FINAL REPORT FOR NCHRP 20-7 (364)

Transcription

1 FINAL REPORT FOR NCHRP 20-7 (364) Revisions of AASHTO PP 74 Test Metod for Optical sizing and Sape Determination of Glass Beads Utilized in Traffic Marings August 2017 TRANSPORTATION RESEARCH BOARD NAS-NRC LIMITED USE DOCUMENT Tis proposal is for use of recipient in selection of a researc agency to conduct wor under te National Cooperative Higway Researc Program. After researc agency as been selected, te rejected proposals are to Pavement Systems, LLC 6105 Maiden Lane, Betesda, MD, Pone (202) i

2 TABLE OF CONTENTS LIST OF TABLES... iv LIST OF FIGURES... v Acnowledgements... 1 CHAPTER 1 INTRODUCTION AND RESEARCH APPROACH Introduction and Bacground Project Objective Project Scope...3 CHAPTER 2 DESIGN AND EXECUTION OF THE STUDY Tas 1 Questionnaire Tas 2 Instructions for Interlaboratory Study Tas 3 Selection of Materials Selection of Particle Analyzers Retsc Tecnology GmbH, Camsizer AnaTec/ Microtrac 3D PartAn Haver-Tyler CPA Sympatec Malvern Canty Tas 4 Selection of Laboratories Tas 5 Sample Preparation and Sipment Tas 6 Glass Bead Properties D10, D50, and D Cumulated Percent Retained Aspect Ratio, b/l or T/L Form Factor Parameter, SPHT Percent Round Percent Non-Round CHAPTER 3 RESEARCH RESULTS Analysis of Results of all Particle Analyzers D10, D50, and D Percent Retained Aspect Ratio Parameter (b/l Ratio) ii

3 3.1.4 Form Factor Parameter (SPHT) Percent Round based on b/l Percent Non-Round based on SPHT Comparison of Results from 2-D and 3-D Analysis Metods Statistics of Camsizer Results D10, D50, D Percent Retained Aspect Ratio Parameter (b/l) Form Factor Parameter (SPHT) Percent Round based on b/l Percent Non-Round based on SPHT Effect of Obscuration of Particles on Measurement Variability Precision Estimates of Glass Bead Properties Basis of Precision Estimates Comparison of te Precision Estimates based on Data from Camsizer and Various Devises Proposed Precision Estimates for AASHTO PP Revisions to te AASHTO PP CHAPTER 4 CONCLUSIONS AND RECOMMNEDATIONS Summary Findings Recommendations REFERENCES Appendix A- Survey Results and analysis Appendix B-ILS Instructions... 1 Appendix C- ILS Data... 4 Appendix D- Precison Estimates of AASHTO PP iii

4 LIST OF TABLES Table 2-1-List of participating laboratories, teir position, and type of teir particle analyzers. 10 Table 3-1- List of properties measured by eac laboratory Table 3-2- Statistics of D10, D50, and D90 parameters (all analyzers) Table 3-3- Statistics of percent retained (all analyzers) Table 3-4- Average and repeatability/ reproducibility statistics of b/l ratio (all analyzers) Table 3-5- Average sand repeatability/reproducibility statistics of SPHT parameters (all analyzers) Table 3-6- Average sand repeatability/reproducibility statistics of percent round based on b/l (all analyzers) Table 3-7- Averages and repeatability/reproducibility statistics of %non-round based on SPHT (all analyzers) Table 3-8-Summary of statistics of Q10, Q50, and Q90 data measured wit Camziser Table 3-9- Summary of statistics of Percent Retained ( 1-Q3) data measured wit Camziser Table Summary of statistics of aspect ratio ( b/l) data measured wit Camziser Table Summary of statistics of SPHT parameter measured wit Camziser Table 3-12-Summary of statistics of % round based on b/l parameter measured wit Camziser 64 Table Summary of statistics of % non-round beads based on SPHT parameter measured wit Camziser Table Precision estimates of te glass bead properties Table Proposed Precision Estimates for Percent Retained and Percent circular properties 75 iv

5 LIST OF FIGURES Figure 2-1 Te Camsizer Dynamic Image Analyzer from Horiba Scientific... 6 Figure 2-2 Te PartAn Dynamic Image Analyzer from Microtrac... 7 Figure 2-3 Te CPA Dynamic Image Analyzer from Haver... 7 Figure 2-4 Te QICPIC Dynamic Image Analyzer from Sympatec... 8 Figure 2-5 Te Mastersizer 3000 Particle Size Analysis Instrument from Malvern... 8 Figure 2-6 Te CANTY Dynamic Image Analyzer... 9 Figure 3-1- Median D10, D50, and D90 and teir associated error bars and and statistics and teir tresold values for Type 1 glass beads Figure 3-2- Median D10, D50, and D90 and teir associated error bars and and statistics and teir tresold values for Type 2 glass beads Figure 3-3-Median D10, D50, and D90 and teir associated error bars and and statistics and teir tresold values for Type 3 glass beads Figure 3-4- Median D10, D50, and D90 and teir associated error bars and and statistics and teir tresold values for Type 4 glass beads Figure 3-5- Type 1 glass beads % retained median and te associated error bars and te and statistics and teir tresold values Figure 3-6- Type 2 glass beads median % retained and te associated error bars and te and statistics and teir tresold values Figure 3-7- Type 3 glass beads median % retained and te associated error bars and te and statistics and teir tresold values Figure 3-8- Type 4 glass beads median % retained and te associated error bars and te and statistics and teir tresold values Figure 3-9- Median b/l ratio and te associated error bars and te and statistics and teir tresold values for Type 1 glass beads Figure Median b/l ratio and te associated error bars and te and statistics and teir tresold values for Type 2 glass beads Figure Median b/l ratio and te associated error bars and te and statistics and teir tresold values for Type 3 glass beads Figure Median b/l ratio and te associated error bars and te and statistics and teir tresold values for Type 4 glass beads Figure Median SPHT and te associated error bars and te and statistics and teir tresold values for Type 1 glass beads Figure Median SPHT and te error bars; te and statistics and teir tresold values for Type 2 glass beads v

6 Figure Median SPHT and te error bars; te and statistics and teir tresold values for Type 3 glass beads Figure Median SPHT and te associated error bars; te and statistics and teir tresold values for Type 4 glass beads Figure Median % round and te associated error bars and te and statistics and teir tresold values for Type 1 glass beads Figure Median % round and te error bars; te and statistics and teir tresold values for Type 2 glass beads Figure Median % round and te error bars; te and statistics and teir tresold values for Type 3 glass beads Figure Median % round and te associated error bars and te and statistics and teir tresold values for Type 4 glass beads Figure Median percent non-round and te associated error bars and te and statistics and teir tresold values for Type 1 glass beads Figure Median percent non-round plus te error bars and te and statistics and teir tresold values for Type 2 glass beads Figure Median percent non-round plus te error bars and te and statistics and teir tresold values for Type 3 glass beads Figure Median percent non-round plus te error bars and te and statistics and teir tresold values for Type 4 glass beads Figure Comparison of statistics of D10, D50, D90 data from PartAn 2D and 3D analysis 51 Figure Comparison of statistics of % Retained data from PartAn 2D and 3D analysis Figure Comparison of statistics of aspect ratio data from PartAn 2D and 3D analysis Figure Comparison of statistics of SPHT data from PartAn 2D and 3D analysis Figure Comparison of statistics of % round data from PartAn 2D and 3D analysis Figure Comparison of statistics of % non-round data from PartAn 2D and 3D analysis Figure Comparison of statistics of D10, D50, D90 data measured wit Camsizer and wit all brands of particle analyzers Figure Comparison of statistics of D10, D50, D90 data measured wit Camsizer and wit all brands of particle analyzers Figure Comparison of statistics of b/l data measured wit Camsizer and wit all brands of particle analyzers Figure Comparison of statistics of SPHT data measured wit Camsizer and wit all brands of particle analyzers Figure Comparison of statistics of percent round based on b/l measured by Camsizer and by all brands of particle analyzers vi

7 Figure Comparison of statistics of percent non-round measured wit all brands of particle analyzers Figure Relationsip between average and standard deviation of te glass bead properties measured wit Camsizer particle analyzer Figure Relationsip between average and standard deviation of te glass bead properties measured wit various brands of particle analyzers vii

8 ACKNOWLEDGEMENTS Te researc reported erein was performed under NCHRP Project 20-07(364) by Pavement Systems LLC. Dr. Hale Azari was te principal investigator on te study. Potters industry and SWARCO REFLEX, LLC supplied te glass beads used in te interlaboratory study. Teir elp is very muc appreciated. Gilson company graciously provided elp wit equipment, facility, and personnel for preparing samples of glass beads for te interlaboratory study. Teir contribution to tis study is deeply regarded. Te autor wises to acnowledge te laboratories wo participated in tis study by providing responses to te survey, offering comments to improve AASHTO PP74, and taing part in te interlaboratory study. Teir willingness to volunteer teir time and resources is greatly recognized. 1

9 CHAPTER 1 INTRODUCTION AND RESEARCH APPROACH 1.1 Introduction and Bacground Te nigttime visibility of traffic marings is determined by te initial retro-reflectivity level of te glass beads in pavement marings, wic is greatly dependent on te size distribution and roundness of glass particles applied on top of te maring material. Use of iger percentages of perfectly round glass particles tremendously improves reflectivity of te pavement marings. Similarly, use of well graded glass particles, not only increases te initial reflectivity but also extends te useful reflectivity life of te pavement maring. Tis necessitates, rapid and reliable screening of te glass particles for acceptable level of roundness and size distribution according to a standard test metod before tey are applied to te road. Recently, use of optical computerized dynamic particle analyzers provides significant improvements in te measurement procedure of te glass particles over te traditional measurements of roundness wit round-o-meter (ASTM D 1155) and size distribution by sieve analysis (ASTM D 1214). Te computerized procedure decreases te subjectivity, wic in turn increases te reliability of te measuring process. It also increases te productivity of screening by reducing te measurement time, wic is essential during te construction season. In 2010, AMRL in coordination wit te AASHTO Subcommittee on Materials Tecnical Section 4c, State DOTs, as well as glass bead and equipment manufacturing communities completed a National Cooperative Higway Researc (NCHRP) Study [1], wic resulted in a standard test metod for use wit computerized, optical particle analyzers. Te developed standard, referred to as AASHTO PP 74, includes specifications tat apply to several types of analyzers for sizing and roundness measurements of a wide range of glass beads. Te test metod also provides a complete set of precision estimates tat allows laboratories to determine te precision of teir measurements troug comparison wit te provided precision criteria. Since te completion of te study, additional manufacturers ave offered automated, ig resolution computerized optical sizing macines wit added features for improved accuracy and precision for te measurements of sape and size distribution of glass beads. Te use of new equipment requires tat AASHTO PP 74 test metod to be revised so tat it would not be limited to te previously evaluated macines. It is also required tat te accuracy and precision of te test metod using measurements wit te new equipment be re-evaluated. For tat reason, Tecnical Section 4c of te Subcommittee on Materials (SOM) as requested conduct of tis study for te modification of AASHTO PP 74 for optical sizing and roundness determination of glass beads utilized in traffic marings. Tis will ensure eeping te AASHTO PP 74 standard updated so tat it will be bot utilized and propagated, wit a furterance tat it will act as a guide for te development of oter AASHTO standards utilizing similar tecnology. Te tass of tis study will include coordination wit te venders of te macines and wit te users and producers of glass beads to initiate an interlaboratory study for updating te precision estimates of te standard practice and to cange te language of te AASHTO PP 74 to include a more variety of computerized optical equipment. Tis requires use of equipment from multiple manufacturers for analysis of several types of glass beads from multiple glass bead suppliers. As a start, tis study will collect input from te practitioners and researcers for te applicability of AASHTO PP 74 standard for teir routine screening of pavement maring glass beads. A 2

10 review of te devices will be conducted to identify te common features and if tere are any major differences between te computerized optical equipment in te maret. As part of an interlaboratory study (ILS), replicate samples of various types of glass beads will be prepared and sent to te laboratories equipped wit computerized optical equipment to compare accuracy and precision of measurements from various devices. Based on te input received from te users and producers of glass beads and te equipment manufacturers and based on te analysis of te ILS data, AASHTO PP 74 will be modified. Te implication of tis study is to maintain uniformity in te measurement of glass bead properties. 1.2 Project Objective Te objectives of tis study are to modify AASHTO PP 74 standard practice to be applicable to optical computerized particle analyzers entered into te maret since te development of te standard in 2011 and to update te precision estimates of te practice by including te measurements made wit te new equipment. 1.3 Project Scope Te scope of te project involves te following major activities: Identify manufacturers of computerized optical particle analyzers, producers of glass beads, and state and private laboratories, wic use optical analyzers in teir routine screening of glass beads for pavement marings. Send a questionnaire to te users and producers in te glass bead industry to obtain teir comments on applicability of AASHTO PP 74 for use wit teir computerized optical equipment. Te willingness of te agencies to participate in an ILS to update te precision estimates of te test will be inquired. Prepare instructions for te ILS based on AASHTO PP 74 and te comments received from te users and producers of glass beads and computerized optical equipment. Obtain Types 1, 2, 3, and 4 glass beads (as described in AASHTO M 247) from two different glass bead suppliers for use in te ILS. For eac source and type combination, prepare representative replicates of glass bead samples to be sent to te participating laboratories. Te mass of te samples is determined based on te type of te glass beads. Sip tree replicates of te four materials along wit te instructions and data seets to te participating laboratories for sape and size measurements. Tis would be a total of 12 samples to eac laboratory. Perform statistical analysis on te sape and size data received from te laboratories to determine precision of te measurements. Update te precision statement of AASHTO PP 74 using te new set of precisions. Modify AASHTO PP 74 to incorporate te relevant comments from te users and producers of glass beads and te computerized optical equipment to extent te applicability of te test metod to te new equipment in te maret. 3

11 CHAPTER 2 DESIGN AND EXECUTION OF THE STUDY 2.1 Tas 1 Questionnaire Te manufacturers of computerized optical particle analyzer, glass bead suppliers, and state and private laboratories equipped wit te particle analyzers were identified. A questionnaire was prepared and sent out to te corresponding agencies to inquire teir input on te applicability of AASHTO PP 74 for te routine use of teir equipment. In te questionnaire, te willingness of eac agency to participate in an ILS for evaluation of te precision of measurements wit optical analyzers was inquired. Te responses of 30 agencies to te survey are provided in Appendix A. A summary of responses to some of te survey questions are as follows: Wat are te most common types of glass beads you test or certify routinely for use wit pavement marings? Type 1 is te most common type bead wit 21 votes and Type 3, Type 4, and Type 2 are te second, tird, and fourt in te ran wit 9, 8, and 7 votes, respectively. Based on tese results, Type 1, 2, 3, and 4 were obtained for te interlaboratory study. Do you use computerized optical equipment for testing gradation and roundness of glass beads? Out of 31 responses, tere were 11 Yes and 20 No. Wat practice do you follow for determining gradation and roundness of glass beads using computerized optical equipment? Please list te standard designation or provide a copy of te practice if not a publised standard. Out of 11 responses, 5 states use PP 74, one state (OK) uses M 247, one state (OH) uses Retsc Tecnology (Camsizer) manual, one state (KS) uses ASTM D 1214/ASTM D 1155, wic are manual metods not computerized, one state (TX) uses TXDOT specification, and Potters Industry uses T Have you followed AASHTO PP 74, Determination of Size and Roundness of Glass Beads Used in Traffic Marings by Means of Computerized Optical Metod, for routine evaluation of glass beads? 7 out of 11 states responded Yes. Has AASHTO PP 74 been applicable to your equipment? 7 out of 7 states responded Yes to tis question. Have you encountered any sortcomings wit te use of AASHTO PP 74 for your equipment? Please explain. Out of 7 responses, 6 answered No and one state (MN) provided a comment tat Section defines b as Xcmin but defines b/l as XFemin / XFemax. Correction as been incorporated to te standard practice to clarify. Please state if any section of AASHTO PP 74 is too specific and not relevant to your computerized optical particle analyzer. 4

