Test Methods for Characterizing Aggregate Shape, Texture, and Angularity (2007)

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.

6Evaluation of Merits and Deficiencies of Test Methods Information gathered from the literature (summarized in Appendix D) was used to compare 21 available test methods and identify test methods for further experimental evaluation in this study. The advantages and disadvantages of the test methods are summarized in Table 2. The test methods were divided into 11 groups based on analysis concept, as shown in Table 3. The four indirect meth- ods in the first group rely on packing of aggregates that flow through a specific-sized orifice. Uncompacted void content of fine aggregates (also known as Fine Aggregate Angularity [FAA] test) and uncompacted void content of coarse aggregates were selected for further evaluation because they are widely used and cheaper and easier to use than other tests in the same group. Janoo and Korhonen (1) concluded that the FAA test was the easiest to use when it compared to time index, rugosity, and particle index. Time index was not selected because it is a time consuming test (1) and was classified as having fair perfor- mance, predictability, precision, and accuracy (2). In the second group of tests, a compacted specimen is exposed to pressure or shear forces. Of these methods, the CAR test is a relatively new test and has not received enough evaluation. Chowdhury and Button (3) concluded that the CAR test method offers much more sensitivity than either the FAA test or the direct shear test. This method also has more advantages than the Florida bearing ratio and direct shear tests; it was selected for evaluation. The percentage of fractured particles in coarse aggregate method (ASTM D 5821) was selected because it is currently included in the Superpave system. Rao and Tutumluer (4) described this method as being time consuming, labor intensive, and subjective. Also, it was classified in another study as having low prediction and precision, with medium practicality (5). Both the ASTM D 4791 test method for measuring flat and elongated coarse aggregates and the multiple ratio shape analy- sis test method were selected. The multiple ratio shape analysis provides more detailed measurements in terms of the distri- bution of the dimensional ratio. ASTM D 4791 was selected because it is included in the Superpave system although it was described as tedious, labor extensive, and time consuming (16, 17) and it does not identify spherical or rounded particles and measure one particle at a time (4, 7). The next group of tests uses one camera to image and evaluate particles. It includes the VDG-40 Videograder, Com- puter Particle Analyzer, Micromeritics OptiSizer PSDA, Video Imaging System (VIS), and Buffalo Wire Works PSSDA. Of these methods only the VDG-40 Videograder and Buffalo Wire Works PSSDA were selected for evaluation. The VDG-40 Videograder was selected because it is capable of analyzing every particle in the sample and it showed good correlation with manual measurements of flat and elongated particles (8, 9). The PSSDA method was selected because of its ability to analyze particles with a wide range of sizes (from passing sieve #200 to 1.5 in.). The Camsizer system uses two cameras to capture images at different resolutions; it evaluates a large number of particles in the sample as they fall in front of a backlight. Using two cameras improves the accuracy of measuring the characteris- tics of both coarse and fine aggregates. The system has the capability of automatically producing the distribution of particles’ size, shape, angularity, and texture. The WipShape system uses two cameras to capture images of aggregates passing on a mini-conveyor or on a rotating circular lighting table. This system was selected because it can analyze large quantities of particles in a short time and has the potential to measure and report various shape factors includ- ing sphericity, roundness, and angularity (10, 11). UIAIA uses three cameras to capture images from three orthogonal directions and build a 3-D shape of each particle; it automatically determines flat and elongated particles, coarse aggregate angularity, coarse aggregate texture, and gradation. The use of three images for each particle allows an accurate C H A P T E R 2 Findings

Test Method Estimated Equipment Cost ($) Measured Aggregate Characteristics segatnavdasiD segatnavdA AASHTO T 304 (ASTM C 1252) Uncompacted Void Content of Fine Aggregate 250 • A combination of angularity, texture, and shape. • Simple. • Inexpensive. • Saeed et al. (2001) selected it to measure the properties of aggregates in unbound layers. • Meininger (1998) selected it to measure the properties of aggregates in PCC pavements. • Janoo and Korhonen (1999) recommended it over time index, rugosity, and particle index. • Used in the current Superpave system. • Lee et al. (1999a) and Chowdhury and Button (2001) reported that the test does not consistently identify angular and cubical aggregates. Also, some fine aggregate with good field performance history did not meet the Superpave criteria. • The results are influenced by shape, angularity, texture, and bulk specific gravity. AASHTO TP56 Uncompacted Void Content of Coarse Aggregate 500 • A combination of angularity, texture, and shape. • Simple. • Inexpensive. • Kandhal and Parker (1998) selected it to measure the properties of aggregates in asphalt pavements. • Meininger (1998) selected it to measure the properties of aggregates in PCC pavements. • The results are influenced by shape, angularity, texture, and bulk specific gravity. ASTM D 3398 Standard Test Method for Index of Aggregate Particle Shape and Texture 400 • A combination of angularity, texture, and shape. • Simple. • Inexpensive. • Saeed et al. (2001) classified this test as having fair performance, predictability, precision, and accuracy. • Meininger (1998) reported that the results have high correlation with the FAA test, which is more practical and easier to use. • Fowler et al. (1996) reported that the method does not provide good correlation with concrete performance. • Results influenced by bulk properties, shape, angularity, and texture. Compacted Aggregate Resistance (CAR) Test 500 • A combination of angularity, texture, and shape. • Simple. • Inexpensive. • Chowdhury and Button (2001) reported that the CAR test method is more sensitive to changes in aggregate characteristics than FAA and direct shear test methods. • The results are influenced by shape, angularity, texture, and bulk properties. Florida Bearing Value of Fine Aggregate 1,000 • A combination of angularity, texture, and shape. • Simple. • The results are influenced by shape, angularity, texture, and bulk properties. • Less practical and involves more steps than the FAA test. • Operates based on the same concept as the CAR test but requires more equipment and time. • Lee et al. (1999b) stated that FAA test has better correlation with HMA performance than this test. Rugosity 500 • A combination of angularity, texture, and shape. • Simple. • Inexpensive. • The results are influenced by shape, angularity, texture, and bulk properties. • It is based on the same concept as the FAA test and the uncompacted voids in coarse aggregates test. However, it requires more time and is less practical than these tests. Time Index 500 • A combination of angularity, texture, and shape. • Simple. • Inexpensive. • The results are influenced by shape, angularity, texture, and bulk properties. • It is based on the same concept as the FAA test and the uncompacted voids in coarse aggregates test. However, it requires more time and is less practical than these tests. AASHTO T 236 (ASTM D 3080) Direct Shear Test 10,000 • A combination of angularity, texture, and shape. • Simple. • Chowdhury and Button (2001) reported that the test method has good correlation with HMA performance. • Expensive. • The results are influenced by shape, angularity, texture, mineralogy, and particle size distribution. • Nonuniform stress distribution causes discrepancies in the measured internal friction. Table 2. Advantages and disadvantages of the testing methods used to measure aggregate characteristics. (continued on next page)

8Test Method Estimated Equipment Cost ($) Measured Aggregate Characteristics segatnavdasiD segatnavdA ASTM D 5821 Determining the Percentages of Fractured Particles in Coarse Aggregate 0 • Angularity. • Simple. • Inexpensive. • Used in the current Superpave system. • Labor intensive and time consuming. • Depends on the operator’s judgment. • Meininger (1998) classified this method as having low prediction, precision, and medium practicality. Flat and Elongated Coarse Aggregates ASTM D 4791 250 • Shape. • Used in current Superpave system. • Able to identify large portions of flat and elongated particles. • Gives accurate measurements of particle dimension ratio. • Found to be related to performance of unbound pavement layers (Saeed et al. 2001). • Tedious, labor extensive, time consuming to be used on a daily basis (Yeggoni et al. 1996, Rao and Tutumluer 2000). • Limited to test only one particle at a time. • Unable to identify spherical, rounded, or smooth particles. • Doesn’t directly predict performance (Meininger 1998, Fowler et al. 1996). Multiple Ratio Shape Analysis 1,500 • Shape. • Simple. • Inexpensive. • Provides the distribution of dimensional ratio in aggregate sample. • Does not address angularity or texture. VDG-40 Videograder 45,000 • Shape. • Measures the shape of large aggregate quantity. • Weingart and Prowel (1999) and Tutumluer et al. (2000) reported good correlation with manual measurements of flat-elongated particles. • Expensive. • Does not address angularity or texture. • Assumes idealized particle shape (ellipsoid). • Uses one camera magnification to capture images of all sizes. Computer Particle Analyzer CPA 25,000 Micromeritics OptiSizer PSDA 50,000 • Shape. • Shape. • Measures the shape of large aggregate quantity. • Measures the shape of large aggregate quantity. • Expensive. • Does not address angularity or texture. • Assumes idealized particle shape (ellipsoid). • Uses one camera magnification to capture images of all sizes. • Expensive. • Does not address angularity or texture. • Assumes idealized particle shape (ellipsoid). • Uses one camera magnification to capture images of all sizes. Video Imaging System (VIS) 60,000 • Measures the shape of large aggregate quantity. Camsizer 45,000 • Measures the shape of large aggregate quantity. • Uses two cameras to capture images at different magnifcations based on aggregate size. WipShape 35,000 • Measures the shape of large aggregate quantity. • Measures the three dimensions of aggregates. University of Illinois Aggregate Image Analyzer (UIAIA) 35,000 • Measures the shape of large aggregate quantity. • Measures the three dimensions of aggregates. Aggregate Imaging System (AIMS) 35,000 • Measures the three dimensions of aggregates. • Uses a mechanism for capturing images at different resolutions based on particle size. • Gives detailed analysis of texture. Laser-Based Aggregate Analysis System 25,000 • Shape. • Shape. • Angularity. • Shape. • Angularity. • Shape. • Angularity. • Texture. • Shape. • Angularity. • Texture. • Shape. • Angularity. • Texture. • Measures the three dimensions of aggregates. • Expensive. • Does not address angularity or texture. • Assumes idealized particle shape (ellipsoid). • Uses one camera magnification to capture images of all sizes. • Expensive. • Assumes idealized particle shape (ellipsoid). • Expensive. • Does not address texture. • Uses same camera magnification to capture images of all sizes. • Expensive. • Uses same camera magnification to capture images of all sizes. • Expensive. • Expensive. • Use the same scan to analyze aggregates with different sizes. Note: Prices listed are estimates based on information from users and vendors. Table 2. (Continued).

computation of the volume of each aggregate particle and provides information about the actual 3-D characteristics of the aggregate. AIMS uses one video camera and a microscope to capture different types of images based on the type of aggregate and the property to be measured. The system measures the three dimensions of the aggregate particles. Images can be captured using different resolutions based on the particle size detected by the system. The system is reported to analyze the charac- teristics of fine and coarse aggregates and provide a detailed analysis of texture for coarse aggregates. LASS uses a laser scan to determine particles’ shape and angularity; although this system was selected initially for eval- uation, it was not available to this study during the experi- mental evaluation period. Aggregate Selection This section includes a description of the aggregates that were selected and used to evaluate the testing methods presented in Table 3. Aggregates were selected to cover a range of origin, rock type, and characteristics. The thirteen coarse aggregates and five fine aggregates described in Table 4 were used in this study. Three coarse sizes and three fine sizes were used to per- form the evaluation (see Table 4). Experienced individuals from the industry and highway agencies assisted in selecting and 9 Test Method Direct (D) or Indirect (I) Method Features of Analysis Concept Consideration for Further Experimental Evaluation Uncompacted Void Content of Fine Aggregates AASHTO T304 seY I Uncompacted Void Content of Coarse Aggregates AASHTO TP56 seY I oN I ytisoguR Time Index I Packing of aggregate that flows through a given sized orifice No Index for Particle Shape and Texture ASTM D3398 I Packing of aggregate in a mold using two levels of compactions No Compacted Aggregate Resistance CAR seY I oN I oitaR gniraeB adirolF Angle of Internal Friction from Direct Shear Test I Exposing a compacted specimen to pressure or shear forces No Percentage of Fractured Particles in Coarse Aggregate ASTM D5821 seY selcitrap fo noitcepsni lausiV D Flat and Elongated Coarse Aggregates ASTM D4791 seY D Multiple Ratio Shape Analysis D Measuring particle dimension using caliper Yes seY D redargoediV 04-GDV oN D rezylanA elcitraP retupmoC oN D ADSP reziSitpO scitiremorciM oN D )SIV( metsyS gnigamI oediV Buffalo Wire Works PSSDA D Using one camera to image and evaluate particles in the sample as they fall in front of a backlight Yes Camsizer D Uses two cameras to image and evaluate particles in the sample as they fall in front of a back light Yes WipShape D Uses two cameras to capture image of aggregates passing on a mini conveyor system Yes University of Illinois Aggregate Image Analyzer (UIAIA) D Uses three cameras to capture three projections of a particle moving on a conveyor belt Yes Aggregate Imaging System (AIMS) D Uses one camera and autofocus microscope to measure the characteristics of coarse and fine aggregates Yes Laser-Based Aggregate Analysis System seY nacs resal a sesU D Table 3. Features and consideration of test methods for experimental evaluation.

