R2 = 0.40) and the Florida Bearing Ratio had the highest correlation with the Lee et al. (64) conducted a study on the effect of FAA on repeated shear at constant height (R2 = 0.29). The authors HMA performance for the Indiana Department of Transpor- concluded that uncompacted voids alone may not be suffi- tation. The study included six fine aggregate sources, which cient to evaluate the fine aggregate contribution to mixture were used to produce 18 9.5-mm NMAS mixtures using dif- rutting performance. It was observed that a mixture having ferent gradations and blends of the fine aggregate. Only one an uncompacted voids content of 43 performed as well as a of the fine aggregate sources was a natural sand. The coarse mixture with an uncompacted voids content of 48. The authors aggregate used for all 18 mixtures was a partially crushed note that this may be due to the confounding effects of gra- (80% one crushed face) gravel. The angularity and texture of dation and compactability (the uncompacted voids content of the fine aggregate sources were evaluated using ASTM 48 represents the slag mixtures). C1252 Method A (FAA test), CAR test, and Florida Bearing Value (Indiana Test Method 201-89). The Florida Bearing Value is a precursor to the CAR test. Instead of using a Mar- 2.4.4.9 Pooled Fund Study 176 shall press, the sample was loaded through the flow of lead shot into a receiving container. The uncompacted voids con- One of the goals of the National Pooled Fund Study No. tent of the fine aggregate ranged from 38.7 to 49.0. Blends of 176, "Validation of SHRP Asphalt Mixture Specifications the six sands were prepared to produce uncompacted void Using Accelerated Testing," was to examine the effect of contents of 46, 45, and 43. Regression analysis indicates an FAA on the rutting performance of Superpave mixtures. Two R2 = 0.70 between the uncompacted voids and CAR peak coarse aggregates--a limestone and granite--and three fine load. The trend indicated an increase in CAR peak load with aggregates--a natural sand, limestone sand, and granite sand-- an increase in uncompacted voids. were used in the study (65). The fine aggregates had uncom- Volumetric mix designs were conducted for each of the pacted void contents of 39, 44, and 50, respectively. The 18 mixtures. The first nine mixtures were produced one each aggregates were combined with a neat PG 64-22 to pro- with the six sands and three blends of those six sands. Nine duce 21 mixture designs: 9 of 9.5-mm NMAS and 12 of additional mixtures were produced, five using a slag sand 19.0-mm NMAS. A trend was observed between the design with varying percentages of natural sand and mineral filler asphalt content and the uncompacted voids content. The rela- and four with a limestone sand (S gradation mix) and differ- tionship indicated that for a given gradation shape (above, ent percentages of natural sand. Rut testing was performed through, or below the maximum density line), optimum asphalt on the mixtures using the PurWheel Laboratory Tracking content increased with increasing uncompacted voids. Device and the SST. The PurWheel device applies loads to The rutting propensities of the mixes were tested with the the slabs of HMA with a rubber wheel having a contact pres- PurWheel, the SST, and Triaxial Tests and in the APT facil- sure of 620 kPa. PurWheel testing was conducted on dry ity. The APT facility is a full-scale, indoor accelerated load- slabs at 60C. SST testing for frequency sweep at constant ing facility managed by Indiana DOT and Purdue University.