12 All of te 6 responses were None. FL commented tat PP 74 is not too specific and wors well for tem. MN responded tat te practice does a good job in defining wic parameters apply to te different instruments. Please state if any section of PP 74 is too general and not providing important specifics for your computerized optical particle analyzer. All of te 5 responses were None. Please provide te manufacturer and model of your computerized optical particle analyzer. All 11 responses were Camsizer by Retsc. Please feel free to mae any comments or suggest any canges to PP 74 to improve its applicability to your equipment. Tree comments were provided: MO stated tat tey lie te metod; FL commented tat It would be nice if te metod provided information on measuring beads wit a ig refractive index; and Potters stated tat tey ave found troug extensive correlations, tat b/l and xcmin are te best metods for testing glass beads. Tis agrees well wit te finding of te NCHRP 20-07(243) study, were it was sown tat b/l and xcmin were more reliable measures of roundness and size tan te oter parameters studied. Bot b/l and xcmin ave been already incorporated in PP 74. Are you willing to participate in NCHRP (364) interlaboratory study to test samples of glass beads wit your computerized optical particle analyzer? Out of 12 responses, 10 states indicated tat tey are willing to participate in te interlaboratory study. In addition to sending te questioner, several agencies were individually contacted and teir comments and participation in te ILS were inquired. Te agencies included Penn DOT, Microtrac, W.S. Tyler, Sympatec, Canty, and Malvern. Comments were received from Penn DOT, W.S. Tyler, Microtrac, and Canty. Discussion of te comments and te revisions to PP74 based on te provided comments are provided in Section Tas 2 Instructions for Interlaboratory Study Based on AASHTO PP 74, instructions for measuring sape and size properties of glass beads using computerized optical particle analyzer were prepared. Appendix B provides te ILS instructions. 2.3 Tas 3 Selection of Materials Based on te responses to te first question of te survey, four of te most common types of glass bead were obtained from two suppliers of glass beads. Type 1 and 2 were provided by Potters industry and Types 3 and Type 4 were provided by SWARCO REFLEX, LLC. Te glass beads were supplied in 50-lbs bags. 5

13 2.4 Selection of Particle Analyzers Six types of particle analyzers were selected to measure te glass bead properties. Te equipment included Camsizer, PartAn 3 D, Haver-Tyler CPA, Canty analyzer, Malvern, and SympaTec analyzer. All particle analyzers, except Malvern, are operating based on flowing particles in front of one or more cameras. Te Malvern particle analyzer is equipped wit magnifying lenses and automatic measurements are performed under a microscope. A description of eac system is as follows: Retsc Tecnology GmbH, Camsizer Horiba's Camsizer Digital Image Processing System [2] analyze size and sape of glass beads in te size range from 20 microns to 30 millimeters. Te Camsizer laboratory instrument uses two cameras for capturing te size and sape of te free flowing glass particles. All te performed measurements are automated trougout te test. Figure 2-1 sows te Camsizer particle analyzer. Figure 2-1 Te Camsizer Dynamic Image Analyzer from Horiba Scientific AnaTec/ Microtrac 3D PartAn Te PartAn, (formally nown as te DIA) is a Dynamic Image Analyzer from Microtrac [3] tat can measure 25 morpological parameters of a material ranging in particle size from 20 microns to 35 mm. PartAn software uses te 3D measurement (from multiple 2D images) and te special algoritm to determine size, sape, and transparency of glass beads. Figure 2-2 sows te PartAn particle analyzer. 6

14 Figure 2-2 Te PartAn Dynamic Image Analyzer from Microtrac Haver-Tyler CPA Te HAVER TYLER CPA 2-1 [4] counts and measures properties, in real time, of individual glass bead particles in te range from 10 microns to 400 mm. Te resulting distribution curve is based on te actual number of particles measured witin a given size range. Te CPA conveys dry, free-flowing material samples to be examined by a camera lens, wic is maing approximately 28,000 scans per second. Figure 2-3 sows te CPA particle analyzer. Figure 2-3 Te CPA Dynamic Image Analyzer from Haver Sympatec System-Partiel-Tecni (SympaTec) [5] is manufacturer of QICPIC, a dynamic image analysis instrument for analysis of size and sape distribution and as well as te caracterization of te sape of individual small particle in a range of 1 microns to 30 mm. Figure 2-4 sows te QICPIC particle analyzer. 7

15 2.4.5 Malvern Figure 2-4 Te QICPIC Dynamic Image Analyzer from Sympatec Te Malvern Mastersizer 3000 laser diffraction particle size analyzer [7] performs operatorindependent measurements to determine particle size and sape distributions over te range of 0.1 to 3500 nanometer. Figure 2-5 sows te Mastersizer particle analyzer. Malvern particle analyzer does not operate based on te same principals as te flowing stream particle analyzers Canty Figure 2-5 Te Mastersizer 3000 Particle Size Analysis Instrument from Malvern Te Canty Process Tecnology [7] determines particle size and sape in te range of 10 microns to 2,400 microns. Figure 2-6 sows te Canty particle analyzer. 8

16 Figure 2-6 Te CANTY Dynamic Image Analyzer 2.5 Tas 4 Selection of Laboratories Twenty-two laboratories were selected to participate in te interlaboratory study. Te selection of laboratories was based on diversity of te types of particle analyzers to be used for measuring te properties of te beads. Twenty-two laboratories agreed to participate in te ILS. Among tem, tere were 9 state DOTs, 5 glass bead/pavement maring suppliers, 7 particle analyzer manufacturers, and a consulting company. A list of participating laboratories is sown in Table 2-1. DOT laboratories and te glass beads and pavement maring producers/suppliers were equipped wit Camsizer particle analyzer (# 1 troug #14 in Table 2-1). Te measurements using oter types of particle analyzers were made by eiter te manufacturers of te equipment or teir distributors. Additionally, one set of measurement was made by Pavement Systems consulting using CPA particle analyzer (#18 in te table), wic was provided as a loan to Pavement Systems by W.S. Tyler for a limited time. 9

17 Table 2-1-List of participating laboratories, teir position, and type of teir particle analyzers Laboratory Status Particle Analyzer Type 1-Illinois User/DOT Retsc Tecnology GmbH, Camsizer 2-Montana User/DOT Retsc Tecnology GmbH, Camsizer 3-Kansas User/DOT Retsc Tecnology GmbH, Camsizer 4-Missouri User/DOT Retsc Tecnology GmbH, Camsizer 5-PA User/DOT Retsc Tecnology GmbH, Camsizer 6-OK User/DOT Retsc Tecnology GmbH, Camsizer 7-Florida User/DOT Retsc Tecnology GmbH, Camsizer 8-Oio User/DOT Retsc Tecnology GmbH, Camsizer 9-Mn User/DOT Retsc Tecnology GmbH, Camsizer 10- Ennis-Flint laboratories (Palestine) 11- Ennis-Flint laboratories (Hillcrest) Pavement Maring Supplier Pavement Maring Supplier Retsc Tecnology GmbH, Camsizer Retsc Tecnology GmbH, Camsizer 12- Ennis-Flint laboratories (Tomasville) Pavement Maring Supplier Retsc Tecnology GmbH, Camsizer 13-Future Glass Bead Supplier Retsc Tecnology GmbH, Camsizer 14-Potters Glass Bead Suppliers Retsc Tecnology GmbH, Camsizer 15-Canty Particle Analyzer Manufacturer LabPlus SolidSizer 16-Malvern Particle Analyzer Manufacturer Mastersizer /Insitec using Laser Diffraction Tecnology 17-Sympatec Particle Analyzer Manufacturer Qicpic Particle analyzer 18-Pavement Systems Consulting Computerized Particle Analysis (CPA) 19-Haver & Boecer (H&B) Particle Analyzer Manufacturer Computerized Particle Analysis (CPA) 20-W.S. Tyler Particle Analyzer Manufacturer Computerized Particle Analysis (CPA) 21-Microtrac (Norway) Particle Analyzer Manufacturer PartAn 2D & 3D 22-Microtrac (Yor) Particle Analyzer Manufacturer PartAn 2D & 3D 10

18 2.6 Tas 5 Sample Preparation and Sipment Te samples for te interlaboratory study were prepared at te Gilson facility in Oio. Gilson provided 4 types of splitters to split eac 50 lbs. of glass beads down into 75-g to 150-g replicate samples. Te bags of glass beads were split following te practices outlined in ASTM B 215. From eac type of glass beads, 72 representative replicate samples were prepared. Te masses of te samples were determined based on te type of glass beads as specified in AASHTO PP 74. Tree replicates of te four types of glass beads, total of 12 samples, were prepared and sent to eac participating laboratory. 2.7 Tas 6 Glass Bead Properties A list of properties to be measured wit te particle analyzers are included in te instruction seet of te ILS (Appendix A). Laboratories were requested to measure as many of te properties as teir equipment are programmed to perform. A description of te main properties are as follows: D10, D50, and D90 Te D10, D50, and D90 decile measurements are te diameter at wic, 10%, 50%, and 90% of a sample is finer in mass. Te notations for D10, D50, and D90 are as follows: D10 (particle size at Q3 = 10% (passing)) D50 (particle size at Q3 = 50% (passing)) D90 (particle size at Q3 = 90% (passing)) Cumulated Percent Retained Percent passing (Q3) and percent retained (1-Q3) determine te size distribution of te glass beads. Since te two properties are calculated based on eac oter, te laboratory were ased to provide eiter one or bot properties Aspect Ratio, b/l or T/L Te Aspect ratio parameters, b/l or T/L, are measures of sape of particles in two or tree dimensions (2-D or 3-D), respectively. Tey are te ratio of a particle s smallest dimension, wic is breat (b) in 2-D and ticness (T) in 3-D to a particle s largest dimension, wic is lengt (l or L in 2-D or 3-D). For a perfect circle/spere, b/l and T/L are equal 1. Any deviation from perfect circle/spere would result in b/l or T/L of less tan 1. A tresold of 0.85 is used for b/l and tresold of 0.83 is used for T/L. Tese values are selected based on te correlation between measurements performed wit x-ray tomograpy and manual roundometer measurements performed as part of te NCHRP study (343) [1]. Any particle wit b/l (T/L) of 0.83 (0.85) and above is considered circular or sperical Form Factor Parameter, SPHT Te SPHT also nown as Form Factor (FF), as described in ISO 9276 [8], is anoter parameter for measuring ow close a particle is to a circle. Tis is te ratio of particle area (multiplied by a coefficient) to te squaring of circumference: SPHT= 4 A/U 2, were A is te area and U is te 11

19 circumference of a projection of a particle. For a perfect circle, SPHT is equal 1. Any deviation from a circle would result in SPHT of less tan 1. A tresold of 0.93 is used for SPHT based on te correlation between roundometer and x-ray tomograpy measurements [1]. Any particle wit SPHT of 0.93 or above is considered to be circular Percent Round Te percent circular/sperical in tis study as been determined based on b/l and T/L parameters. Te percent of particles aving b/l (T/L) greater tan 0.85 (0.83) was considered circular or sperical, respectively. Percent circular/sperical ave been requested to be reported for eac class size and as a weigted average of all sieve sizes for eac glass bead type Percent Non-Round Te percent of non-circular can be determined based on SPHT parameter. Te percent of particles wit SPHT value smaller tan 0.93 are considered percent of non-circular. Similar to percent circular, tis property can be determined on eac individual sieve size or as a weigted average of all sieve sizes for eac bead type. 12

20 CHAPTER 3 RESEARCH RESULTS Twenty-four sets of data on size and sape of te four types glass beads were received from 22 participating laboratories. Te number of sets measured by eac equipment is as follows: 14 sets by Camsizer particle analyzer, tree sets by Haver-Tyler CPA, one set by Canty, one set by Malvern, and one set by Sympatec. Eac of te two Microtrac laboratories provided two sets of data, one set of 2-D measurements and anoter set of 3-D measurements. All data sets were obtained using a flowing stream digital image analyzers, except data from Malvern, wic was measured using laser diffraction tecnology. Only te data obtained wit flowing stream digital image analyzers were used for developing precision estimates. Two sets of precision estimates were prepared. One set was prepared based on te data measured by all flowing stream digital image analyzers. Te second set of precision estimates was prepared based on te analysis of Camsizer data. Tis served two purposes: 1) to compare ow te data measured wit oter types of particle analyzers are compared wit tose measured wit Camsizer, wic is being prevailingly used in te pavement maring industry today and 2) ow te variability of size and sape measurements wit Camsizer ave canged since te development of precision estimates of AASHTO PP 74, wic were prepared based on Camsizer data. For developing te precision estimates, only tose properties tat were measured by 5 or more laboratories were used. Six main size and sape properties of te four glass bead types were statistically analyzed to assess te variability of te measurements. A list of properties and te number of laboratories/equipment tat provided data on eac property are sown in Table Analysis of Results of all Particle Analyzers Te statistical analysis of te glass bead properties involves determining repeatability and reproducibility of te measurements following ASTM E 691 [10]. Te outlier data were eliminated by calculating te and statistics outlined in ASTM E 691. Te statistic represents between-laboratory variability and te statistic represents te witin-laboratory variability. Any data set wit and values exceeding te tresold values are considered outliers and are eliminated from te variability analysis. Te following sections provide te results of analysis of te size and sape measurements collected by different types of flowing stream particle analyzers D10, D50, and D90 For te four types of glass beads, between 21 to 24 sets of data were submitted for D10, D50, and D90 parameters. Te data are provided in Appendix C. Figure 3-1 troug Figure 3-4 provide grapical representations of te median D10, D50, and D90 values of te four types of glass beads along wit teir associated error bars. Te statistical parameters and and teir tresold values for identifying te outlier data are also sown in te figures. Te red bars sow te tresold and values corresponding to te number of data sets analyzed. Te observations from figure regarding te variability of te data are as follows: 13

21 Table 3-1- List of properties measured by eac laboratory Lab D10, D50, D90 % Passing/ Retained b/l per class size b/l Overall SPHT Per class size SPHT, Overall % Circular per class size (based on b/l) Overall % Circular (based on b/l) % Non- Circular Per class size (based on (SPHT) % Overall Non- Circular (based on (SPHT) 1-Illinois - 2-Montana 3-Kansas Missouri PA - 6-OK Florida Oio Mn 10- Ennis-Flint (Palestine) 11- Ennis-Flint (Hillcrest) 12- Ennis-Flint (Tomasville) Future Lab Potters Canty Malvern Sympatec Pavement Systems 20-Haver & Boecer(H&B) W.S. Tyler Microtrac (Norway)-2D 22-Microtrac (Yor)-2D 23-Microtrac (Norway)-3D 24-Microtrac (Yor)-3D

22 Q90, mm Q50, mm Q10, mm Figure 3-1 sows tat for Type 1 glass beads, in terms of between-laboratory variability, te - value of data sets #23 and #15 (corresponding to D10 and D50) exceed te tresold -value. In terms of repeatability, -values of data set #16, #22 and #24 (corresponding to D10, D30, and D50), exceeds te tresold -values. After removing te above sets as te outliers, 20 sets of data are remaining for furter evaluation. Type 1- D10, Median Type 1- D10, Values Type 1- D10, Values critical Type 1- D50, Median Type 1- D50, Values Type 1- D50, Values critical Type 1- D90, Median Type 1- D90, Values Type 1- D90, Values critical Figure 3-1- Median D10, D50, and D90 and teir associated error bars and and statistics and teir tresold values for Type 1 glass beads 15

23 Q90, mm Q50, mm Q10, mm Figure 3-2 sows tat for Type 2 glass beads, in terms of between-laboratory variability, te - value of data set #15 (corresponding to D10) exceeds te tresold value. In terms of singleoperator variability, te -values of data set #16 (corresponding to D10, D50, and D90) exceed te tresold -value. After removing te above data sets as te outliers, 21 sets of data are remaining for furter evaluation. Type 2- D10, Median Type 2- D10, Values Type 2- D10, Values critical Type 2- D50, Median Type 2- D50, Values Type 2- D50, Values critical Type 2- D90, Median Type 2- D90, Values Type 2-D90, Values critical Figure 3-2- Median D10, D50, and D90 and teir associated error bars and and statistics and teir tresold values for Type 2 glass beads 16