10 providing these aggregates (pictures of representative samples are provided in Appendix E). Mineralogical content of the thirteen aggregates was deter- mined using X-ray diffraction (XRD)—a technique that uses X-rays of a single wavelength for establishing the structures of crystalline solids. The sample analyzed was in a powder form, consisting of fine grains of single crystalline material. Aggregates of the size 9.5 to 4.75 mm (3/8″ to sieve #4) were ground to a powder form (smaller than 0.075 mm and passes sieve #200). A few grams of the powder sample was placed in a holder, and then the sample was illuminated with X-rays of a fixed wave length in the diffractometer. The intensity of the reflected radiation was recorded. These data were then analyzed for the reflection angle to calculate the interatomic spacing (d-value in angstrom units of 10−7 cm). The intensity was mea- sured to discriminate the various d-spacing, and the results were compared to specific tables to identify possible matches with mineral phases. The mineralogical content of the aggre- gates used in this study is presented in Table 5. The ASTM C 702 test procedure was followed to obtain rep- resentative aggregate samples. Randomization was employed in dividing the aggregate into smaller representative samples to reduce bias due to unforeseen factors that would affect mea- surements. Aggregates selected for evaluation were sieved, reduced to smaller samples, and washed according to ASTM and AASHTO standard procedures. The same sample used in the nondestructive tests was used by all operators and for all test replicates. Each sample of a coarse aggregate size was 1 kg, while each sample of a fine aggre- gate was 0.5 kg. In conducting the tests, the operators were asked to return the aggregates to the sample after running each test, and mix the sample before running the following test using the same method or a different method. Aggregate Sizes Label Source Aggregate Description 25.4 - 19.0 mm (1- 3/4”) 12.5 - 9.5 mm (1/2- 3/8”) 9.5 - 4.75 mm (3/8”- #4) 4.75 - 2.36 mm (#4 - #8) 2.36 - 1.18 mm (#8 - #16) 0.6 - 0.3 mm (#30 - #60) 1 Montgomery AL Uncrushed River Gravel and Sand X X X X X X 2 Montgomery AL Crushed River Gravel and Sand X X X X X X 3 Childersburg AL Limestone X X X 4 Auburn AL Dolomite X X X 5 Birmingham AL Slag X X X X X X 6 Brownwood TX Limestone X X X X X X 7 Fairfield OH Crushed Glacial Gravel X X X 8 Fairfield OH Uncrushed Glacial Gravel X X X 9 Forsyth GA Granite X X X 10 Ruby GA Granite X X X X X X 11 Knippa TX Traprock X X X 12 San Antonio TX Limestone X X X 13 Augusta GA Granite X X X Table 4. Aggregate sources and sizes. Aggregate Aggregate Description Minerals Present 1 Uncrushed River Gravel and Sand Quartz, Dolomite (trace) 2 CrushedRiver Gravel and Sand Quartz 3 Limestone Calcite, Dolomite, Quartz 4 Dolomite Dolomite 5 Slag Akermanite, Calcite, Quartz 6 Limestone Calcite, Quartz, Dolomite 7 Crushed Glacial Gravel Dolomite, Calcite, Quartz 8 Uncrushed Glacial Gravel Dolomite, Calcite, Quartz 9 Granite Quartz, Biotite, Albite, Labradorite 10 Granite Quartz, Chlorite, Albite, Amesite, Anorthite, Phlogophite (Mica), Muscovite 11 Traprock Tephrite, Diopside, Augite, Anorthite 12 Limestone Calcite 13 Granite Quartz, Albite, Calcite, Anorthite, Microcline, Kaolinite Table 5. Mineralogical content of aggregates.

Experimental Design and Statistical Analysis This section documents the experimental evaluation of the test methods. The evaluation covered the repeatability, repro- ducibility, accuracy, cost, and operational characteristics. The first three characteristics were evaluated through statistical analysis of the characteristics of a wide range of aggregates from different sources with various characteristics. The accu- racy analysis was conducted for the parameters employed in the test methods, and for the test methods themselves includ- ing the hardware components. The information that pertains to cost and operational characteristics was collected from vendors, researchers, and operators who have dealt with these systems. As indicated earlier, some of the selected methods have been in practice for years and they are usually performed using standard procedures. However, for the methods that have been developed recently, the manufacturer’s or the developer’s instructions were followed to perform the testing. It was nec- essary in some cases to perform the standard tests with minor modifications in order to conduct the tests on the selected aggregate sizes. A summary of aggregate sizes and parameters obtained from each of the selected test methods is shown in Table 6. Descriptions of the testing procedures and modi- fications, if any, and aggregate properties are provided in Appendix D. Evaluation of Repeatability and Reproducibility Repeatability and reproducibility of test methods were evaluated through measuring the characteristics of aggregate samples several times by single and multiple operators. The operators were uniformly trained on the application of the test methods and were provided with the same set of instructional guidelines. One coarse aggregate size (12.5 − 9.5 mm [1/2 − 3/8″ ]), and one fine aggregate size (2.36 − 1.18 mm [sieve #8 − #16]) were used for the repeatability analysis. Each of the operators measured the properties of these aggregate sizes three times. Reproducibility was assessed by measuring the shape char- acteristics (as applicable to the test method, see Table 6) for aggregate sizes listed in Table 4 by each of the three opera- tors. All operators conducted measurements using the same samples. Standard deviation and coefficient of variation were used to quantify repeatability and reproducibility. Analysis of vari- ance (ANOVA) was used in the statistical analysis according to the ASTM procedures (ASTM E 177, ASTM C 802, and ASTM C 670). The repeatability and reproducibility statistical parameters were calculated for each test method as follows: • Repeatability calculations: For each material and operator, the average of replicates is given by Equation 1, and the variation in measurements is calculated by Equation 2. Where n is the number of measurements by an operator for one material and x–i is the average of the measurements of operator i, and S 2i is the variance for operator i. Table 7 shows the arrangement of variation data within and between operators for one single material using one test method. The repeatability of a test method is evaluated for each aggregate material and all operators by Equation 3: where p = 3 is the number of operators. • Reproducibility Calculations: The average of measure- ments made by all operators for a singe material is given by Equation 4 and the variation between operators is given by Equation 5. Variations between operators are calculated by: Then, reproducibility of a test method is given by: Repeatability and reproducibility of the test method on all aggregates were estimated by pooling standard deviations and coefficients of variations over all materials according to the guidelines of ASTM C 802. Because each of the selected test methods measures aggregate characteristics using different S S S pooledR L mm 2 2 2 7= + ( ) ( ) S S S pooled nL x mm m 2 2 2 6= − ( )[ ] ( ) s x p x p x i m m 2 2 2 1 5= − ( ) −( ) ∑ ( ) x x p m i = ∑ ( )4 S pooled s p m i i p 2 2 1 3( ) = =∑ ( ) S x nx n i if i j n 2 2 1 2 1 2= − ⎛⎝⎜ ⎞⎠⎟ −( ) = ∑ ( ) x x n i ij j n = = ∑ 1 1( ) 11

12 Aggregate Size Characteristics Test C 1 C 2 C 3 F 1 F 2 F 3 Shape (Abbreviation) Angularity (Abbreviation) Texture (Abbreviation) Uncompacted Void Content of Fine Aggregates AASHTO T 304 X X % Loose Uncompacted Void Content (UCVCF) Uncompacted Void Content of Coarse Aggregates AASHTO TP 56 X X % Loose Uncompacted Void Content (UCVCC) Compacted Aggregate Resistance CAR X X X Max Shear Resistance (CAR) Percentage of Fractured Particles in Coarse Aggregate ASTM D 5821 X X X % of Fractured Faces (PFF) Flat and Elongated Coarse Aggregates ASTM D 4791 X X X Flat Elongated Ratio (FER) Multiple Ratio Shape Analysis X X X Dimensional Ratio (MRA) VDG-40 Videograder X X X X Flat Ratio (VDG- 40 FLAT) & Slenderness ratio (VDG-40 SLEND) Buffalo Wire Works PSSDA- Large X X X Roundness (PSSDA-Large ROUND) Roundness (PSSDA-Large ROUND) Buffalo Wire Works PSSDA- Small X X X Roundness (PSSDA-Small ROUND) Roundness (PSSDA-Small ROUND) Camsizer X X X X X Sphericity (CAMSPHT), Symmetry (CAMSYMM), Ratio of Length to Breadth (CAML/B) Convexity (CAMCONV) WipShape X X X Dimensional Ratio (WSFER) Minimum Average Curve Radius (WSMACR) University of Illinois Aggregate Image Analyzer (UIAIA) X X X X Flat Elongated Ratio (UIFER) Angularity Index (UIAI) Surface Texture Index (UISTI) Aggregate Imaging System (AIMS) X X X X X X Sphericity (AIMSSPH) & Form 2-D Index (AIMSFORM) Gradient Angularity Index (AIMSGRAD), Radius Angularity Index (AIMSRAD) Texture Index (Wavelet) (AIMSTXTR) Aggregate sizes: C1 = 25.4 – 19.0 mm (1 – 3/4"); C2 = 12.5 – 9.5 mm (1/2 – 3/8"); C3 = 9.5 – 4.75 mm (3/8" – #4); F1 = 4.75 – 2.36 mm (#4 – #8); F2 = 2.36 – 1.18 mm (#8 – #16); F3 = 0.6 – 0.3 mm (#30 – #60). Table 6. Aggregate size and characteristics measured using the test methods.