OCR for page 26
30 The primary goal of the Phase I testing was to evaluate the the PurWheel and the APT facility. However, it should be sensitivity of the various test methods to the study factors noted that the air void contents of the natural sand and gran- (66). Based on screening tests performed with the PurWheel ite fine aggregate sections were close to the 2.5% level iden- device in Phase I of the study, four mixtures were selected tified by Cross and Brown (10, 17) below which mixtures for APT facility testing. A limestone coarse aggregate was were less sensitive to aggregate properties. The air void con- used to produce 19.0-mm NMAS mix designs using all three tents of the limestone fine aggregate sections (FAA 44) were sands. The natural sand (FAA 39) and limestone sand (FAA approximately 2.5 percentage points higher than the natural 44) were used to produce coarse-graded mixes (below the sand and granite fine aggregate sections. These variations maximum density line). The limestone sand and granite sand were not planned but are part of the variation associated with were used to produce fine-graded mixes (above the maximum full-scale test sections. Thus, although the limestone fine density line). These four mix designs were placed at both low aggregate indicated the best rutting performance for the high- and high in-place densities. density sections, this result may be more related to the higher The results of the APT facility testing are shown in Table 7. in-place air voids and lower asphalt contents of those mix- It is apparent that both mixtures produced with the limestone tures than to the performance of the fine aggregate. This sand (FAA 44) had design asphalt contents that were approx- emphasizes the fact that screening tests for FAA and texture imately 1 percentage point less than the mixtures produced cannot by themselves ensure mixture performance. with the natural or granite sand. For the low-density sections, In Phase II of Pooled Fund Study 176, an additional 6 mix- the crushed limestone sand (FAA 44) produced both the best tures were tested in the APT facility for a total of 10 mixtures and worst rutting results in the APT facility; however, the dry and in excess of 20 sections (considering varying densities PurWheel results ranked both of the limestone sand mixtures and asphalt contents). Stiady et al. (67) discussed the findings as performing the best. For the high-density (low air void) relative to aggregate. Based on Figure 5, the rounded natural sections, the limestone sand mixtures performed best in both sand (FAA 39) produced the worst rutting performance; TABLE 7 INDOT/Purdue APT facility results from Phase I of Pooled- Fund Study 176 (65) Mixture Design Average As- APT Rut PURWheel (FAA, Asphalt Constructed Depth, mm Dry Test Gradation) Content, Wheel Path (Adjusted for Ranking % Air Voids, 76-mm layer % thickness) Low Density Sections 44 ARZ 4.6 8.8 5.3 1 50 ARZ 5.9 6.4 6.3 3 39 BRZ 5.5 5.2 9.4 4 44 BRZ 4.6 6.4 11.8 2 High Density Sections 44 ARZ 4.6 5.3 4.3 1 44 BRZ 4.6 5.7 8.0 2 50 ARZ 5.9 2.9 9.3 3 39 BRZ 5.5 2.6 15.7 4 Figure 5. APT facility rutting versus uncompacted voids content by gradation type (67).

OCR for page 26
31 however, the limestone fine aggregate (FAA 44) performed object (i.e., erosion) followed by replacement of these pixels as well or better than the granite fine aggregate (FAA 50). (i.e., dilation) to simplify the form. The surface parameter is The mix designs produced with the granite fine aggregate had believed to be a measure of angularity and is calculated as the consistently higher asphalt contents. Analysis of variance percentage of the particle area lost after six cycles of erosion (ANOVA) performed on the triaxial shear strength test results followed by six cycles of dilation (56). Fractal behavior is from the 21 mixtures indicated that the uncompacted void defined "as the self-similarity exhibited by an irregular contents for the fine aggregates in the mixtures were a sig- boundary when captured at different magnifications" (56). nificant factor (66). Fractal length increases with an increase in aggregate angular- ity. Form factor describes an object's dimensions, particularly surface irregularity. The form factor of a perfectly circular 2.4.4.10 Evaluation of Superpave FAA object is 1; therefore, form factor decreases with increasing Specification by Chowdhury et al. surface irregularity. The VDG-40 Videograder was developed by LCPC, the Chowdhury et al. (54) conducted a study to evaluate vari- French national road and bridges laboratory (68). The device ous measures of FAA and texture and their relationship to rut- was developed primarily to measure aggregate grading of ting performance. The study was conducted for the Interna- particles larger than 1 mm (No. 16 sieve), but it can also mea- tional Center for Aggregate