24 Q90, mm Q50, mm Q10, mm Figure 3-3 Sows tat for Type 3 glass beads, in terms of between-laboratory variability te - value of data sets #22 and #19 (corresponding to D50 and D90) exceed te critical -values. In terms of repeatability, te -values of data set #15, # 16, and #22 (corresponding to D10, D50, and D90) exceed te tresold values. After removing te above data sets as te outliers, 20 sets of data are remaining for furter evaluation. Type 3- D10, Median Type 3- D10, Values Type 3- D10, Values critical Type 3- D50, Median Type 3- D50, Values Type 3- D50, Values critical Type 3- D90, Median Type 3- D90, Values Type 3- D90, Values critical Figure 3-3-Median D10, D50, and D90 and teir associated error bars and and statistics and teir tresold values for Type 3 glass beads Error! Reference source not found.figure 3-4 sows tat for Type 4 glass beads, in terms of between-laboratory variability te value of data set # 19 (corresponding to D90) exceeds te 17

25 Q90, mm Q50, mm Q10, mm tresold values. In terms of repeatability, te values of data set #16 and #19 (corresponding to D10, D50, and D90) exceed te tresold values. After removing te above data sets as te outliers, 22 sets of data are remaining for furter evaluation. Type 4- D10, Median Type 4- D10, Values Type 4- D10, Values critical Type 4- D50, Median Type 4- D50, Values Type 4- D50, Values critical Type 4- D90, Median Type 4- D90, Values Type 4- D90, Values critical Figure 3-4- Median D10, D50, and D90 and teir associated error bars and and statistics and teir tresold values for Type 4 glass beads Te results of statistical analysis of D10, D50, and D90 data after removing te outlier data are provided in Table 3-2. As indicated from te table, te average of te D10, D50, and D90 are witin te range of decile values for te four types of glass beads. It is also indicated tat te repeatability and reproducibility coefficients of variation are smaller for te larger glass beads (Types 3 & 4) tan for te smaller glass beads (Types 1 & 2). Tis migt indicate tat particle analyzers are more capable of measuring size of te larger glass beads. 18

26 Table 3-2- Statistics of D10, D50, and D90 parameters (all analyzers) Type Property # of Labs Expected Diameter, Range (mm) Measured Diameter, Average (mm) Repeat. STD, S r Repeat. CV% Reprod. STD, S R Reprod. CV% D to D to D to D to D to D to D to D to D to D to D to D to Percent Retained Twenty-four sets of percent passing/retained data for glass beads of Types 1 and 4, twenty-tree sets on Type 3 glass beads, and twenty sets of data on Type 2 glass beads were submitted. For eac glass bead type, te data tat were significantly different from te rest of te data were not included in te analysis. Figure 3-5 troug Figure 3-8 provide grapical representation of te median percent retained values of te four types of glass beads along wit teir associated error bars. Te statistical parameters and and te tresold values for identifying te outlier data are also sown in te figures. Te red bars sow te tresold and values corresponding to te number of data sets analyzed. Te observations from te figures regarding te variability of te data are as follows: 19

27 Retained, % Retained, % Retained, % Figure 3-5 sows tat for Type 1 glass beads, in terms of between-laboratory variability, te - value of data set #23 (corresponding to all sizes) exceeds te tresold value. In terms of singleoperator variability, te -values of data sets #22, #24, #21(corresponding to various class sizes) exceed te tresold -value. After removing te above data sets as te outliers, 18 sets of percent retained data pertaining to Type 1 glass beads are remaining for furter evaluation. Type 1- % Retained #30, Median Type 1- % Retained #30, Values Type 1- % Retained #30, Values Type 1- % Retained #50, Median Type 1- % Retained #50, Values Type 1- % Retained #50, Values critical Type 1- % Retained #100, Median Type 1- % Retained # 100, Values Type 1- % Retained #100, Values critical Figure 3-5- Type 1 glass beads % retained median and te associated error bars and te and statistics and teir tresold values 20

28 % Retained % Retained % Retained % Retained Figure 3-6 sows tat for Type 2 glass beads, in terms of between-laboratory variability, te - value of data set #23 (corresponding to #25 glass beads) exceeds te tresold value. In terms of single-operator variability, te -values data sets #22 and #24 (corresponding to #50 and #100 beads) exceed te tresold -value. After removing te above data sets as te outliers, 15 sets of percent retained data pertaining to Type 2 glass beads are remaining for furter evaluation. Type 2- % Retained # 25, Median Type 2- % Retained #25, Values Type 2- % Retained #25, Values Type 2- % Retained #35, Median Type 2- %Retained, #35, Values Type 2- % Retained #35, Values critical Type 2- % Retained #50, Median Type 2- % Retained #50, Values Type 2- % Retained #50, Values critical Type 2- % Retained #100, Median Type 2- % Retained #100, Values Type 2- % Retained #100, Values critical Figure 3-6- Type 2 glass beads median % retained and te associated error bars and te and statistics and teir tresold values 21

29 Retained, % Retained, % Retained, % Retained, % Figure 3-7 sows tat for Type 3 glass beads, in terms of between-laboratory variability, te - values of data sets #22 and # 23 (corresponding to various size classes) exceed te tresold value. In terms of single-operator variability, te -values of data sets #4 and #23 (corresponding to #18 and #25 class sizes) exceed te tresold -value. After removing te above data sets as te outliers, 20 sets of percent retained data pertaining to Type 3 glass beads are remaining for furter evaluation. Type 3- %Retained #16, Median Type 3- %Retained #16, Values Type 3- %Retained #16, Values critical Type 3- %Retained #18, Median Type 3- %Retained #18, Values Type 3- %Retained #18, Values critical Type 3- %Retained #20, Median Type 3- %Retained #20, Values Type 3- %Retained #20, Values critical Type 3- %Retained #25, Median Type 3- %Retained #25, Values Type 3- Retained #25, Values critical Figure 3-7- Type 3 glass beads median % retained and te associated error bars and te and statistics and teir tresold values 22

30 Retained, % Retained, % Retained, % Retained, % Figure 3-8 sows tat for Type 4 glass beads, in terms of between-laboratory variability, te - value of data set #23 (corresponding to #20 beads) exceeds te tresold value. In terms of single-operator variability, te -values of data set #10 (corresponding to #16 beads) exceeds te tresold -value. After removing te above data sets as te outliers, 19 sets of percent retained data pertaining to Type 4 glass beads are remaining for furter evaluation. Type 4- %Retained #14, Median Type 4- %Retained #14, Values Type 4- %Retained # 14, Values critical Type 4- %Retained #16, Median Type 4- %Retained #16, Values Type 4- %Retained #16, Values critical Type 4- %Retained #18, Median Type 4- %Retained #18, Values Type 4- %Retained #18, Values critical Type 4- %Retained #20, Median Type 4- Retained #20, Values Type 4- Retained #20, Values critical Figure 3-8- Type 4 glass beads median % retained and te associated error bars and te and statistics and teir tresold values Te results of statistical analysis of percent retained after removing te outlier data are provided in Table 3-3. Tree observations can be made from te table. First, te average of te percent retained are witin te acceptable range of percent retained values for te four glass bead types. Second, a lower coefficient of variation corresponds to te sieve sizes wit iger percentage of beads. Tird, te variability of measurements is smaller for te larger bead sizes (Types 3 and 4). 23

31 Table 3-3- Statistics of percent retained (all analyzers) Type Property # of Labs Expected Retained, Range (%) Measured Retained, Average (%) Repeatability STD, Sr, % Repeatability CV% Reproducibility STD, SR, % Reproducibility CV% # to # to # to # to # to # to # to # to # to # to # to # to # to # to # to Aspect Ratio Parameter (b/l Ratio) Te median b/l values, te associated error bars, and te and statistics for te four types of glass beads are sown in Figure 3-9 troug Figure Te resulted statistics of te b/l ratio after removing te outlier data are provided in Table 3-4. Te observations from te table and figures are as follows: 1. For Type 1 glass beads, Figure 3-9 sows tat in terms of between-laboratory variability, none of te -values exceed te tresold -value. In terms of single-operator variability, te -values of data sets #21, #7, and #24 (corresponding to #50 and #100 beads) exceed te tresold -value. After removing data sets #21, #7, and #24 from te analysis, te resulted statistics in Table 3-4 indicate tat te average overall b/l ratio is 0.88 and it ranges between 0.87 to 0.91 for te individual sieve sizes. Te overall b/l repeatability coefficient of variation is 0.3%, ranging from 0.1 % to 0.4 % for te individual sieve sizes and te overall reproducibility coefficient of variation is 9.3%, ranging from 5.1% to 11.3% for te individual sieve sizes. 2. For Type 2 glass beads, Figure 3-10 sows tat in terms of reproducibility, none of te values exceeds te tresold values. In terms of repeatability, te values of data sets #5, #16, #21, #14, and #24 corresponding to various class sizes exceed te tresold values. After removing te above data sets from te analysis, te resulted statistics in Table 3-4 indicates tat te overall b/l ratio is 0.88 and for te individual sieve sizes te range is 0.85 to Te overall b/l repeatability coefficient of variation is 0.2%, 24

32 ranging from 0.2% to 0.7% for te individual sieve sizes. Te overall reproducibility coefficient of variation is 6.6%, ranging from 3.9% to 12.6% for te individual sieve sizes. 3. For Type 3 glass beads, Figure 3-11 sows tat in terms of reproducibility, none of te values exceeds te tresold values. In terms of repeatability, te values for data sets #21, #24, #7, and #19 corresponding to various class sizes exceed te tresold values. After removing te above data sets from te analysis, te resulted statistics in Table 3-4 indicate tat te overall b/l ratio is 0.94 and for te individual sieve sizes te range is 0.90 to Te overall b/l repeatability coefficient of variation is 0.1% and ranges between 0.1 % to 0.7 % for te individual sieve sizes. Te overall reproducibility coefficient of variation is 2.9 % and ranges between 2.0% to 5.4% for te individual sieve sizes. 4. For Type 4 glass beads, Figure 3-12 sows tat in terms of reproducibility, te overall value for Data Set #6 (corresponding to #12 beads) exceeds te tresold value. In terms of repeatability, values for Sets #23, #21, #24, #13, #15 exceed te tresold values corresponding to various class sizes. After removing te above data sets, te resulted statistics in Table 3-4 indicate tat te overall b/l ratio is 0.94 and it ranges between 0.89 to 0.94 for te individual sieve sizes. Te overall b/l repeatability coefficient of variation is 0.1% and ranges between 0.1 % to 0.4% for te individual sieve sizes. Te overall reproducibility coefficient of variation is 2.3 %, ranging from 2.8% to 5.0% for te individual sieve sizes. Several observations can be made from Table 3-4. First, te average b/l ratio for all class sizes are larger tan te tresold b/l value of 0.85 (specified in AASHTO PP74). Second, te b/l of te smaller glass beads (Types 1 and 2) are smaller tan tose of te larger glass beads (Types 3 and 4). Tird, bot single operator and multilaboratory variability of b/l measurements are larger for te smaller glass beads tan for te larger glass beads. Tis could indicate better capability of te particle analyzers in measuring properties of larger sized glass beads. 25

33 b/l b/l b/l b/l Type 1- b/l #30, Median Type 1- b/l #30, Values Type 1- b/l #30, Values critical Type 1- b/l #50, Median Type 1- b/l #50, Values Type 1- b/l #50, Values critical Type 1- b/l #100, Median Type 1- b/l # 100, Values Type 1- b/l #100, Values critical Type 1- Overall b/l, Median Type 1- Overall b/l, Values Type1- Overall b/l, Values critical Figure 3-9- Median b/l ratio and te associated error bars and te and statistics and teir tresold values for Type 1 glass beads 26

34 b/l b/l b/l b/l b/l Type 2- b/l #25, Median Type 2- b/l #25, Values Type 2- b/l #25, Values critical Type 2- b/l #35, Median Type 2- b/l #35, Values Type 2- b/l #35, Values critical Type 2- b/l #50, Median Type 2- b/l #50, Values Type 2- b/l #50, Values critical Type 2- b/l #100, Median Type 2- b/l #100, Values Type 2- b/l #100, Values critical Type 2- Overall b/l, Median Type 2- Overall b/l, Values Type 2- Overall b/l, Values critical Figure Median b/l ratio and te associated error bars and te and statistics and teir tresold values for Type 2 glass beads 27

35 b/l b/l b/l b/l b/l Type 3- b/l #14, Median Type 3- b/l #14, Values Type 3- b/l #14, Values 1.2 critical Type 3- b/l #16, Median Type 3- b/l #16, Values Type 3- b/l #16, Values critical Type 3- b/l #18, Median Type 3- b/l #18, Values Type 3- b/l #20, Values critical Type 3- b/l #20, Median Type 3- b/l #20, Values Type 3- b/l #18, Values critical Type 3- Overall b/l, Median Type 3- Overall b/l, Values Type 3- Overalll b/l, Values critical Figure Median b/l ratio and te associated error bars and te and statistics and teir tresold values for Type 3 glass beads 28

36 b/l b/l b/l b/l b/l Type 4- b/l #12, Median Type 4- b/l #12, Values Type 4- b/l #12, Values critical Type 4- b/l #14, Median Type 4- b/l #14, Values Type 4- b/l #14, Values 1.2 critical Type 4- b/l #16, Median Type 4- b/l #16, Values Type 4- b/l #16, Values critical Type 4- b/l #18, Median Type 4- b/l #18, Values Type 4- b/l #18, Values critical Type 4- Overall b/l, Median Type 4- Overall b/l, Values Type 4- Overall b/l, Values critical Figure Median b/l ratio and te associated error bars and te and statistics and teir tresold values for Type 4 glass beads 29

37 Table 3-4- Average and repeatability/ reproducibility statistics of b/l ratio (all analyzers) Type Sieve Size # of Labs in Analysis Measured b/l, Average (%) Repeatability STD, S r, % Repeatability CV% Reproducibility STD, S R, % Reproducibility CV% # # # Overall # # # # Overall # # # # Overall # # # # Overall Form Factor Parameter (SPHT) Te median SPHT values, te associated error bars, and te and statistics for te four types of glass bead are sown in Figure 3-13 troug Figure Te resulted statistics of SPHT after removing te outlier data are provided in Table 3-5. Te observations from te figures and te table are as follows: 1. For Type 1 glass beads, Figure 3-13 sows tat in terms of reproducibility, none of te values exceed te tresold values. In terms of repeatability, values for Sets #2, #23, #24, and #22 corresponding to Sieve #30, #50, and #100 exceed te tresold value. After excluding te above data sets, te resulted statistics in Table 3-5 indicate tat te average of overall SPHT is 0.93 and it ranges between 0.93 to 0.96 for te individual sieve sizes. Te overall repeatability coefficient of variation for SPHT is 0.2% and te repeatability of SPHT for te individual sieve sizes is eiter 0.1 % or 0.2%. Te overall reproducibility coefficient of variation is 4.0% and ranges from 1.9% to 4.9% for te individual sieve sizes. 30