analysis parameters or indices with different scales, repeat- ability and reproducibility were assessed independently for each parameter. The final results of repeatability and repro- ducibility for all test methods are reported for each character- istic and for coarse and fine aggregates separately in Tables 8 and 9, respectively. The abbreviations of the parameters provided in the manuals and standards of test methods are used here. In interpreting the results, the following factors should be taken into consideration: (1) The methods differ significantly in the level of detail pro- vided in the results. While the indirect methods provide only an average index, direct methods can provide the distribution of characteristics in an aggregate sample. This advantage of direct methods has not been considered because the calculations are based on average values in order to analyze all test methods using the same statisti- cal methods. (2) The test methods differ in the range of results. Some methods have analysis parameters with narrow ranges that make it difficult to distinguish between aggregates, while others have wide ranges. (3) Measurements from a test method were all conducted using a single device and well-trained operators. (4) The high sensitivity of some test methods to variations in aggregate characteristics, which is an advantage, can increase variation and reduce the repeatability and reproducibility. Considering all these factors, it is recommended to differ- entiate among test methods based on the levels of variability shown in Tables 8 and 9. The percentage of fractured faces test had very high vari- ability compared to all other test methods as also reported by Meininger (5) and Saeed et al. (2). According to the results in Table 9, the uncompacted void content test for fine aggregate had low variability. Saeed et al. (2) rated this test as having a fair precision (ability to repeatedly provide correct results). The results of this test were analyzed using the same specific gravity for each aggregate. The variability of the test results is mainly due to error in measuring the specific gravity. Therefore, it is expected that the variability of the uncompacted void con- tent test would increase significantly when the variability in specific gravity measurements is considered. The image analysis methods had high variability when the percentage of particles with a dimensional ratio of 5:1 was considered. This was mainly due to the small percentages of particles that exhibited this characteristic, such that any slight variation in accounting for these particles was manifested as high coefficient of variation. Therefore, the variability was evaluated based on the percentage of particles with a dimen- sional ratio smaller or larger than 3:1. The image analysis methods (UIAIA, AIMS, Camsizer, PSSDA, WipShape) had low to medium variability in terms of angularity and texture measurements. The AIMS angular- ity indices had low variability, while the texture indices had medium variability. As will be discussed later, automation of the AIMS top lighting intensity would reduce the variability. Evaluation of Accuracy The accuracy of the test methods can be evaluated by correlating the measurements from these tests with the mea- surements obtained from standards or reference tests that are considered to be accurate. The three dimensions of coarse par- ticles can be measured using a digital caliper—an accurate, but slow method. However, because test methods that are accepted to be accurate in quantifying texture and angularity are not available, the following approach was adopted to assess the accuracy of the test methods: • The accuracy was evaluated based on the procedure rec- ommended by standards and/or by the developers, and for the analysis methods (mathematical functions and indices) employed in the imaging-based systems. This approach allowed evaluation of the accuracy of the analysis methods irrespective of the characteristics of the image acquisition setup. 13 Operator Data (replicates) ijx Average ix Within Operator Variance 2 iS 1 I II III 1x 2 1S 2 I II III 2x 2 2S 3 I II III 3x 2 3S Table 7. Arrangement of variation in measurements within and between operators for one aggregate.

14 • Accuracy of analysis methods in imaging-based systems was evaluated through: – Analysis of diagrams of particles with different charac- teristics. These diagrams were developed by geologists in the past to describe and quantify the two-dimensional shape and angularity of sediments. They were plotted based on actual observations of sediments and manual measurements of their shape and angularity. This task provided an initial screening test for the analysis methods by determining whether the analysis methods are capable of (1) identifying clear differences between particle pro- jections or (2) separating the different characteristics (shape, angularity, and texture). – Analysis of the uniqueness of test methods. It was nec- essary to evaluate the correlations among the different test methods to identify analysis methods that are able Coefficient of Variation (CV) Characteristics Test Method Parameter Abbreviation Measured Parameter as Reported by Test Method Repeatability Reproducibility Uncompacted Void Content of Coarse Aggregate UCVCC Percent Uncompacted Void Content L L 0 Fractured Faces H H 1 Fractured Face M H Percent Fractured Faces PFF 2 Fractured Faces M H Camsizer CAMCONV Conv3 L L WipShape WSMACR Minimal Average Curve Radius L L University of Illinois Aggregate Imaging System UIAIA UIAI Angularity Index L L AIMSGRAD Gradient Angularity L L Aggregate Imaging System AIMS AIMSRAD Radius Angularity L L Angularity Buffalo Wire Works PSSDA-Large PSSDA-Large ROUND Average Roundness L L University of Illinois Aggregate Imaging System UIAIA UISTI Mean Surface Texture Index L L Aggregate Imaging System AIMS AIMSTXTR Texture Index M M Camsizer CAMCONV Conv3 L L Uncompacted Void Content of Coarse Aggregate UCVCC Percent Uncompacted Void Content L L Texture WipShape WSMACR Minimal Average Curve Radius L L CAMSPHT SPHT3 L L Camsizer CAMSYMM Symm3 L L AIMSFORM Form 2-D L L Aggregate Imaging System AIMS AIMSSPH Sphericity L L Shape/Parameter Buffalo Wire Works PSSDA-Large PSSDA-Small ROUND Average Roundness L L Flat and Elongated Ratio FER Percent of Flat and Elongated Particles L H <Wt 2:1 L L Wt 2:1– 3:1 L L Wt 3:1– 4:1 H H Multiple Ratio Analysis MRA MRA Wt 4:1– 5:1 M H VDG-40 SLEND Slenderness Ratio L L VDG-40 Videograder VDG-40 FLAT Flatness Factor L L Camsizer CAML/B l/b3 L L <2:1 L M <3:1 M H WipShape WSFER <4:1 H H < 3:1 L L University of Illinois Aggregate Imaging System UIAIA UIFER 3:1 – 5:1 H H <3:1 L L Shape/ Dimensional Ratio Aggregate Imaging System AIMS AIMSFER 3:1 – 5:1 H H Low (L) CV<=10%, Medium (M) 10%< CV<=20%, High (H) CV>20% Table 8. Classification of coarse aggregate test methods based on repeatability and reproducibility.

to capture the same characteristics. Consequently, the method that is easier to implement and interpret was to be recommended. • Accuracy of test methods is evaluated through: – Comparison between the shape measurements using the test methods and the measurements of particles’ dimen- sions using a digital caliper. – Comparison between the texture and angularity visual rankings of aggregates by experienced individuals and results of test methods. This comparison identified test methods that are not capable of ranking aggregates with extreme differences in angularity and texture character- istics (e.g., uncrushed river gravel vs. crushed gravel, uncrushed river gravel vs. crushed granite). Accuracy of Analysis Methods Comparison with geological projections. The two dimen- sional image analysis methods listed in Table 10 were used to analyze the particle projections shown in Figure 2 (a detailed description of these analysis techniques is presented in Appen- dix C). These particle projections were developed by geologists in the past to describe and quantify the 2-D shape and angu- larity of sediments. These shapes were plotted based on actual 15 Coefficient of Variation (CV) Characteristics Test Method Parameter Abbreviation Measured Parameter as Reported by Test Method Repeatability Reproducibility Uncompacted Void Content of Fine Aggregates UCVCF Percent Uncompacted Void Content L L Camsizer CAMCONV Conv3 L L AIMSGRAD Gradient Angularity L L Aggregate Imaging System AIMS AIMSRAD Radius Angularity L L Buffalo Wire Works PSSDA-Small PSSDA-Small ROUND Average Roundness M M Angularity Compacted Aggregate Resistance CAR CAR Aggregate Resistance L L CAMSPHT SPHT3 L L CAMSYMM Symm3 L L Camsizer CAML/B l/b3 L L Aggregate Imaging System AIMS AIMSFORM Form 2-D L L Shape Buffalo Wire Works PSSDA-Small PSSDA-Small ROUND Average Roundness M M Low (L) CV<=10%, Medium (M) 10%<CV<=20% Table 9. Classification of fine aggregate test methods based on repeatability and reproducibility. Analysis Method Description Texture Index Using Wavelet Used by AIMS analysis Software (AIMSTXTR) Gradiant Angularity Index Used by AIMS analysis Software (AIMSGRAD) Radius Angularity Index Used by AIMS analysis Software (AIMSRAD) 2-D Form Index Used by AIMS analysis Software (AIMSFORM) Sphericity Used by AIMS analysis Software (AIMSSPH) Texture Index (Fourier) (FRTXTR) Angularity Index (Fourier) (FRANG) Form Index (Fourier) (FRFORM) Flat & Elongated Ratio Used By University of Illinois System (UIFER) Angularity Using Outline Slope Used By University of Illinois System (UIAI) Surface Texture Using Erosion- Dilation Technique Used By University of Illinois System (UISTI) Aspect Ratio Used in Image Pro Software (ASPTPRO) Fractal Dimension Used in Image Pro Software (FRCTLPRO) Roundness Used in Image Pro Software (ROUNDPRO) Note: Analysis methods are described in Appendix D. Table 10. Methods used in analyzing aggregate images.

16 observations of sediments and manual measurements of their shape and angularity. Figure 2(a) was developed by Rittenhouse (13) based on an earlier version developed by Wadell (14, 15) to measure 2-D shape; it is considered a standard and accurate method for evaluating shape (16, 17). Figure 2(b) was devel- oped by Krumbein (18) to evaluate angularity. Correlations between analysis method parameters and visual numbers by Rittenhouse and Krumbein (Figure 2) were analyzed using the Pearson and Spearman coefficients. The Pearson coefficient (r) is defined as in Equation 8: r x x y y x x y y i i i n i i i n i n = −( ) −( ) −( ) −( ) = == ∑ ∑∑ 1 2 2 11 ( )8 where x and y represent two p-dimensional observations (items) x = [x1, x2, . . . , xp] and y = [y1, y2, . . . , yp]. x represents the values measured by the image analysis methods on the projections, and y represents the visual numbers assigned to the projections in Figures 2a and 2b. The Spearman coefficient is defined exactly as the Pearson coefficient in Equation 8, but x and y represent the ranking of the image analysis results and visual numbers, respectively, instead of the actual values. The correlation results are shown in Tables 11 and 12. Examples of the correlations of image analysis methods with angularity visual numbers are shown in Figure 3. Rittenhouse (13) and Krumbein (18) projections can be used to identify analysis methods capable of capturing changes in shape and angularity, respectively. The correlation results shown in Tables 11 and 12 suggest that: • The following methods can be used only to describe shape without being affected by angularity of a particle: (a) Flat Elongated Ratio used by University of Illinois test method (UIFER), (b) Form Index measured using Fourier Series (a) Rittenhouse (1943) (b) Krumbein (1941) Figure 2. Charts used by geologists in the past for visual evaluation of granular materials. Analysis Method Parameter Pearson Correlation Coefficient Spearman Correlation Coefficient Applicability AIMSGRAD 0.458 -0.54 N AIMSRAD -0.868 -0.894 Y* AIMSFORM -0.98 -0.991 Y* FRFORM -0.918 -0.993 Y FRANG -0.814 -0.99 Y* FRTXTR -0.858 -0.999 Y* UIFER -0.938 -0.993 Y UIAI -0.388 -0.368 N UISTI 0.273 0.425 N ASPTPRO -0.938 -0.995 Y FRCTLPRO 0.256 -0.322 N ROUNDPRO -0.941 -0.996 Y* * Method correlates with two characteristics. Analysis Method Parameter Pearson Correlation Coefficient Spearman Correlation Coefficient Applicability AIMSGRAD -0.886 -0.983 Y AIMSRAD -0.964 -0.967 Y* AIMSFORM -0.958 -0.967 Y* FRFORM -0.016 -0.033 N FRANG -0.908 -0.883 Y* FRTXTR -0.942 -0.967 Y* UIFER 0.486 -0.317 N UIAI -0.959 -0.983 Y UISTI -0.957 -0.983 Y ASPTPRO -0.414 0.317 N FRCTLPRO -0.869 -0.867 Y ROUNDPRO -0.959 -0.967 Y* * Method correlates with two characteristics. Table 11. Pearson and Spearman correlation coefficients of Rittenhouse sphericity. Table 12. Pearson and Spearman correlation coefficients of Krumbein roundness.