Research. The study evaluated sure shape properties. Aggregates are backlit as they fall in 23 fine aggregates using seven different procedures: uncom- front of a linear charged couple device camera, which pro- pacted voids content (AASHTO T304), DST (ASTM D3080), duces a line scan image of the aggregate. The aggregates fall CAR test, three different methods of digital image analysis, off a rotating wheel, which prevents them from tumbling as and visual inspection. The image analysis techniques included they fall in front of the camera. An ellipse having the same the Hough Transform by the University of Arkansas, which length and area is fit to each particle. The ratio of the length was discussed previously; unified image analysis by Wash- to the width of each particle is reported as the slenderness ington State University; and the VDG-40 Videograder con- ratio (SR). The SR may be determined as a distribution or an ducted by the Virginia Transportation Research Council. average. The flatness factor is a property for the group of The samples tested by Washington State University were aggregates tested; it is related to the ratio of the average sieved, and only the material passing the 1.18-mm sieve and width to average thickness of the particles. retained on the 0.600-mm sieve was used for analysis. The Based upon the data presented in the paper (54), a correla- aggregates were stained black to improve their contrast with tion matrix was developed between the indices for angularity the background prior to capturing the images. An optical determined with each test method (Table 8). (See Chowdhury microscope linked to an image analyzer was used to capture et al. for some of the correlations [54, 69].) Regression analy- images of the fine aggregate. Three techniques were used sis was performed using Minitab statistical software. The to analyze the binary image: surface erosion-dilation, frac- upper number in the cell is the coefficient of determination tal behavior, and form factor (56). Surface erosion-dilation (R2 ) and the lower number is the significance level ( p-value) involves removing layers of image pixels on the fringe of the based on the ANOVA. TABLE 8 Correlation matrix for fine aggregate test results using data from Chowdhury et al. (54) Test UV, Angle Log University University VDG-40 Procedure AASHTO of CAR of of Slenderness T304 Internal Stability Arkansas Washington Ratio Method Friction K-index Surface (SR) A (AIF) Parameter ASTM (SP) D3080 UV 1.001 0.07 0.17 0.76 0.72 0.47 0.0002 0.222 0.050 0.000 0.000 0.000 AIF 1.00 0.53 0.06 0.05 0.22 0.000 0.000 0.244 0.292 0.028 Log CAR 1.00 0.20 0.16 0.72 0.000 0.031 0.061 0.000 K-index 1.00 0.69 0.50 0.000 0.000 0.000 SP 1.00 0.43 0.000 0.001 SR 1.00 0.000 1 Coefficient of determination (R2) 2 ANOVA level of significance (p-value)

OCR for page 26
32 The uncompacted voids content correlated well with two specifications. Based on laboratory results, it is possible to of the digital imaging methods, K-index (R2 = 0.76) and sur- design mixes using fine aggregate that fails the uncompacted face parameter (R2 = 0.72), and had a fair correlation with the voids criteria but produces acceptable rutting performance. SR (R2 = 0.47). The relationships between uncompacted Regression analysis using data provided by Chowdhury et al. voids and all three direct measures of fine aggregate particle (54) did indicate a good relationship between uncompacted shape were significant based on the ANOVA. The authors voids and VMA (R2 = 0.70). This suggests that uncompacted noted that four crushed limestone aggregates that have good voids may also identify fine aggregates that will assist in field performance histories showed high values of K-index meeting minimum VMA requirements. even though their uncompacted voids contents were less than Angle of internal friction, as tested by ASTM D3080, pro- 45 (54). Kandhal and Parker (2) also found a good relation- duced the best relationship (R2 = 0.69) with the APA rut ship between the EAPP (i.e., ellipse-based area of the object depths (54). Log of CAR stability and the VDG-40 SR pro- divided by the perimeter squared) and uncompacted voids duced fair correlations (R2 = 0.46 and 0.42, respectively). No content (R2 = 0.76) as measured by the University of correlation (R2 = 0.07) was found with the Washington State Arkansas Hough Transform. Uncompacted voids content cor- University surface parameter discussed previously, but a fair related poorly with both angle of internal friction (AIF) and correlation (R2 = 0.58) was found with a second parameter, the Log of CAR stability. fractal length. There was a fair correlation between the two shear mea- surements, AIF and Log of CAR stability (R2 = 0.53). AIF did not correlate well with any other test, although the rela- 2.4.4.11 Evaluation of Superpave Criteria tionship with the