38 2. For Type 2 glass beads, Figure 3-14 sows tat in terms of reproducibility, none of te values exceeds te tresold values. In term of repeatability, te values for Sets #23, #22, #24, and #16 corresponding to sieves #35, #50, #100 exceed te tresold values. After excluding te above data sets from te analysis, te resulted statistics in Table 3-5 indicate tat te average of overall SPHT is 0.94 and it ranges from 0.92 to 0.96 for te individual sieve sizes. Te repeatability coefficient of variation is 0.3% for overall SPHT and it is eiter 0.1% or 0.3% for te individual sieve sizes. Te reproducibility coefficient of variation is 3.0 % for te overall SPHT and it ranges from 2.0% to 3.8% for te individual sieve sizes. 3. For Type 3 glass beads, Figure 3-15 sows tat in terms of of reproducibility, none of te values exceeds te tresold values. In terms of repeatability, te values for data set #16 corresponding to several class sizes, exceed te tresold values. After removing te above data sets from te analysis, te resulted statistics in Table 3-5 indicate tat te average of te overall SPHT is 0.97 and it ranges from 0.96 to 0.98 for te individual sieve sizes. Te repeatability coefficient of variation for te overall SPHT is 0.2% and it is eiter 0.2% or 0.4 % for te individual sieve sizes. Te reproducibility coefficient of variation is 1.9 % for te Overall SPHT and ranges from 1.9% to 2.1% for te individual sieve sizes. 4. For Type 4 glass beads, Figure 3-16 sows tat in terms of reproducibility, none of te values exceeds te tresold values. In terms of repeatability, values for data set #16 corresponding to different class sizes exceed te tresold value. After removing data set #16, te resulted statistics in Table 3-5 indicate tat te average overall SPHT is 0.97 and te average for te individual sieve sizes ranges from 0.95 to Te repeatability coefficient of variation for overall SPHT is 0.1% and ranges from 0.0% to 0.3% for te individual sieve sizes. Te reproducibility coefficient of variation for overall SPHT is 2.1% and ranges from 2.0% to 2.4% for te individual sieve sizes. Several observations can be made from Table 3-5 and its comparison wit Table ) as it was expected, te average SPHT are larger tan te average b/l values corresponding to te four glass bead types. 2) te average SPHT parameter of te fours glass bead types are all larger tan te SPHT tresold value of 0.93 (specified in AASHTO PP 74). 3) te average SPHT values are sligtly larger for te larger glass beads (Type 3 and Type 4) tan for te smaller beads (Type 11 and Type 2). 4) te reproducibility coefficient of variation is sligtly smaller for te larger glass beads tan for te smaller glass beads. 5) te repeatability coefficients of variation of te smaller beads (Type 1 and Type 2) are equivalent to tose of larger beads (Type 3 and Type 4). 6) larger number of laboratories reported b/l parameter tan te SPHT parameter. 7) bot repeatability and reproducibility coefficient of variations of SPHT are smaller tan tose of b/l ratio; te reason for small repeatability is tat several laboratories reported te same SPHT value for te tree replicates. Te reason for small reproducibility is tat te SPHT values reported by various laboratories were witin a very narrow range. 31

39 SPHT SPHT SPHT SPHT Type 1- SPHT #30, Median Type 1- SPHT #30, Values Type 1- SPHT #30, Values critical Type 1- SPHT #50, Median Type 1- SPHT #50, Values Type 1- SPHT #50, Values critical Type 1- SPHT #100, Median Type 1- SPHT # 100, Values Type 1- SPHT #100, Values critical Type 1- Overall SPHT, Median Type 1- Overall SPHT, Values Type 1- Overall SPHT, Values critical Figure Median SPHT and te associated error bars and te and statistics and teir tresold values for Type 1 glass beads 32

40 SPHT SPHT SPHT SPHT SPHT Type 2- SPHT #25, Median Type 2- SPHT #25, Values Type 2- SPHT #25, Values critical Type 2- SPHT #35, Median Type 2- SPHT #35, Values Type 2- SPHT #35, Values critical Type 2- SPHT #50, Median Type 2- SPHT #50, Values Type 2- SPHT #50, Values critical Type 2- SPHT #100, Median Type 2- SPHT #100, Values Type 2- SPHT #100, Values critical Type 2- Overall SPHT, Median Type 2- Overall SPHT, Values Type 2- Overall SPHT, Values 1.2 critical Figure Median SPHT and te error bars; te and statistics and teir tresold values for Type 2 glass beads 33

41 SPHT SPHT SPHT SPHT SPHT Type 3- SPHT #16, Median Type 3- SPHT #16, Values Type 3- SPHT #16, Values critical Type 3- SPHT #18, Median Type 3- SPHT #18, Values Type 3- SPHT #18, Values critical Type 3- SPHT #20, Median Type 3- SPHT #20, Values Type 3- SPHT #20, Values critical Type 3- SPHT #25, Median Type 3- SPHT #25, Values Type 3- SPHT #25, Values 1.2 critical Type 3- Overall SPHT, Median Type 3- Overall SPHT, Values Type 3- Overall SPHT, Values critical Figure Median SPHT and te error bars; te and statistics and teir tresold values for Type 3 glass beads 34

42 SPHT SPHT SPHT SPHT SPHT Type 4- SPHT #14, Median Type 4- SPHT #14, Values Type 4- SPHT #14, Values critical Type 4- SPHT #16, Median Type 4- SPHT #16, Values Type 4- SPHT #16, Values critical Type 4- SPHT #18, Median Type 4- SPHT #18, Values Type 4- SPHT #18, Values 1.2 critical Type 4- SPHT #20, Median Type 4- SPHT #20, Values Type 4- SPHT #20, Values critical Type 4- Overall SPHT, Median Type 4- Overall SPHT, Values Type 4- Overall SPHT, Values critical Figure Median SPHT and te associated error bars; te and statistics and teir tresold values for Type 4 glass beads 35

43 Table 3-5- Average sand repeatability/reproducibility statistics of SPHT parameters (all analyzers) Type Sieve Size # of Labs Average SPHT, % Repeatability STD, S r, % Repeatability CV% Reproducibility STD, S R, % Reproducibility CV% # # # Overall # # # # Overall # # # # Average # # # Percent Round based on b/l # Overall Te median of percent of particles close to a circle (round) based on b/l parameter for te four types of glass beads, te associated error bars, and te and statistics for determining te outlier data are sown in Figure 3-17 troug Figure Te statistics of percent round particles are provided in Table 3-6. Te observations from te table and figures are as follows: For Type 1 glass beads, 15 sets of te % round data on overall gradation and 13 sets of % round on individual sieves were received. Figure 3-17 sows tat in terms of reproducibility, te values for data sets #21 and #23 (corresponding to sieves #50 and #100 and te overall gradation) exceed te tresold values. In terms of repeatability, values for data set #23 (corresponding to Sieve #100 and te overall gradation) exceed te tresold values. After excluding te outlier data, 14 sets of % round data for te overall gradation and 12 set of % round data on te individual sieves are remained for te analysis. Te results of statistical analysis are provided in 1. Table 3-6. As indicated from te table, te average of overall percent round is 84.0 % and it ranges between 83.0% to 86.4% for te individual sieve sizes. Te resulted repeatability coefficient of variation is 0.7% for te overall gradation and it ranges between 0.8% to 1.3% for te individual sieve sizes. Te computed reproducibility 36

44 coefficient of variation is 5.7% for te overall gradation and ranges from 2.5% to 7.8% for te individual sieve sizes. For Type 2 glass beads, 13 sets of % round data for te overall gradation and 12 sets of % round for te individual sieves were received. Figure 3-18 sow tat in terms of reproducibility, te values for data sets #23 and #17 and #21 corresponding to Sieves #35, #50, and #100 exceed te tresold values. In terms of repeatability, te values for data sets #2, #14, and #17 (corresponding to Sieve #25, #50, and #100) exceed te tresold values. After excluding te outlier data, 13 sets of % round data for te overall gradation and 10 or 11 sets of % round data on te individual sieves are remained for te analysis. Te resulted statistics are provided in 2. Table 3-6. As indicated from te table, te average percent round is 79.5% for te overall gradation and it ranges from 79.3% to 83.1% for te individual sieve sizes. Te repeatability coefficient of variation is % for te overall % round and ranges from 0.7% to 2.2% for te individual sieve sizes. Te reproducibility coefficient of variation is 5.2 % for te overall percent round and it ranges from 2.5% to 10.1% for percent round on individual sieve sizes. For Type 3 glass beads, 15 sets of te % round data for te overall gradation and 12 and 13 sets of % round on te individual sieves were received. Figure 3-19 sows tat in terms of reproducibility, none of te values exceeds te tresold values. In terms of repeatability, te values for data Sets #10 and #2, corresponding to various sieve sizes, exceed te tresold values. After excluding te outlier data, 14 sets of overall % round data and 11, 12, and 13 sets of % round data on te individual sieves are remained for te analysis. Te resulted statistics in 3. Table 3-6 indicate tat te average of te percent circularity is 94.0% for te overall gradation and it ranges from 85.4% to 95.4% for te individual sieve sizes. Te resulted repeatability coefficient of variation for te overall % round is 0.1% and ranges from 0.2 % to 1.2 % for te individual sieve sizes. Te reproducibility coefficient of variation is 2.1% for te overall percent circularity and ranges from 1.9 % to 9.2% for te individual sieve sizes. For Type 4 glass beads, 16 sets of % round data for te overall gradation and 14 sets of data on individual sieves were received. Prior to and analysis, data sets #6 and #8 were excluded since tey were significantly different from oter results. Figure 3-20 sows tat for te remaining data, none of te and values exceed te tresold and. Terefore, 15 data sets data on te overall % round and 13 sets of % round data on te individual sieves were analyzed. Te resulted statistics are provided in 4. Table 3-6. As sown in te table, te average percent round is 92.2% for te overall gradation and it ranges from 87.7% to 93.9% for te individual sieve sizes. Te computed repeatability coefficient of variation of percent round for te overall gradation is 0.4% and it ranges from 0.3% to 2.5% for te individual sieve sizes. Te reproducibility coefficient of variation for % round is 2.0 % on te overall gradation and it ranges between 3.2% to 7.6% on individual sieve sizes. Some general observations can also be made from Table 3-6 and Figure 3-17 troug Figure Tese are as follows: 1. More data sets were provided for te overall gradation tan on te individual sieve sizes. Terefore, te repeatability and reproducibility statistics of percent round would be more reliable if prepared based on te data on overall gradation tan based on te data on individual sieves. 37

45 2. Type 3 and Type 4 glass beads ave more percent circular beads tan Type 1 and Type Bot repeatability and reproducibility coefficient of variation for % round is smaller for te larger beads (Types 3 and 4) tan for te smaller beads (Types 1 and 2). 4. majority of te data sets provided on percent round were measured by Camsizer instrument. Oter sets provided were measured by Sympatec, PartAn 2D and PartAn 3D. Tis migt indicate tat oter particle analyzers are not calibrated for measuring percent round based on b/l ratio. 5. Figure 3-17 troug Figure 3-19 sow tat and values corresponding to te percent round data measured by Camsizer (laboratories 1 troug 14) are smaller tan tose provided by oter particle analyzers. Tis could indicate tat for measuring % round based on b/l, Camsizer units are calibrated similarly, wic is different from calibration of oter particle analyzers. 38

46 % Round % Round % Round % Round Type 1- b/l % Round #30, Median Type 1- b/l % Round #30, Values Type 1- b/l % Round #30, Values critical Type 1- b/l % Round #50, Median Type 1- b/l % Round #50, Values Type 1- b/l % Round #50, Values critical Type 1- b/l % Round #100, Median Type 1- b/l % Round # 100, Values Type 1- b/l % Round #100, Values critical Type 1- Overall b/l % Round, Median Type 1- Overall b/l % Round, Values Type 1- Overall % Round b/l, Values critical Figure Median % round and te associated error bars and te and statistics and teir tresold values for Type 1 glass beads 39

47 % Round % Round % Round % Round % Round Type 2- b/l % Round #25, Median Type 2- b/l % Round #25, Values Type 2- b/l % Round #25, Values critical Type 2- b/l % Round #35, Median Type 2- b/l % Round #35, Values Type 2- b/l % Round #35, Values critical Type 2- b/l % Round #50, Median Type 2- b/l % Round #50, Values 3.30 Type 2- b/l % Round #50, Values critical Type 2- b/l % Round #100, Median Type 2- b/l % Round #100, Values Type 2- b/l % Round #100, Values critical Type 2- Overall b/l % Round, Median Type 2- Overall b/l % Round, Values Type 2- Overall b/l % Round, Values critical Figure Median % round and te error bars; te and statistics and teir tresold values for Type 2 glass beads 40

48 b/l % Round % Round % Round % Round Type 3- b/l % Round #16, Median Type 3- b/l % Round #16, Values Type 3- b/l % Round #16, Values critical Type 3- b/l % Round #18, Median Type 3- b/l % Round #18, Values Type 3- b/l % Round #18, Values critical Type 3- b/l % Round #20, Median Type 3- b/l % Round #20, Values Type 3- b/l % Round #20, Values critical Type 3- b/l % Round #25, Median Type 3- b/l % Round #25, Values Type 3- b/l % Round #25, Values critical Type 3- Overall b/l % Round, Median Type 3- Overall b/l Percent Round, Values Type 3- Overall % Round b/l, Values critical Figure Median % round and te error bars; te and statistics and teir tresold values for Type 3 glass beads 41

49 % Round % Round % Round % Round % Round Type 4- b/l % Round #14, Median Type 4- b/l % Round #14, Values Type 4- b/l % Round #14, Values critical Type 4- b/l % Round #16, Median Type 4- b/l % Round #16, Values Type 4- b/l % Round #16, Values critical Type 4- b/l % Round #18, Median Type 4- b/l % Round #18, Values Type 4- b/l % Round #18, Values critical Type 4- b/l % Round #20, Median Type 4- b/l % Round #20, Values Type 4- b/l % Round #20, Values critical Type 4- Overall b/l % Round, Median Type 4- Overall % Round b/l, Values Type 4- Overall b/l % Round, Values critical Figure Median % round and te associated error bars and te and statistics and teir tresold values for Type 4 glass beads 42

50 Table 3-6- Average sand repeatability/reproducibility statistics of percent round based on b/l (all analyzers) Type Sieve Size # of Data Sets Measured Circularity, Average (%) Repeatability STD, S r, % Repeatability CV% Reproducibility STD, S R, % Reproducibility CV% # # # Overall # # # # Overall # # # # Overall # # # # Overall Percent Non-Round based on SPHT Te statistics of percent particles away from a circle (non-round) based on SPHT parameter for te four types of glass beads are provided in Table 3-7. Te median % non-round values, te associated error bars, and te and statistics for determining te outlier data prior to analysis of precision estimates are sown in Figure 3-21 troug Figure Te observations from te table and figures are as follows: 1. For Type 1 glass beads, 9 sets of data were received on te overall gradation and 7 sets of data were received on individual sieve sizes. Figure 3-21 sows tat in terms of reproducibility, te value for data set #17 corresponding to sieve #50 exceeds te tresold value and tat data was also excluded from te analysis. In terms of repeatability, value for data set #23 corresponding to te overall gradation exceeds te 43

51 tresold value and tat data set was excluded from te analysis. Based on te analysis of te remaining 8 sets of % non-round data on te overall gradation and te 6 sets of data on te individual sieves, as provided in Table 3-7, te average of percent non-round is 14.1 % for te overall gradation and it ranges from 12.0% to 14.1% for te individual sieve sizes. Te repeatability coefficients of variation are 4.2% for te overall gradation and ranges from 6.7% to 7.8% for te individual sieve sizes. Te reproducibility coefficient of variation is 28.5% for te overall gradation and ranges from 17.4% to 56.4% for te individual sieve sizes. 2. For Type 2 glass beads, 9 sets of data were received on te overall gradation and 7 sets of data were received on individual sieves. Figure 3-22 sows tat in terms of reproducibility, te values of all sets are smaller tan te tresold values; terefore, no data were excluded based on statistic. In terms of repeatability, te value for data set #17 corresponding to sieve #100 exceeded te tresold value and it was excluded from te analysis. Based on te analysis of te remaining 9 data sets on te overall gradation and 6 or 7 data sets on te individual sieves (as sown in Table 3-7), te average of percent non-circular is 15.5% for te overall gradation and ranges between 12% to 15.4% for te individual sieve sizes. Te repeatability coefficient of variation of percent non-circularity is 4.7% for te overall gradation and ranges from 2.9% to 11.5% for te individual sieve sizes. Te reproducibility coefficient of variation is 30 % for te overall percent non-circular and it ranges from 30.4% to 93.2% for te individual sieves. 3. For Type 3 glass beads, 10 sets of non-round data were received on te overall gradation and 5 or 6 sets of data were received on te individual sieves. Figure 3-23 sows tat in terms of reproducibility, none of te values exceeds te tresold values. in terms of repeatability, te values for data sets #5, #2, and #23 exceed te tresold values corresponding to Sieves #16, #25, and te overall gradation, respectively; terefore, tose data sets were excluded from te analysis. Based on te remaining 9 data sets on te overall gradation and 4 to 6 data sets on te individual sieves, as provided in Table 3-7, te average of percent non-round on te overall gradation is 3.8% and it ranges between 2.1% to 5.0% for te individual sieve sizes. Te repeatability coefficient of variation for te overall % non-circularity is 4.2% and ranges from 2.5 % to 10.8 % for te individual sieve sizes. Te reproducibility coefficient of variation is 45.4% for te Overall percent non-circularity and ranges from 66.8% to 110.1% for te individual sieve sizes. 4. For Type 4 glass beads, 10 sets of non-round data on te overall gradation and 6 sets of data on te individual sieves were received. Figure 3-24 sows tat in terms of reproducibility, te values of none of te data sets exceed te tresold value; terefore, no data were excluded based on statistic. In terms of repeatability, values for data set #17 corresponding to sieves #20 and te value for data set #10 corresponding to te overall gradation exceed te tresold values and tose data sets were excluded from te analysis. Based on te analysis of te remaining 9 data sets on te overall gradation and 5 or 6 data sets on individual sieves, as provided in Table 3-7, te average percent non-round on te overall gradation is 3.6% and it ranges between 2.9% to 4.7% on te individual sieves. Te repeatability coefficient of variation for te overall percent non-circular is 3.4% and it ranges from 4.9% to 23.1% for te individual sieves. Te reproducibility coefficient of variation for overall non-circularity is 54.8 % and it ranges between 67.4% to 95% for te individual sieve sizes. 44