(FRFORM), and (c) Aspect Ratio measured using Image Pro software (ASPTPRO). • The following methods can be used to describe angularity without being affected by shape: (a) Gradient Angularity used in the Aggregate Imaging System AIMS (AIMSGRAD), (b) Angularity Index used by the University of Illinois test method (UIAI), (c) Fractal technique used in Image Pro software (FRCTLPRO). • Roundness measured using Image Pro (ROUNDPRO), and Texture Index using Fourier (FRTXTR), Angularity Index using Fourier (FRANG), Form Index Using AIMS (AIMSFORM), and Radius Angularity using AIMS (AIMSRAD) have good correlation with Rittenhouse sphericity numbers and Krumbein roundness numbers. This indicates that these methods are not as unique as the other methods in distinguishing between angularity and shape of particles. The Angularity Index (UIAI) and Tex- ture Index (UISTI) have high correlations with each other. This could be attributed to the nature of the projections in Figure 2b as they might have been created to have the same levels of surface irregularities at the angularity and texture scales. In other words, there was no distinction between angularity and texture in the projections in Figure 2b. Uniqueness of test methods. This task was performed to examine the uniqueness of the analysis methods in capturing aggregate characteristics. A simple setup of a camera and a microscope was used to capture images of 50 randomly selected coarse particles (12.5 − 9.5 mm; 1/2 − 3/8 in.), and 50 fine par- ticles (2.36 − 1.18 mm; sieve #8–#16) of each aggregate type at specific resolution. The setup was equipped with top lighting to capture gray images for texture analysis and a backlighting to capture black and white images for angularity analysis. The resulting images were analyzed using standard image analysis techniques, some of which are employed in the imaging-based tests evaluated in this study. Using the capabilities of SPSS software, the analysis results from the 50 images of the coarse aggregate size of each aggre- gate type were used to cluster the analysis methods. The analy- sis methods were clustered on the basis of similarities or distances using Ward’s Linkage method. Two types of similarities were used. The Pearson correlation coefficient, given by Equation 8, was used as a measure of proximity when variables (analysis methods) were grouped, and the Euclidean distance, given by Equation 9, was used to cluster aggregates. where x and y represent two p-dimensional observations (items) x = [x1, x2, . . . , xp] and y = [y1, y2, . . . , yp]. Ward’s Linkage method was applied with Pearson cor- relation proximity measure to the analysis results to identify clusters of analysis methods. The results of the cluster analysis are shown in Table 13. For each aggregate type, the test methods that have the same number (1, 2, 3, or 4) are more correlated with each other than with other test methods and are consid- ered clustered. For example, the data from AIMSTXTR analy- sis of CA-1 is statistically different than the data from all the other test methods, indicating that this analysis method cap- tures an aggregate characteristic different than what is captured by all the other methods. The percentage of aggregates that a test method is clustered with other test methods is shown in Table 14. For example, the AIMSTXTR method is clustered alone in 54 percent of aggregates, clustered with another method in 31 percent of aggregates, and with two other methods in 9 percent of aggregates. The increase in percentage in the cells toward the left of the table indicates an increase in the uniqueness of the characteristic measured using this method. Based on the results in Tables 15 and 16, AIMSTXTR is the most unique among the texture parameters, AIMSGRAD and d x y x yi i i p , ( )( ) = −( ) = ∑ 2 1 9 17 R2 = 0.92 0 100 200 300 400 500 600 700 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Krumbein Visual Number UI AI V al ue (a) R2 = 0.94 0 1000 2000 3000 4000 5000 6000 0 0.2 0.4 0.6 0.8 1 Krumbein Visual Number A IM SG RA D Va lu e (b) Figure 3. Examples of the correlations of image analysis methods with visual numbers of angularity.

UIAI are the most unique among the angularity parameters, and AIMSSPH is the most unique among the shape parameters. The UIAI and UISTI methods are clustered together for 12 of the 13 aggregates. Clustering of aggregates based on the results of analysis methods. Ward’s Linkage method was used to cluster aggre- gates based on the angularity and texture measured using each of the analysis methods; results are shown in Tables 15 and 16. As shown in Table 15, both the FRTXTR and FRACTLPRO texture parameters place aggregates CA-1 (uncrushed gravel) and aggregates CA-9 and CA-10 (both are granite) in the same cluster, indicating the inability of methods to detect significant differences between these aggregates. Similarly, UISTI places both aggregates CA-2 (crushed gravel) and CA-10 (granite) in the same texture cluster. The results in Table 16 show that AIMSRAD, FRANG, and ROUNDPRO methods cluster the uncrushed (CA-1) and crushed gravel (CA-2) in the same group, indicating the inability of these methods to distinguish the difference in angularity. Table 17 summarizes the characteristics of the analysis methods. Accuracy of Test Methods A digital caliper was used to measure the three dimensions of 100 particles selected randomly from each of the aggregates 18 Coarse Aggregate Analysis Method 1 2 3 4 5 6 7 8 9 10 11 12 13 AIMSTXTR 1 1 1 1 1 1 1 1 1 1 1 1 1 AIMSGRAD 2 2 2 2 2 2 2 2 2 2 1 2 2 AIMSRAD 2 2 3 2 2 3 2 1 2 3 2 3 3 AIMSFORM 2 2 2 2 2 3 2 1 2 3 2 3 3 AIMSSPH 3 3 3 3 3 1 3 3 3 1 3 1 4 UIFER 4 4 4 4 4 4 4 3 4 4 4 4 1 UIAI 4 4 4 4 4 4 4 2 4 4 4 4 2 UISTI 4 4 4 4 4 4 4 2 4 4 4 4 1 FRFORM 2 2 2 2 2 3 2 1 2 3 2 3 3 FRANG 2 2 2 2 2 3 2 1 2 3 2 3 3 FRTXTR 2 2 2 2 2 3 2 4 2 3 2 3 3 ASPCTPRO 2 2 2 2 2 3 2 1 2 3 2 3 3 FRCTLPRO 2 2 3 2 2 3 3 1 2 3 2 3 3 ROUNDPRO 2 2 2 2 2 3 2 1 2 3 2 3 3 Table 13. Clustering of analysis methods (4 clusters) based on Pearson correlation. Number of Methods to Cluster With Analysis Method 0 1 2 6 7 8 AIMSTXTR 54% 31% 8% 8% 8% AIMSGRAD 23% 15% 8% 8% 8% 38% AIMSRAD 8% 54% 38% AIMSFORM 8% 54% 38% AIMSSPH 54% 38% 8% UIFER 8% 92% UIAI 8% 92% UISTI 100% FRFORM 8% 54% 38% FRANG 8% 54% 38% FRTXTR 8% 8% 46% 38% ASPCTPRO 8% 54% 38% FRCTLPRO 8% 8% 46% 38% ROUNDPRO 8% 54% 38% Table 14. Percentage of clustered aggregates for each analysis method.

with sizes passing a 12.5 mm (1⁄2 in.) sieve and retained on a 9.5 mm (3⁄8 in.) sieve. The percentage of particles with a longest to shortest ratio dimension of 3:1 or more sphericity was calculated; the results are shown in Table 18. The correlations between the caliper measurements and results of test methods were estimated in terms of the coefficient of multiple deter- minations (R2). R2 is a statistic that measures how successful the fit is in explaining the variation of the data. It is defined as the ratio of the sum of squares of the regression (SSR) and the total sum of squares (also known as sum of squares about the mean [SST]) and is expressed as The MRA method had the highest correlation with the digi- tal caliper. The UIFER test method was not able to measure all R SSR SST y y y y i i n i i n 2 2 1 2 1 10= = −( ) −( ) = = ∑ ∑ ) ( ) 19 Method Class 1 Class 2 Class 3 Class 4 AIMSTXTR 1, 2, 12 3, 5, 10, 11, 13 4, 6, 7, 8 9 UISTI 1, 8 2, 3, 7, 10, 11, 13 4, 6, 9, 12 5 FRTXTR 1, 7, 9, 10, 12 2, 4 3, 5, 6, 11, 13 8 FRACTLPRO 1, 4, 9, 10, 12 2, 3, 6, 11, 13 5, 7 8 Method Class 1 Class 2 Class 3 Class 4 AIMSGRAD 1, 8 2, 4, 6, 7, 12 5, 9, 10 3, 11, 13 AIMSRAD 1, 2, 9 3, 4, 11, 13 5, 6, 7, 10, 12 8 UIAI 1 2, 6, 9 3, 4, 5, 7, 10, 11, 12, 13 8 FRANG 1, 2, 3, 6, 9, 11, 12 4, 5, 7, 10 8 13 FRACTLPRO 1, 4, 9, 10, 12 2, 3, 6, 11, 13 5, 7 8 ROUNDPRO 1, 2, 6, 12 3, 4, 5, 7, 9, 10, 11 8 13 Table 15. Coarse aggregates in texture classes estimated using Ward’s Linkage. Table 16. Coarse aggregates in angularity classes estimated using Ward’s Linkage. Analysis Method Features AIMSTXTR • Capable of separating aggregates with different texture characteristics. • Most unique among the texture parameters. AIMSGRAD • Capable of separating aggregates with different angularity characteristics. • Capable of separating angularity from shape. AIMSRAD • Captures angularity but it is not capable of separating 2-D shape from angularity. AIMSFORM • Captures 2-D shape but it is not capable of separating shape from angularity. AIMSSPH • Capable of separating aggregates with different characteristics. • Captures unique characteristics of aggregates. FRTXTR • Does not separate angularity from shape. • Clusters aggregates with distinct characteristics. This can be improved if different image resolutions are used. FRANG • Does not separate angularity from shape. • Clusters aggregates with distinct characteristics. This can be improved if different image resolutions are used. FRFORM • Capable of separating shape from angularity. • Clusters aggregates with distinct characteristics. This can be improved if different image resolutions are used. Used by University of Illinois System (UIFER) • Capable of separating aggregates with different characteristics. • Capable of separating shape from angularity. UISTI • Capable of separating aggregates with different aggregate characteristics. • Clusters aggregates similar to UIAI. This can be improved if different image resolutions are used. ASPTPRO • Separates angularity from shape. FRCTLPRO • Separates angularity from shape. • Clusters aggregates with distinct characteristics in the same group. This can be improved if different image resolutions are used. ROUNDPRO • Separates angularity from shape. • Clusters aggregates with distinct characteristics in the same group. This can be improved if different image resolutions are used. Table 17. Features of methods used in analyzing aggregate images.

aggregates used in this study due to their dark color. Both AIMS and PSSDA-Large provide a sphericity value. The sphericity measured using the digital caliper had very good agreement with AIMS and PSSDA-Large measurements. Figure 4 shows a comparison between AIMS measurements and digital caliper measurements for sphericity. Measurements of angularity and texture of coarse aggregates were compared with visual rankings of aggregates made by five evaluators with backgrounds in asphalt pavements, con- crete pavements, geology, and petrographic analysis. These evaluators were provided with a form to fill with the rankings. R2 values between the evaluators for texture and angularity rankings are shown in Table 19. The rankings made by the evaluators were more correlated for texture than for angularity. The evaluators suggested that the main difficulty was in visually separating angularity from texture. Therefore, aggregates were ranked based on surface irregularity that combines both angularity and texture. There- fore, it was decided for this study to establish a visual ranking of surface irregularity; the correlation between rankings is shown in Table 19. The experimental measurements were compared to the visual rankings of surface irregularity and texture. The comparison with surface irregularity is useful since the evaluated tests them- selves do not use the same methods to analyze angularity and texture. In fact, the definition of angularity in a certain test method can be similar to the definition of texture in another test method. Very good correlation was found between the eval- uators ranking aggregates based on surface irregularity; average rankings are shown in Table 20. Similarly, the evaluators ranked fine aggregate angularity by examining their shape under a microscope; visual rankings are shown in Table 21. The correlations between the measurements and the cor- responding visual ranking were used to rank the test methods, as described later. Cost and Operational Characteristics of Test Methods Information about cost and operational characteristics was collected from vendors, researchers, and operators who have familiarity with these systems for use in ranking the test methods. The information included cost, ease of use, portabil- ity, ability of interpreting data, readiness for implementation 20 Coarse Aggregate Average Sphericity 3:1 & Higher (%) CA 1 0.717 8 CA 2 0.740 2 CA 3 0.675 18 CA 4 0.662 30 CA 5 0.731 2 CA 6 0.711 6 CA 7 0.624 42 CA 8 0.706 6 CA 9 0.643 38 CA 10 0.697 18 CA 11 0.659 22 CA 12 0.666 18 CA 13 0.638 38 Table 18. Aggregate sphericity from longest to shortest dimensions from digital caliper results. 0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00 0.00 0.20 0.40 0.60 0.80 1.00 Sphericity-AIMS Sp he ric ity -M an ua l M ea su re m en ts Figure 4. Comparison between sphericity measurements of AIMS and the digital caliper.