VDG-40 SR was significant (p-value = for VMA and FAA for Florida DOT 0.028). Log of CAR stability correlated well with the VDG-40 by Roque et al. SR (R2 = 0.72). There was a fairly good correlation between K-index and surface parameter (R2 = 0.69); both methods had Roque et al. (51) conducted a study on FAA for the Florida moderate correlations with the VDG-40 SR. The authors DOT. A total of nine fine aggregates were included in the noted (54): study: six limestone sources, two granite sources, and a gravel source. The fine aggregates were evaluated using AASHTO The CAR test appears to separate uncrushed and crushed T304 Methods A, B, and C as discussed previously; using the aggregates much better than the FAA test. This could be, in part, due to the high filler content of the crushed materials as ASTM D3080 (DST); and visually. Two alternative grada- compared to the sands. tions, other than that specified in AASHTO T304 Method A, were also evaluated (51, 59). These gradations were selected A laboratory rutting study was conducted with four of the to represent the range of fine aggregate gradations used in the fine aggregates: three crushed materials and one natural sand. study. The authors concluded that "material type had a far Two blends of materials were also produced using two of the greater effect on FAA than did gradation. Furthermore, all crushed materials, one with 15% and the other with 30% of the three gradations appeared to result in the same relative FAA natural sand. A single limestone coarse aggregate and a coarse rankings for the fine aggregates tested" (51). A poor correla- 19.0-mm NMAS gradation were used for all of the mixtures. tion (R2 = 0.32) was observed between the uncompacted The binder grade was not reported. Superpave mix designs voids content and direct shear strength when both tests were were performed for each of the six blends. The mixtures pro- conducted using the AASHTO T304 Method A gradation. duced using the natural sand and blend with 30% natural sand The trend indicates decreasing shear strength with increasing did not meet the Superpave minimum VMA requirements. uncompacted voids content. This may be due to the packing Cylindrical samples at 4 1% air voids were tested in the characteristics of the fine aggregates with higher uncom- APA at 64C with a 445-N (100-lb) vertical load and 694 kPa pacted voids contents. The authors conclude that "although (100 psi) hose pressure. Regression analysis indicated a fair FAA had some influence on the shear strength, aggregate to poor relationship (R2 = 0.37) between uncompacted voids toughness and gradation appeared to overwhelm its effects, and APA rut depth (54). The mix with 100% natural sand confirming that FAA alone was not a good predictor of fine fines (FAA = 39.0) had the highest rut depth (9.2 mm) fol- aggregate shear strength" (51). lowed closely by the mix with the crushed river gravel fines Five of the fine aggregate sources, three limestone sources, (FAA = 44.3, rut depth = 9.1 mm). The mix containing the a granite source, and a gravel source were used to evaluate crushed river gravel had the highest asphalt content of all of the effect of fine aggregate on mixture performance. A sin- the mixes evaluated (tied with granite/natural sand blend). gle limestone source was used as the coarse aggregate and to The mix with the granite fines (FAA = 48.0) had the least develop a reference coarse and fine gradation commonly used amount of rutting (4.0 mm), followed closely by the mix with in Florida. The four other fine aggregates were used to volu- the limestone fines (FAA = 43.5, rut depth = 4.4 mm). This metrically replace the reference aggregate. The material pass- illustrates the concern with the current uncompacted voids ing the 4.75-mm sieve was replaced for the coarse gradation,

OCR for page 26
33 and the material passing the 2.36-mm sieve was replaced 2.4.4.13 NCHRP Project 4-19(2) for the fine gradation. Volumetric replacement was done to account for any differences in specific gravities between the Ongoing research as part of NCHRP Project 4-19(2), materials. "Validation of Performance-Related Tests of Aggregates for Superpave mixture designs were performed for each of the Use in Hot-Mix Asphalt Pavements," is examining the rela- 10 blends using Ndesign = 109 compaction level. The binder tionship between uncompacted voids tests and rutting through grade was not reported. Six of the ten mix designs failed one accelerated testing using the Indiana prototype APT facility. or more Superpave criteria. Two of the three limestone Six fine aggregates were initially selected for the fine aggre- sources failed minimum VMA (14% minimum for 12.5-mm gate characterization portion of the study: crushed gravel, NMAS). The granite source failed the voids filled with granite, dolomite, traprock sands, and two natural