52 Several general observations can also be made from Table 3-7 and Figure 3-21 troug Figure Tese are as follows: 1. less number of laboratories provided %non-round data based on SPHT tan percent round based on b/l (10 sets of non-round vs. 15 sets of round); Tis could indicate tat b/l is a more recognized parameter for determining percent round/ non-round tan SPHT. 2. bot witin- and between-variability of percent round based on b/l is smaller tan tose of percent non-round based on SPHT. It could be concluded tan b/l is a more robust parameter tan SPHT for determining percent of round/non-round particles. 3. Te average percent non-round particles are larger in smaller glass bead types tan te large glass bead types (14.1% and 15.5%. for Types 1 and 2 vs. 3.8% and 3.6% for Types 3 and 4). 4. Te repeatability coefficients of variation of percent round measurements are very small and comparable for all glass bead types. 5. Te reproducibility coefficient of variation of percent non-round are very large for various glass bead types and te values are larger for te larger glass bead types (e.g., 54.8% for Type 4 vs. 30% for Type 2 glass beads). 45

53 Table 3-7- Averages and repeatability/reproducibility statistics of %non-round based on SPHT (all analyzers) Type Sieve Size # of Data Sets Non- Circularity, Average (%) Repeatability STD, S r, % Repeatability CV% Reproducibility STD, S R, % Reproducibility CV% # # # Overall # # # # Overall # # # # Overall # # # # Overall

54 % Non-Round % Non-Round % Non-Round % Non-Round Type 1- SPHT % Non-Round #30, Median Type 1- SPHT % Non-Round #30, Values Type 1- SPHT % Non-Round #30, Values critical Type 1- SPHT % Non-Round #50, Median Type 1- SPHT % Non-Round #50, Values Type 1- SPHT % Non-Round #50, Values critical Type 1- SPHT % Non-Round #100, Median Type 1- SPHT % Non-Round # 100, Values Type 1- SPHT % Non-Round #100, Values critical Type 1- Overall SPHT % Non-Round, Median Type 1- Overall SPHT % Non-Round, Values Type 1- Overall SPHT % Non-Round, Values critical Figure Median percent non-round and te associated error bars and te and statistics and teir tresold values for Type 1 glass beads 47

55 % Non-Round % Non-Round % Non-Round % Non-Round % Non-Round Type 2- SPHT % Non-Round #25, Median Type 2- SPHT % Non-Round #25, Values Type 2- SPHT % Non-Round #25, Values critical Type 2- SPHT % Non-Round #35, Median Type 2- SPHT % Non-Round #35, Values Type 2- SPHT % Non-Round #35, Values 28.0 critical Type 2- SPHT % Non-Round #50, Median Type 2- SPHT % Non-Round #50, Values Type 2- SPHT % Non-Round #50, Values 30.0 critical Type 2- SPHT % Non-Round #100, Median Type 2- SPHT % Non-Round #100, Values Type 2- SPHT % Non-Round #100, Values critical Type 2- Overall SPHT % Non-Round, Median Type 2- Overall SPHT % Non-Round, Values Type 2- Overall SPHT % Non-Round, Values critical Figure Median percent non-round plus te error bars and te and statistics and teir tresold values for Type 2 glass beads 48

56 % Non-Round % Non-Round % Non-Round % Non-Round % Non-Round Type 3- SPHT % Non-Round #16, Median Type 3- SPHT % Non-Round #16, Values Type 3- SPHT % Non-Round #16, Values critical Type 3- SPHT % Non-Round #18, Median Type 3- SPHT % Non-Round #18, Values 3.30 Type 3- SPHT % Non-Round #18, Values critical Type 3- SPHT % Non-Round #20, Median Type 3- SPHT % Non-Round #20, Values Type 3- SPHT % Non-Round #20, Values critical Type 3- SPHT % Non-Round #25, Median Type 3- SPHT % Non-Round #25, Values Type 3- SPHT % Non-Round #25, Values critical Type 3- Overall SPHT % Non-Round, Median Type 3- Overall SPHT Percent Non-Round, Values Type 3- Overall SPHT % Non-Round, Values critical Figure Median percent non-round plus te error bars and te and statistics and teir tresold values for Type 3 glass beads 49

57 % Non-Round % Non-Round % Non-Round % Non-Round % Non-Round Type 4- SPHT % Non-Round #14, Median Type 4- SPHT % Non-Round #14, Values Type 4- SPHT % Non-Round #14, Values 13.0 critical Type 4- SPHT % Non-Round #16, Median Type 4- SPHT % Non- Round #16, Values Type 4- SPHT % Non-Round #16, Values 13.0 critical Type 4- SPHT % Non-Round #18, Median Type 4- SPHT % Non-Round #18, Values Type 4- SPHT % Non-Round #18, Values critical Type 4- SPHT % Non-Round #20, Median Type 4- SPHT % Non-Round #20, Values Type 4- SPHT % Non-Round #20, Values 13.0 critical Type 4- Overall SPHT % Non-Round, Median Type 4- Overall SPHT %Non- Round, Values Type 4- Overall SPHT % Non-Round, Values critical Figure Median percent non-round plus te error bars and te and statistics and teir tresold values for Type 4 glass beads 3.2 Comparison of Results from 2-D and 3-D Analysis Metods Microtrac PartAn particle analyzer provides bot 2-D and 3-D analysis of te glass beads size and sape measurements, wic maes possible comparison of te results from 2-D and 3-D metods. In addition to widt (b) and lengt (l) of te particles, PartAn 3D reports ticness (T), as te tird dimension, by tracing te particles as tey tumble down te measurement area. T 50

58 would be te smallest dimension of te smallest particle project, wile b is te smallest dimension of te largest particle projection. Te T parameter is used instead of b in te calculation of aspect ratio (b/l) to provide better estimate of te particles sape. Two Microtrac laboratories in Norway and USA (Yor) provided 2D and 3D measurements. Te comparison of te 2D and 3D results are discussed as follows. Figure 3-25 sows te grapical comparison of te averages of D10, D50, and D90 determined by te 2D and 3D analysis metods. Te first two columns of eac category represent 2D analysis and te second two columns represent te 3D analysis. As indicated from te figure, in general, sligtly smaller decile values are determined by te 3D analysis tan by te 2D analysis. D10, D50, D90-Average Types 1 & 2 D10, D50, D90-Average Types 3 & D 10 D 50 D 90 D 10 D 50 D D 10 D 50 D 90 D 10 D 50 D 90 Type 1 Type 2 Type 3 Type 4 Microtrac-Norway-2D Microtrac-Yor-2D Microtrac-Norway-3D Microtrac-Yor-3D Microtrac-Norway-2D Microtrac-Yor-2D Microtrac-Norway-3D Microtrac-Yor-3D Figure Comparison of statistics of D10, D50, D90 data from PartAn 2D and 3D analysis Figure 3-26 sows te grapical comparison of te averages of % retained determined by PartAn 2D and 3D analysis metods. Te first two columns of eac category represents 2D analysis and te second two columns represent te 3D analysis. As indicated from te figure, using 3D analysis, sligtly less amount of particles is sown to be retained in te middle class sizes and sligtly more particles are retained in te next smaller class sizes. Tis observation is more evident for Types 1 and 2 glass beads. % Retained-Average Types 1 & 2 % Retained-Average Types 3 & #30 #50 #100 #18 #25 #35 # #16 #18 #20 #25 #14 #16 #18 #20 Type 1 Type 2 Type 3 Type 4 Microtrac-Norway-2D Microtrac-Yor-2D Microtrac-Norway-3D Microtrac-Yor-3D Microtrac-Norway-2D Microtrac-Yor-2D Microtrac-Norway-3D Microtrac-Yor-3D Figure Comparison of statistics of % Retained data from PartAn 2D and 3D analysis Figure 3-27 sows te grapical comparison of te averages aspect ratio (b/l and T/L) determined by PartAn 2D and 3D analysis metods. Te first two columns of eac category represents 2D analysis and te second two columns represent te 3D analysis. As indicated from 51

59 te figure, for eac class size of eac glass bead type, smaller aspect ratio is measured by te 3D analysis tan by te 2D analysis. Tis is expected since in 3D analysis te ticness (T), wic is smaller tan te widt (b), is used in calculating te aspect ratio. Aspect Ratio-Average Types 1 & 2 Aspect Ratio-Average Types 3 & #30 #50 #100 Average #25 #35 #50 #100 Average 0.7 #16 #18 #20 #25 Average #14 #16 #18 #20 Average Type 1 Type 2 Type 3 Type 4 Microtrac-Norway-2D Microtrac-Yor-2D Microtrac-Norway-3D Microtrac-Yor-3D Microtrac-Norway-2D Microtrac-Yor-2D Microtrac-Norway-3D Microtrac-Yor-3D Figure Comparison of statistics of aspect ratio data from PartAn 2D and 3D analysis Figure 3-28 sows te grapical comparison of te averages of SPHT parameter determined by PartAn 2D and 3D analysis metods. Te first two columns of eac category represents 2D analysis and te second two columns represent te 3D analysis. Figure sows tat te 2D and 3D SPHT values are equivalent and tere is no specific trend of increase or decrease in % round wen 2D or 3D analysis are used (compare te first and tird columns and compare te second and fourt columns for eac class size of eac glass bead type). SPHT-Average Types 1 & 2 SPHT-Average Types 3 & #30 #50 #100 Average #25 #35 #50 #100 Average #16 #18 #20 #25 Average #14 #16 #18 #20 Average Type 1 Type 2 Type 3 Type 4 Microtrac-Norway-2D Microtrac-Yor-2D Microtrac-Norway-3D Microtrac-Yor-3D Microtrac-Norway-2D Microtrac-Yor-2D Microtrac-Norway-3D Microtrac-Yor-3D Figure Comparison of statistics of SPHT data from PartAn 2D and 3D analysis Figure 3-29 sows te grapical comparison of te average % round determined by PartAn 2D and 3D analysis metods. One laboratory (Norway) provided complete sets of 2D and 3D for all class sizes of te four types of glass beads. Te oter laboratory (Yor) provided only te % round data for te overall gradation and not for te individual sieve sizes. Figure sows tat te 2D and 3D % round values are equivalent and tere is no specific trend of increase or decrease in % round wen 2D or 3D analysis are used. 52

60 % Round-Average Types 1 & 2 % Round-Average Types 3 & #30 #50 #100 Average #25 #35 #50 #100 Average #16 #18 #20 #25 Average #14 #16 #18 #20 Average Type 1 Type 2 Type 3 Type 4 Microtrac-Norway-2D Microtrac-Yor-2D Microtrac-Norway-3D Microtrac-Yor-3D Microtrac-Norway-2D Microtrac-Yor-2D Microtrac-Norway-3D Microtrac-Yor-3D Figure Comparison of statistics of % round data from PartAn 2D and 3D analysis Figure 3-30 sows te grapical comparison of te average % non-round based on SPHT parameter determined by PartAn 2D and 3D analysis metods. Similar to % round data, te % non-round data were provided in complete sets of 2D and 3D for all class sizes of te four types of glass beads by te laboratory in Norway. However, % non-round data by te laboratory in Yor were provided for only te overall gradation and not for te individual sieve sizes. As indicated from te figure, tere is no specific trend of increase or decrease in % non-round wen 2D or 3D analysis are used. % Non-Round-Average Types 1 & 2 % Non-Round-Average Types 3 & #30 #50 #100 Average #25 #35 #50 #100 Average #16 #18 #20 #25 Average #14 #16 #18 #20 Average Type 1 Type 2 Type 3 Type 4 Microtrac-Norway-2D Microtrac-Yor-2D Microtrac-Norway-3D Microtrac-Yor-3D Microtrac-Norway-2D Microtrac-Yor-2D Microtrac-Norway-3D Microtrac-Yor-3D Figure Comparison of statistics of % non-round data from PartAn 2D and 3D analysis 3.3 Statistics of Camsizer Results Te Camsizer results were pooled togeter and analyzed separately to assess if tere are any canges in te current precision estimates of PP 74, wic were originally developed based on te Camsizer measurements in Separate analysis of Camsizer data also maes possible comparison of te statistics measured wit various particle analyzers wit tose measured by Camsizer, wic is currently te main instrument for measuring properties of glass beads in te pavement maring industry. Te Camsizer results are discussed in te following sections D10, D50, D90 Table 3-8 provides te statistics of D10, D50, D90 measured by te Camsizer particle analyzer. As indicated from te table, te measured D10, D 50, D 90 are witin te expected range of te values corresponding to eac glass bead type. Te repeatability and reproducibility coefficients of variation (CV) for te four types of glass beads are in a range of 0.1% to % and 0.5% to 1.2 %, respectively. 53

61 Table 3-8-Summary of statistics of Q10, Q50, and Q90 data measured wit Camziser Type Property # of Labs Expected Diameter, Range (mm) Measured Diameter, Average (mm) Repeat. STD, S r, % Repeat. CV% Reproduc. STD, S R, % Reproduc. CV% Q to Q to Q to Q to Q to Q to Q to Q to Q to Q to Q to Q to Figure 3-31 sows te grapical comparisons of te statistics of D10, D50, and D90 data measured using Camsizer wit tose measured wit all types of streaming flow particle analyzers. As sown in te figure, te average D10, D50, and D90 are te same for te two groups of data. However, te repeatability CVs of te Camsizer data, particularly tose corresponding to te smaller bead types (Type 1 and 2), are smaller tan tose of all brands of particle analyzers. Te reproducibility coefficients of variations of D10, D50, and D90 wen measured wit Camsizer are all smaller tan tose measured wit various brands of particle analyzers. 54

62 Reproducibility CV, % Repeatability CV, % Average Diameter, mm Average D10, D50, D D10 D50 D90 D10 D50 D90 D10 D50 D90 D10 D50 D90 Camsizer All Brands Bead Type/Gradation 2.5 D10, D50, D90 Repeatability CV, % Camsizer 0.5 All Brands 0 D10 D50 D90 D10 D50 D90 D10 D50 D90 D10 D50 D Bead Type/Gradation D10, D50, D90 Reproducibility CV, % D10 D50 D90 D10 D50 D90 D10 D50 D90 D10 D50 D90 Camsizer All Brands Bead Type/Gradation Figure Comparison of statistics of D10, D50, D90 data measured wit Camsizer and wit all brands of particle analyzers 55