in central laboratories and field laboratories, and applicability of test method to measure different aggregate types and sizes. Table 22 lists this information. Ranking of Test Methods Using the Analytical Hierarchy Process The Analytical Hierarchy Process (AHP) has been adapted to rank the test methods according to their repeatability, repro- ducibility, accuracy, cost, and operational characteristics. The process is presented as computational software to expedite conducting the calculations and provides the user with flexi- bility in specifying the objectives, ranking criteria, and relative importance or priorities of the different criteria elements. Background on the Process AHP is a decision making process that transforms complex decision making into a series of one-on-one comparisons and then combines the results to help arrive at the best, most jus- tified decision. The process incorporates both subjective and objective evaluation measures such that the bias in decision making is reduced and has been used in several applications dealing with the selection of alternatives, investment distri- bution, and energy allocation (21). The AHP method is based on decomposing the goal into its component parts, moving from the general to the specific (i.e., proceeding from the goal to objectives and criteria sub- objectives down to the alternative courses of action). After structuring the hierarchy of all criteria, the next step is to assign a relative weight to each criterion. Weights are assigned 21 Evaluator I II III IV V I 1 0.57 0.58 0.37 0.3 II 1 0.9 0.41 0.57 III 1 0.41 0.46 IV 1 0.41 Angularity V 1 Evaluator I II III IV V I 1 0.91 0.92 0.89 0.82 II 1 0.95 0.79 0.84 III 1 0.84 0.82 IV 1 0.74 Texture V 1 Evaluator I II III IV V I 1 0.77 0.80 0.76 0.59 II 1 0.95 0.69 0.79 III 1 0.70 0.72 IV 1 0.71 Surface Irregularity V 1 Table 19. Coefficients of multiple determinations (R2) between the rankings of evaluators. Aggregate Texture Surface Irregularity CA-1 1.6 1.8 CA-2 4.4 4.2 CA-3 6.8 8.6 CA-4 7.4 8.1 CA-5 12.8 9.8 CA-6 5.2 5.8 CA-7 5.8 6.0 CA-8 1.4 1.2 CA-9 11.4 9.9 CA-10 11.6 10.4 CA-11 9.0 10.3 CA-12 3.6 4.2 CA-13 10.0 10.7 Notes: 1- CA= coarse aggregate; 2- Higher rank is associated with higher angularity and/or texture. Table 20. Average visual rankings of coarse aggregates by evaluators. Aggregate Visual Ranking FA-1 2 FA-2 4 FA-5 5 FA-6 1 FA-9 3 Notes: 1- FA= fine aggregate; 2- Higher rank is associated with higher angularity. Table 21. Visual ranking of fine aggregate angularity by evaluators.

by the user based on a pairwise comparison judgment scale of 1 to 9 (also known as standard preference table). Then the user calculates priorities, using a simple mathematical procedure, to arrive at overall priorities for the alternatives. The sum of all the criteria beneath a given parent criterion in each level of the model must equal one. Each priority list shows its relative importance within the overall structure. From the overall pri- ority list, the decision maker can choose among alternatives by selecting the highest priority alternative. The mathematical functions involved in AHP can be found in Saaty (21). Program Description Computational software was developed to make the calcula- tion process easier and faster. The program provides the user flexibility in changing objectives or selection criteria weights before making the final selection from available alternatives. The software was created using VC++ programming language that can be run on any computer irrespective of the operating system. The program uses the crude estimate, specified by Saaty (21), to calculate the priority vector through the process of averaging over normalized columns technique. The elements of each column are divided by the sum of that column, and then the elements in each resulting row are added and divided by the sum of the numbers in that row. The program uses a graphical interface environment; the process is summarized in the following steps (for illustration, fine aggregate angularity is used to describe the operation steps of the new program): (1) The user enters the number of testing methods being compared and the characteristics determining the performance of the test method (Figure 5a). (2) Generic text boxes are generated and the user inputs the names of each of the characteristics and testing methods (Figure 5b). (3) The user enters the weights assigned to test methods when pairwise comparison is conducted with respect to each char- acteristic (Figure 6a). Note that because the lower triangle of these matrices is the reciprocal of the upper triangle with ones along the diagonal, the user inputs the upper half of the matrix and the other values are updated automatically. (4) The user is prompted to enter the weights comparing the various char- acteristics with respect to overall satisfaction with a method in a new interface (Figure 6b). (5) The program calculates the priority vectors for each of the matrices and displays them in a new interface window (Figure 7a). (6) The program also 22 Applicability to Aggregate Type and Size(d) Test Method Estimated Price ($) Readiness for Implementation(a) Ability to Interpret Data(b) Ease of Use by Technician(b) Portability(c) Coarse Fine Uncompacted Void Content of Fine Aggregates AASHTO T 304 250 1 1 1 1 N/A 1 Uncompacted Void Content of Coarse Aggregates AASHTO T P56 500 1 1 1 1 1 N/A Compacted Aggregate Resistance (CAR) 500 1 1 1 1 N/A 1 Percentage of Fractured Particles in Coarse Aggregate ASTM D 5821 0 1 1 1 1 (N/A) 1 N/A Flat and Elongated Coarse Aggregates ASTM D 4791 250 1 1 1 1 1 N/A Multiple Ratio Shape Analysis 1,500 2 2 2 1 1 N/A VDG-40 Videograder 40,000 - 50,000 2 3 2 2 1 1 Buffalo Wire Works PSSDA -Large 30,000 - 40,000 3 3 3 2 1 N/A Buffalo Wire Works PSSDA -Small 30,000 - 40,000 2 3 2 2 N/A 1 Camsizer 40,000 - 50,000 2 3 2 2 2 1 WipShape 30,000 - 40,000 2 3 3 2 1 N/A University of Illinois Aggregate Image Analyzer (UIAIA) 30,000 - 40,000 3 3 3 2 3 N/A Aggregate Imaging System (AIMS) 30,000 - 40,000 2 3 3 2 1 1 Notes: (a) 1: Available commercially. Wide use in laboratories. 2: Available commercially. Limited use in laboratories. 3: Not available commercially. Limited use in research laboratories. Can be made available commercially. (b) 1: Very Easy, 2: Easy, 3: Intermediate, 4: Difficult. (c) 1: Can be used in central and field laboratories. Requires less than 1 hr to move it. 2: Can be used in central and field laboratories. Requires less than 4 hrs to move it. 3: Not portable. Cannot be used in central and field laboratories. (d) 1: Measure all aggregate sizes and types, 2: Measure all aggregate types but not all sizes, 3: Measure all sizes but not very dark colored aggregates, N/A: Not Applicable. Table 22. Rating of test methods’ operational characteristics.

23 (a) Number of Characteristics and Test Methods (b) Names of Characteristics and Test Methods Figure 5. Screens of interface to enter numbers and names of characteristics and test methods.

24 (a) Weights Comparing Test Methods to Characteristics (b) Weights Comparing Characteristics Figure 6. Screens of interface to enter weights comparing test methods to characteristics, and characteristics with respect to overall satisfaction with method.

25 (a) Priority Vectors (b) Overall Ranking Figure 7. Screens of priority vectors and overall ranking of test methods.

calculates the overall ranking of the test methods by multiply- ing the priority matrix of the methods by the priority vector of the characteristics and displays it in a separate interface window (Figure 7b). The program also has features that enable the user to: (1) extract the priority vectors and the overall ranking from a text file and (2) examine the influence of changes in the weights or importance of one or more of the characteristics without changing the remaining ones (i.e., the software has to be exe- cuted several times with just one matrix change) without a need to re-enter the unchanged matrices. AHP Ranking of Test Methods The ranking of test methods depends on the desired out- comes from the test. This section provides an example of how the AHP can be used to determine the ranking of test methods measuring fine aggregate angularity, and texture and shape of coarse aggregates. The first level in AHP is the overall goal, which is the satis- faction with test methods. The second level consists of the cri- teria elements by which this satisfaction is measured. These characteristics are repeatability, reproducibility, accuracy, price, readiness for implementation, ability to interpret data and results, ease of use by technician, portability, and applicability to measure different aggregate types and sizes. The third level consists of the test methods that are under evaluation. Figure 8 illustrates a basic hierarchy for the ranking process. The ranking is determined using pairwise comparisons of the characteristics (level 2) and the test methods (level 3). The first pairwise comparison is conducted among the character- istics in the second level using the comparison scale given in Table 23 and results are listed in Table 24. The number in each cell of the table is a weight that reflects the relative importance of the characteristic in the horizontal list compared with the one in the vertical list. If this number is higher than one, it means that the characteristic listed in the row is more impor- tant than the characteristic listed in the column. For example, accuracy is considered three times as important as repeatability and reproducibility and five times as important as all the other characteristics. All other characteristics are considered to be equal in their importance. Weights that compare test methods based on each of the characteristics are based on the measurements and data pre- sented in Chapter 2; these are listed in Table 25. The compar- ison scale values shown in Table 28 were selected based on the importance of each of the desired characteristics as follows: • Repeatability/Reproducibility: Repeatability and reproduc- ibility are categorized into three main categories as Levels 1, 2, and 3. Levels 1 and 2 can be considered as acceptable scales and some of the test methods can move from Level 2 to 1 with some minor improvements. However, Level 3 is un- acceptable because it covers high ranges of coefficient of vari- ations. Therefore, the difference between Levels 3 and 2 is less desirable than the difference between Levels 1 and 2. • Accuracy: Accuracy of test methods was assessed based on the correlation between the test method and a reference method. The scale for accuracy was established by dividing the R2 values into four categories as shown in Table 26. The ratio between the numbers assigned to each accuracy group is then used to assign the accuracy scale. • Price: The price scale is assigned taking into consideration that the lowest price of a test method is about $250, while the highest price is about $45,000 ($250 is taken as the basis for the cost ratio). 26 AIMS Method n Method 2 Method 1 Satisfaction with Method (Overall Objective) R ep ea ta bi lit y Ea se o f U se In te rp re t D a ta R ep ro du ci bi lit y A cc u ra cy Pr ic e R ea di ne ss Po rt a bi lit y A pp lic a bi lit y Figure 8. An example of basic analytical hierarchy process (AHP). Verbal Judgment of Preference Numerical Rating Equally Important or Preferred 1 Weakly More Important 3 Moderately More Important 5 Strongly More Important 7 Absolutely More Important 9 Weakly Less Important 1/3 Moderately Less Important 1/5 Strongly Less Important 1/7 Absolutely Less Important 1/9 Table 23. Rating scale.