sands. The uncompacted void contents (Method A) for these sands asphalt (VFA) requirements on the high side because of a ranged from 40.3 to 49.1 (23). Later, alternative dolomite and high VMA (16%). The authors noted that "the FAA did traprock sands were included that produced HMA mixtures appear to identify substandard VMA mixtures" (51). with better volumetric properties (uncompacted void con- Rutting tests were performed with the APA. Test tempera- tents of 46.8% and 49.2%). ture and loads used in the APA were not reported. The results The study tracked the measured uncompacted void con- for the fine mixtures are reported by Roque et al. (51). The tents from the HMA mix design through field construction. authors state that the rutting results agree with the direct shear On average, a 1.8% reduction in voids was observed between results, aggregate toughness, and known field performance. the HMA mix design value and material recovered from HMA The trend between uncompacted voids and APA rut depths samples taken at the asphalt plant. Rismantojo states that "the indicated decreased rutting with increasing uncompacted degradation was significantly correlated with the initial UVA voids. Two fine aggregates with uncompacted voids less than [uncompacted voids] values. Fine aggregates with high ini- 45 and high toughness (LA abrasion < 35%) exhibited a rut tial UVA values appeared to degrade more than those with depth equivalent to a fine aggregate with an uncompacted low UVA values" (23). voids content in excess of 45. Roque et al. (51) recommend Mixture designs were performed with all eight fine aggre- including aggregate toughness as part of the AASHTO T304 gates using a single uncrushed gravel coarse aggregate to acceptance criteria. Aggregates with uncompacted voids amplify the effect of the fine aggregate. The original dolomite between 42 and 50 would be acceptable with LA abrasion and traprock sources produced VMA values that were exces- values of the parent rock less than 35%. If the LA abrasion sively high (17.4% and 18.0% at Ndesign = 100 gyrations). This of these fine aggregates were to exceed 35%, their rutting resulted in failing VFA values (exceeding 75%). The mix- performance may not be adequate. tures produced using the other original fine aggregates and two replacement aggregates met all of the Superpave crite- ria. Correlations were performed between the volumetric prop- erties and measured fine aggregate properties. Uncompacted 2.4.4.12 Evaluation of the Effect of FAA on voids produced a significant correlation (R2 = 0.59) with den- Compaction and Shearing Resistance of sity at Ninitial (23). A model was developed to relate uncom- Asphalt Mixtures by Stackston et al. pacted voids and dust proportion to VMA. As expected, VMA increased with increasing uncompacted voids and decreasing Stackston et al. (70) conducted a study to evaluate the dust proportion (23). effect of FAA on compaction effort and rutting resistance. The six mixtures with passing Superpave volumetric prop- Three aggregate sources were used in the study. Twenty-four erties were tested in the full-scale Indiana APT facility. The Superpave mix designs were developed using blends of the results indicate that uncompacted voids Methods A and B as three materials and two gradation shapes: fine and s-shaped. well as the uncompacted voids from Virginia Test Method 5 The response of the mixtures was evaluated using Superpave (VTM 5) were significantly related to the total rut depth after volumetric properties and the gyratory load plate assembly. 1,000 passes. The R2 = 0.65 for Method A was slightly less The gyratory load plate assembly measures the force on the than for the other two methods. AASHTO T304 Method A sample at three points. This force is converted to a force per produced the best relationship with the total rut depth after cycle. Testing indicated that the density at Ninitial decreases 20,000 passes (R2 = 0.51); however, the relationship was not with increasing uncompacted voids content. This indicates significant (p-value = 0.286) (23). The author noted that the that mixes with higher uncompacted voids contents would be decrease in rut depth with increasing uncompacted voids less likely to be tender mixes. Data from the gyratory load occurs to a lesser extent above 45% voids. Rismantojo (23) plate assembly indicated that mixes with higher uncompacted concludes that the results of the current study are similar to voids contents are harder to compact. The authors reported those reported by Kandhal and Parker (2), including that fine- that the effect of uncompacted voids content was not consis- graded mixtures with uncompacted voids contents (Method tent in terms of rutting resistance as measured by the gyra- A) between 42% and 46% demonstrate similar levels of rut- tory load plate assembly (70). ting resistance.