63 3.3.2 Percent Retained Table 3-9 provides te statistics of percent retained on eac sieve size for te four types of glass beads measured wit Camsizer analyzer. As indicated from te table, te average percent retained on eac sieve size closely agrees wit te expected range of te percent retained for eac bead type. As indicated from te table, te repeatability and reproducibility coefficients of variation of percent retained, corresponding to te sieves wit te igest percent retained, are sown saded for eac of te four glass bead types. Tese are te values used in te preparation of precision statements since tey are associated wit te most number of beads. Te values range from 0.5% to 1.7% for te repeatability CV and 1.3% to 3.1% for te reproducibility CV. Table 3-9- Summary of statistics of Percent Retained ( 1-Q3) data measured wit Camziser Type Property # of Labs Expected Retained, Range (%) Measured Retained, Average (%) Repeat. STD, S r, % Repeat. CV% Reproduc. STD, S R, % Reproduc. CV% # to # to # to # to # to # to # to # to # to # to # to # to # to # to # to Figure 3-32 sows te grapical comparison of te statistics of Percent Retained measured by Camsizer and tose measured by all brands of particle analyzers. As indicated from te figure, te average percent retained measured by Camsizer and wit all brands of particle analyzers are very similar. However, te majority of repeatability CVs are larger wen measured wit all particle analyzers tan wen measured by only Camsizer, particularly te CVs associated wit te small bead types. In terms of reproducibility, te CVs are generally larger wen measured wit various particle analyzers tan wen measured wit Camsizer. 56

64 #30 #50 #100 #25 #35 #50 #100 #16 #18 #20 #25 #14 #16 #18 #20 Reproducibility CV, % #30 #50 #100 #25 #35 #50 #100 #16 #18 #20 #25 #14 #16 #18 #20 Repeatability CV, % #30 #50 #100 #25 #35 #50 #100 #16 #18 #20 #25 #14 #16 #18 #20 Retained, % Average % Retained Camsizer All Brands Bead Types/Gradation Percent Retained Repeatability CV, % Camsizer All Brands Bead Types/Gradation Percent Retained Reproducibility CV, % Camsizer All Brands Bead Types/Gradation Figure Comparison of statistics of D10, D50, D90 data measured wit Camsizer and wit all brands of particle analyzers 57

65 3.3.3 Aspect Ratio Parameter (b/l) Table 3-10 provides te statistics of b/l on eac sieve size for te four types of glass beads measured by Camsizer only. In addition to te b/l for eac sieve size, te b/l for te overall gradation, wic is te weigted average of all sieve sizes is sown in te table. As indicated, te average b/l on eac sieve size and an average overall b/l for te four types of beads are in a range of 0.91 to 0.97, wic are above te tresold value of 0.85 for roundness. In general, smaller coefficients of variation are associated wit te overall gradation or wit te sieves wit larger amount of beads, referred to as dominant class sizes. Te coefficients of variation of b/l corresponding to te dominant class sizes of te four types of glass beads are 0.1% or 0.2% witin laboratory and ranges from 0.6% to 2.4% between laboratory. Te coefficients of variation of b/l corresponding to te overall gradation is 0.1% or 0.2% witin laboratory and in a range of 0.7% to 1.1% between laboratory. 58

66 Table Summary of statistics of aspect ratio ( b/l) data measured wit Camziser Type Sieve Size # of Labs Measured b/l, Average (%) Repeatability STD, S r, % Repeatability CV% Reproducibility STD, S R, % Reproducibility CV% # # # Overall # # # # Overall # # # # Overall # # # # Overall Figure 3-33 sows te grapical comparison of te statistics of b/l data measured by Camsizer and te data from all brands of particle analyzers. As sown in te figure, Camsizer instruments measured larger average b/l values (iger roundness). Tis is especially evident for b/l ratio of te smaller beads (Types 1 and Type 2). In terms of variability, te repeatability CV values of te two groups are, in general, equivalent. Te repeatability CVs of te two groups are all below 0.7%. However, te reproducibility CV values are significantly increased wen measurements performed wit different brands of particle analyzers. For smaller glass beads, Type 1 and 2, te reproducibility CVs of te data from all instrument brands 5 times larger tan tose from Camsizer. 59

67 #30 #50 #100 Overall #25 #35 #50 #100 Overall #16 #18 #20 #25 Overall #14 #16 #18 #20 Overall Reproducibility CV, % #30 #50 #100 Overall #25 #35 #50 #100 Overall #16 #18 #20 #25 Overall #14 #16 #18 #20 Overall Repeatability CV, % #30 #50 #100 Overall #25 #35 #50 #100 Overall #16 #18 #20 #25 Overall #14 #16 #18 #20 Overall Average b/l Average b/l Camsizer All Brands Bead Types/Gradation b/l Repeatability CV, % Camsizer All Brands Bead Types/Gradation b/l Reproducibility CV, % Camsizer All Brands Bead Types/Gradation Figure Comparison of statistics of b/l data measured wit Camsizer and wit all brands of particle analyzers 60

68 3.3.4 Form Factor Parameter (SPHT) Table 3-11 sows te statistics of SPHT parameter measured by Camsizer for eac sieve size of te four types of glass beads and for te overall gradation, wic is te weigted average of SPHT of individual sieve sizes. As indicated by te igligted cells in te table, te average SPHT for te dominant class sizes are in a range of 0.96 to 0.99 and for te overall gradation are in a range of 0.95 to 0.98 for te four types of beads. Te repeatability coefficients of variation of SPHT corresponding to te dominant class sizes of te four types of glass beads are 0.0% or 0.1% and in te range of 0.0% to 0.2% for te Overall gradation. Te reproducibility coefficients of variation of SPHT corresponding to te dominant class size is in te a range of 0.3% to 1.7% and for te overall gradation is in te range of 0.5% to 1.2%. Table Summary of statistics of SPHT parameter measured wit Camziser Type Property # of Labs Average SPHT Repeatability STD, S r, % Repeatability CV% Reproducibility STD, S R, % Reproducibility CV% # # # Overall # # # # Overall # # # # Overall # # # # Overall Figure 3-34 sows grapical comparison of te statistics of SPHT measured by Camsizer and tose measured by all types of particle analyzers. As indicated from te grap, larger SPHT values are measured by Camsizer tan by all brands of particle analyzers (iger roundnss). In terms of variability, te repeatability CVs are equivalent; owever, te reproducibility CVs of SPHT data measured by different types of particle analyzers are in majority of cases significantly larger tan tose measured by Camsizer only. 61

69 #30 #50 #100 Overall #25 #35 #50 #100 Overall #16 #18 #20 #25 Overall #14 #16 #18 #20 Overall Reproducibility CV, % #30 #50 #100 Overall #25 #35 #50 #100 Overall #16 #18 #20 #25 Overall #14 #16 #18 #20 Overall Repeatability CV, % #30 #50 #100 Overall #25 #35 #50 #100 Overall #16 #18 #20 #25 Overall #14 #16 #18 #20 Overall Average b/l Average SPHT Camsizer All Brands Bead Type/Gradation SPHT Repeatability CV, % Camsizer All Brands Bead Type/Gradation SPHT Reproducibility CV, % Camsizer All Brands Bead Type/Gradation Figure Comparison of statistics of SPHT data measured wit Camsizer and wit all brands of particle analyzers 62

70 3.3.5 Percent Round based on b/l Table 3-12 sows te statistics of percent round, determined by percentage of particles wit b/l ratio larger tan For eac type of glass beads, laboratories provided percent round values for eac sieve size and/or for te overall gradation, wic is te weigted average of percent round of all sieve sizes computed by te particle analyzer. As indicated from te table, te larger glass beads (Types 3 and 4) ave iger percentage of particles wit b/l ratio above 0.85 tan te small glass beads (Types 1 and 2). Te average overall percent round is 84.8% and 79.9% for Types 1 and Type 2 glass beads and 93.5% and 91.3 % for Types 3 and 4 glass beads, respectively. Tis migt be due of presence of more of flay and irregular saped particles among te smaller glass bead types. In terms of variability, Table 3-12 sows tat te repeatability and reproducibility coefficients of variations for te two larger glass bead types (Type 3 and 4) are smaller tan tose for Types 1 and 2 glass beads (reproducibility CV of 0.5% for Type 3 vs. 1.2% for Type 1). Te repeatability CV of percent round corresponding to te dominant class sizes of te four types of glass beads are in a range of 0.2% to 0.6% and for te Overall gradation are in a range of 0.1% to 0.8%. For reproducibility, te CVs corresponding to te dominant class sizes are in a range of % to 2.2% and for te overall gradation are in a range of 0.5% to %. Figure 3-35 sows te grapical comparison of te statistics of percent round data measured based on b/l using Camsizer and based on te measurements by all brands of particle analyzers. As indicated from te grap, te average percent round computed from te two groups of data are very similar. Te repeatability CVs of te Camsizer percent round data is also equivalent wit te repeatability CVs of all particle analyzers data. However, te reproducibility CVs are significantly larger wen data are obtained by all different particle analyzers tan wen measured by Camsizer only. Te CVs of te Camsizer data are less tan 6% and tose of all types of particle analyzers are below 10%. 63

71 Table 3-12-Summary of statistics of % round based on b/l parameter measured wit Camziser Type Property # of Labs Measured b/l Roundness, Average (%) Repeatability STD, S r, % Repeatability CV% Reproducibility STD, S R, % Reproducibility CV% # # # Overall # # # # Overall # # # # Overall # # # # Overall

72 #30 #50 #100 Overall #25 #35 #50 #100 Overall #16 #18 #20 #25 Overall #14 #16 #18 #20 Overall Reproducibility CV, % #30 #50 #100 Overall #25 #35 #50 #100 Overall #16 #18 #20 #25 Overall #14 #16 #18 #20 Overall Repeatability CV, % #30 #50 #100 Overall #25 #35 #50 #100 Overall #16 #18 #20 #25 Overall #14 #16 #18 #20 Overall % Circularity Percent Round, Average, % Camsizer All Brands Bead Type/Gradation 3.0 Percent Round, Repeatability CV, % Camsizer All Brands Bead Type/Gradation Percent Round, Reproducibility CV, % Camsizer All Brands Bead Type/Gradation Figure Comparison of statistics of percent round based on b/l measured by Camsizer and by all brands of particle analyzers 65

73 3.3.6 Percent Non-Round based on SPHT Table 3-13 sows te statistics of percent non-round, as percentage of particles wit SPHT smaller tan Less number of laboratories provided percent non-round based on SPHT for eiter individual sieve sizes or for te overall gradation tan provided percent round based on b/l. Tis was probably because teir instruments were not setup to provide percent round/non-round based on SPHT parameter. As it is indicated from te table, larger glass beads (Type 3 and 4) ave smaller percentages of beads wit SPHT of below 0.93 tan smaller glass bead (Types 1 and 2). Te average percent non-round is in a range of 12.1% to 21.9% for Type 1 and 2 glass beads wile te range is 3.1% to 9.7% for Type 3 and 4 glass beads. Tis trend agrees wit wat was observed wit percent round based on b/l parameter. In terms of variability, te repeatability CV for te percent non-round corresponding to te dominant size class is in a range of 4.0% to 6.4% and for te overall gradation is in a range of 1.7% to 4.9%. Te reproducibility COVs for corresponding to te dominant size class 5.5% to 17.6 % and for te overall gradation is in a range of 12.7% to 17.7%. 66

74 Table Summary of statistics of % non-round beads based on SPHT parameter measured wit Camziser Type Property # of Labs Measured Non- Round, Average (%) Repeatability STD, S r, % Repeatability CV% Reproducibility STD, S R, % Reproducibility CV% # # # Overall # # # # Overall # # # # Overall # # # # Overall Figure 3-36 sows grapical comparison of te statistics of percent non-round data measured based on SPHT by Camsizer and tose measured by all brands of particle analyzers. As indicated from te grap, te average of percent non-round measured by Camsizer is in majority of cases larger tan tose measured by all brands of particle analyzers. Similar to te earlier observations from te percent round data, te percent non-round of smaller glass beads (Types 1 and 2) are larger tan tose of te larger glass bead types (Type 3 and 4); most probably due to flay and irregular saped particles among te smaller glass bead types. In terms of variability, te repeatability CVs of te Camsizer s % non-round data appear to be equivalent to te repeatability CVs of all particle analyzers data. However, te reproducibility CVs of te Camsizer s data are significantly smaller tan reproducibility CVs of te data measured by all different types of particle analyzers. Te CVs of te Camsizer data are less tan 22%; However, wen data are measured wit all equipment brands te CV could increase to around 120%. 67

75 #30 #50 #100 Overall #25 #35 #50 #100 Overall #16 #18 #20 #25 Overall #14 #16 #18 #20 Overall Reproducibility CV, % #30 #50 #100 Overall #25 #35 #50 #100 Overall #16 #18 #20 #25 Overall #14 #16 #18 #20 Overall Repeatability CV, % #30 #50 #100 Overall #25 #35 #50 #100 Overall #16 #18 #20 #25 Overall #14 #16 #18 #20 Overall %Non- Circular Non-Round, Average, % Camsizer All Brands Bead Type/Gradation 25.0 Non-Round, Repeatability CV, % Camsizer All Brands Bead Type/Gradation Non-Round, Reproducibility CV, % Camsizer All Brands Bead Type/Gradation Figure Comparison of statistics of percent non-round measured wit all brands of particle analyzers 68

76 3.4 Effect of Obscuration of Particles on Measurement Variability One of te reasons for te observed differences between te measurement results from various particle analyzers could be te obscuration of te particles, wic would result in overlapping and toucing particle projections. Te number of toucing/overlapping particle projections captured by eac instrument and te metod eac instrument uses to analyze te images of te particle projections could ave an effect on te measured properties. In every static or dynamic imaging system wit measurements of more tan one particle tere is a cance of particles overlapping or toucing. Overlapping particles may be seen as one particle because te viewer, te camera or evaluation software is not able to differentiate between an image wit two particles toucing or two beads melting togeter into one single oblong particle. Some instruments use analysis software tat contains algoritms for separating te touced/overlapping particles and some oter devices consider te toucing/overlapping particles as a non-circular particle and analyze tem as so. Te differences in te approaces used by eac particle analyzer could result in differences in te measured glass bead properties. In terms of size measurement, te effect from te obscuration of particles could be insignificant. Wen size of two toucing particles is measured by te widt of te particle projection (xc min = smallest of all maximum cords from different measurement directions of a static 2D image of a particle), te correct size is still measured. In oter words, te results of size measurements of two toucing particles or two single particles are identical wen te size definition xc min is used. Tis could be te reason tat te size measurements from various analyzers in tis study were not significantly different. However, te sape parameters lie SPHT, b/l, or T/L are sensitive against obscuration. In case two particles are toucing eac oter, tere are going to be lower values of SPHT and aspect ratio b/l or T/L. In a condition were te vibrating feeding system of an instrument ave an inomogeneous feeding, te obscuration influence can even be iger. A mecanical device, suc as a guidance seet or feed guide after te particles leave te vibrating feeder, could ave a positive effect of te omogeneity of te feeding of te particles into te measurement zone. 69

77 3.5 Precision Estimates of Glass Bead Properties Basis of Precision Estimates Figure 3-37 and Figure 3-38 sow te relationsip between averages and repeatability/reproducibility standard deviations of various glass bead properties measured by Camsizer and by various brands of particle analyzers. As indicated from te figures, te relationsips between average and standard deviation for majority of properties are eiter not significant (R 2 < 50%) or te direction of te relationsips are not consistent. For example, for te percent retained data by Camsizer, te direction of te relationsip between average and SR standard deviation for Type 1 and Type 2 glass beads is opposite of eac oter. It was terefore concluded tat tere are no prevailing relationsips between averages and standard deviations of te size and sape properties of te glass beads and terefore, as specified in ASTM C 670 [11], te basis of te precision estimates for te glass bead properties would be standard deviation Comparison of te Precision Estimates based on Data from Camsizer and Various Devises Two sets of precision estimates for te size and sape properties of te four glass bead types were prepared. One set was prepared by pulling te Camsizer data and anoter set was prepared by pulling all data from various instruments. Te data from Malvern instrument was not included in te precision estimate analysis since its measurement mecanism is not te same as tat of te flowing stream particle analyzers. Te precision estimates are provided in Table As indicated from te table, wile te witin-laboratory standard deviations (Sr) of te two sets are comparable, te between-laboratory standard deviations (SR) are significantly larger wen data are pulled from all particle analyzers tan wen only Camsizer s data were pulled. Te differences are more pronounced for te sape properties (percent round and percent non-round) tan for te size properties (decile values and % retained). Te reproducibility standard deviations of percent round and percent non-round data from various particle analyzers are up to 6.5 times larger tan tose of te Camsizer data. Te large between-laboratory variability of te sape parameters could be caused by using te same tresold values by various devices for roundness determination. Different instruments may require different tresold values for b/l and SPHT in determining percent round/non-round. Tis is because of te differences in te resolution of te cameras, te intensity of instrument ligt source, and use of different image analysis filters for separating te touced particles. Use of different algoritm for evaluation of roundness is anoter possible factor causing disparity of te sape results from different analyzers. 70