• Readiness/Portability: The scale for readiness reflects the preference for a test method that has been used by research and testing laboratories and thus methods that are not avail- able commercially are considered slightly less desirable than those that are available. However, this point is not highly emphasized in the scale (the maximum possible ratio is only 5) because any of the methods can be made available commercially in the future. The same applies for porta- bility, as the portability of those methods that are given a scale of “3 Not portable” can be improved with some design changes. • Interpretation of Data and Ease of Use: The values assigned in Table 22 are based on current knowledge of the test methods regarding their use in routine analysis of aggregates. Except for the methods labeled (4:difficult), technical train- ing can improve the assigned value from (3:intermediate) to (2:easy) or even (1:very easy) indicating that the change from 3 to 4 is less desirable than the change from 1 to 3. • Applicability to Measure Different Aggregate Types and Sizes: Test methods are expected to measure all aggregate types and sizes. If the method fails to measure some sizes or some aggregate types, or both, its applicability rating should be reduced. The values assigned for the applicability of test method to measure different aggregate types and sizes listed in Table 22 are based on current knowledge and experience with the test methods. The assigned values assume that it is weakly more important (assigned a value of 3) to have a method that can measure all aggregate types and sizes than a method that can measure all aggregate types but not all aggregate sizes or to have a method that can measure some aggregate sizes for all aggregate types than a method that can measure all sizes for some aggregate types. It is considered moderately more important (assigned a value of 5) to have a method that can measure all aggregate types and sizes than a method that can measure all aggregate sizes but not all aggregate types. A few examples are provided to highlight the process for ranking the test methods. Fine Aggregate Angularity AHP was used to rank the test methods that measure fine aggregate angularity: uncompacted void content of fine aggre- gate (UCVCF), compacted aggregate resistance (CAR), Cam- sizer, Buffalo Wire Works (PSSDA-Small), and AIMS. In this example, the same weights (1) were assigned for all the charac- teristics in the second level (i.e., characteristics were consid- ered equally important). This means that all cells in Table 27 will have a value of 1. A pairwise comparison of all test methods according to one characteristic was then conducted using numerical ratings selected from Table 23. These ratings are used in Table 28 in order to compare a test method from the horizontal list to that of the vertical list based on the characteristics under consideration. Once the values in Tables 27 and 28 are assigned, the next step consists of the computation of priority lists of test methods for each of the desirable characteristics. In mathematical terms, the principal eigen vector is computed for each matrix which gives the vector of priority ordering. Saaty (21) proposed some crude estimates that can be easily followed to calculate these vectors. One good estimate method is to divide the ele- ments of each column in the matrix by the sum of that column (i.e., normalize the column). Then elements in each resulting row are added then divided by the number of elements in the row. This is a process of averaging over the normalized column. The resulting priority vectors from each matrix in Table 29 are then combined to create a matrix that represents priority of test method by each characteristic. In order to obtain the overall ranking of the test methods, the priority matrix of the methods by each characteristic will be multiplied by the 27 Characteristics of Test Methods R ep ea ta bi lit y R ep ro du ci bi lit y A cc u ra cy Pr ic e R ea di ne ss In te rp re t D at a Ea se o f U se Po rt ab ili ty A pp lic ab ili ty Repeatability 1 1 0.33 1 1 1 1 1 1 Reproducibility 1 1 0.33 1 1 1 1 1 1 Accuracy 3 3 1 5 5 5 5 5 5 Cost 1 1 0.2 1 1 1 1 1 1 Readiness 1 1 0.2 1 1 1 1 1 1 Interpret Data 1 1 0.2 1 1 1 1 1 1 Ease of Use 1 1 0.2 1 1 1 1 1 1 Portability 1 1 0.2 1 1 1 1 1 1 Applicability 1 1 0.2 1 1 1 1 1 1 Table 24. Example of the relative importance of the test methods characteristics.

priority vector of the characteristics resulting from Table 27. In other words, the overall ranking of a method can be obtained by multiplying the weight indicating the rank of a test method with respect to the characteristic by the weight of that char- acteristic then add them up for all characteristics. The result- ing priority vectors and the overall ranking of test methods used to measure fine aggregate angularity are presented in Table 29. 28 Comparison Scale Criterion Characteristic 1 3 5 7 9 2:1 X 3:2 X Repeatability/ Reproducibility 3:1 X 2:1 X 3:1 X 3:2 X 4:1 X 4:2 X Accuracy Coarse-Shape (Ratio of R2 groups) 4:3 X 2:1 X 3:1 X 3:2 X 4:1 X 4:2 X Accuracy Coarse- Irregularity (Ratio of R2 groups) 4:3 X 2:1 X 3:1 X 3:2 X 4:1 X 4:2 X Accuracy Coarse-Texture (Ratio of R2 groups) (Rankings) 4:3 X 2:1 X 3:1 X 3:2 X 4:1 X 4:2 X Accuracy Fine-Angularity (Ratio of R2 groups) 4:3 X X X X X Price (Ratio of Cost) X X X Readiness X X X Portability X X X X X Data Interpretation X X X X X Ease of Use X X X Applicability <6 >6 <20 >20 <50 >50 <80 >80 2:1 3:2 3:1 2:1 3:2 3:1 2:1 3:2 3:1 4:3 4:2 2:1 3:2 3:1 4:3 4:2 2:1 3:2 3:1 X Table 25. Weights that compare test methods based on each of the characteristics. R2 Category > 0.70 1 0.6 – 0.7 2 0.5 – 0.6 3 < 0.5 4 Table 26. Accuracy categories based on R2 values.

The results of this example show that when all characteristics were assumed to be equally important, the uncompacted void content of fine aggregate (UCVCF) method was at the top of the priority list, mainly due to the low cost of this test. How- ever, this priority order will change if the weights assigned to the characteristics or to the methods are changed. In another example, the accuracy of the test method was considered more important than the other characteristics, and was thus assigned a value of 5 (based on the scale provided in Table 23). The new matrix together with the calculated pri- ority vector are presented in Table 30. Multiplying the new characteristic’s priority vector by the matrix of priority vectors (resulting from comparing method with respect to the charac- teristics presented in Table 28) will result in the overall ranking of test methods presented in Table 31 for different accuracy levels of preference. It is apparent from Table 31 that when only accuracy is considered moderately or absolutely more important than the other characteristics, the ranking of test methods has changed; AIMS ranked first in the priority ordering list (with more significant difference in the latter case). The results from the two examples clearly indicate that the selected weights can have a significant influence on the over- all ranking of test methods. Therefore, it is very important that the weights should be selected based on expert opinion and judgment of the process. Coarse Aggregate Texture AHP was used in this example to rank the test methods that are used to measure coarse aggregate texture (UCVCC, Camsizer, WipShape, UIAIA, and AIMS). In this example, accuracy was considered moderately more important than applicability of a test method to measure all aggregate sizes and types (assigned a value of 5) and absolutely more important than all other remaining characteristics (assigned a value of 9). Also, applicability to different aggregate types and sizes was considered moderately more important than other methods (assigned a value of 5). The priority list for all the character- istics based on this consideration and the resulting priority vector are presented in Table 32. Using the weights provided in Table 25, the process described for fine aggregate angularity was followed; the resulting prior- ity vectors presented in Table 33 for testing methods with respect to characteristics were obtained. The overall ranking of test methods used to measure coarse aggregate texture, presented in Table 34, clearly shows that AIMS has the highest rank among all methods. As dis- cussed in the previous section, the wavelet method that AIMS uses in analyzing coarse aggregate texture was found to be unique and most accurate; it contributed significantly to this ranking although some imaging methods have compa- rable characteristics. Because imaging methods will become more practical and easy to use with some reasonable training, only repeatability, reproducibility, accuracy, and applicability should be con- sidered in comparing test methods. When this criterion was applied, the overall ranking of test methods measuring coarse aggregate texture shown in Table 34 was obtained, placing AIMS on the top of the priority list and thus it would be the user’s first choice for measuring coarse aggregate texture. The overall rankings of test methods presented in Table 34 show that UCVCC method has high priority when all character- istics are considered, but it becomes less favorable when price becomes of less concern. Coarse Aggregate Shape AHP was used in this example to rank test methods that measure coarse aggregate shape parameters and dimensional 29 Characteristics of Test Methods R ep ea ta bi lit y R ep ro du ci bi lit y A cc u ra cy Pr ic e R ea di ne ss In te rp re t D at a Ea se o f U se Po rt ab ili ty A pp lic ab ili ty Repeatability 1 1 1 1 1 1 1 1 1 Reproducibility 1 1 1 1 1 1 1 1 1 Accuracy 1 1 1 1 1 1 1 1 1 Cost 1 1 1 1 1 1 1 1 1 Readiness 1 1 1 1 1 1 1 1 1 Interpret Data 1 1 1 1 1 1 1 1 1 Ease of Use 1 1 1 1 1 1 1 1 1 Portability 1 1 1 1 1 1 1 1 1 Applicability 1 1 1 1 1 1 1 1 1 Table 27. Comparison of the characteristics based on overall satisfaction with methods.

30 Test Method Characteristic Test Method UCVCF CAR PSSDA-Small Camsizer AIMS UCVCF 1 1 3 1 1 CAR 1 1 3 1 1 PSSDA-Small 0.33 0.33 1 0.33 0.33 Camsizer 1 1 3 1 1 Repeatability AIMS 1 1 3 1 1 UCVCF 1 1 3 1 1 CAR 1 1 3 1 1 PSSDA-Small 0.33 0.33 1 0.33 0.33 Camsizer 1 1 3 1 1 Reproducibility AIMS 1 1 3 1 1 UCVCF 1 1 1 0.143 0.11 CAR 1 1 1 0.143 0.11 PSSDA-Small 1 1 1 0.143 0.11 Camsizer 7 7 7 1 0.33 Accuracy AIMS 9 9 9 3 1 UCVCF 1 1 9 9 9 CAR 1 1 7 9 7 PSSDA-Small 0.11 0.14 1 1 1 Camsizer 0.11 0.11 1 1 1 Price AIMS 0.11 0.14 1 1 1 UCVCF 1 1 3 3 3 CAR 1 1 3 3 3 PSSDA-Small 0.33 0.33 1 1 1 Camsizer 0.33 0.33 1 1 Readiness AIMS 0.33 0.33 1 1 1 UCVCF 1 1 5 5 5 CAR 1 1 5 5 5 PSSDA-Small 0.20 0.20 1 1 1 Camsizer 0.20 0.20 1 1 1 Interpretation of Data AIMS 0.20 0.20 1 1 1 UCVCF 1 1 3 3 5 CAR 1 1 3 3 5 PSSDA-Small 0.33 0.33 1 1 3 Camsizer 0.33 0.33 1 1 3 Ease of Use AIMS 0.20 0.20 0.33 0.33 1 UCVCF 1 1 3 3 3 CAR 1 1 3 3 3 PSSDA-Small 0.33 0.33 1 1 1 Camsizer 0.33 0.33 1 1 1 Portability AIMS 0.33 0.33 1 1 1 1 1 1 1 1 CAR 1 1 1 1 1 PSSDA-Small 1 1 1 1 1 Camsizer 1 1 1 1 1 Applicability UCVCF AIMS 1 1 1 1 1 Table 28. Comparison of test methods measuring fine aggregate angularity.