78 Sr Standard Deviation,% SR Standard Deviation, % Sr Standard Deviation, % SR Standard Deviation, % Sr Standard Deviation SR Standard Deviation Sr Standard Deviation SR Standard Deviation Sr Standard Deviation, % SR Standard Deviation, % Sr Standard Deviation, mm SR Standard Deviation, mm D10, D50, D90 Repeatability STD vs. Average Camsizer Data D10, D50, D90 Reproducibility STD vs. Average Camsizer Data R² = #N/A R² = R² = R² = 115 Type 1 Type 2 Type 3 Type R² = R² = R² = R² = Type 1 Type 2 Type 3 Type Average, mm Average, mm % Retained, Repeatability STD vs. Average Camsizer Data % Retained, Reproducibility vs. Average Camsizer Data R² = R² = R² = R² = Type 1 Type 2 Type 3 Type R² = R² = R² = R² = 7.4E-05 Type 1 Type 2 Type 3 Type Average, % Average, % b/l, Repeatability STD vs. Average Camsizer Data b/l, Reproducibility STD vs. Average Camsizer Data R² = R² = R² = Type 1 Type 2 Type R² = R² = Type 1 Type 2 2 R² = Type R² = Type 3 5 R² = Type Average Average 6 SPHT, Repeatability STD vs. Average Camsizer Data SPHT, Reproducibility STD vs. Average Camsizer Data R² = Type R² = Type 1 3 R² = Type R² = Type R² = R² = Type 3 Type R² = R² = Type 3 Type Average Average 2.5 % Circular, Repeatability STD vs. Average Camsizer Data 5.0 % Circular, Reproducibilty STD vs. Average Camsizer Data 2.0 R² = Type R² = Type R² = Type R² = Type 2 R² = Type R² = Type R² = Type 4 R² = Type Average, % Average, % 2.50 % Non-Circular, Repeatability STD vs. Average Camsizer Data 5.0 % Non-Circular, Reproducibilty STD vs. Average Camsizer Data 2.00 R² = 655 Type R² = Type 1 R² = Type R² = Type 2 0 R² = R² = Type 3 Type R² = R² = Type 3 Type Average, % Average, % Figure Relationsip between average and standard deviation of te glass bead properties measured wit Camsizer particle analyzer 71

79 Sr Standard Deviation, % SR Standard Deviation, % Sr Standard Deviation, % SR Standard Deviation, % Sr Standard Deviation SR Standard Deviation Sr Standard Deviation SR Standard Deviation Sr Standard Deviation, % SR Standard Deviation, % Sr Standard Deviation, mm SR Standard Deviation, mm D10, D50, D90 Repeatability STD vs. Average All Brands Data D10, D50, D90 Reproducibility STD vs. Average All Brands Data R² = R² = 676 R² = 115 R² = Type 1 Type 2 Type 3 Type R² = R² = R² = R² = Type 1 Type 2 Type 3 Type Average, mm Average, mm % Retained, Repeatability STD vs. Average All Brands Data % Retained, Reproducibility STD vs. Average All Brands Data R² = Type 1 R² = Type 2 R² = Type 3 R² = Type Type 1 R² = 837 R² = Type 2 Type 3 R² = Type 4 R² = Average, % Average, % 8 b/l, Repeatability STD vs. Average All Brands Data 0.12 b/l, Reproducibility STD vs. Average All Brands Data R² = Type 1 R² = Type 2 R² = Type 3 R² = Type R² = R² = R² = R² = Type 1 Type 2 Type 3 Type Average Average 5 SPHT, Repeatability STD vs. Average All Brands Data 0.10 SPHT, Reproducibility STD vs. Average All Brands Data 4 R² = 6E-05 Type R² = Type 1 3 R² = 416 Type R² = Type R² = R² = Type 3 Type R² = R² = Type 3 Type Average Average 2.5 % Circular, Repeatability STD vs. Average All Brands Data 10.0 % Circular, Reproducibility STD vs. Average All Brands Data 2.0 R² = Type R² = Type R² = Type R² = Type 2 R² = R² = Type 3 Type R² = R² = Type 3 Type Average, % Average, % 2.0 % Non-Circular, Repeatability STD vs. Average All Brands Data 10.0 % Non-Circular, Reproducibility STD vs. Average All Brands Data 1.5 R² = 654 Type R² = Type R² = R² = R² = Type 2 Type 3 Type R² = 17 R² = R² = Type 2 Type 3 Type Average, % Figure Relationsip between average and standard deviation of te glass bead properties measured wit various brands of particle analyzers Average, % 72

80 Table Precision estimates of te glass bead properties Camsizer Various Particle Analyzers Property Material Type Sr d2s-sr SR d2s-sr Sr d2s-sr SR d2s-sr Type D10 (mm) Type Type Type Type D50 (mm) Type Type Type Type D90 (mm) Type Type Type Type Retained, % Type Type Type Type b/l Type Type Type Type SPHT Type Type Type Type Round, % Type Type Type Type Non-Round, % Type Type Type

81 3.5.3 Proposed Precision Estimates for AASHTO PP 74 To update precision estimates of PP 74, te current precision estimates are compared wit te two sets of precisions developed in tis study. A comparison of te existing precision estimates in PP wit te ones developed in tis study based on te Camsizer data, indicates significant improvement in variability of measurements of te size and sape of te pavement maring glass beads. For example, for single-operator, te allowable difference between two replicate percent retained measurements for Type 1 glass beads as canged from 3.8% to 0.97%. For te percent round, te allowable difference corresponding to single-operator canges from 2.8% to 1.2%. Tere are also improvements in multilaboratory precision. For example, for Type 3 glass beads, te d2s for percent retained as canged from 5.9% to 3.0% and te d2s for percent round as improved from 3% to 1.2%. Te improvement in precision estimates could come from several factors: improved guidelines for operation of particle analyzers due to te development of AASHTO PP 74, enancement in te particle analyzer correlation models and software, and improvements in te operators sills. A comparison was also made between te existing precision estimates wit te ones developed in tis study based on te data from various particle analyzers. Te results indicate tat for percent retained, tere are improvements for bot single-operator and multilaboratory precisions wen data are pulled from all particle analyzers. For percent round, wile tere are improvements in single-operator precision, te multilaboratory precision, from various analyzers, are larger tan tose in PP For example, for single operator, te d2s of percent retained for Type 1 as canged from 3.8% to 1.5% and te d2s of percent round as canged from 2.8% to 1.7%. However, for multilaboratory d2s of % round, te allowable difference as canged from 4.5% to 13.6%. Tis migt indicate tat wile various particle analyzers are armonized for size measurements, tey require to use a more unified calibration process for sape measurements. Based on te above observations, it could be concluded tat updating te current precision estimates based on te data from various particle analyzers is not sensible. Te reason is tat various particle analyzers need to establis better correlations between percent round/ non-round and te b/l or SPHT tresold values. Since at current time, all te state DOT and te pavement maring/ glass bead suppliers/producers are using Camsizer for size and roundness measurements of glass beads, increasing te allowable differences between replicate measurements and between laboratory results based on te data from oter particle analyzers, does not seem advantageous to te industry. At a later time, wen particle analyzers ave establised a more unified approac for correlating percent round/non-round data based on b/l and SPHT tresold values, te precious estimates can be reevaluated. Terefore, te precision estimates based on te Camsizer measurements collected in tis study, are suggested as an update for te precision estimates in AASHTO PP Table 3-15 provides te proposed precision estimates for size and sape measurements of pavement maring glass beads. 74

82 Table Proposed Precision Estimates for Percent Retained and Percent circular properties Single-Operator Precision Multilaboratory Precision Type Index and Test Property Standard Deviation (1s) a Acceptable Range of Two Test Results (d2s) a Standard Deviation (1s) a Acceptable Range of Two Test Results (d2s) a % retained: Type 1 Type Type 3 Type Type % Round: Type 1 Type Type 3 Type Type

83 3.6 Revisions to te AASHTO PP 74 In addition to precision estimates, a series of revisions ave been made to AASHTO PP 74 based on te comments received from te users and producers of particle analyzers. Te revisions are purposed to mae AASHTO PP 74 more applicable to various brands of particle analyzers and to improve te language of te standard practice. Te comments and te resulted revisions to te latest version of PP 74 (2013) are as follows: Section (Comment 1) Comment: T and L are te smallest and largest dimensions for 3-D analysis. Revision: T, L, and T/L are replaced by Aspect ratio (widt/lengt) Section (Comment 2) Comment: Does te camera actually capture te maximum dimensions of eac particle or is tere variation due to randomness of te fall? Revision: te largest and smallest dimensions can be measured is replaced by te largest and smallest dimensions can be estimated Section 3.15 (Comment 1) Comment: Multiple layers of all particles is a confusing definition. Revision: multiple orientations of a single particle Section (Comment 2) Comment: NSP is a manufacturer defined term and not internationally agreed Revision: NSP is removed Figure 1 Caption. Comment: Editorial correction of caption for better description of Figure 1 Revision: Scematic diagram of particles as tey fall is replaced by Scematic diagram of a particle as it falls Note 2, Section 3.1.7, and Section Comments: Te requirement for 2 cameras only applies to a specific manufacturer. Revisions (Note 2): Two metods of image focus and capture can be employed. Te resulted images from te two metods are used for 2-D or 3-D analysis of te particles. Te 2-D analysis requires one or more fixed-position adjustable focus camera (s) and te 3-D analysis requires one adjustable position/focus camera. Revision (Section 3.1.7): One is added to te beginning of te section. Revisions (Section 3.1.8): te existing section is replaced wit te following: One or two fixed-position focus cameras (2-D) Te ig-resolution digital cameras provide two different levels of magnifications, te iger magnification to measure te smaller particles, in a smaller frame area, and te lower resolution to measure larger particles in a larger frame area. Tis metod uses a normalization algoritm to put togeter two 76

84 Section distributions of sizes and sapes of particles captured at two different magnifications. Because of tis algoritm te iger-magnification camera does not need to capture all of te small particles. Because of te guide plate tere are no out-of-focus particles. Comment: sould leave open to oter illumination strategies-front ligt, side ligt, etc. Revision: ligt source for strong and omogeneous illumination wen camera(s) capture an image. Section Comment: definition of cord is not needed due to simplifying of te definitions of xf min and xf max in Section and Section 3.2.5). Revision: Definition of cord is deleted. Section Comment: NSP is a manufacturer defined term and is not internationally agreed Revision: NSP is removed from te standard practice Section Comment: Form Factor (FF) is an ISO accepted term for te calculation of 4πA/P 2 referred to as SPHT. Revision: Section is canged to SPHT a sape parameter also referred to as Form Factor (FF), Circularity, or Spericity. SPHT= 4πA/P 2, were A is te measured area, and P is te measured perimeter. For an ideal circle, SPHT is 1; oterwise, it is smaller tan 1. Te tresold value for deciding if a particle is close to a circle or not is approximately Section and Note 3 Comment: move te description of T/L and Note 3 after Section (description of b/l) Revision: Description of T/L and Note 3 are moved to te end of Section 3.2. Section Comment: definition of Feret Diameter can be simplified. Revision: Section is revised as follows: XF or XFe Feret diameter, (xc) a set of diameters (maximum cord) by scanning particles in a defined number of directions (See Figure 2). For a convex particle, te mean Feret diameter (mean value of all directions) is equal to te diameter of a circle wit te same circumference or perimeter. Section Comment: te set of maximum cords is not clear 77

85 Revision: Section as been revised as follows: XF min sortest Feret diameter of te measured set of Feret diameters of a particle projection (for close correlation to sieve analysis - See Figure 2). Section Comment: Wat is te relationsip between xc max and xf max? Revision: xc max, xf max, and l are different terms for te same property. xc max is added to te terms. Section 3.2.4, Section , and Section Comment: Wat is te relationsip between XFe min and Xc min?" Revision: XFe min, xc min, w, and b are different terms of te same property. Tese terms ave been added to Section 3.2.4, Section , and Section for more clarifications. Trougout Section 3.2 Comment: description of terms need to be according to ISO standards as muc as possible Revision: te terms corresponding to Feret Diameter (XFe, XFe min, XFe max) trougout Section 3.2 are replaced by (xf, xf min, xf max) to reflect te conventions in ISO ISO Figure 2 Comment: te terms in Figure 2 and its caption need to be modified to agree wit te ISO accepted terms. Revision: te terms corresponding to Feret Diameter (XFe, XFe min, XFe max) are modified in Figure 2 and its caption to reflect te conventions in ISO (xf, xf min, xf max). Section 4.1. Comment #1: Is 60 images required? Will 50 or 40 frames/sec not wor? Revision #1: at a minimum rate of 60 images/s is replaced by at a rate specific to eac particle analyzer Comment #2: tere are no guaranties tat images of eac particle are captured from different directions Revision #2: from different directions is removed Section 4 and Section 5 Comment: according to ISO, roundness as a specific definition, wic is different from its purpose in tis standard. Revision: Replaced Roundness wit sape trougout te document 78

86 Section 6.1 Comment #1: Migt be better to say 'multiple instruments wit varying measuring strategies may. Revision #1: tis sentence is added Multiple instruments wit varying measuring strategies may be used. Comment #2: 3D measuring system does not require a guide plate Revision #2: 2-D measuring system is added after guide plate. Section (Comment 1) Comment: prepare two specimens te following section of te document does not identify wat is done wit results of te two specimens. Revision: Note 4 as been added to Section 11. Note 4 Testing specimens in duplicates is to ensure tat tey meet single-operator and multilaboratory precision tolerances specified in Section 14. Section (Comment 2) Comment: Table 1 is unnecessarily restrictive and complex. Te reason tat M247 restricts sample size per bead type is tat during manual sieving, sieves wit fine mes sizes are easily overloaded / clogged. Te M247 sample size restriction is not really applicable to te optical analysis because te reason for te restriction is no longer applicable. I suggest getting rid of te size restrictions per bead type. Response: Te weigt/mass of a sample in relationsip wit its main particle size is a good indicator for te amount of particles, wic is an important factor for reproducible measurement wit a poto-optical analyzer. If samples consist of perfectly round particles, te sape measurement will not be influenced by te number of particles (since direction of te camera does not affect te measurement results). However, for particles tat are not all close to a circle, mass and terefore number of particles sould be sufficient to ensure reliable measurement of te sape parameters. Revision: Note 5 as been added to Section : Te mass of a sample in conjunction wit its main particle size is a good indicator of te number of particles. Regulating te mass and terefore te number of particles could be important for repeatable/reproducible measurements wit a particle analyzer. Section (Comment 3) Comment: Required sample weigts is typically 50 grams. I understand tat reduced sample weigts could introduce additional statistical error. For various reasons, can te sample weigt be increased? Revision: Notes 6 and Note 7 ave been added to Section : Note 6 Wile using masses larger tan te minimum masses recommended in Table 1 sould not affect te properties of te glass beads, it would increase te measurement and analysis time. it is recommended to limit te sample mass to eep te measurement time in a reasonable range. Note 7 Te maximum mass of a specimen is macine specific and is determined by te capacity of te dosage funnel of te particle analyzers. 79

87 Section Comment: Record te mass of eac specimen. For wat reason? Te equipment doesn t require it nor does Section 13 Report ave any indication tat te mass of te specimen is on te final report. Revision: Note 11 as been added: Note 11 Altoug mass of a specimen is not required to be reported (Section 13), it is a useful piece of information tat can be recorded as part of te output file. In an event tere is a confusion on te type of glass beads tested, mass of a specimen in conjunction wit te number of particles is a good indicator of te type of te glass beads tested. Section Comment: Terms need to be corrected. Revision: XF min, Xc min, or b replaces XFe min or t Section Comment #1: mae correction to XFe min/xfe max based on ISO convention Revision #1: XFe min/xfe max is replaced by XF min/xf max Comment #2: According to ISO Roundness as a different definition tan used in PP 74. Revision #2: for roundness measurement is replaced by aspect ratio as a sape parameter Section , Section & Section Comment: NSP is not an internationally defined term and according to ISO Roundness as a different definition tan used in PP 74. Revision: NSP is deleted and for roundness measurement is replaced by for sape measurement Section Comments: D10, D50, D90 are not necessary in a sample of mixed bead sizes. Tese decile measurements are possibly useful if one is evaluating a single size of bead. Te decile requirement sould be deleted from section 11 and from section 13 (reporting). Response: D10 and D50 and D90 are indicators for bot normally distributed samples and for multi-peaed poly-dispersed materials to get information about te particle size distribution. Section 13.1 Comment: Section 13 is unnecessarily detailed and does not reflect te needs of te individual DOTs. Revisions: Tis as been added to Section 13.1: According to te needs of individual agencies, te report of te analysis could include some or all of te following information: 80

88 Section Comment: If a state specification is based on % passing, te report sould not ave to include % retained, te extra information is confusing. Revisions: and percent retained is replaced by or percent retained Section Comment: Percent round is not an internationally agreed term Revisions: Section is revised to: Percent of particles close to a circle by b/l or T/L in eac class size and/or te weigted average for te wole sample; Section Comment: Percent round is not an internationally agreed term Revisions: Section is revised to: Percent of particles close to a circle by SPHT parameter in eac class size and/or te weigted average for te wole sample; Section Comment: NSP is not an internationally defined term Revision: NSP is deleted Section Comment: Te report sould indicate instrument model and Mfr. Revisions: Added as Section : Instrument model and manufacturer Section 14 Based on te new data collected in tis study, te precision statement of PP 74 is updated. 81

89 CHAPTER 4 CONCLUSIONS AND RECOMMNEDATIONS Te major goal of tis study was to update te AASHTO PP 74 Standard Practice in support and guide of te users and reflect te state-of-te-art in tecnology and industry for analysis of sape and size of pavement maring glass beads. A summary of findings of te study and recommendations for furter improvement of te practice are discussed as follows: 4.1 Summary 4.2 Findings Tis study was aimed on improving te language of AASHTO PP 74 standard practice to broaden te application of te standard to te computerized optical particle analyzers from various manufacturers. A questioner was prepared and sent to te state DOTs inquiring teir experience wit PP 74. Te questioner was also sent to te producers and suppliers of te pavement maring glass beads and te manufacturers of particle analyzers. Te opinions of te participants on te applicability of PP 74 for use wit teir instruments and teir comments on improving te standard practice was obtained. Twelve states and one supplier provided responses to te questioner. Te questioner and responses are provided in Appendix A. A summary of responses is given in Section Te comments suggested by te participants, are discussed in Section 3.5 and incorporated for te revision of AASHTO PP 74. In te questioner, te willingness of te entities to participate in an interlaboratory study for updating te precision estimates of PP 74 was also requested. 22 laboratories (including users and producers of particle analyzers) agreed to participate in te interlaboratory study. Two participating laboratories of a same producer eac provided two sets of results, one from 2-D and te oter from 3-D analysis of te measurement data. Samples and instructions were prepared for te ILS. Tree replicates of 4 types of glass beads totaling in 180 samples were prepared at te Gilson Facility. Instructions included list of size and sape properties to be measured. Instructions and te samples were sent to te 22 participating laboratories, were tey measured various properties of te glass beads according to te instructions. Among te laboratories, 14 were equipped wit Retsc Camsizer. Oter laboratories were equipped wit te following: tree wit Tyler CPA, two wit Microtrac FPA, one wit Malvern Mastersizer, one wit Canty particle analyzer, and one wit Sympatec QICPIC. Laboratories only provided data for tose properties tat tey were mostly familiar and were routinely measuring using teir instruments. Te data for six of te mostly measured properties, wic were provided by 5 or more laboratories, were analyzed for updating PP 74 precision statements and for comparison of te statistics of te properties measured by various particle analyzers. Analysis of te data collected in tis study ave sown several facts regarding te data collected by various brands of particle analyzers. Te findings are as follows: 1. Te precision of size and sape measurements of glass bead properties ave significantly improved since te development of AASHTO PP 74 in Te allowable singleoperator and multilaboratory variability of PP74-13 as significantly lowered based on te Camsizer data collected in tis study. Tis progress could indicate te importance of AASHTO PP 74 Standard Practice in standardization of computerized optical size and 82

90 sape measurements of pavement glass beads. Te improvements in precision of measurements could also be due to upgrade in ardware/software of te Camsizer particle analyzer and from increased experience of te operators. 2. Tere are significant differences in te variability of te sape measurements performed by Camsizer and tose performed by various manufacturers. Tis is especially true for multilaboratory precision. Te possible reasons for te differences could be 1) ligt intensity and resolution of cameras used wit eac device, 2) tresold values used for analysis of particle images. 3) matematical fit for te correlation between optical size measurements and sieve/circularity measurements. 4) te number of toucing/overlapping particle projections 4) use of different analysis metods for processing te touced/overlapping particles. 3. Te percent circular data based on b/l ratio is significantly less variable tan percent noncircular based on SPHT parameter. It could be concluded tat b/l parameter is more reliable tan te SPHT parameter in determining ow close particles are to a circle. 4. Te recommended tresold value of 0.85 for b/l in AASHTO PP74, for determining ow close particles are to a circle, were selected for te Camsizer settings. However, it migt not be appropriate for oter brands of particle analyzers due to te differences in pysical resolution of teir cameras. 5. Similarly, te tresold value of 0.93 for te SPHT parameter was selected for te Camsizer settings. Te tresold for SPHT migt be different for oter particle analyzers. 4.3 Recommendations Based on te te findings of tis study, te following recommendations are made: 1. A series of canges to AASHTO PP74, based on te comments received from users and producers of pavement glass beads and manufacturers of particle analyzers, ave been prepared. It is recommended tat tese canges, wic expand te applicability of PP 74 for use wit various particle analyzers to be incorporated to te standard practice. 2. It is recommended tat te existing precision statement in AASHTO PP 74 be revised using te precision estimates developed in tis study based on te data collected by te Camsizer particle analyzer. Te reason is tat tere are significant improvements in te precision estimates of percent retained and percent circular wen only Camsizer data were analyzed. 3. Tere is an immense need to armonize te measurements from various particle analyzers to reduce te multilaboratory precision. It is recommended tat a researc study to be planned to ensure use of consistent calibration metods, correct correlation models, constant filtering tecniques, and correct tresold values for maing decision on good and bad particles. 4. Te obscuration of particles could a major reason for te variability of sape measurements. It is recommended to compare various instruments in terms of te number of obscured particles captured and te metods eac instrument use to control 83

91 obscuration. Tis would elp in identifying sources of te problem witin eac instrument and to develop practices to reduce obscuration. 6. A study for defining more reliable parameters for determining Non-circular particles is recommended. 7. Tresold values tat provide te same size and sape measurements sould be establised for different particle analyzer units. Te adjustment of tresold values is equipment dependent and may be controlled automatically. It is recommended tat in a coordinated researc study, te tresold values tat result in correct and precise measurement of size and sape of standard glass beads to be identified. 8. Translucency or transparency of te glass beads is an important property for reflecting ligt. Several state DOTs ave indicated te need for tese properties as part of measuring bead qualities. A researc study could address te transparency measurements. 84

92 REFERENCES 1. Azari, H. and Garboczi, E., Optical Sizing and Roundness Determination of Glass Beads Used in Traffic Marings, NCHRP Web-Only Document 346, 2010, ttp://onlinepubs.trb.org/onlinepubs/ncrp/ncrp_w346.pdf 2. AASHTO PP74, Determination of Size and Sape of Glass Beads Used in Traffic Marings by Means of Computerized Optical Metod, Standard Specifications for Transportation Materials and Metods of Sampling and Testing, Tirty Sixt Edition, American Association of State Higway and Transportation Officials, Wasington, DC. 3. Retsc Tecnology GmbH, ttp:// 4. Microtrac, ttp:// 5. Haver-Tyler CPA, cpa.wstyler.com/particle/analyzers 6. ttp://wstyler.com/product/cpa-particle-size-and-sape-analyzer/ 7. Sympatec, 8. Malvern, / 9. Canty, ISO , Representation of Results of Particle Size Analysis Part 6: Descriptive and Quantitative Representation of Particle Sape and Morpology 11. ASTM E691, Standard Practice for Conducting an Interlaboratory Study to Determine te Precision of a Test Metod, Vol , Quality Control Standards, West Consoocen, PA, ASTM, Designation C 670, Standard Practice for Preparing Precision and Bias Statements for Test Metods for Construction Materials, Annual Boo of ASTM Standards, Volume 4.02, ASTM, West Consoocen, PA,

93 APPENDIX A- SURVEY RESULTS AND ANALYSIS 86

94 SOM Questionnaire - TS 4c Determination of Size and Roundness of Glass Beads Used in Traffic Marings by Means of Computerized Optical Metod Wednesday, September 16,

95 Executive Summary Tis report contains a detailed statistical analysis of te results to te survey titled SOM Questionnaire - TS 4c Determination of Size and Roundness of Glass Beads Used in Traffic Marings by Means of Computerized Optical Metod. Te results analysis includes answers from all respondents wo too te survey in te 50 day period from Monday, July 27, 2015 to Tuesday, September 15, completed responses were received to te survey during tis time. 2

96 Survey Results & Analysis Name of Your Organization: Bead Type Oter Utilizes Computerized Optical Equipment Alabama Department of Transportation X X X Arizona DOT X X X X Aransas State Higway & Transportation Dept. X California Dept. of Transportation X ConnDOT Delaware Dept of Transportation X X FL Dept of Transportation X X YES Georgia Department of Transportation X X X X Illinois Department of Transportation X YES Kansas Department of Transportation X X X YES Louisiana DOTD X X X X Maine Department of Transportation X Minnesota Department of Transportation X YES 3

97 Mississippi Department of Transportation X Missouri Department of Transportation X X YES Montana Department of Transportation X X YES NC Department of Transportation Nevada Department of Transportation X X NYSDOT X X Oio Department of Transportation X X YES OK Dept of Transportation X X YES Rode Island Department of Transportation - Materials and Quality Assurance Sout Carolina Department of Transportation Sout Daota DOT X X X YES X Texas Dept of Transportation X X YES Uta Department of Transportation X Virginia DOT X Wasington State DOT X West Virginia Dept of Transportation X 4

98 Potters Industries LLC X X X X X X YES 5

99 Oter Responses: AZ AZ 708 CA CA State Specification # GA Are type 5 still manufactured? IL IDOT Specification Article KS NC NV Blended Beads We ave our own spec were te gradation NDOT Specification OH Oio DOT specification item C, D. RI RI Std Type II grad is sim to Type 1 UT WV M 247 wit specialized Gradation Based on Warranty Potters TTB 1325 D, 6

100 7

101 3) Wat standard do you follow for obtaining representative glass bead samples from a bul container of glass beads? Please list te standard designation or provide a copy of te procedure if not a publised standard. AL AASHTO M AR AASHTO T 346 AZ CA CT DE FL GA IL Sample Tief California Test #406 for field samples. See weblin below; ttp:// Not specified. Manufacturer provides representative sample of batc. T346 ASTM D1214, ASTM D346 All of te long line striping is contracted out by maintenance. We rarely sample except for Qualified Products List. Latest applicable ASTM or AASHTO Std. KS AASHTO TP 97 LA ME MS MN We use AASHTO T 346 section 4 for barrels or oter bul containers. We also use a splitter to ensure a representative sample for bags sent to te lab. No standard metod - sample is scooped from container. Mississippi DOT, Materials Division Inspection Testing and Certification Manual. Lin: ttp://sp.mdot.ms.gov/materials/manuals/inspection,%20testing,%20and%20certification%20manual/entire%20manual%20witout%20ap pendices.pdf unnown MO We use te sampling metod in AASHTO T346. 8

102 MT We do not tae samples for acceptance from bul containers on construction jobs. We sample at te point of delivery, i.e. from te nozzle or bead gun. Our maintenance forces follow ASTM B 215 for sampling wen comparing samples from a tief to samples at te nozzle. 9

103 NC NV NY OH Te witnessing of sampling sall be performed by an approved representative of te Department of Transportation. Te sample(s) sall be of sufficient amount to perform all tests described in te Department of Transportation Specification for drop-on and intermix glass beads. Tree bul containers are selected (company picing seet) at random from te production run. Te Inspector also signs and dates te pic seet. A tieving rod sall be inserted into te container five times. (Ref: Kansas TS modified by NCDOT) Once in eac corner and once in te center and sall be pused troug as far as possible. Te five tievings are ten split using a 1:1 splitter to approximately TWO one quart amounts. Te witnessing party ten carefully seals eac container wit te NCDOT Sealing tape or similar metods. Designate Sample A as te company retain. Te company submits Sample B to te Independent Lab address above. NDOT taes traffic bead samples at te last point of introduction. Bead samples are taen from te striping truc bead gun just prior to application. Standard split sampling of 60 lbs. sample size. Oio DOT Supplemental Spec: ttp:// 08_ _for_2013.PDF OK AASHTO M 247 RI SC SD TX Do not sample bul containers. We do not currently sample from bul containers. Tese sipments are accepted based on manufacturer's certification but I would be interested in learning wat oters are doing. SDDOT does not sample from bul containers. see attaced lin: ttp:// we sample from individual bags per te following: ttp://ftp.dot.state.tx.us/pub/txdot-info/cst/tms/800-b_series/pdfs/ctm830.pdf we do not typically sample from bul containers, altoug we do get samples from te production line UT See MOI 932 section 02 VA AASHTO T 346 WA We relied on AASHTO M version, but te updated 2013 version eliminated te sampling and testing language so tere are no current procedures. 10

104 WV NA. All Traffic Paint and Glass Beads are andled tru warranty PO contracts. Potters AASHTO T ) Do you use computerized optical equipment for testing gradation and roundness of glass beads? 11

105 12

106 5) Wat practice do you follow for determining gradation and roundness of glass beads using computerized optical equipment? Please list te standard designation or provide a copy of te practice if not a publised standard. Wat practice do you follow for determining gradation and roundness of glass beads using computerized optical equipment? Please list te standard designation or provide a copy of te practice if not a publised standard. FL IL AASHTO PP74 AASHTO PP74 KS ASTM D 1214/ASTM D 1155 MN AASHTO - PP 74 13

107 MO AASHTO PP MT PP 74 OH Retsc Tecnology Camsizer - Particle Analyzer Programmed for OHioDOT glass beads. OK AASHTO M 247 SC TX N/A Gradation is based on te 2D projected area of bead to te CCD sensor, te total pixel count is converted to sieve size. Roundness is calculated based on aspect ration b/l and spereicity Area/circumference. Potters AASHTO T

108 6) Wat mass of eac glass bead type do you use for particle size and roundness measurements wit your computerized optical equipment? Wat mass of eac glass bead type do you use for particle size and roundness measurements wit your computerized optical equipment? FL IL KS MN MO We use te mass specified in table 1 of PP grams for small gradation 100 grams for large gradation grams for all types Type 1-50g Type 1 50 grams Type grams MT As described in PP 74, 50 g for type 1 and 70 g for type 2. OH SC Around 50 grams grams TX 100 grams for type 2 and 150 grams for type 3 Potters 50g +/- 20% 15

109 7) Have you followed AASHTO PP 74, Determination of Size and Roundness of Glass Beads Used in Traffic Marings by Means of Computerized Optical Metod, for routine evaluation of glass beads? 16