31 Priority Vectors for Test Methods with Respect to Characteristics R ep ea ta bi lit y R ep ro du ci bi lit y A cc u ra cy Pr ic e R ea di ne ss In te rp re t D at a Ea se o f U se Po rt ab ili ty A pp lic ab ili ty Priority Vector of Characteristics with Respect to Overall Satisfaction with Method Overall Ranking Test Method UCVCF 0.231 0.231 0.051 0.444 0.333 0.385 0.342 0.333 0.20 0.111 Repeatability 0.283 UCVCF CAR 0.231 0.231 0.051 0.402 0.333 0.385 0.342 0.333 0.20 0.111 Reproducibility 0.279 CAR PSSDA- Small 0.077 0.077 0.051 0.052 0.111 0.077 0.130 0.111 0.20 × 0.111 Accuracy = 0.098 PSSDA- Small Camsizer 0.231 0.231 0.306 0.049 0.111 0.077 0.130 0.111 0.20 0.111 Price 0.161 Camsizer AIMS 0.231 0.231 0.540 0.052 0.111 0.077 0.056 0.111 0.20 0.111 Readiness 0.179 AIMS 0.111 Interpret Data 0.111 Ease of Use 0.111 Portability 0.111 Applicability Table 29. Resulting priority vectors and overall ranking of test methods measuring fine aggregate angularity (assuming characteristics are equally important). Characteristic R ep ea ta bi lit y R ep ro du ci bi lit y A cc u ra cy Co st R ea di ne ss In te rp re t D at a Ea se o f U se Po rt ab ili ty A pp lic ab ili ty Priority Vector Repeatability 1 1 0.2 1 1 1 1 1 1 0.077 Reproducibility 1 1 0.2 1 1 1 1 1 1 0.077 Accuracy 5 5 1 5 5 5 5 5 5 0.385 Cost 1 1 0.2 1 1 1 1 1 1 0.077 Readiness 1 1 0.2 1 1 1 1 1 1 0.077 Interpret Data 1 1 0.2 1 1 1 1 1 1 0.077 Ease of Use 1 1 0.2 1 1 1 1 1 1 0.077 Portability 1 1 0.2 1 1 1 1 1 1 0.077 Applicability Note: Accuracy is moderately more important than other characteristics. 1 1 0.2 1 1 1 1 1 1 0.077 Accuracy Level of Preference Test Method 1 = Equally Important 5 = Moderately Important 9 = Absolutely Important UCVCF 0.28 0.21 0.17 CAR 0.28 0.21 0.17 PSSDA-Small 0.10 0.08 0.08 Camsizer 0.16 0.21 0.23 AIMS 0.18 0.29 0.35 Table 30. Comparison of characteristics with respect to overall satisfaction with method. Table 31. Overall ranking of test methods measuring fine aggregate angularity for different accuracy levels of preference.

32 Characteristic R ep ea ta bi lit y R ep ro du ci bi lit y A cc u ra cy Co st R ea di ne ss In te rp re t D at a Ea se o f U se Po rt ab ili ty A pp lic ab ili ty Priority Vector Repeatability 1 1 0.11 1 1 1 1 1 0.2 0.046 Reproducibility 1 1 0.11 1 1 1 1 1 0.2 0.046 Accuracy 9 9 1 9 9 9 9 9 5 0.465 Cost 1 1 0.11 1 1 1 1 1 0.2 0.046 Readiness 1 1 0.11 1 1 1 1 1 0.2 0.046 Interpret Data 1 1 0.11 1 1 1 1 1 0.2 0.046 Ease of Use 1 1 0.11 1 1 1 1 1 0.2 0.046 Portability 1 1 0.11 1 1 1 1 1 0.2 0.046 Applicability 5 5 0.2 5 5 5 5 5 1 0.211 Note: Accuracy is moderately more important than applicability and absolutely more important than other characteristics. Priority Vectors for Test Methods with Respect to Characteristics Test Method R ep ea ta bi lit y R ep ro du ci bi lit y A cc u ra cy Co st R ea di ne ss In te rp re t D at a Ea se o f U se Po rt ab ili ty A pp lic ab ili ty UCVCC 0.231 0.231 0.036 0.650 0.442 0.556 0.496 0.442 0.280 Camsizer 0.231 0.231 0.183 0.084 0.165 0.111 0.238 0.165 0.107 WipShape 0.231 0.231 0.036 0.088 0.165 0.111 0.089 0.165 0.281 UIAIA 0.231 0.231 0.372 0.088 0.063 0.111 0.089 0.165 0.051 AIMS 0.077 0.077 0.372 0.088 0.165 0.111 0.089 0.165 0.281 Test Method All Characteristics Considered Only Repeatability, Reproducibility, Accuracy, and Applicability Considered UCVCC 0.22 0.10 Camsizer 0.16 0.13 WipShape 0.13 0.10 UIAIA 0.23 0.21 AIMS 0.27 0.24 Table 32. Comparison of characteristics with respect to overall satisfaction with method. Table 33. Priority vectors of test methods measuring coarse aggregate texture. Table 34. Overall ranking of test methods measuring coarse aggregate texture.

ratio (FER, MRA, VDG-40 Videograder, Camsizer, Wip- Shape, UIAIA, AIMS, and Buffalo Wire Works [PSSDA- Large]). The criterion that was used in the coarse aggregate texture example was used here. Therefore, the priority list for all the characteristics in the second level and the result- ing priority vector presented in Table 35 will apply for this example. The results presented in Table 36 show that, when all char- acteristics are considered, the MRA has the highest rank among all methods. The method’s high accuracy, ease of use, and low cost contributes to this ranking. It is expected that the imag- ing methods will become more practical and easy to use after being in practice for some time and thus only repeatability, reproducibility, accuracy, and applicability should be consid- ered in comparing test methods. The weighting factors assigned to the accuracy catego- ries can influence the ranking of test methods. For exam- ple, the threshold for the highest accuracy category is 0.7 (Table 26). If the analysis is conducted for an R2 of 0.69 (instead of 0.7) for the highest accuracy level, a somewhat different ranking will result (as shown in the last column of Table 36). X-Ray Computed Tomography of Aggregates Traprock, limestone, and crushed river gravel aggregates were analyzed in this part of the study. Particles smaller than 12.5 mm (1⁄2 in.) but larger than 9.5 mm (3/8 in.) were placed in a plastic sample container 100 mm (4 in.) in diameter and 150 mm (6 in.) in height that was then filled with wax to eliminate any disturbance to the particle arrangement during scanning. X-ray computed tomography (CT)—a nondestruc- tive technique to image the interior of the sample—was used to produce images (examples are shown in Figure 9) that were analyzed to quantify the characteristics of the granular materials. The 3-D shape of particles was quantified based on measure- ments conducted on 3-D X-ray CT images using the Spherical Harmonic Series (SHS) presented by Garboczi (22). The results of the X-ray CT images analysis are shown in Figure 10 and summarized in Table 37. These results show gravel to be the most spherical material and that traprock has the highest angularity and texture, followed by limestone and then gravel. 33 Priority Vectors for Test Methods with Respect to Characteristics Test Method R ep ea ta bi lit y R ep ro du ci bi lit y A cc u ra cy Co st R ea di ne ss In te rp re t D at a Ea se o f U se Po rt ab ili ty A pp lic ab ili ty FER 0.143 0.019 0.041 0.496 0.328 0.408 0.356 0.270 0.180 MRA 0.143 0.183 0.213 0.244 0.125 0.213 0.158 0.270 0.180 VDG-40 0.143 0.183 0.213 0.052 0.125 0.076 0.158 0.105 0.180 Camsizer 0.143 0.183 0.213 0.052 0.125 0.076 0.158 0.105 0.066 WipShape 0.143 0.066 0.019 0.052 0.125 0.076 0.057 0.105 0.180 UIAIA 0.143 0.183 0.088 0.052 0.046 0.076 0.057 0.105 0.034 AIMS 0.143 0.183 0.213 0.052 0.125 0.076 0.057 0.105 0.180 PSSDA- Large 0.143 0.183 0.088 0.052 0.046 0.076 0.057 0.105 0.180 Table 35. Priority vectors of test methods measuring coarse aggregate shape with respect to characteristics. Test Method All Characteristics Considered Only Repeatability, Reproducibility, Accuracy, and Applicability Considered Only Repeatability, Reproducibility, Accuracy, and Applicability Considered* FER 0.15 0.06 0.06 MRA 0.20 0.15 0.15 VDG-40 0.18 0.15 0.15 Camsizer 0.15 0.13 0.13 WipShape 0.08 0.06 0.06 UIAIA 0.08 0.06 0.13 AIMS 0.17 0.15 0.15 PSSDA-Large 0.11 0.09 0.09 *Using different values for accuracy categories. Table 36. Overall ranking of test methods for measuring coarse aggregate shape.

The SHS based on the images supplied by 3-D imaging techniques (such as X-ray CT) can be used to reconstruct the 3-D particle profiles. These reconstructed profiles can be used in simulation programs that incorporate real 3-D particle representations (22, 23). Figure 11 shows 3-D reconstructed profiles of gravel, limestone, and traprock materials. These profiles show some digital layering resulting from the relatively low resolution used to capture the X-ray CT images (0.8 mm/ 34 (a) Gravel (b) Limestone (c) Traprock Figure 9. Examples of X-ray CT images. These 2-D images are 1024  1024 pixels in size, with each pixel representing a physical distance of about 0.1 mm. The slice-to-slice resolution in the out of plane direction was 0.8 mm per voxel length. The images to the left were obtained using X-ray CT; and the images to the right were thresholded to highlight aggregate particles. 0 10 20 30 40 50 60 70 80 90 100 0 0.2 0.4 0.6 0.8 1 Sphericity Index Pe rc en ta ge o f P ar tic le s, % Gravel Limestone Traprock 0 10 20 30 40 50 60 70 80 90 100 0 0.05 0.1 0.15 0.2 Form, SHS Signature Pe rc en ta ge o f P ar tic le s, % Gravel Limestone Traprock (a) (b) 0 10 20 30 40 50 60 70 80 90 100 0.030.020.010 Angularity, SHS Signature Pe rc en ta ge o f P a rti cl es , % Gravel Limestone Traprock 0 10 20 30 40 50 60 70 80 90 100 0.00060.00040.00020 Texture, SHS Signature Pe rc en ta ge o f P a rti cl es , % Gravel Limestone Traprock (c) (d) Figure 10. Results of the analysis of images obtained using X-ray CT.

35 Mean* St. Dev.** Analysis Method Statistical Distribution Model TR. LS. GR. TR. LS. GR. Sphericity Index Normal 0.61 0.61 0.65 0.16 0.17 0.16 Shape, SHS Signature LogNormal -1.11 (0.0773) -1.21 (0.0610) -1.30 (0.0496) 0.26 0.28 0.36 Angularity, SHS Signature LogNormal -1.98 (0.0106) -2.06 (0.00868) -2.33 (0.00466) 0.26 0.27 0.46 Texture, SHS Signature LogNormal -3.64 (2.29×10-4) -3.66 (2.17×10-4) -3.76 (1.76×10-4) 0.22 0.21 0.24 *The mean values for the LogNormal models are provided for the log scale and between brackets for the arithmetic scale. **The standard deviation values for the LogNormal model are provided for the log scale. Table 37. Summary of the statistical analysis of X-ray CT images. (a) Gravel (b) Limestone (c) Traprock Figure 11. Reconstruction of three-dimensional profiles of particles using spherical harmonic series. slice). However, the reconstructed profiles show the smooth- ness of the gravel particles. The findings of the X-ray CT of aggregate shape analysis are summarized as follows: • SHS analysis indicated that traprock had the highest angu- larity and texture, followed by limestone, and then gravel. • Analysis of X-ray CT images was capable of discriminating among the angularity and texture of the different aggregates. • 3-D X-ray CT stores the 3-D shapes in a computer for fur- ther computer simulations. • The image processing techniques used in separating the particles in X-ray CT require substantial manual manip- ulation of images. These segmentation techniques could

influence and alter the measurements of angularity and texture. • While the X-ray CT is a powerful research tool, it is pre- mature to use it as a practical tool for routine measurements of aggregate shape. Statistical-Based Methodology for Classification of Aggregates The ease of interpretation of test results is an essential part to facilitate the implementation in practice. The imaging-based tests discussed in this report provide measurements of a large number of particles. These measurements are valuable to detect differences between aggregates based on sound statistical methods. Therefore, it is essential to develop a methodology to summarize the measurements and present them to the user in a simple form that facilitates implementation. This section contains a methodology presented in the visual basic program of an Excel workbook to summarize the aggre- gate characteristics and classify aggregates based on these char- acteristics. The program includes graphical presentations of the results, helps to compare the results from different aggregates, and combines the results of multiple analyses of the same aggregate source. Aggregates’ shape, angularity, and texture were measured using the three analysis methods that are part of the AIMS software: (1) sphericity as a 3-D measure of coarse aggregates, (2) gradient angularity for coarse and fine aggregates, and (3) texture of coarse aggregates quantified by the wavelet method. Measurements from 195 tests on coarse aggregates and 75 tests on fine aggregates were used in developing the methodology. On average, a coarse aggregate test involved 56 particles and a fine aggregate test involved about 300 par- ticles. All these data were used in the development of the new classification system. The use of different operators and repeated measurements ensured that the classification methodology accounted for variations in measurements among operators. Cluster analysis was used to develop groups (or clusters) of aggregates based on the distribution of their characteris- tics. In this study, the usual metric of Euclidean distance (Equation 9) and Ward’s Linkage method were used. The clus- tering method was applied to all characteristics obtained from AIMS. Three methods for grouping the analysis results were used with the objective of determining whether common group limits can be obtained for aggregates irrespective of their size. In one method, group limits were selected for each aggregate characteristic based on measurements by all operators for each size separately. In another method, the group limits were determined by averaging those obtained for the three sizes. The third method was to group the analysis results obtained for each characteristic using data from all operators and for all sizes combined. Results of clustering using the three different methods are shown in Figure 12. Figure 12a shows the groups’ limits of the coarse aggregate texture for each size, the average for the limits of three sizes (“Avg. Sizes” label in Figure 12a), and for all sizes combined (“All” label in Figure 12a). The results show that the groups’ limits obtained using the three were very close. The same conclusion was reached by examining the results in Figures 12b and 12d for the other characteristics. Further analysis was also conducted to determine whether it is feasible to unify the angularity groups’ limits of both the fine and coarse fractions. The groups’ limits for the angular- ity of fine and coarse aggregates were determined, plotted in Figure 12, showing slight differences between the limits of fine and coarse fractions, with the largest difference being in the third group. This difference, however, is small compared to the actual angularity values, and thus could be unified limits. The new aggregate shape classification limits are shown in Figure 13. Analysis and Results The AIMS software was used to calculate the percentages of each aggregate that belong to the different groups in Fig- ure 13; the results are shown in Figures 14 and 15. These figures show the distribution of a certain shape property in a number of aggregate samples. The variability in the char- acteristics within and between aggregates indicates that com- paring or classifying aggregates based on percent of particles in a single group could be misleading. This is also true for the classification based on average values, especially when an aggregate sample includes a small percent of particles that have extremely high or low values. As such, the new classifi- cation methodology considers the distribution rather than an average value. The discussion provided in the following sections highlights the implications of using the developed methodology on aggregate shape classification with emphasis on examining the effects of different factors such as crushing on aggregate characteristics. Aggregate Texture versus Angularity The classification methodology incorporates measurements of texture and angularity for coarse aggregates, but it uses angularity measurements only for fine aggregates. A study by Masad et al. (24) clearly showed that a high correlation exists between angularity (measured on black and white images) and texture (measured on gray-scale images) of fine aggregates. This finding led to focusing on fine aggregates angularity mea- sured on black and white images. This is an easier task than 36

37 0 100 200 300 400 500 600 700 800 900 Polished Smooth Low Roughness Moderate Roughness High Roughness Upper Limits for Texture Classes Te xt u re In de x All Avg Sizes 3/4 3/8 #4 (a) Coarse Aggregates Texture 0 2000 4000 6000 8000 10000 12000 Rounded Sub-Rounded Sub-Angular Angular Upper Limits for Gradient Angularity Classes G ra di en t A n gu la rit y All Avg Sizes 3/4 3/8 #4 (b) Coarse Aggregates Angularity Figure 12. Limits of groups (clusters) of individual and combined aggregates. (continued on next page)

38 0 2000 4000 6000 8000 10000 12000 Rounded Sub-Rounded Sub-Angular Angular Upper Limits for Gradient Angularity Classes G ra di en t A ng ul a rit y All Avg Sizes #8 #16 #60 (c) Fine Aggregates Angularity 0 0.2 0.4 0.6 0.8 1 Flat Elongated Low Sphericity Moderate Sphericity High Sphericity Upper Limits for Sphericity Classes Sp he ric ity All Avg Sizes 3/4 3/8 #4 (d) Coarse Aggregates Shape (Sphericity) Figure 12. (Continued).

02000 4000 6000 8000 10000 12000 Rounded Sub-Rounded Sub-Angular Angular Upper Limits for Gradient Angularity Classes G ra di en t A ng u la rit y Coarse-ALL Fine-ALL Coarse-Avg Fine-Avg (e) Coarse and Fine Aggregates Angularity R ou nd ed Su b R ou nd ed Su b A ng u la r A ng u la r2100 4000 5400 Fl at / El on ga te d Lo w Sp he ric ity M od er at e Sp he ric ity H ig h Sp he ric ity0.6 0.7 0.8 1.0 Po lis he d Sm oo th Lo w R ou gh ne ss M od er at e R ou gh ne ss H ig h R ou gh ne ss 165 275 350 460 Angularity Shape Texture Figure 12. (Continued). Figure 13. Aggregate characteristics classification chart.

40 0% 20% 40% 60% 80% 100% 3 4 521 6 7 8 9 10 11 12 13 Aggregate Label Pe rc en ta ge in T ex tu re G ro up Polished Smooth Low Roughness Moderate Roughness High Roughness 0% 20% 40% 60% 80% 100% 321 6 754 8 9 10 11 12 13 Aggregate Label Pe rc en ta ge in A ng ul ar ity G ro up Rounded Sub Rounded Sub Angular Angular 0% 20% 40% 60% 80% 100% 321 5 6 74 8 9 10 11 12 13 Aggregate Label Pe rc en ta ge in F or m G ro up Flat Elongated Low Sphericity Moderate Sphericity High Sphericity (b) Angularity in Coarse Aggregate (c) Form in Coarse Aggregate (a) Texture in Coarse Aggregate Figure 14. Distributions of coarse aggregate characteristics.

capturing the surface texture of fine aggregates rapidly and accurately using a computer-automated system. In the case of coarse aggregates, it was found that there is a distinct differ- ence between angularity and texture, and these two properties have different effects on performance (24, 25). As can be seen from Figure 16a, which shows the average texture and corre- sponding angularity for each of the coarse aggregate samples, aggregates could have high angularity but low texture. This is even true for individual particles, as shown in Figure 16b. Particles from aggregates CA-2 and CA-9 (see Table 4) had comparable angularity values but there was a significant dif- ference in texture. The cumulative distribution of texture in the coarse aggre- gate samples shown in Figure 17 indicates that the texture of these aggregate samples was spread over a wide range; none of the other characteristics had such a wide range. Texture also had higher variability than angularity within an aggregate sample (see Figure 14a and Figure 14b). Effect of Crushing and Size on Shape Properties The developed methodology can be used to examine the influence of crushing on shape. Two types of crushed and uncrushed aggregates were used in this study: river gravel (CA-1 and CA-2) and glacial gravel (CA-7 and CA-8). CA-1 and CA-8 were uncrushed, while CA-2 and CA-7 were crushed. The results in Figures 14a and 14b show that crushing the gravel did not influence texture, but significantly increased their angularity. Texture measurements were conducted on different sizes of the same aggregate type in order to investigate the influence of aggregate size on texture. Examples of results are shown in Figure 18. Aggregate size did not have a noticeable influence on texture. However, aggregate angularity changed with aggre- gate size. The analysis methods also captured the influence of crush- ing on shape or proportions of particle dimensions. The effect of aggregate size on sphericity varied from one aggregate to another. For example, the sphericity of the crushed river gravel was higher than for uncrushed gravel, indicating that aggregate crushing made the particles more equi-dimensional. How- ever, the crushed glacial gravel (CA-7) showed less sphericity than the uncrushed material (CA-8). Crushing the natural sand FA-1 to become FA-2 increased angularity, as depicted in Figure 15. FA-1 is an example of high quality natural sand that had angularity comparable to some manufactured sands. For example, FA-1 had higher angularity than crushed limestone (FA-6). Shown in Figure 19 is an example of the effect of size on fine aggregate angularity. Angularity increased as particle size decreased due to crushing. Identifying Flat, Elongated, or Flat and Elongated Particles The sphericity value gives a very good indication of the proportions of particle dimensions. However, one cannot determine whether an aggregate has flat, elongated, or flat and elongated particles using the sphericity alone. To this end, the chart shown in Figure 20 is included in the AIMS software to distinguish among flat, elongated, and flat and elongated par- ticles. Superimposed on this chart are the 3:1 and 5:1 limits for the longest to shortest dimension ratio and the results from CA-2 and CA-4. The figure shows that both aggregates pass 41 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% 2 5 61 10 Aggregate Label Pe rc en ta ge in A ng ul ar ity G ro up Rounded Sub-Rounded Sub-Angular Angular Figure 15. Distributions of fine aggregate angularity.

42 0 500 1000 1500 2000 2500 3000 3500 4000 0 100 200 300 400 500 600 Texture Index G ra di en t A ng u la rit y In de x CA-1 CA-2 CA-12 CA-8 CA-7 CA-4 CA-6 CA-11 CA-13 CA-5 CA-3 CA-10 CA-9 (a) Average Texture and Angularity of Coarse Aggregates 0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000 0 100 200 300 400 500 600 700 800 900 Texture Index G ra di en t A ng u la rit y In de x CA-2 CA-9 (b) Texture and Angularity of Coarse Aggregate Particles Figure 16. Variations in texture and angularity properties in coarse aggregates.

43 0 10 20 30 40 50 60 70 80 90 100 0 200 400 600 800 Texture Index Pe rc en ta ge o f P ar tic le s, % CA-1 CA-2 CA-3 CA-4 CA-5 CA-6 CA-7 CA-8 CA-9 CA-10 CA-11 CA-12 CA-13 High RoughnessModerateRoughness Low RoughnessSmoothPolished 0 10 20 30 40 50 60 70 80 90 100 0 200 400 600 800 Pe rc en ta ge o f P ar tic le s, % CA-1-3/4 CA-1-3/8 CA-1-#4 CA-7-3/4 CA-7-3/8 CA-7-#4 CA-10-3/4 CA-10-3/8 CA-10-#4 High Roughness Moderate Roughness Low RoughnessSmoothPolished Texture Index Figure 17. Texture index for different coarse aggregate types. Figure 18. Examples of the effect of coarse aggregate size on texture.

the 5:1 requirement (both had less than 10 percent particles with dimensional ratio of 5:1), but have distinct distributions in terms of flat and elongated particles. Such analysis reveals valuable information about the distribution that would not have been obtained if aggregates were classified based on the ratio of 5:1 only. This information will help to understand the influence of aggregate characteristics on asphalt and concrete mix properties. 44 0 10 20 30 40 50 60 70 80 90 100 0 1000 2000 3000 4000 5000 6000 7000 8000 Angularity "Gradient Method" Pe rc en ta ge o f P ar tic le s, % FA-2-#8 FA-2-#16 FA-2-#60 AngularSub- AngularRounded Sub- Rounded 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Short/Intermediate = Flatness Ratio In te rm ed ia te /L on g = E lo ng at io n Ra tio CA-2 CA-4 SP=0.5 SP=0.6 SP=0.7 SP=0.8 1 : 5 1 : 3 Ratio of shortest to longest axes Particles Become Less Flat Pa rti cl es B ec om e Le ss E lo ng at ed Figure 19. Example of the effect of fine aggregate size on angularity. Figure 20. Chart for identifying flat, elongated, or flat and elongated aggregates.